Biomimetics and Marine Technology - Marine Technology Society
Biomimetics and Marine Technology - Marine Technology Society
Biomimetics and Marine Technology - Marine Technology Society
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Journal<br />
The International, Interdisciplinary <strong>Society</strong> Devoted to Ocean <strong>and</strong> <strong>Marine</strong> Engineering, Science, <strong>and</strong> Policy<br />
Volume 45 Number 4 July/August 2011<br />
<strong>Biomimetics</strong> <strong>and</strong> <strong>Marine</strong> <strong>Technology</strong>
INTRODUCING<br />
The Next Generation<br />
in Underwater Acoustic Modems<br />
The ATM-900 Series Modems<br />
from Teledyne Benthos<br />
• Advanced Data Logger Capabilities<br />
– Increased from 704K to 6MB (flash file format)<br />
– Searchable by record number <strong>and</strong> time<br />
• Dual Serial Port – Interface with more than one<br />
sensor at a time (optional)<br />
• Integrated real-time clock for data time<br />
stamping<br />
ATM-900 Series<br />
• New Comm<strong>and</strong> Structure & Modem Parameters<br />
• Pinger Mode<br />
• Arbitrary Waveform Generator (optional)<br />
Available April 2011<br />
See us at Ocean Business - Booth #S1.<br />
A Member of Teledyne <strong>Marine</strong><br />
49 Edgerton Drive, North Falmouth, MA 02556 USA<br />
T: +1 508.563.1000 • F: +1 508.563.6444<br />
E: benthos@teledyne.com<br />
www.benthos.com
Volume 45, Number 4, July/August 2011<br />
<strong>Biomimetics</strong> <strong>and</strong> <strong>Marine</strong> <strong>Technology</strong><br />
Guest Editors: Frank E. Fish <strong>and</strong> Donna M. Kocak<br />
Turn to page 7 for a key to the cover images.<br />
Text: SPi<br />
Cover <strong>and</strong> Graphics:<br />
Michele A. Danoff, Graphics By Design<br />
The <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal<br />
(ISSN 0025-3324) is published by the <strong>Marine</strong> <strong>Technology</strong><br />
<strong>Society</strong>, Inc., 5565 Sterrett Place, Suite 108, Columbia,<br />
MD 21044.<br />
MTS members can purchase the printed Journal for $27<br />
domestic <strong>and</strong> $50 (plus $50 S&H) international.<br />
Non-members <strong>and</strong> library subscriptions are $420 online only,<br />
$124 print—domestic, $140 (plus $50 S&H) print—<br />
international, $435 print <strong>and</strong> online (worldwide);<br />
Single-issue (hardcopy) is $20 plus $7.50 S&H (domestic),<br />
$24.50 S&H (international); Pay-per-view (worldwide):<br />
$15/article. Postage for periodicals is paid at Columbia, MD,<br />
<strong>and</strong> additional mailing offices.<br />
POSTMASTER:<br />
Please send address changes to:<br />
<strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal<br />
5565 Sterrett Place<br />
Suite 108<br />
Columbia, Maryl<strong>and</strong> 21044<br />
Copyright © 2011 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong>, Inc.<br />
In This Issue<br />
8<br />
<strong>Biomimetics</strong> <strong>and</strong> <strong>Marine</strong> <strong>Technology</strong>:<br />
An Introduction<br />
Frank E. Fish, Donna M. Kocak<br />
14<br />
Biomimicking <strong>Marine</strong> Mechanisms <strong>and</strong><br />
Organizational Principles<br />
Commentary by Yoseph Bar-Cohen<br />
16<br />
Sink <strong>and</strong> Swim: Clues From Nature for<br />
Aquatic Robotics<br />
Commentary by Jeannette Yen<br />
19<br />
Developing Bioinspired Autonomous<br />
Systems<br />
Commentary by Thomas M. McKenna<br />
24<br />
GhostSwimmer AUV: Applying<br />
<strong>Biomimetics</strong> to Underwater Robotics for<br />
Achievement of Tactical Relevance<br />
Commentary by Michael Rufo,<br />
Mark Smithers<br />
31<br />
Autonomous Robotic Fish as Mobile<br />
Sensor Platforms: Challenges <strong>and</strong><br />
Potential Solutions<br />
Xiaobo Tan<br />
41<br />
Robotic Models for Studying Undulatory<br />
Locomotion in Fishes<br />
George V. Lauder, Jeanette Lim,<br />
Ryan Shelton, Chuck Witt, Erik Anderson,<br />
James L. Tangorra<br />
56<br />
Thrust Production in Highly Flexible<br />
Pectoral Fins: A Computational Dissection<br />
Srinivas Ramakrishnan, Meliha Bozkurttas,<br />
Rajat Mittal, George V. Lauder<br />
65<br />
Learning From the Fins of Ray-Finned<br />
Fish for the Propulsors of Unmanned<br />
Undersea Vehicles<br />
James L. Tangorra, Timo Gericke,<br />
George V. Lauder<br />
74<br />
Bioinspired Design Process for an<br />
Underwater Flying <strong>and</strong> Hovering Vehicle<br />
Jason D. Geder, John S. Palmisano,<br />
Ravi Ramamurti, Marius Pruessner,<br />
Banahalli Ratna, William C. S<strong>and</strong>berg<br />
83<br />
A Twistable Ionic Polymer-Metal<br />
Composite Artificial Muscle for<br />
<strong>Marine</strong> Applications<br />
Kwang J. Kim, David Pugal,<br />
Kam K. Leang<br />
99<br />
Batoid Fishes: Inspiration for the Next<br />
Generation of Underwater Robots<br />
Keith W. Moored, Frank E. Fish,<br />
Trevor H. Kemp, Hilary Bart-Smith<br />
110<br />
Bioinspired Propulsion Mechanisms<br />
Based on Manta Ray Locomotion<br />
Keith W. Moored, Peter A. Dewey,<br />
Megan C. Leftwich, Hilary Bart-Smith,<br />
Alex<strong>and</strong>er J. Smits<br />
119<br />
Inspired by Sharks: A Biomimetic<br />
Skeleton for the Flapping, Propulsive<br />
Tail of an Aquatic Robot<br />
John H. Long, Jr., Tom Koob,<br />
Justin Schaefer, Adam Summers,<br />
Kurt Bantilan, Sindre Grotmol,<br />
Marianne Porter
In This Issue<br />
130<br />
Lateral-Line-Inspired Sensor Arrays for<br />
Navigation <strong>and</strong> Object Identification<br />
Vicente I. Fern<strong>and</strong>ez, Audrey Maertens,<br />
Frank M. Yaul, Jason Dahl,<br />
Jeffrey H. Lang, Michael S. Triantafyllou<br />
147<br />
A Conserved Neural Circuit-Based<br />
Architecture for Ambulatory <strong>and</strong><br />
Undulatory Biomimetic Robots<br />
Joseph Ayers, Anthony Westphal,<br />
Daniel Blustein<br />
153<br />
A Hybrid Class Underwater Vehicle:<br />
Bioinspired Propulsion, Embedded<br />
System, <strong>and</strong> Acoustic Communication<br />
<strong>and</strong> Localization System<br />
Michael Krieg, Peter Klein,<br />
Robert Hodgkinson, Kamran Mohseni<br />
165<br />
Modeling of Artificial Aurelia aurita<br />
Bell Deformation<br />
Keyur B. Joshi, Alex Villanueva,<br />
Colin F. Smith, Shashank Priya<br />
181<br />
Swimming <strong>and</strong> Walking of an<br />
Amphibious Robot With Fin Actuators<br />
Naomi Kato<br />
198<br />
<strong>Marine</strong> Applications of the Biomimetic<br />
Humpback Whale Flipper<br />
Frank E. Fish, Paul W. Weber,<br />
Mark M. Murray, Laurens E. Howle<br />
208<br />
Shark Skin Separation Control<br />
Mechanisms<br />
Amy Lang, Philip Motta,<br />
Maria Laura Habegger, Robert Hueter,<br />
Farhana Afroz<br />
216<br />
Can Biomimicry <strong>and</strong> Bioinspiration<br />
Provide Solutions for Fouling Control<br />
Emily Ralston, Geoffrey Swain<br />
228<br />
BOOK REVIEW: Sex, Drugs, <strong>and</strong> Sea<br />
Slime: The Oceans’ Oddest Creatures<br />
<strong>and</strong> Why They Matter<br />
by Ellen Prager<br />
Reviewed by Jason Goldberg
Deep Tracks!<br />
Teledyne RD Instruments’ new PAVS150 DVL keeps your<br />
vehicle on course up to 500 meters above the seafloor.<br />
The 150 kHz Phased Array Velocity Sensor (PAVS150)<br />
Doppler Velocity Log delivers over 2x the range of our<br />
industry-st<strong>and</strong>ard Workhorse Navigator. Now that’s<br />
deep tracks!<br />
The PAVS150’s 1000M depth rating, combined with its<br />
increased range allows you to submerge your manned<br />
or unmanned vehicle sooner <strong>and</strong> transit longer, saving<br />
you time, money, <strong>and</strong> peace of mind.<br />
Stay the course with PAVS!<br />
Full details at:<br />
http://www.rdinstruments.com/PAVS.aspx<br />
Poway, CA USA • Tel. +1-858-842-2600<br />
www.rdinstruments.com
Learn how MTS can help your company<br />
reach the technology leaders <strong>and</strong> innovators<br />
in the marine technology industry.<br />
Opportunities include advertising in Currents, the MTS Journal,<br />
sponsorships, online <strong>and</strong> custom-tailored programs.<br />
MTS —<br />
here for the marine technology community,<br />
here for you!<br />
Email Mary Beth Loutinsky at<br />
mbloutinsky@gmail.com or<br />
call 703-629-3810
<strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Officers<br />
BOARD OF DIRECTORS<br />
President<br />
Jerry Boatman<br />
QinetiQ North America – <strong>Technology</strong> Solutions<br />
Group<br />
President-elect<br />
Drew Michel<br />
ROV Technologies, Inc.<br />
Immediate Past President<br />
Elizabeth Corbin<br />
VP—Section Affairs<br />
Lisa Medeiros<br />
OceanWorks International<br />
VP—Education <strong>and</strong> Research<br />
Jill Z<strong>and</strong>e<br />
MATE Center<br />
VP—Industry <strong>and</strong> <strong>Technology</strong><br />
Jerry C. Wilson<br />
Fugro Pelagos, Inc.<br />
VP—Publications<br />
Karin Lynn<br />
Treasurer <strong>and</strong> VP—Budget <strong>and</strong> Finance<br />
Debra Kill<br />
International Submarine Engineering<br />
VP—Government <strong>and</strong> Public Affairs<br />
Justin Manley<br />
Liquid Robotics<br />
SECTIONS<br />
Canadian Maritime<br />
Vacant<br />
Florida<br />
Vacant<br />
Gulf Coast<br />
Laurie Jugan<br />
Consultant<br />
Hampton Roads<br />
Raymond Toll<br />
SAIC<br />
Hawaii<br />
Stewart Burley<br />
Strategic Theories Unlimited<br />
Houston<br />
Robert Keith<br />
Phoenix International Holdings, Inc.<br />
Japan<br />
Prof. Toshitsugu Sakou<br />
Tokai University<br />
Monterey<br />
Jill Z<strong>and</strong>e<br />
MATE Center<br />
New Engl<strong>and</strong><br />
Chris Jakubiak<br />
UMASS Dartmouth-SMAST<br />
Newfoundl<strong>and</strong> <strong>and</strong> Labrador<br />
Bill O’Keefe<br />
Surmount Technologies, Inc.<br />
Oregon<br />
TBD<br />
Puget Sound<br />
Fritz Stahr<br />
University of Washington<br />
San Diego<br />
Scott Mau<br />
Southwest Fisheries Sciences Center<br />
South Korea<br />
Dr. Seok Won Hong<br />
Maritime & Ocean Engineering Research Inst.<br />
(MOERI/KORDI)<br />
Washington, D.C.<br />
Brent Evers<br />
Hadal Technologies, Inc.<br />
PROFESSIONAL COMMITTEES<br />
Industry <strong>and</strong> <strong>Technology</strong><br />
Buoy <strong>Technology</strong><br />
Dr. Walter Paul<br />
Woods Hole Oceanographic Institution<br />
Cables <strong>and</strong> Connectors<br />
Helmut H. Portmann<br />
National Data Buoy Center<br />
Deepwater Field Development <strong>Technology</strong><br />
Dr. Benton Baugh<br />
Radoil, Inc.<br />
Diving<br />
David C. Berry<br />
Subsea Construction <strong>and</strong> Diving Consultant<br />
Dynamic Positioning<br />
Howard Shatto<br />
Shatto Engineering<br />
Manned Underwater Vehicles<br />
William Kohnen<br />
SEAmagine Hydrospace Corporation<br />
Moorings<br />
Jack Rowley<br />
SAIC<br />
Oceanographic Instrumentation<br />
Dr. Jim Irish<br />
University of New Hampshire<br />
Offshore Structures<br />
Dr. Peter W. Marshall<br />
MHP Systems Engineering<br />
Remotely Operated Vehicles<br />
Drew Michel<br />
ROV Technologies, Inc.<br />
Renewable Energy<br />
Rich Chwaszczewski<br />
SAIC<br />
Ropes <strong>and</strong> Tension Members<br />
Evan Zimmerman<br />
Delmar Systems, Inc.<br />
Seafloor Engineering<br />
Herb Herrmann<br />
Naval Seafloor Cable Protection Office<br />
Underwater Imaging<br />
Dr. Fraser Dalgleish<br />
Harbor Branch Oceanographic Institute<br />
Unmanned Maritime Vehicles<br />
Rafael M<strong>and</strong>ujano<br />
Vehicle Control Technologies, Inc.<br />
Education <strong>and</strong> Research<br />
<strong>Marine</strong> Archaeology<br />
Dan Warren<br />
C & C Technologies<br />
<strong>Marine</strong> Education<br />
Erica Moulton<br />
MATE Center<br />
<strong>Marine</strong> Geodetic Information Systems<br />
Dave Zilkoski<br />
NOAA<br />
<strong>Marine</strong> Materials<br />
Vacant<br />
Ocean Exploration<br />
Guillermo Söhnlein<br />
OceanGate<br />
Physical Oceanography/Meteorology<br />
Dr. Richard L. Crout<br />
National Data Buoy Center<br />
Remote Sensing<br />
Herb Ripley<br />
Hyperspectral Imaging Limited<br />
Government <strong>and</strong> Public Affairs<br />
<strong>Marine</strong> Law <strong>and</strong> Policy<br />
Montserrat Gorina-Ysern<br />
Healthy Children–Healthy Oceans Foundation<br />
<strong>Marine</strong> Mineral Resources<br />
Dr. John C. Wiltshire<br />
University of Hawaii<br />
<strong>Marine</strong> Security<br />
Dallas Meggitt<br />
Sound & Sea <strong>Technology</strong><br />
Ocean Economic Potential<br />
James Marsh<br />
University of Hawaii<br />
Ocean Observing Systems<br />
Donna Kocak<br />
HARRIS CapRock Communications<br />
Ocean Pollution<br />
Jacob Sobin<br />
NOAA Coastal Services Center<br />
STUDENT SECTIONS<br />
Duke University<br />
Counselor: Douglas Nowacek, Ph.D.<br />
Florida Atlantic University<br />
Counselor: Douglas A. Briggs, Ph.D.<br />
Florida Institute of <strong>Technology</strong><br />
Counselor: Stephen Wood, Ph.D., P.E.<br />
Long Beach City College<br />
Counselor: Scott Fraser<br />
Massachusetts Institute of <strong>Technology</strong><br />
Counselor: Alex<strong>and</strong>ra Techet, Ph.D.<br />
Monterey Peninsula College/Hartnell College<br />
Counselor: Jeremy R. Hertzberg<br />
Texas A&M University—College Station<br />
Counselor: Patrick Lynett<br />
Texas A&M—Corpus Christi<br />
Counselor: Lea-Der Chen, Ph.D.<br />
Texas A&M University—Galveston<br />
Counselor: Frank Warnakula, Ph.D.<br />
United States Naval Academy<br />
Counselors: Capt. Joseph T. Arcano<br />
(USN Ret), Ph.D.<br />
Cmdr. David J. Robillard<br />
University of Hawaii<br />
Counselor: R. Cengiz Ertekin, Ph.D.<br />
University of Houston<br />
Counselors: Raresh Pascali, P.E.,<br />
Chuck Richards<br />
University of North Carolina—Charlotte<br />
Counselor: James Conrad, Ph.D.<br />
University of Southern Mississippi<br />
Counselor: Stephen Howden, Ph.D.<br />
Webb Institute<br />
Counselor: Matthew Werner<br />
HONORARY MEMBERS<br />
†Robert B. Abel<br />
†Charles H. Bussmann<br />
John C. Calhoun, Jr.<br />
John P. Craven<br />
†Paul M. Fye<br />
David S. Potter<br />
†Athelstan Spilhaus<br />
†E. C. Stephan<br />
†Allyn C. Vine<br />
†James H. Wakelin, Jr.<br />
†deceased
Editorial Board<br />
Brian Bingham, Ph.D.<br />
Editor<br />
University of Hawaii at Manoa<br />
Corey Jaskolski<br />
Hydro Technologies<br />
Donna Kocak<br />
HARRIS CapRock Communications<br />
Scott Kraus, Ph.D.<br />
New Engl<strong>and</strong> Aquarium<br />
Dhugal Lindsay, Ph.D.<br />
Japan Agency for <strong>Marine</strong>-Earth Science<br />
& <strong>Technology</strong><br />
Justin Manley<br />
Liquid Robotics<br />
Stephanie Showalter<br />
National Sea Grant Law Center<br />
Jason Stanley<br />
Schilling Robotics<br />
Edith Widder, Ph.D.<br />
Ocean Research <strong>and</strong> Conservation<br />
Association<br />
Jill Z<strong>and</strong>e<br />
MATE Center<br />
Editorial<br />
Karin Lynn<br />
VP of Publications<br />
Brian Bingham, Ph.D.<br />
Editor<br />
Amy Morgante<br />
Managing Editor<br />
Administration<br />
Jerry Boatman<br />
President<br />
Richard Lawson<br />
Executive Director<br />
Jeanne Glover<br />
Membership <strong>and</strong> Marketing Manager<br />
Michael Hall<br />
Member Groups Manager<br />
Chris Barrett<br />
Director of Professional Development<br />
<strong>and</strong> Meetings<br />
Suzanne Voelker<br />
Subscription Manager<br />
The <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> is<br />
a not-for-profit, international professional<br />
society. Established in 1963, the <strong>Society</strong>’s<br />
mission is to promote the exchange of<br />
information in ocean <strong>and</strong> marine engineering,<br />
technology, science, <strong>and</strong> policy.<br />
Please send all correspondence to:<br />
The <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong><br />
5565 Sterrett Place, Suite 108<br />
Columbia, MD 21044<br />
(410) 884-5330 Tel.<br />
(410) 884-9060 FAX<br />
MTS Journal: morganteeditorial@verizon.net<br />
Publications: publications@mtsociety.org<br />
Membership: Jeanne.Glover@mtsociety.org<br />
Programs: Michael.Hall@mtsociety.org<br />
Director: Rich.Lawson@mtsociety.org<br />
Online: www.mtsociety.org<br />
MEMBERSHIP INFORMATION<br />
may be obtained by contacting the <strong>Marine</strong><br />
<strong>Technology</strong> <strong>Society</strong>. Benefits include:<br />
■ Free subscription to the online <strong>Marine</strong><br />
<strong>Technology</strong> <strong>Society</strong> Journal, with highly<br />
reduced rates for the paper version<br />
■ Free subscription to the bimonthly news -<br />
letter, Currents, covering events, business<br />
news, science <strong>and</strong> technology, <strong>and</strong><br />
people in marine technology<br />
■ Member discounts on all MTS publications<br />
■ Reduced registration rates to all MTS <strong>and</strong><br />
MTS-sponsored conferences <strong>and</strong> workshops<br />
■ Member-only access to an expansive Job<br />
Bank <strong>and</strong> Member Directory<br />
■ Reduced advertising rates in MTS<br />
publications<br />
■ National recognition through our Awards<br />
Program<br />
Individual dues are $75 per year. Life<br />
membership is available for a one-time<br />
fee of $1,000. Patron, Student, Emeritus,<br />
Institutional, Business, <strong>and</strong> Corporate<br />
memberships are also available.<br />
ADVERTISING<br />
Advertising is accepted by the <strong>Marine</strong><br />
Tech nology <strong>Society</strong> Journal. For more information<br />
on MTS advertising <strong>and</strong> policy,<br />
please contact Mary Beth Loutinsky,<br />
mbloutinsky@att.net.<br />
COPYRIGHT<br />
Copyright © 2011 by the <strong>Marine</strong> <strong>Technology</strong><br />
<strong>Society</strong>, Inc. Authorization to photocopy<br />
items for internal or personal use, or the<br />
internal or personal use of specific clients, is<br />
granted by the <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong>,<br />
provided that the base fee of $1.00 per copy,<br />
plus .20 per page is paid directly to Copyright<br />
Clearance Center, 222 Rosewood Dr.,<br />
Danvers, MA 01923.<br />
For those organizations that have been<br />
granted a photocopy license by CCC, a separate<br />
sys tem of payment has been arranged.<br />
The fee code for users of the Transactional<br />
Reporting Service is 0025-3324/89 $1.00 +<br />
.20. Papers by U.S Government employees<br />
are declared works of the U.S. Government<br />
<strong>and</strong> are therefore in the public domain.<br />
The <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> cannot be<br />
held responsible for the opinions given <strong>and</strong><br />
the statements made in any of the articles<br />
published.<br />
ABSTRACTS<br />
MTS Journal article abstracts, if available,<br />
can be accessed for free at http://www.<br />
ingentaconnect.com/content/mts/mtsj.<br />
Print <strong>and</strong> electronic abstracts of MTS Journal<br />
articles are also available through GeoRef , Aquatic Sciences <strong>and</strong><br />
Fisheries Abstracts, published by Cambridge<br />
Scientific Abstracts , <strong>and</strong> Geobase’s<br />
Oceanbase published by Elsevier Science.<br />
CONTRIBUTORS<br />
Contributors can obtain an information <strong>and</strong><br />
style sheet by contacting the managing editor.<br />
Sub missions that are relevant to the concerns<br />
of the <strong>Society</strong> are welcome. All papers are subjected<br />
to a stringent review procedure directed<br />
by the editor <strong>and</strong> the editorial board. The<br />
Journal focuses on technical material that may<br />
not otherwise be available, <strong>and</strong> thus technical<br />
papers <strong>and</strong> notes that have not been published<br />
previously are given priority. General commentaries<br />
are also accepted <strong>and</strong> are subject to review<br />
<strong>and</strong> approval by the editorial board.<br />
6 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
Key to Cover Images<br />
Front cover: GhostSwimmer AUVs from Boston<br />
Engineering Advanced Systems Group <strong>and</strong> Olin College<br />
Intelligent Vehicle Lab. Image courtesy of HARRIS CapRock<br />
Communication; underwater photo courtesy of Dr. Tamara Frank.<br />
Back cover: (Image courtesy of HARRIS CapRock Communications)<br />
1. Propeller with tubercles (propeller courtesy of Laurens Howls<br />
<strong>and</strong> morphing courtesy of John A. Lever)<br />
2. CephaloBot prototype hybrid vehicle<br />
3. Robotic Turtle, “RT-I”<br />
4. Mantabot<br />
5. Artificial Aurelia aurita bell deformation (photo courtesy of<br />
Alex Villanueva, CIMSS, Virginia Tech)<br />
6. Autonomous robotic fish (photo courtesy of Xiaobo Tan)<br />
7. Bluegill Sunfish (Lepomis macrochirus) with robotic pectoral<br />
fins (photo courtesy of George Lauder <strong>and</strong> James Tangorra)<br />
8. Four-fin vehicle based on Bird wrasse (Gomphosus varius)<br />
9. Tadro4 modeled after living electric ray Narcine<br />
(photo courtesy of Dr. Steve Kajiura)<br />
10. Lobster-based robot (robot photo courtesy of Brian Tucker,<br />
Bresnahan Photography; lobster photo courtesy of<br />
Dr. Tamara Frank)
INTRODUCTION<br />
<strong>Biomimetics</strong> <strong>and</strong> <strong>Marine</strong> <strong>Technology</strong>:<br />
An Introduction<br />
Frank E. Fish<br />
West Chester University<br />
Donna M. Kocak<br />
HARRIS CapRock Communications<br />
I<br />
nspiration for the development of new technologies is at the heart of the biomimetic<br />
approach. As there is a wide diversity of biological forms, particular attributes can be targeted<br />
that provide innovative solutions to engineering problems. Biology can provide new<br />
technological possibilities <strong>and</strong> enhance performance of existing technologies. Biomimicry is<br />
a tool for solving problems in the conceptual <strong>and</strong> embodiment phases of design (Reap<br />
et al., 2005). The goal of biomimetics is to use biological inspiration to engineer machines<br />
that emulate the performance of animals (Kumph & Triantafyllou, 1998; Taubes, 2000),<br />
particularly in instances where the animal’s performance exceeds current technology. The<br />
natural experimentation that has occurred through the evolutionary process has produced<br />
the plethora of organisms, both living <strong>and</strong> extinct. Within the phylogenetic lineages of<br />
these organisms, there has been essentially a “cost-benefit analysis” where particular designs<br />
for specific functions have been optimized to perform with respect to the rigors of their<br />
environment (Fish, 2006).<br />
Biologists are well acquainted with the specific adaptations present in animals, which may<br />
be of interest to engineers. For biologists, an adaptationist program has allowed for the identification<br />
of novel features of organisms based on engineering principles, whereas for engineers,<br />
identification of such novel features is necessary to exploit them for biomimetic<br />
development. This new synergy between biologists <strong>and</strong> engineers can be beneficial in advancing<br />
technology by looking to nature to provide solutions to current problems.<br />
For marine technologies, the biomimetic approach particularly holds the promise of enhanced<br />
performance <strong>and</strong> increased efficiency for operation in the aquatic realm. It was in the<br />
oceans that life first evolved <strong>and</strong> where complex animals have thrived for over 600 million<br />
years. There are marine representatives from every major animal phylum. <strong>Marine</strong> animals<br />
survive in environments as diverse as tropical coral reefs, polar ice-capped oceans, <strong>and</strong> the<br />
lightless abyssal depths. The diversity of habitats available in marine systems has led to a<br />
vast array of body designs <strong>and</strong> physiological <strong>and</strong> behavioral mechanisms. These adaptations<br />
that evolved in animals are used to overcome the biotic <strong>and</strong> abiotic challenges in the ocean.<br />
To deal with the rigors of the marine environment, animals have developed specialized<br />
sensory systems (e.g., echolocation, electroreception), mechanisms to deal with pressure<br />
8 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
(e.g., buoyancy control), strategies to economize on energy (e.g., fusiform body design,<br />
schooling, burst-<strong>and</strong>-glide swimming), armor (e.g., bony scales, mollusk shells), stability<br />
mechanisms (e.g., paired <strong>and</strong> median fins), maneuverability (e.g., flexible bodies, vectored<br />
thrust), speed (e.g., high-aspect-ratio oscillatory propulsors, jet propulsion), stealth (e.g.,<br />
camouflage, low acoustic signature), <strong>and</strong> use of compliant materials (e.g., collagen, protein<br />
rubbers, mucous). In using such specializations to enhance their own survival, animals attempt<br />
to function in a manner to minimize their total energy budget while maximizing the<br />
performance of the specialization. Animals are doing the type of optimization that engineers<br />
seek to incorporate into designs (Vincent, 1990), <strong>and</strong> specifically, marine animals<br />
are dealing with the very problems that are of concern for marine engineers.<br />
<strong>Marine</strong> technology is well suited for the application of bioinspired design, as there is a need<br />
for exploitation of the oceans for new sources of food, energy, <strong>and</strong> minerals. Exploration of<br />
the world’s oceans is expensive. Ship time, support personnel, maintenance facilities, <strong>and</strong><br />
energy costs can be prohibitively costly. Added to these costs are the expanse of the ocean<br />
surface (3.6 × 10 8 km 2 ) to be explored <strong>and</strong> the dangers associated with working in the<br />
deep-water environment (average depth = 3,650 m). The Challenger Deep (depth =<br />
10,902 m; pressure = 16,500 psi or 113,764 kPa) in the Mariana Trench was only visited<br />
once by a manned vehicle, Trieste, in 1960. Since that time, only two unmanned expeditions,<br />
robotic deep-sea probe Kaikō <strong>and</strong> HROV Nereus, have returned to the deepest surveyed point<br />
in the ocean.<br />
Historically, marine animals have served as the inspiration for technological design. During<br />
the Renaissance, animals were identified as streamlined bodies for drag reduction that<br />
could be applied to manufactured devices. Between 1505 <strong>and</strong> 1508, Leonardo da Vinci<br />
was particularly interested in flow in water, as revealed in his notebooks, Codex Leicester<br />
(Ball, 2009). Da Vinci wrote on the function of streamlined bodies in reducing drag <strong>and</strong><br />
noted the streamlined shape of a fish (Anderson, 1998). He argued that the fish could<br />
move through the water with little resistance because its shape allowed the water to flow<br />
smoothly over the afterbody without prematurely separating. Da Vinci recognized <strong>and</strong> demonstrated<br />
a similar design with the hull shape of ships.<br />
Giovanni Borelli in 1680 made an examination of the swimming motions of animals with<br />
their application to submarine technology (Borelli, 1680). In his book De Motu Animalium<br />
(The Movement of Animals), Borelli likened swimming to flying in that both were accomplished<br />
by the displacement of fluids, although he noted the differences in density of air <strong>and</strong><br />
water <strong>and</strong> their effects on stability <strong>and</strong> buoyancy. Borelli described the design of an early submarine<br />
that incorporated ideas based in part on animals for buoyancy regulation <strong>and</strong> propulsion.<br />
The submarine would submerge using a hydrostatic mechanism based on the swim<br />
bladder of a fish by filling goatskin bags, located inside the submarine, through holes in<br />
July/August 2011 Volume 45 Number 4 9
the sides of the boat. Propulsion would be accomplished by oars projecting through the hull<br />
<strong>and</strong> fitted with watertight seals. When the submarine was on the bottom, it was envisioned<br />
that the oars would push off the s<strong>and</strong>y substrate to move the boat along. In mid-water, the<br />
oars would paddle like the feet of frogs or geese. During the rearward power stroke, a flexible<br />
paddle at the end of the oar would exp<strong>and</strong> to work on a large mass of fluid. During the forward<br />
recovery stroke, the paddle would fold passively to reduce the frontal area <strong>and</strong> drag on the oar.<br />
However, Borelli considered that propulsion of the boat would be easier if a flexible oar were<br />
positioned at the stern, emulating the motion of a fish tail. Despite the elaborate design for its<br />
time, it is doubtful if this early biomimetic experiment was successfully used.<br />
Cayley examined the streamlined body shapes of a trout <strong>and</strong> a dolphin in 1809 as solids of<br />
least resistance design (Gibbs-Smith, 1962). Cayley unsuccessfully attempted to apply these<br />
natural designs to the hull of a boat for moving on the water surface (Vogel, 1998). The<br />
rounded configuration of the hull was unstable with respect to roll, <strong>and</strong> low drag did not<br />
occur. The design of fish <strong>and</strong> dolphins is similar to the optimal shape for drag reduction<br />
of submerged bodies, such as modern submarines. These natural swimmers <strong>and</strong> submarines<br />
have fusiform body shapes with a rounded leading edge <strong>and</strong> slowly tapering tail. The forerunner<br />
for hulls used by modern nuclear submarines, the USS Albacore, was built in 1953<br />
with a fusiform shape.<br />
Both Cayley’s hull <strong>and</strong> the USS Albacore demonstrate limitations due to a misuse of the<br />
biomimetic approach. For Cayley, strict adherence to copying biological designs without<br />
proper insight into the function <strong>and</strong> limitations of those designs proved disastrous (Vogel,<br />
1998; Fish, 2006). While the shapes of fish <strong>and</strong> dolphins are appropriate for movement<br />
underwater, their shapes are not effective at the water surface. Hulls with broad beams <strong>and</strong><br />
greater buoyancy, like those displayed by waterfowl, provide enhanced stability <strong>and</strong> opportunity<br />
for greater speed at the water surface (Aigeldinger & Fish, 1995).<br />
InthecaseoftheUSSAlbacore, the design is only analogous with fish <strong>and</strong> dolphins.<br />
Although the designs are convergent, there was no information exchange to determine design.<br />
The Albacore was likened to the shape of a fish, but the submarine’s design was not biologically<br />
inspired (Harris 1997; Largess & M<strong>and</strong>elblatt, 1999). This was similar to the description<br />
of the fictional submarine, the Nautilus, in Jules Verne’s Twenty Thous<strong>and</strong> Leagues under<br />
the Sea:<br />
We were lying upon the back of a sort of submarine boat, which appeared (as far as I could<br />
judge) like a huge fish of steel.<br />
The hull shape of the Albacore was based on the “Lyon form” of airship models (Largess &<br />
M<strong>and</strong>elblatt, 1999). Originally, submarine hulls were designed more as surface ships due to<br />
10 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
the limited amount of time that they could operate submerged. The streamlined hull of the<br />
Albacore made it the fastest <strong>and</strong> most maneuverable submarine of its time.<br />
The convergent designs of the Albacore <strong>and</strong> marine animals reflect selection, both artificial<br />
<strong>and</strong> natural, respectively, for optimizing similar performance parameters (i.e., speed, drag reduction,<br />
maneuverability). However, similarity of design does not necessarily always translate<br />
into identical performance <strong>and</strong> assumptions of performance expectations should be approached<br />
with caution. Similar shapes can have different functions or have limitations to a<br />
function when viewed in isolation without consideration of the whole system. Despite its<br />
enhanced maneuverability compared to other submarines, the Albacore’s rateofturnat<br />
2° s −1 is poor when compared with similarly shaped dolphins, which can turn at 453° s −1<br />
<strong>and</strong> in a confined space of 20% of body length (Fish, 2002). The rigid hull of the submarine<br />
in concert with yaw control mainly from an aft-positioned rudder limits agility <strong>and</strong> maneuverability<br />
compared to flexible-bodied dolphins with multiple fore <strong>and</strong> aft mobile control<br />
surfaces (Fish, 2002; Fish & Nicastro, 2003).<br />
The biomimetic approach dem<strong>and</strong>s first careful observation of the whole biological system<br />
to identify the principles <strong>and</strong> attributes of the system. Thus, major limitations <strong>and</strong> constraints<br />
of any biological design can be defined before translation to an engineered system.<br />
The association between biologists <strong>and</strong> engineers becomes paramount as biomimetic technologies<br />
are developed.<br />
This special issue of the <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal brings together both biologists<br />
<strong>and</strong> engineers who are currently involved with the development of biomimetic devices for<br />
applications in the marine environment. The intent is to assess the current state of technology<br />
<strong>and</strong> incite new applications that apply these <strong>and</strong> other innovative concepts as technology advances.<br />
The research presented in this issue mimics specific aspects of fish, skates, rays, sharks,<br />
lobsters, jellyfish, squids, <strong>and</strong> sea turtles. In many of these papers, the goal is focused on locomotion<br />
or propulsion in robotic counterparts so they can maneuver more efficiently in the<br />
animal’s designed-for environment. This may be acted out either individually or in multiples<br />
as schools (or swarms) of fish. Swimming <strong>and</strong> walking using amphibious designs are also considered.<br />
A recent student competition sponsored by the Office of Naval Research (ONR) was<br />
held for the first time (involving several of the authors in this issue) to evaluate the performance<br />
of “mantabots,” which aspire to perform as gracefully as the sleek <strong>and</strong> elegant manta<br />
rays they attempt to copy (Pennisi, 2011). A commentary by McKenna provides an overview<br />
of this <strong>and</strong> other biomimetic work sponsored by ONR. Other papers in this issue identify a<br />
single unique aspect <strong>and</strong> strive to achieve the same benefit in an engineered device. Simulating<br />
the lateral line system found in most aquatic vertebrates using pressure sensors is one example<br />
that can be used to supplement vision <strong>and</strong> sonar in turbid waters. Deriving a control<br />
mechanism based on shark skin properties to increase vehicle swimming speeds, imitating the<br />
July/August 2011 Volume 45 Number 4 11
tubercles of whale fins to enhance hydrodynamics, <strong>and</strong> engineering a bioinspired solution<br />
for natural antifouling mechanisms are other examples. Whether your interest is in underwater<br />
vehicles, imaging, dynamic positioning, marine materials, ocean pollution, remote<br />
sensing, oceanographic instrumentation, ocean observing, marine science, or education,<br />
this special issue should provide valuable content.<br />
References<br />
Aigeldinger, T.L., & Fish, F.E. 1995. Hydroplaning by ducklings: Overcoming limitations to swimming at the<br />
water surface. J Exp Biol. 198:1567-4.<br />
Anderson, J.D. 1998. A History of Aerodynamics. Cambridge: Cambridge University Press.<br />
Ball, P. 2009. Flow. Oxford: Oxford University Press.<br />
Borelli, G.A. 1680. De Motu Animalium Pars (The Movement of Animals, translated by P. Maquet (1989)).<br />
Berlin: Springer-Verlag.<br />
Fish, F.E. 2002. Balancing requirements for stability <strong>and</strong> maneuverability in cetaceans. Integr Comp Biol.<br />
42:85-93. doi: 10.1093/icb/42.1.85.<br />
Fish, F.E. 2006. Limits of nature <strong>and</strong> advances of technology in marine systems: What does biomimetics have<br />
to offer to aquatic robots Appl Bionics Biomech. 3:49-60. doi: 10.1533/abbi.2004.0028.<br />
Fish, F.E., & Nicastro, A.J. 2003. Aquatic turning performance by the whirligig beetle: constraints on<br />
maneuverability by a rigid biological system. J Exp Biol. 206:1649-56. doi: 10.1242/jeb.00305.<br />
Gibbs-Smith, C.H. 1962. Sir George Cayley’s Aeronautics 1796-1855. London: Her Majesty’s Stationery<br />
Office.<br />
Harris, B. 1997. The Navy Times Book of Submarines: A Political, Social, <strong>and</strong> Military History. New York:<br />
Berkley.<br />
Kumph, J.M., & Triantafyllou, M.S. 1998. A fast-starting <strong>and</strong> maneuvering vehicle, the ROBOPIKE.<br />
In: Proceedings of the International Symposium on Seawater Drag Reduction, ed. Meng, J.C.S.,<br />
pp. 485-90. Newport, Rhode Isl<strong>and</strong>.<br />
Largess, R.P., & M<strong>and</strong>elblatt, J.L. 1999. U.S.S. Albacore: Forerunner of the Future. Portsmouth, NH:<br />
The Portsmouth <strong>Marine</strong> <strong>Society</strong>.<br />
Pennisi, E. 2011. Manta Machines, 332:28-9, Science, 27 May 2001, www.sciencemag.org. Retrieved on<br />
June 14, 2011.<br />
Reap, J., Baumeister, D., & Bras, B. 2005. Holism, biomimicry <strong>and</strong> sustainable engineering. In: Proceedings<br />
of IMECE2005. November 5-11, 2005. Orl<strong>and</strong>o, FL.<br />
12 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
Taubes, G. 2000. Biologists <strong>and</strong> engineers create a new generation of robots that imitate life. Science. 288:80-3.<br />
doi: 10.1126/science.288.5463.80.<br />
Vincent, J. 1990. Structural Biomaterials. Princeton: Princeton Univ. Press.<br />
Vogel, S. 1998. Cat’s Paws <strong>and</strong> Catapults. New York: W. W. Norton.<br />
July/August 2011 Volume 45 Number 4 13
COMMENTARY<br />
Biomimicking <strong>Marine</strong> Mechanisms<br />
<strong>and</strong> Organizational Principles<br />
AUTHOR<br />
Yoseph Bar-Cohen<br />
Jet Propulsion Laboratory,<br />
California Institute of <strong>Technology</strong><br />
Nature is effectively a giant laboratory<br />
where trial-<strong>and</strong>-error evolutionary<br />
experiments are taking place. As<br />
nature performs its experiments, all<br />
the fields of science <strong>and</strong> engineering<br />
are employed, including physics,<br />
chemistry, mechanical engineering,<br />
<strong>and</strong> materials science. The processes<br />
range in scale from nano <strong>and</strong> micro<br />
(e.g., viruses <strong>and</strong> bacteria) to macro<br />
<strong>and</strong> mega (e.g., our life scale, elephants,<br />
<strong>and</strong> whales). To address the<br />
many survival challenges, biological<br />
systems came up with superb solutions.<br />
The constraints in addressing<br />
these challenges are similar to those<br />
that human engineers are facing, including<br />
the need to maximize the<br />
functionality of their design <strong>and</strong> produce<br />
systems that use minimal resources<br />
(e.g., materials, energy, cost,<br />
etc.). Humans have always made efforts<br />
to use nature as a model for<br />
inspiring innovation <strong>and</strong> problem<br />
solving. However, biological <strong>and</strong><br />
botanical systems have superior<br />
capabilities, including producing<br />
materials—they use their body temperature,<br />
the materials are recyclable,<br />
<strong>and</strong> the process does not involve<br />
pollution.<br />
To take advantage of the capabilities<br />
of nature, the field of biomimetics<br />
involves seeking to underst<strong>and</strong> <strong>and</strong><br />
use of the capabilities as a model for<br />
copying, adapting, <strong>and</strong> inspiring concepts<br />
<strong>and</strong> designs (Bar-Cohen, 2005;<br />
Bar-Cohen, 2011; Benyus, 1998;<br />
Vincent, 2001). For this purpose,<br />
scientists are seeking rules, concepts,<br />
mechanisms, <strong>and</strong> principles to inspire<br />
new possibilities. Some of the benefits<br />
that resulted from biomimetic approaches<br />
have led to improved structures,<br />
actuators, sensors, interfaces,<br />
control algorithms, software, drugs,<br />
defense, <strong>and</strong> intelligence, <strong>and</strong> may<br />
help to improve our ability to recycle<br />
materials <strong>and</strong> protect the environment.<br />
Some of the biomimetic characteristics<br />
that are being developed include shape<br />
morphing, self-repair, self-replication,<br />
<strong>and</strong> self-reconfiguration (Bar-Cohen<br />
& Breazeal, 2003; Bar-Cohen, 2011).<br />
Increasingly, researchers are working<br />
towards adapting the capabilities of<br />
many creatures to perform tasks in<br />
hard-to-reach areas <strong>and</strong> in conditions<br />
that are too harsh or dangerous for<br />
humans.<br />
<strong>Marine</strong> biosystems are quite rich<br />
in capabilities that have <strong>and</strong> can further<br />
benefit humans from biomimicking.<br />
There are many examples that<br />
one can list, including one as simple<br />
as the fins that are used by swimmers<br />
<strong>and</strong> divers that significantly enhance<br />
their performance. While it may be<br />
arguable that the fins were a biologically<br />
inspired invention, one can state<br />
that it is common knowledge that<br />
swimming creatures (e.g., geese,<br />
swans, seagulls, seals, <strong>and</strong> frogs) have<br />
feet with membranes that help them<br />
swim. The stability, maneuvering,<br />
<strong>and</strong> swimming performance of underwater<br />
animals are determined by the<br />
morphology, position, <strong>and</strong> mobility<br />
of their control surfaces. For cetaceans<br />
(i.e., whales, dolphins, <strong>and</strong> porpoises)<br />
the pectoral flippers are mobile hydrofoils<br />
that generate lift similar to engineered<br />
hydrofoils. The flippers have<br />
various shapes <strong>and</strong> are used to perform<br />
lateral turning, dive, surface,<br />
brake, <strong>and</strong> other mobility-related<br />
functions. Studies of the 3-D geometry<br />
<strong>and</strong> hydrodynamic performance<br />
of cetacean flippers with various<br />
morphologies help provide insight<br />
into the maneuverability, drag <strong>and</strong><br />
lift performance at high Reynolds<br />
numbers (Fish et al., 2011). Wagging<br />
the body <strong>and</strong> tail is the main propulsion<br />
method of marine swimmers<br />
(e.g., billfish <strong>and</strong> sailfish), allowing<br />
them to reach significant speeds of<br />
over 75 km/h. Submarines that<br />
could perform efficiently as marine<br />
swimmers using a flexible body<br />
would be an important exploration<br />
tool for scientists <strong>and</strong> would potentially<br />
have many military applications.<br />
While significant advances were<br />
reached via mimicking <strong>and</strong> the inspiration<br />
of marine biology, there are many<br />
capabilities in nature that are still far<br />
superior to engineered capabilities;<br />
examples include the following:<br />
■ The sonar of marine animals is far<br />
superior to any existing marine<br />
sonar (Muller & Hallam, 2004).<br />
■ Sea shells <strong>and</strong> skeletons of marine<br />
invertebratesarefarstronger<strong>and</strong><br />
lighter than human-made materials,<br />
<strong>and</strong> their fabrication does not create<br />
pollution concerns.<br />
14 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
■ Muscles stick to rocks even though<br />
the adhesion is done in water, <strong>and</strong><br />
they sustain sticksion in spite of<br />
the strong impacts of ocean waves.<br />
On the other h<strong>and</strong>, most humanmade<br />
adhesives fail when the adhesion<br />
is done in on a wet surface.<br />
■ The chiton, which is a diminutive<br />
mollusk, has very strong teeth that<br />
it uses to munch on rocks <strong>and</strong> extract<br />
food (Weaver et al., 2010).<br />
This capability may be used to inspire<br />
the development of effective<br />
lightweight bits for in situ planetary<br />
exploration sampling drills<br />
(Figure 1).<br />
One hopes that engineers will be<br />
able to rapidly prototype biological<br />
capabilities as fast as it is now possible<br />
to graphically edit photos of biological<br />
systems, as illustrated in Figure 2. In<br />
this figure, a photo of a shark was edited<br />
to create an imaginary image of a<br />
U.S. Navy vessel that is shark-like.<br />
While we are somewhat far from this<br />
capability, significant advances have<br />
been made by scientists <strong>and</strong> engineers<br />
whoseektomimicmarinebiology.<br />
This special issue is dedicated to describing<br />
<strong>and</strong> discussing the latest advances<br />
that were inspired by marine<br />
mechanisms <strong>and</strong> their organizational<br />
principles.<br />
FIGURE 2<br />
An example is shown where software was<br />
used to turn a photo of a natural shark into<br />
a vessel-like naval system. The photograph<br />
(a) was taken by the author; the image<br />
below (b) shows the modification to a Navy<br />
marine vehicle <strong>and</strong> is courtesy of David<br />
Hanson, Hanson Robotics LLC, TX.<br />
Acknowledgment<br />
Some of the research reported in<br />
this article was conducted at the Jet<br />
Propulsion Laboratory, California Institute<br />
of <strong>Technology</strong>, under a contract<br />
with the National Aeronautics<br />
<strong>and</strong> Space Administration.<br />
Author:<br />
Yoseph Bar-Cohen<br />
Jet Propulsion Laboratory,<br />
California Institute of <strong>Technology</strong><br />
4800 Oak Grove Drive,<br />
Pasadena, CA 91109-8099<br />
Email: yosi@jpl.nasa.gov<br />
Benyus, J.M. 1998. Biomimicry: Innovation<br />
Inspired by Nature. New York, NY: Harper-<br />
Collins Publishers. pp. 1-302.<br />
Fish, F.E., Smits, A.J., Haj-Hariri, H.,<br />
Iwasaki, T., & Bart-Smith, H. 2011.<br />
Biomimetic swimmer inspired by the manta<br />
ray, Chapter 17. In <strong>Biomimetics</strong>: Nature-<br />
Based Innovation, ed. Bar-Cohen, Y. Boca<br />
Raton, FL: CRC Press, Taylor & Francis<br />
Group.<br />
Muller, R., & Hallam, J.C.T. 2004. From<br />
bat pinnae to sonar antennae: augmented<br />
obliquely truncated horns as a novel parametric<br />
shape model. In: Proceeding of the 8th<br />
International Conference on the Simulation<br />
of Adaptive Behavior, SAB’04. Massachusetts,<br />
USA: The International <strong>Society</strong> for Adaptive<br />
Behavior.<br />
Vincent, J.F.V. 2001. Stealing ideas from<br />
nature, Chapter 3. In: Deployable Structures,<br />
ed. Pellegrino, S. Vienna, Austria: Springer-<br />
Verlag. pp. 51-58.<br />
Weaver, J.C., Wang, Q., Miserez, A.,<br />
Tantuccio, A., Stromberg, R., Bozhilov,<br />
K.N.P., … Kisailus, D. 2010. Analysis of an<br />
ultra hard magnetic biomineral in chiton<br />
radular teeth. Mater Today, 13(1-2):<br />
pp. 42-52.<br />
FIGURE 1<br />
A front view of the Eudoxochiton nobilis<br />
(“Noble” chiton). Courtesy of Iain Anderson.<br />
References<br />
Bar-Cohen, Y. (Ed.). 2005. <strong>Biomimetics</strong>—<br />
Biologically Inspired Technologies. Boca<br />
Raton, FL: CRC Press. pp. 1-527.<br />
Bar-Cohen, Y. (Ed.). 2011. <strong>Biomimetics</strong>:<br />
Nature-Based Innovation. Boca Raton, FL:<br />
CRC Press, Taylor & Francis Group.<br />
pp. 1-788.<br />
Bar-Cohen, Y., & Breazeal, C. (Eds.). 2003.<br />
Biologically Inspired Intelligent Robots.<br />
Bellingham, WA: SPIE Press. pp. 1-393.<br />
Vol. PM122.<br />
July/August 2011 Volume 45 Number 4 15
COMMENTARY<br />
Sink <strong>and</strong> Swim: Clues From Nature<br />
for Aquatic Robotics<br />
AUTHOR<br />
Jeannette Yen<br />
Georgia Institute of <strong>Technology</strong><br />
There are many ways to travel by<br />
water. Watercraft can use propellers<br />
<strong>and</strong> sails while living organisms exhibit<br />
quite a variety of modes of transport.<br />
They sink <strong>and</strong> swim, flap <strong>and</strong><br />
glide, stroke <strong>and</strong> jet. From this, we<br />
see a distinction between the limited<br />
human solutions <strong>and</strong> the diverse natural<br />
solutions. This natural diversity<br />
continues to inspire inventions in robotics.<br />
Leonardo da Vinci envisioned<br />
walking on water (Figure 1), something<br />
water striders could do millions<br />
of years ago (Hu et al., 2007). Bathtub<br />
toys such as the TwiddleFish,<br />
designed by Chuck Pell of Duke<br />
University (Figure 2), taught us the<br />
importance of stiffness in how fish<br />
FIGURE 1<br />
Leonardo da Vinci: Shoes for walking on water<br />
(image from http://en.wikipedia.org/wiki/<br />
Walking_on_water).<br />
swim so well. Pell commented, “In<br />
the h<strong>and</strong>s of kids…they can really<br />
feel what’s goingon.” To the delight<br />
of pre-K–12 children <strong>and</strong> their parents,<br />
there are propulsive toys on the<br />
market with remarkably functional<br />
body parts, serving as inspiration <strong>and</strong><br />
outreach to youngsters.<br />
Recent efforts have focused on<br />
flapping by fish <strong>and</strong> mantas (e.g.,<br />
Tangorra et al., 2011; Fish et al.,<br />
2011a) or jetting by jellyfish <strong>and</strong><br />
squid (e.g., Moslemi & Krueger,<br />
2011). By replicating nature via robotics,<br />
we underst<strong>and</strong> the significance<br />
of the number of joints (Dean et al.,<br />
2009), the shape of the fin (Curet<br />
et al., 2011), <strong>and</strong> the direction of a<br />
tail swish (Long et al., 2006) for controlling<br />
movement. Using these robots<br />
to repeatedly vary the timing of<br />
undulations of a fish or jellyfish, we<br />
discover that the frequency of the<br />
real organism is optimized to capture<br />
FIGURE 2<br />
TwiddleFish by Charles Pell, Bio-Design Studio,<br />
Duke University (cited by Guterl, 1996). Photo<br />
courtesy of F. Fish.<br />
the energy shed in the vortices left by<br />
the previous pulse (Triantafyllou &<br />
Triantafyllou, 1995; Fish, 2006;<br />
Ruiz et al., 2010). This is not something<br />
you could ask a fish or a jellyfish<br />
to do over <strong>and</strong> over again (though as<br />
biologists, we have patiently waited<br />
<strong>and</strong> recorded our aquatic creatures<br />
performing these behaviors over <strong>and</strong><br />
over <strong>and</strong> over again; see Catton et al.,<br />
2011). We identify the body parts that<br />
are important for propulsion, <strong>and</strong> we<br />
figure out how many propulsors are<br />
needed, along with their placement.<br />
Control systems coordinate the multiple<br />
fins. Sensors integrating different<br />
sensory modalities are sought to develop<br />
navigation systems that reliably<br />
guide the robot to reach its destination.<br />
We test out new actuators that<br />
accelerate unsteadily to increase thrust.<br />
We use new materials for joints or<br />
bodies with a compliance to achieve<br />
the flexibility needed for realistic undulations<br />
<strong>and</strong> squeezes (Lauder et al.,<br />
2007; Cutkosky & Kim, 2009; Ruiz<br />
et al., 2010). By matching nature as<br />
closely as possible, we gain a greater<br />
underst<strong>and</strong>ing of how nature works.<br />
But is matching reality necessary<br />
What if instead of underst<strong>and</strong>ing<br />
how shape <strong>and</strong> structure are optimized<br />
for best propulsion, we determine<br />
how to achieve stealth by<br />
designing robots that blend in with<br />
the background turbulence or match<br />
the disturbance made by the other<br />
members of the school A robot like<br />
that would enable us to spy on<br />
schools from the inside out, by becoming<br />
a schoolmate. Can we study<br />
16 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
how the sensing system is integrated<br />
with the propulsors to enhance the<br />
acuity of information capture What<br />
principles can we abstract from natural<br />
aquatic propulsion to improve<br />
how we save energy or save materials<br />
or improve performance, as all surviving<br />
species do in a variety of ways to<br />
suit the constraints of the environment<br />
in which they have evolved<br />
Perhaps we can take advantage of the<br />
environment <strong>and</strong> glide <strong>and</strong> surf where<br />
possible, using the free energy of the<br />
sea. Oceanographic gliders like the<br />
Slocum glider Scarlet 27 (Figure 3;<br />
Schofield et al., 2010) are beautifully<br />
designed, wherein one version varies its<br />
ballast by a phase change in response to<br />
ambient ocean temperature (Webb et al.,<br />
2001). By using whale- or copepodlike<br />
buoyancy control (Clarke, 1978;<br />
Pond & Tarling, 2011) <strong>and</strong> harvesting<br />
environmental energy in the ocean<br />
temperature gradient, this glider traveled<br />
great distances with much less<br />
energy than other forms of propulsion.<br />
Simple modifications of shape on<br />
key control surfaces can lead to large<br />
variations in the balance of lift <strong>and</strong><br />
drag essential for tight maneuvering,<br />
as exemplified by the tubercles of<br />
whale fins that now enable wind turbines<br />
to capture energy at lower wind<br />
speeds (Fish et al., 2011b). Can we further<br />
achieve an economy of materials<br />
by adopting the streamlined form of<br />
the shark <strong>and</strong> applying the hierarchical<br />
structure of its denticles to fine<br />
tune that drag reduction (Dean &<br />
Bhushan, 2010)<br />
Indeed, looking at the familiar<br />
players in the sea points out many<br />
cases of convergent evolution in<br />
terms of undulatory or jet-like propulsion.<br />
A closer look reveals other<br />
unusual adaptations for aquatic mobility<br />
such as the parachutes of pteropods<br />
or the floats of siphonophores or<br />
FIGURE 3<br />
(A) Autonomous robots that follow the routes of swimming penguins are collecting information<br />
that could help scientists underst<strong>and</strong> why the birds’ populations are dropping rapidly. The underwater<br />
robots, called gliders, are programmed to record ocean conditions as they follow the<br />
tracks of Adelie penguins swimming in the Southern Ocean surrounding Antarctica. (B) Diagram<br />
of Teledyne-Webb Corporation’s Slocum Glider (coastal model). The Front Main Housing<br />
Section glider’s ballast, <strong>and</strong> consequently its flight, is controlled by moving water into or out<br />
of the Fore Wet Section. From Kahl et al. (2010). (C) The 221-day path taken by Scarlet Knight<br />
(adapted from Schofield et al., 2010; Kahl et al., 2010; images courtesy of O. Schofield: http://<br />
rucool.marine.rutgers.edu).<br />
the multiple oars of copepods, krill,<br />
<strong>and</strong> polychaetes. Analyses of their<br />
transport mechanisms may again<br />
open our eyes to novel designs of future<br />
underwater vehicles that can<br />
steadily hover, smoothly cruise, rapidly<br />
escape, quietly sink or perform any<br />
gait as needed in tight quarters. With<br />
July/August 2011 Volume 45 Number 4 17
the rapid evolution occurring in materials<br />
science research <strong>and</strong> in control<br />
systems, we may find ourselves traveling<br />
through fluids in unexpected<br />
ways.<br />
Author:<br />
Jeannette Yen<br />
School of Biology,<br />
Center for Biologically Inspired Design<br />
Georgia Institute of <strong>Technology</strong>,<br />
Atlanta, GA. 30332-0230<br />
Email: jeannette.yen@biology.gatech.<br />
edu<br />
References<br />
Catton, K.B., Webster, D.R., Kawaguchi, S.,<br />
& Yen, J. 2011. The hydrodynamic disturbances<br />
of two species of krill: Implications for aggregation<br />
structure. J Exp Biol. 214:1845-56.<br />
Clarke, M.R. 1978. Buoyancy control as a<br />
function of the spermaceti organ in the sperm<br />
whale. J Mar Biol Assoc UK. 58:27-71.<br />
doi: 10.1017/S0025315400024395.<br />
Curet, O.M., Patankar, N.A., Lauder, G.V.,<br />
& MacIver, M.A. 2011. Aquatic manoeuvering<br />
with counter-propagating waves: A novel<br />
locomotive strategy. J Roy Soc Interface.<br />
8(60):1041-50. doi: 10.1098/rsif.2010.0493.<br />
Cutkosky, M.R., & Kim, S. 2009. Design<br />
<strong>and</strong> fabrication of multi-material structures for<br />
bioinspired robots. Philos T R Soc A.<br />
367:1799-813. doi: 10.1098/rsta.2009.0013.<br />
Dean, B., & Bhushan, B. 2010. Shark-skin<br />
surfaces for fluid-drag reduction in turbulent<br />
flow: A review. Philos T R Soc A. 368(1929):<br />
4775-806. doi: 10.1098/rsta.2010.0201.<br />
Dean, M.N., Swanson, B.O., & Summers,<br />
A.P. 2009. Biomaterials: Properties, variation<br />
<strong>and</strong> evolution. Integr Comp Biol. 49(1):<br />
15-20. doi: 10.1093/icb/icp012.<br />
Fish, F.E. 2006. Limits of nature <strong>and</strong> advances<br />
of technology: What does biomimetics have to<br />
offer to aquatic robots Appl Bionics Biomech.<br />
3(1):49-60. doi: 10.1533/abbi.2004.0028.<br />
Fish, F.E., Nichols, R.H., Dudas, M.A.,<br />
Moored, K.W., & Bart-Smith, H. 2011a.<br />
Kinematics of swimming in the manta ray<br />
(Manta birostris): 3D analysis of open<br />
water maneuverability. Integr Comp Biol.<br />
51(Suppl. 1):E42.<br />
Fish, F.E., Weber, P.W., Murray, M.M., &<br />
Howle, L.E. 2011b. The tubercles on<br />
humpback whale’s flipper: Application of<br />
bio-inspired technology. Integr Comp Biol.<br />
51(1):203-13. doi: 10.1093/icb/icr016.<br />
Guterl, F. 1996. Playthings of Science.<br />
http://discovermagazine.com/1996/dec/<br />
playthingsofscie946. (accessed 15 July 2011).<br />
Hu, D., Chan, B., & Bush, J.W.M. 2007.<br />
Water-walking devices. Exp Fluids. 43:<br />
769-78. doi: 10.1007/s00348-007-0339-6.<br />
Kahl, L., Oscar Schofield, A., & Fraser, W.R.<br />
2010. Autonomous gliders reveal features<br />
of the water column associated with foraging<br />
by adelie penguins. Integr Comp Biol.<br />
50(6):1041-50. doi: 10.1093/icb/icq098.<br />
Lauder, G.V., Anderson, E.J., Tangorra, J., &<br />
Madden, P.G. 2007. Fish biorobotics: Kinematics<br />
<strong>and</strong> hydrodynamics of self-propulsion.<br />
J Exp Biol. 210:2767-80. doi: 10.1242/<br />
jeb.000265.<br />
Long, J.H.J., Koob, T.J., Irving, K., Combie,<br />
K., Engel, V., Livingston, N., … Schumacher,<br />
J. 2006. Biomimetic evolutionary analysis:<br />
Testing the adaptive value of vertebrate tail<br />
stiffness in autonomous swimming robots.<br />
J Exp Biol. 209:4732-46. doi: 10.1242/<br />
jeb.02559.<br />
Moslemi, A.A., & Krueger, P.S. 2011. The<br />
effect of Reynolds number on the propulsive<br />
efficiency of a biomorphic pulsed-jet<br />
underwater vehicle. Bioinspir Biomim.<br />
6(2):026001. doi: 10.1088/1748-<br />
3182/6/2/026001.<br />
Pond, D.W., & Tarling, G.A. 2011. Phase<br />
transitions of wax esters adjust buoyancy<br />
in diapausing Calanoides acutus. Limnol<br />
Oceanogr. 56(4):1310-8.<br />
Ruiz, L.A., Whittlesey, R.W., & Dabiri, J.O.<br />
2010. Vortex-enhanced propulsion.<br />
J Fluid Mech. 668:5-32. doi: 10.1017/<br />
S0022112010004908.<br />
Schofield, O., Kohut, J., Glenn, S., Morell,<br />
J., Capella, J., Corredor, J., … Boicourt, W.<br />
2010. A regional Slocum glider network in the<br />
Mid-Atlantic coastal waters leverages broad<br />
community engagement. Mar Technol Soc J.<br />
44(6):64-74.<br />
Tangorra, J., Phelan, C., Esposito, C., &<br />
Lauder, G.V. 2011. Use of biorobotic models<br />
of highly deformable fins for studying the<br />
mechanics <strong>and</strong> control of fin forces in fishes.<br />
Integr Comp Biol. 51(1):176-89. doi:<br />
10.1093/icb/icr036.<br />
Triantafyllou, M.S., & Triantafyllou, G.S.<br />
1995. An efficient swimming machine.<br />
Sci Am. 272:64-70. doi: 10.1038/<br />
scientificamerican0395-64.<br />
Webb, D.C., Simonetti, P.J., & Jones, C.P.<br />
2001. SLOCUM: An underwater glider<br />
propelled by environmental energy.<br />
IEEE J Oceanic Eng. 26(4):447-52.<br />
doi: 10.1109/48.972077.<br />
18 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
COMMENTARY<br />
Developing Bioinspired Autonomous Systems<br />
AUTHOR<br />
Thomas M. McKenna<br />
Office of Naval Research<br />
Research that seeks to identify the<br />
principles, strategies, <strong>and</strong> mechanisms<br />
used by motile aquatic animals offers<br />
opportunities to develop undersea<br />
vehicles that exceed current capabilities<br />
<strong>and</strong> enable the Navy to exp<strong>and</strong><br />
the operational envelope of autonomous<br />
undersea vehicles. Autonomous<br />
undersea vehicles serve in a number of<br />
important current <strong>and</strong> emerging roles<br />
in Navy missions, including surveillance<br />
for anti-submarine warfare,<br />
mine countermeasures, Improvised<br />
Explosive Device (IED) detection<br />
<strong>and</strong> localization, force protection (e.g.,<br />
counter-diver missions) in harbors, <strong>and</strong><br />
riverine exploration <strong>and</strong> characterization.<br />
However, current unmanned undersea<br />
vehicles (UUVs) have technical<br />
gaps in areas such as mission duration<br />
(due to power constraints), excessive<br />
noise generation, speed, limited maneuverability,<br />
lack of tight integration<br />
of sensing <strong>and</strong> maneuver, propulsion<br />
<strong>and</strong> maneuver dead zones at low<br />
speeds, <strong>and</strong> inability to operate in the<br />
surf zone. Considering the capabilities<br />
of sea creatures for efficient propulsion<br />
over a large range of speeds, their ability<br />
to operate in or even exploit energetic<br />
ocean environments like the surf<br />
zone, extraordinary maneuverability,<br />
<strong>and</strong> the special sensing evolved for predation,<br />
schooling <strong>and</strong> navigation,<br />
there are many lessons for technologists.<br />
The basic research programs in<br />
Bio-Inspired Autonomous Systems at<br />
the Office of Naval Research (ONR)<br />
have supported research using four<br />
approaches: (1) the identification,<br />
modeling, <strong>and</strong> emulation of the biomechanics<br />
<strong>and</strong> fluid mechanics of<br />
underwater propulsion <strong>and</strong> control<br />
in swimming organisms, (2) identification<br />
<strong>and</strong> exploration of the<br />
sensorimotor control of animals <strong>and</strong><br />
integrated closed loop control using<br />
biosensing (e.g., biosonar, electrosense,<br />
lateral line sensors, optic flow,<br />
magnetic sense), (3) the development<br />
of muscle-like actuators <strong>and</strong> fin designs<br />
that exploit these materials,<br />
<strong>and</strong> (4) design <strong>and</strong> development of<br />
swimming prototype Autonomous<br />
Underwater Vehicle (AUVs) as proof<br />
of principle <strong>and</strong> for performance<br />
evaluation. These studies take their<br />
inspiration from diverse sea creatures<br />
<strong>and</strong> include high-performance swimmers<br />
like bluefin tuna, squid, rays<br />
<strong>and</strong> seals; animals that thrive in the<br />
surf zone like lobsters <strong>and</strong> crabs; <strong>and</strong><br />
animals with special senses like<br />
biosonar (i.e., dolphins) <strong>and</strong> electrosense<br />
(i.e., sharks <strong>and</strong> ribbonfish).<br />
More recently, opportunities have<br />
emerged for microrobotic underwater<br />
systems capable of sensing, reporting,<br />
<strong>and</strong> remediation of environmental<br />
chemicals that exploit the convergence<br />
of synthetic biology, nanotechnology,<br />
<strong>and</strong> electro-optic systems.<br />
This has prompted renewed interest<br />
in the propulsion biology of organisms<br />
that use cilia <strong>and</strong> flagella for<br />
locomotion.<br />
Key Science <strong>and</strong><br />
<strong>Technology</strong> Issues<br />
There are a number of key issues<br />
for the development of bioinspired<br />
autonomous systems.<br />
1. Developing high-efficiency<br />
propulsion that exceeds the capability<br />
of propellers. This is particularly<br />
important for achieving long-duration<br />
missions. Although there is a sizable<br />
Navy effort to develop new energy<br />
sources for underwater vehicles, introduction<br />
of more efficient propulsion<br />
systems would reduce the power requirements<br />
<strong>and</strong> thereby lengthen mission<br />
durations. Early efforts to mimic<br />
the caudal fins of high-performance<br />
swimmers like tuna did not produce<br />
performance that exceeds propellers<br />
(but recent work on the Ghostswimmer<br />
TM , discussed in Rufo <strong>and</strong> Smithers’<br />
commentary in this issue, looks promising),<br />
<strong>and</strong> B<strong>and</strong>yopadhyay (2005) has<br />
produced a meta-analysis showing that<br />
animals do not have an advantage over<br />
man-made systems in cruise, but animals<br />
do show greater maneuverability<br />
relative to man-made underwater<br />
vehicles. One promising new form of<br />
propulsion is the use of bioinspired<br />
high-lift foil propulsors. High-lift propulsion<br />
was introduced in the biological<br />
context in the analysis of fly wing<br />
lift (B<strong>and</strong>yopadhyay, 2009; Ellington,<br />
1984; Dickinson et al., 1999). Fly<br />
wings, which pitch <strong>and</strong> heave with a<br />
90° phase difference, can achieve lift<br />
coefficients substantially greater than<br />
rigid foils that have a constant angle<br />
of attack, <strong>and</strong> high-lift foil propulsors<br />
should be capable of an order of magnitude<br />
greater lift than traditional<br />
naval propellers. High-lift foils also<br />
produce substantially less noise than<br />
traditional propellers. B<strong>and</strong>yopadhyay<br />
et al. (2008) have produced a series of<br />
rigid high-lift foils (roughly analogous<br />
to penguin pectoral fins), characterized<br />
July/August 2011 Volume 45 Number 4 19
their propulsion properties, <strong>and</strong><br />
mounted them on an undersea vehicle.<br />
This first version was called bioinspired<br />
autonomous undersea vehicle (BAUV)<br />
<strong>and</strong> a second, lower-diameter version<br />
was called self-propelled line array<br />
(SPLINE) (Figures 1 <strong>and</strong> 2). High-lift<br />
flapping foils are efficient, <strong>and</strong> combining<br />
six multiple foils on a vehicle can<br />
achieve extraordinary maneuverability<br />
<strong>and</strong> hover <strong>and</strong> exhibit low-noise emission,<br />
but the flapping foil configuration<br />
is not consistent with producing high<br />
speeds. Recently, B<strong>and</strong>yopadhay has<br />
designed a new propulsor called<br />
“Slosher” that has multiple foils with<br />
variable angles of attack arranged radially<br />
that can operate as a high-lift propulsor<br />
at low speeds or, with the foils<br />
locked, as a traditional propeller at<br />
high speeds. This avoids the deadb<strong>and</strong><br />
of controllability at low speeds.<br />
Moreover, hybrid vehicles, such as the<br />
RAZOR (Figure 3), have been developed<br />
that combine four high-lift<br />
foils <strong>and</strong> two props. At low speeds,<br />
the high-lift foils provide high maneuverability<br />
<strong>and</strong> hover, but when the<br />
props provide higher speeds for transit<br />
of the vehicle, the foils become control<br />
surfaces.<br />
FIGURE 1<br />
BAUV developed by Dr. Promode B<strong>and</strong>yopadyay<br />
at the Naval Undersea Warfare Center, Newport,<br />
RI (NUWC-NPT). The BAUV has six high-lift<br />
foils <strong>and</strong> a controller based on a model of the<br />
olivo-cerebellar circuits of mammals.<br />
FIGURE 2<br />
SPLINE. This is a refinement of the BAUV, designed<br />
for long-duration testing of its ability<br />
to pull a load, maintain accurate position, <strong>and</strong><br />
keep a line taut that has been fixed at the distal<br />
end while maneuvering (NUWC-NPT).<br />
FIGURE 3<br />
RAZOR vehicle. This vehicle was developed by<br />
Richard Berube <strong>and</strong> Promode B<strong>and</strong>yopadyay<br />
at NUWC-NPT. It has four high-left foils <strong>and</strong><br />
two rotating propulsors. At low speeds, it<br />
uses the foils for high maneuverability.<br />
2. Developing adaptive controllers<br />
for high-degree-of-freedom<br />
bioinspired propulsors. Many of<br />
the bio-inspired propulsion systems<br />
exhibit high degrees of freedom. For<br />
the case of the high-lift flapping<br />
foils,roll,pitch,<strong>and</strong>frequencyare<br />
motion parameters to be specified to<br />
achieve efficient fin kinematics, <strong>and</strong><br />
for multiple fin vehicles, the phasing<br />
of the fins is critical. In addition to<br />
comm<strong>and</strong>ing these parameters, for a<br />
given vehicle maneuver, the fin controller<br />
must respond to perturbations.<br />
Fortunately, there are bioinspired solutions<br />
to such control issues. One of<br />
the key components of the mammalian<br />
motor control system is the olivocerebellar<br />
system. Coupled neurons<br />
in the inferior olive generate a 10-Hz<br />
rhythm, <strong>and</strong> animal muscle contractions<br />
are initiated from particular<br />
phases of this rhythm. The inferior<br />
olive has multiple oscillating domains,<br />
with phase shifts between these domains,<br />
<strong>and</strong> these domains are under<br />
the control of the cerebellar cortex,<br />
which is a recipient of sensory inputs<br />
from many modalities. Llinas et al.<br />
(2004) developed a model whereby<br />
sequences of motor comm<strong>and</strong>s can<br />
be generated by the olivo-cerebellar<br />
system. The biophysics of coupling<br />
of these neurons also promotes a<br />
phase reset property for rapid synchronization<br />
(Kazantsev et al.,<br />
2004). B<strong>and</strong>yopadhyay (2008) has<br />
implemented this model in analog circuits<strong>and</strong>hasshownittobeaneffective<br />
controller of power <strong>and</strong> thrust in<br />
thehigh-liftfoils<strong>and</strong>theabilityto<br />
rapidly return to normal following<br />
perturbations. This model is able to<br />
synchronize across multiple fins in<br />
order to achieve optimum gaits that<br />
minimize pitching <strong>and</strong> rolling of the<br />
BAUV.Thiscontrollerenabledthe<br />
BAUV to precisely hold a line load<br />
for 20 days. Based on these laboratory<br />
results, the BAUV vehicle is predicted<br />
to support a mission duration of<br />
3 weeks with current battery technology,<br />
although this assumes most of<br />
that time is spent in hover.<br />
For fish <strong>and</strong> eels that use whole<br />
body or caudal fin propulsion,the<br />
oscillations are generated by central<br />
pattern generators. However, the modulation<br />
of these neural generators<br />
during locomotion to achieve desired<br />
thrust <strong>and</strong> vectors has not been fully<br />
characterized for animals, <strong>and</strong> using<br />
this approach in artificial systems requires<br />
learning or genetic algorithms<br />
to tune them to particular maneuvers.<br />
20 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
3. Exploitation of fish swimming<br />
modes for underwater vehicle propulsion<br />
<strong>and</strong> maneuvers. Fish have a<br />
number of swimming modes that are<br />
worthy of consideration for emulation.<br />
Fish swimming types can be<br />
classified as either body <strong>and</strong>/or caudal<br />
fin movements vs. those that use median<br />
<strong>and</strong>/or paired fin propulsion.<br />
One review of fish swimming modes<br />
(Sfakiotakis et al., 1999) singled out<br />
lunate tail propulsion (e.g., tuna), undulating<br />
fins (e.g., some rays), <strong>and</strong><br />
labriform (oscillatory pectoral fin)<br />
swimming mechanisms as having the<br />
greatest potential for exploitation in<br />
artificial systems. ONR is currently<br />
supporting research to exploit all<br />
three of these mechanisms in addition<br />
to the gymnotiform mode that involves<br />
undulations of a long ventral<br />
median fin. Batoid fishes utilize one<br />
of two modes of locomotion, employing<br />
either undulatory (passing multiple<br />
waves down the fin orbody)<br />
or oscillatory (flapping) kinematics<br />
(Rosenberger, 2001). An ambitious<br />
effort to build batoid vehicles, using<br />
predominately oscillating kinematics,<br />
is being undertaken by a team led<br />
by Hillary Bart-Smith (described in<br />
Moored, Fish, Kemp, & Bart-Smith,<br />
this issue) supported by an ONR program<br />
managed by Robert Brizzolara.<br />
Two prototype manta vehicles were<br />
recently demonstrated in a student<br />
competition (Pennisi, 2011).<br />
Exploitation of fish locomotion<br />
using lunate tail fins began with<br />
the seminal studies of Triantafyllou<br />
(Barrett et al., 1996; Anderson et al.,<br />
1998) analyzing how bio-inspired<br />
foils create <strong>and</strong> exploit vortex structures.<br />
His laboratory also developed<br />
the interesting Robotuna (based on<br />
the bluefin tuna) <strong>and</strong> Robopike prototypes,<br />
which were propelled using<br />
caudal fins. Triantafyllou’s group<br />
demonstrated very high-propulsion<br />
efficiencies (up to 91%) for the<br />
Robotuna; however, the Robotuna<br />
did not achieve very high speeds.<br />
One of the creators of Robotuna,<br />
David Barrett at Olin College, is now<br />
teamed with Boston Engineering to<br />
develop the Ghostswimmer vehicle<br />
(described in Rufo’s commentary<br />
in this issue). The objective of the<br />
Ghostswimmer project is to exceed<br />
the performance of current similar<br />
size UUVs on speed, maneuver, mission<br />
duration, noise, rapid response<br />
<strong>and</strong> cost. The goal is to demonstrate<br />
the capability to conduct fully autonomous<br />
missions.<br />
Another active research area seeks<br />
to exploit the principles of fish pectoral<br />
fins for propulsion <strong>and</strong> maneuver.<br />
Fish pectoral fins are highly flexible<br />
<strong>and</strong> enable a high degree of precise<br />
maneuver <strong>and</strong> station keeping in<br />
currents. Sunfish fins have flexible<br />
rays, <strong>and</strong> attached membranes exhibit<br />
a cupping motion on the forward<br />
stroke that produces upper <strong>and</strong> lower<br />
leading edge vortices. These fins generate<br />
positive thrust throughout the fin<br />
beat, <strong>and</strong> turning involves asymmetric<br />
use of these fins. Small prototype fins<br />
with this ray <strong>and</strong> membrane structure<br />
have been shown in the laboratory,<br />
but no free swimming vehicles have<br />
yet been developed (Gottlieb et al.,<br />
2010; Tangorra & Lauder, 2011, this<br />
issue). Rigid pectoral fins have been<br />
implemented on a number of swimming<br />
vehicles, but these are mainly<br />
used as control surfaces.<br />
Aquatic animals that are propelled<br />
by jetting can also provide inspiration<br />
for novel propulsion mechanisms.<br />
Mohseni (Krieg & Mohseni, 2010;<br />
Krieg et al., this issue) has characterized<br />
the biomechanics <strong>and</strong> hydrodynamics<br />
of squid propulsion <strong>and</strong><br />
developed new bioinspired thrusters,<br />
<strong>and</strong> Priya <strong>and</strong> his colleagues (see<br />
Joshi et al., this issue) have built prototype<br />
jellyfish that closely mimic<br />
their swimming modes.<br />
4. Development of muscle-like<br />
actuators. To fully exploit animallike<br />
locomotion <strong>and</strong> sensorimotor<br />
control mechanisms, it is essential to<br />
develop muscle-like actuators, linear<br />
actuators, adaptive compliant structural<br />
materials, <strong>and</strong> elastic skins with<br />
properties much closer to biological<br />
systems than current technology.<br />
There are substantial inefficiencies<br />
with implementing bio-inspired locomotion<br />
using rotary motors <strong>and</strong> complex<br />
power transmission systems. Two<br />
recent efforts to develop fins based on<br />
ionic polymer-metal composites are<br />
described in the papers by Kim et al.<br />
<strong>and</strong> Tan in this issue.<br />
5. Closed-loop control of bioinspired<br />
underwater vehicles. In<br />
crustaceans, there are detailed accounts<br />
of sensorimotor reflexes for control of<br />
tail <strong>and</strong> legs, <strong>and</strong> robotic lobsters have<br />
been built with many levels of bioinspiration<br />
(Ayers & Crisman, 1992;<br />
Ayers et al., this issue). However, for<br />
fish, the neural mechanisms by which<br />
the sensory afferent information on<br />
motion, flows, <strong>and</strong> hydroacoustic<br />
pressure is processed, integrated <strong>and</strong><br />
relayed to motor neurons is not<br />
known. One reason for this is that<br />
electrophysiological recording in<br />
swimming fish is technically challenging.<br />
One open question is whether fish<br />
can sense vortices <strong>and</strong> maneuver or<br />
time fin movements to exploit them<br />
or avoid them. The ONR is currently<br />
supporting a number of efforts in<br />
closed-loop control <strong>and</strong> bionavigation.<br />
These projects include identifying the<br />
role of hydroacoustic receptors in the<br />
lateral line <strong>and</strong> fin <strong>and</strong>bodymechano-reception<br />
in flow <strong>and</strong> vortex<br />
sensing <strong>and</strong> tracing the connectivity<br />
July/August 2011 Volume 45 Number 4 21
of these receptors into spinal sensorimotor<br />
circuits involved in locomotion<br />
<strong>and</strong> fin control (Green et al., 2011;<br />
Green & Hale, 2011). This project is<br />
also developing robotic fish prototypes<br />
(see Tangorra et al., this issue; Lauder<br />
et al., this issue) with detailed pectoral<br />
fins. This work entails some basic research<br />
on fish neurophysiology, since<br />
circuit-level neurophysiology is much<br />
more challenging in fish than in terrestrial<br />
vertebrates, <strong>and</strong> many of the<br />
spinal motor reflexes <strong>and</strong> circuits involved<br />
in limb control that were well<br />
established for mammals by the<br />
1980s are still in nascent stage for the<br />
pectoral fins of fish.<br />
6. Exploitation of special senses<br />
of aquatic animals. Anumberof<br />
fish species <strong>and</strong> sharks are able to navigate<br />
<strong>and</strong> search for prey using electroreception;<br />
some animals also exploit<br />
geomagnetic signals for navigation,<br />
<strong>and</strong> the biosonar of dolphins supports<br />
a range of behaviors. The ribbon fish<br />
can sense prey in murky waters using<br />
electrosense <strong>and</strong> then use its ventral<br />
ribbon fin to maneuver to this prey.<br />
Ongoing ONR projects are performing<br />
system identification of visual<br />
<strong>and</strong> eletroreceptive feedback control of<br />
locomotion in the ribbon fish (Roth<br />
et al., 2011; Mitchell et al., 2011) to<br />
build a working electrosense <strong>and</strong> electromagetosense<br />
module for navigation<br />
<strong>and</strong> target localization <strong>and</strong> to build a<br />
prototype ribbon fish (“Ghostbot”)<br />
(Curet et al., 2011). A vehicle designed<br />
to sense electric <strong>and</strong> magnetic anomalies<br />
could enable new mine countermeasures<br />
systems or enable navigation<br />
that does not depend on GPS or<br />
expensive Inertial Measurement Units<br />
(IMUs).<br />
Sharks exhibit exquisite sensitivity<br />
to perturbations of electric fields <strong>and</strong><br />
one current project seeks to develop<br />
new highly sensitive electromagnetic<br />
<strong>and</strong> hydroacoustic sensors based on<br />
shark sensor biophysics (Kalmijn,<br />
Scripps).<br />
Dolphins have extraordinary abilities<br />
to recognize objects using biosonar.<br />
ONR has supported research characterizing<br />
this ability in terms of psychoacoustics,<br />
<strong>and</strong> several dolphin-inspired<br />
sonars have been developed <strong>and</strong> demonstrated.<br />
Recently, Forsythe et al.<br />
(2008) demonstrated closed-loop<br />
control of his bio-inspired autonomous<br />
underwater vehicle using a simple<br />
dolphin-inspired biosonar in the exploration<br />
of acoustic targets. Dolphins,<br />
however, have the disadvantage that<br />
one cannot analyze directly the neural<br />
circuits involved in their biosonar.<br />
Hence, from the earliest days of its existence,<br />
ONR has supported the study<br />
of bat biosonar. The neural circuits of<br />
bat biosonar have been well characterized.<br />
Recently research has focused on<br />
how bats use multistatic biosonar for<br />
obstacle avoidance. Bats not only have<br />
an extraordinary ability to navigate<br />
around obstacles <strong>and</strong> strike prey<br />
using their own sonar, but they can<br />
also perform these feats using the returns<br />
from calls emitted by other bats<br />
in a swarm. This has implications for<br />
the cooperative behavior of multiple<br />
AUVs.<br />
7. Group behaviors. The ONR<br />
Science of Autonomy program is supporting<br />
efforts to identify the principles<br />
of fish schooling <strong>and</strong> applying them to<br />
multiple AUVs conducting ocean surveys.<br />
Such studies have addressed the<br />
minimal sensing requirements of fish<br />
used in schooling, including vision.<br />
Additionally, for AUVs, one could conceivably<br />
use man-made communications<br />
like acoustic modems or lasers to<br />
enhance coordinated actions, but successful<br />
emulation of sense modalities<br />
used in nature (i.e., fish hydroacoustic<br />
sensing) might greatly simplify coordination<br />
of multiple vehicles—in particular<br />
when a large number of vehicles<br />
is operating at close quarters. There is<br />
also an effort to emulate the group<br />
hunting behaviors of dolphins in intelligent<br />
control systems for cooperative<br />
autonomous vehicles.<br />
8. The Navy is interested in distributed,<br />
persistent sensing systems<br />
for tasks such as anti-submarine warfare.<br />
One bioinspired approach to<br />
this is to develop small sensor platforms<br />
that float in the water column<br />
with limited mobility, similar to jellyfish.<br />
An ambitious effort to exploit<br />
multifunctional materials <strong>and</strong> build<br />
such jellyfish colonies (Priya) is underway.<br />
Additionally, the emerging convergence<br />
of nanotechnology, synthetic<br />
biology <strong>and</strong> micro-optoelectronics has<br />
fostered the nascent development of<br />
very small autonomous systems. Some<br />
of the new microrobots being designed<br />
have components consisting of colonies<br />
of micro-organisms <strong>and</strong> hybrid<br />
biomolecules, with motility being<br />
provided by a multitude of cilia or<br />
flagella. These new systems could<br />
function as distributed sensors of hazardous<br />
materials, but with integrated<br />
mobility, mitigation <strong>and</strong> reporting<br />
capabilities.<br />
In summary, there are a number of<br />
promising opportunities for the Navy<br />
to extend the capabilities of autonomous<br />
undersea platforms using bioinspired<br />
technologies in propulsion,<br />
control <strong>and</strong> sensing.<br />
Author:<br />
Dr. Thomas McKenna<br />
Office of Naval Research<br />
Division of Human <strong>and</strong><br />
Bioengineered Systems<br />
875 N. R<strong>and</strong>olph St., Suite 1425<br />
Arlington, VA 22203-1995<br />
Email: tom.mckenna@navy.mil<br />
22 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
References<br />
Anderson, J.M., Streitlien, K., Barrett, D.S.,<br />
& Triantafyllou, M.S. 1998. Oscillating<br />
foils of high propulsive efficiency. J Fluid<br />
Mech. 360:41-72. doi: 10.1017/<br />
S0022112097008392.<br />
Ayers, J., & Crisman, J. 1992. The lobster as<br />
a model for an omnidirectional robotic<br />
ambulation control architecture. In: Biological<br />
Neural Networks in Invertebrate Neuroethology<br />
<strong>and</strong> Robotics, eds. Beer, R., Ritzmann, R.,<br />
& McKenna, T., 287-316. San Diego:<br />
Academic Press.<br />
B<strong>and</strong>yopadhyay, P.R. 2005. Trends in biorobotic<br />
autonomous undersea vehicles. IEEE J<br />
Oceanic Eng. 30(1):109-39. doi: 10.1109/<br />
JOE.2005.843748.<br />
B<strong>and</strong>yopadhyay, P.R. 2009. Swimming <strong>and</strong><br />
flying in nature—The route toward applications:<br />
The Freeman Scholar lecture. ASME J Fluid<br />
Eng. 131:031801-29. doi: 10.1115/1.3063687.<br />
B<strong>and</strong>yopadhyay, P., Beal, D.N., & Menozzi,<br />
A. 2008. Biorobotic insights into how animals<br />
swim. J Exp Biol. 211:206-14. doi: 10.1242/<br />
jeb.012161.<br />
B<strong>and</strong>yopadhyay, P.R. 2008. Synchronization<br />
of animal-inspired multiple high-lift fins<br />
in an underwater vehicle using olivo-cerebellar<br />
dynamics. IEEE J Oceanic Eng. 33(4):<br />
563-78. doi: 10.1109/JOE.2008.2005356.<br />
Barrett, D., Grosenbaugh, M., & Triantafyllou,<br />
M. 1996. The optimal control of a flexible<br />
hull robotic undersea vehicle propelled by an<br />
oscillating foil. In: Proc. 1996 IEEE AUV<br />
Symp. 1-9.<br />
Curet, O.M., Patankar, N.A., Lauder, G.V.,<br />
& MacIver, M.A. 2011. Mechanical properties<br />
of a bio-inspired robotic knifefish with an<br />
undulatory propulsor. Bioinspir Biomim.<br />
6(2). doi: 10.1088/1748-3182/6/2/026004.<br />
Dickinson, M.H., Lehman, F., & Sane, S.P.<br />
1999. Wing Rotation <strong>and</strong> the aerodynamic<br />
basis of insect flight. Science. 284:1954-60.<br />
doi: 10.1126/science.284.5422.1954.<br />
Ellington, C.P. 1984. The aerodynamics<br />
of hovering insect flight. IV. Aerodynamic<br />
mechanisms. Philos T R Soc Lond B.<br />
305:79-113.<br />
Forsythe, S.E., Leinhos, H.A., &<br />
B<strong>and</strong>yopadhyay, P.R. 2008. Dolphininspired<br />
combined maneuvering <strong>and</strong> pinging<br />
for short-distance echolocation. J Acoust<br />
Soc Am. 124(4):255-61.<br />
Gottlieb, J.R., Tangorra, J.L., Esposito,<br />
C.J., & Lauder, G.V. 2010. A biologically<br />
derived pectoral fin for yaw turn maneuvers.<br />
Appl Bionics Biomech. 7:41-55.<br />
doi: 10.1080/11762320903537782.<br />
Green, M.H., Ho, R.K., & Hale, M.E. 2011.<br />
Movement <strong>and</strong> function of the pectoral fins<br />
of the zebrafish (Danio rerio) during slow<br />
swimming. J Exp Biol. in review.<br />
Green, M.H., & Hale, M.E. 2011. Motoneurons,<br />
muscles <strong>and</strong> movement: Neural<br />
control of the pectoral fins during slow <strong>and</strong><br />
fast swimming. Integr Comp Biol. E51.<br />
Kazantsev, V.B., Nekorkin, V.I., Makarenko,<br />
V.I., & Llinas, R. 2004. Self-referential phase<br />
reset based on inferior olive oscillator dynamics.<br />
P Natl Acad Sci. 101(52):18183-8.<br />
doi: 10.1073/pnas.0407900101.<br />
Krieg, M., & Mohseni, K. 2010. Dynamic<br />
modeling <strong>and</strong> control of biologically inspired<br />
vortex ring thrusters for underwater robot<br />
locomotion. IEEE T Robot. 26(3):542-54.<br />
doi: 10.1109/TRO.2010.2046069.<br />
Llinas, R., Leznick, E., & Makarenko, V.I.<br />
2004. The olivo-cerebellar circuit as a universal<br />
motor control system. IEEE J Ocean Eng.<br />
29(3):631-9. doi: 10.1109/JOE.2004.833212.<br />
Mitchell, T., Fortune, E.S., & Cowan, N.J.<br />
2011. Reweighting of vision <strong>and</strong> electrosensation<br />
during locomotion in the Glass Knifefish.<br />
PLoS, to be submitted.<br />
Pennisi, E. 2011. Manta machines. Science.<br />
332:28-9. doi: 10.1126/science.332.6033.1028.<br />
Rosenberger, L.J. 2001. Pectoral fin locomotion<br />
in batoid fishes: Undulation versus<br />
oscillation. J Exp Biol. 204:379-94.<br />
Roth, E., Zhuang, K., Stamper, S., Fortune,<br />
E.S., & Cowan, N.J. 2011. Stimulus predictability<br />
mediates a switch in locomotor<br />
smooth pursuit performance for Eigenmannia<br />
virescens. J Exp Biol. 214:1170-80.<br />
doi: 10.1242/jeb.048124.<br />
Sfakiotakis, M., Lane, D.M., & Davies,<br />
J.B.C. 1999. Review of fish swimming modes<br />
for aquatic locomotion. IEEE J Oceanic Eng.<br />
24(2):237-52. doi: 10.1109/48.757275.<br />
Tangorra, J.L., & Lauder, G.V. 2011. Biorobotic<br />
models of highly deformable fins for<br />
Studies of the mechanics <strong>and</strong> control of<br />
fin forces. Integr Comp Biol. in press.<br />
doi: 10.1093/icb/icr036.<br />
July/August 2011 Volume 45 Number 4 23
COMMENTARY<br />
GhostSwimmer AUV: Applying <strong>Biomimetics</strong><br />
to Underwater Robotics for Achievement<br />
of Tactical Relevance<br />
AUTHORS<br />
Michael Rufo<br />
Mark Smithers<br />
Boston Engineering Corporation<br />
Introduction<br />
I<br />
t is clear that unmanned underwater<br />
vehicles (UUVs), autonomous<br />
underwater vehicles (AUVs), <strong>and</strong><br />
other waterborne robots are successfully<br />
addressing critical capability<br />
requirements for many customers, including<br />
those in defense <strong>and</strong> oil <strong>and</strong><br />
gas. There remains, however, despite<br />
the best efforts of many entities, capability<br />
gaps for these systems.<br />
While UUVs share many of the<br />
same navigation, power, <strong>and</strong> logistical<br />
challenges as their unmanned ground<br />
vehicle (UGV) <strong>and</strong> unmanned aerial<br />
vehicle (UAV) counterparts, they are<br />
subject to additional challenges including<br />
limited communications <strong>and</strong> harsh<br />
environments.<br />
Developing biologically-inspired<br />
(or biomimetic) technologies can provide<br />
some guidance for next generation<br />
systems. The Advanced Systems<br />
Group (ASG) at Boston Engineering<br />
(Waltham, MA) is developing the<br />
GhostSwimmer<br />
<br />
AUV, for example,<br />
which endeavors to attack many of the<br />
problems facing current UUVs. The<br />
increasing interest in the use of longrange/long-duration<br />
UUVs for littoral<br />
observation, military surveillance, <strong>and</strong><br />
other missions dem<strong>and</strong>s alternative or<br />
new technology development. As per<br />
the U.S. Navy UUV Master Plan<br />
(2004), the areas of autonomy, sensors,<br />
<strong>and</strong> communications are<br />
among the leading areas of interest.<br />
Energy <strong>and</strong> propulsion are also main<br />
areas of interest as per this document,<br />
where it states that “advanced energy<br />
<strong>and</strong> propulsion, in combination with<br />
other UUV technologies, will enable<br />
the use of smaller vehicles (reducing<br />
cost) in the long term, <strong>and</strong> will provide<br />
greater performance.” (The<br />
Navy UUV Master Plan, 2004).<br />
GhostSwimmer is a tactical, biomimetic<br />
autonomous “artificial fish”<br />
UUV that employs the mechanics<br />
<strong>and</strong> dynamics of biological systems to<br />
create efficient swimming <strong>and</strong> high<br />
maneuverability while remaining responsive<br />
to the needs of current riverine<br />
<strong>and</strong> littoral missions (among other<br />
possibilities). Its importance lies in its<br />
ability to provide advanced mobility<br />
in a system that employs payloads.<br />
Considerable work has been completed<br />
on underst<strong>and</strong>ing the propulsive<br />
characteristics of individual fish<br />
fins (extensive information on oscillating<br />
foil propulsion exists). The<br />
GhostSwimmer<br />
<br />
builds upon this<br />
<strong>and</strong> is modeled after a tuna. It is propelled<br />
by a composite construction lunate<br />
caudal tail fin (takingitsname<br />
from its crescent moon-like profile).<br />
In the biological tuna, almost all of<br />
theusablethrustisgeneratedbythe<br />
oscillating tail fin. Magnuson describes<br />
it as “a tapered hydrofoil with high<br />
aspect ratio, curved leading edge <strong>and</strong><br />
moderate sweep back. In cross section,<br />
the fin is shaped like a thin symmetrical<br />
airfoil with a rounded anterior, or<br />
leading edge, <strong>and</strong> a sharp posterior or<br />
trailing edge” (Magnuson, 1978).<br />
This oscillating foil has the capability<br />
of producing thrust at high efficiencies.<br />
Triantafyllou <strong>and</strong> Triantafyllou<br />
(1993) stated “Oscillating foils are<br />
known under certain conditions to<br />
produce substantial thrust at high propulsive<br />
efficiency. Fish <strong>and</strong> cetaceans<br />
have developed through evolution a<br />
presumed optimal manner of propulsion<br />
primarily through flapping of<br />
the posterior part of their body. It is<br />
shown that thrust production depends<br />
significantly on the dynamics of the<br />
unstable wake formed behind the foil<br />
or fish tail: thrust develops through<br />
the formation of a reverse Karman<br />
street, whose preferred Strouhal numbers<br />
are between 0.25 <strong>and</strong> 0.35. Experimental<br />
data from flapping airfoils<br />
<strong>and</strong> data from fish observations confirm<br />
the theoretical predictions.”<br />
Triantafyllou <strong>and</strong> Barrett were able to<br />
construct a precision oscillating foil<br />
test apparatus with which propulsive<br />
efficiencies in the range of 80%<br />
were repeatedly measured (Barrett,<br />
1996).<br />
While based on this concept,<br />
GhostSwimmer<br />
<br />
strives to further<br />
the underst<strong>and</strong>ing of how fish use all<br />
their fins to enhance propulsive performance,<br />
maneuverability, <strong>and</strong> stealth<br />
characteristics (all the while providing<br />
a capability for the Warfighter).<br />
24 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
Overcoming the Challenges<br />
for the Future AUVs<br />
At this time, UUVs <strong>and</strong> AUVs<br />
struggle with complex underwater<br />
environments such as shallow <strong>and</strong><br />
obstacle-filled waters. These vehicles<br />
are subject to an intrinsic paradox;<br />
they require stable dynamic behavior<br />
for maintaining intended path/heading<br />
without changing control surfaces or<br />
propulsion force. However, in complex<br />
environments, the vehicle must also be<br />
able to quickly change both path <strong>and</strong><br />
speed safely. Dynamic stability <strong>and</strong> response<br />
aspects such as overshoot, rise<br />
time, <strong>and</strong> peak time become critical<br />
(Azarsinal et al., 2007). Based on their<br />
prowess in these areas, it makes sense to<br />
investigate biological sources of inspiration<br />
in the development of new<br />
man-made systems hoping to excel in<br />
these environments.<br />
The Boston Engineering Advanced<br />
Systems Group’s GhostSwimmer<br />
<br />
AUV (Figure 1) is intended to leverage<br />
this biological inspiration. Existing<br />
AUVs are generally propelled by rotary<br />
propellers driven by electric motors<br />
with energy stored in batteries. Small<br />
diameter propellers typically operate<br />
at low efficiencies <strong>and</strong> can suffer serious<br />
lag times in transient response<br />
(Barrett, 1996). Cost effective propeller<br />
improvements are often limited by<br />
a maximum practical diameter that can<br />
be mounted on a UUV. In complex<br />
underwater areas, rotating propellers<br />
FIGURE 1<br />
Boston Engineering’s GhostSwimmer PH I<br />
AUV (sponsor: ONR Code 341).<br />
also present a significant snagging<br />
risk. Energy technology, despite recent<br />
progress, is awaiting a breakthrough to<br />
provide significantly larger (<strong>and</strong> safer)<br />
power densities; therefore, producing<br />
a technology that can address both<br />
efficiency <strong>and</strong> control, independent<br />
of battery technology, has benefits.<br />
Additional challenges that UUVs<br />
are subject to range from being as simple<br />
as packaging electronics to be water<br />
tight at depth (versus splash-proof ),<br />
to using materials that are resistant to<br />
aggressive corrosion, to being able to<br />
localize underwater without the aid of<br />
conventional techniques (such as the<br />
GPS or radio frequency [RF] communications).<br />
As discussed later in this<br />
commentary, biomimetics can offer<br />
guidance in these areas as well.<br />
<strong>Biomimetics</strong> <strong>and</strong><br />
Pragmatism<br />
<strong>Biomimetics</strong> (from the Greek bios<br />
[life] + mimesis [to imitate]) is not<br />
new. Mankind has been mimicking<br />
nature in products for centuries. Simple<br />
everyday items such as salad tongs<br />
(based on bird beaks) through complex<br />
sensors such as sonar (based on<br />
dolphins <strong>and</strong> bats) have each been<br />
generated by mimicking biology.<br />
Even human terminology mimics nature,<br />
consider saw “teeth”, computer<br />
“viruses” <strong>and</strong> “worms”, orlaptops<br />
that “hibernate.” Engineers <strong>and</strong> researchers<br />
have even mimicked plants<br />
in robotic systems (turning to face<br />
the sun for charging, for example)<br />
(Bar-Cohen, 2006). It should be<br />
noted that this discussion is not<br />
about synthetic life (involving using<br />
biological components to build biological<br />
systems).<br />
Nature evolves by responding to<br />
needs for surviving to the next generation.<br />
It benefits from millions of<br />
variations <strong>and</strong> trial <strong>and</strong> error experiments<br />
over the millennia. Engineers<br />
trying to develop tactically relevant<br />
technologies obviously must “evolve”<br />
much faster <strong>and</strong> in a pragmatic manner.<br />
As such, developing biomimetic<br />
technologies is as much about deciding<br />
what not to imitate as it is deciding<br />
what to imitate (Figure 2). This is because<br />
direct <strong>and</strong> absolute mimicry is<br />
often not appropriate or necessary.<br />
Consider that mankind flies without<br />
flapping wings as birds do. While underst<strong>and</strong>ing<br />
<strong>and</strong> mimicking the control<br />
surface design has direct benefits<br />
to engineer fast <strong>and</strong> high flying aircraft,<br />
engineers diverted from nature’s design<br />
with great success. However, the<br />
definition of the mission must always<br />
be at the core of any engineering effort.<br />
If the goal were to perch on a power<br />
line, even mankind’s best aircraft<br />
would fail.<br />
Evolutionary improvements in nature<br />
also do not always exactly coincide<br />
with the reasons for mimicking them.<br />
This should be considered when deciding<br />
whether the model is appropriate<br />
or ideal for mimicking in an engineered<br />
system designated for use in<br />
critical applications. Clearly, biologists<br />
<strong>and</strong> others trying to learn about “how<br />
biology works” are concerned with<br />
mimicking exactly how a biological<br />
FIGURE 2<br />
Early attempts at biomimetics often missed<br />
the mark (Fuller).<br />
July/August 2011 Volume 45 Number 4 25
model operates, its exact structure, its<br />
mechanical parameters, <strong>and</strong> more.<br />
The focus of this commentary is on<br />
the development of field-capable technologies<br />
for performing unmanned<br />
tasks in relevant mission spaces. In<br />
this sense, it is important to recognize<br />
that animals were evolved for survival<br />
in a particular environment with specific<br />
predators <strong>and</strong> other influences.<br />
In the GhostSwimmer<br />
<br />
case, extant<br />
tuna appeared roughly 60 million<br />
years ago (Dickson & Graham, 2004).<br />
These early tunas lived in a large circumtropical<br />
waterway that encircled<br />
theentireplanet(theTethysSea).<br />
This waterway existed for roughly<br />
50 million years during which the development<br />
of tunas was influenced by<br />
changes in the oceanography, induced<br />
primarily by tectonic activity. These<br />
changes increased productivity, exp<strong>and</strong>ed<br />
food webs, <strong>and</strong> opened potential niches<br />
(Dickson & Graham, 2004).<br />
Graham <strong>and</strong> Dickson suggest that<br />
the appearance of these more extensive<br />
ocean areas with high productivity <strong>and</strong><br />
diversified food webs provided the<br />
catalyst for tuna evolution: enhanced<br />
locomotor performance <strong>and</strong> favored<br />
migratory behavior. Tuna’s specializations,<br />
thunniform swimming, capacity<br />
for regional endothermy, <strong>and</strong> an elevated<br />
aerobic capacity, are thought to<br />
be based on the need for extended<br />
swimming (Graham & Dickson,<br />
2004). Despite the fact that this is different<br />
than the pressing missions of a<br />
UUV (such as mine counter measures),<br />
this extended swimming implies<br />
endurance <strong>and</strong> propulsion that<br />
isofvaluetoapragmaticmission.<br />
This leads to the question the roboticist<br />
must ask: What aspects of a<br />
fish’s “design” or mechanics is applicable<br />
to the desired mission<br />
Another consideration is the different<br />
resources that are available. Evolution<br />
<strong>and</strong> robotics engineers have<br />
distinctly different tool sets, materials,<br />
actuators, <strong>and</strong> power <strong>and</strong> control systems<br />
at their disposal. In certain<br />
areas, engineers have an advantage; exotic<br />
materials such as titanium may<br />
allow advances that nature could not<br />
provide. However, nature still has the<br />
advantage (at least currently) in actuator<br />
power density (when considering<br />
b<strong>and</strong>width <strong>and</strong> other factors), regenerative<br />
capability, sensing, <strong>and</strong> control<br />
prowess. For example, many biomechanics<br />
sources have noted that muscle<br />
tendon series elasticity is critical for<br />
energetic <strong>and</strong> efficiency purposes<br />
(Paluska & Herr, 2006). Many advances<br />
are in the works in the areas of<br />
artificial muscles, neural networks, <strong>and</strong><br />
similar technologies, but to date they<br />
lag behind their biological counterparts<br />
in key areas. Some areas in need<br />
of advancement for actuation include<br />
power to weight ratio, efficiency, fatigue<br />
life, <strong>and</strong> controllability.<br />
Where the Challenges<br />
of Underwater Operation<br />
<strong>and</strong> <strong>Biomimetics</strong> Intersect<br />
The challenges of achieving autonomy<br />
with unmanned systems are<br />
detailed in many articles <strong>and</strong> publications.<br />
From one perspective, autonomy<br />
intrinsically dem<strong>and</strong>s that<br />
engineered systems act like biological<br />
systems. For instance, these systems<br />
need obstacle detection <strong>and</strong> avoidance<br />
or situational awareness (“what<br />
is happening around me”), decisionmaking<br />
capacity (“should I respond<br />
aggressively or passively”), <strong>and</strong> health<br />
monitoring (“am I OK”). Interesting<br />
work in developing biologically<br />
inspired autonomy solutions based<br />
ontheseprinciplesiscurrentlyoccurring<br />
at Jet Propulsion Lab (JPL)<br />
(space mission systems) (Huntsberger,<br />
2001), Office of Naval Research<br />
(ONR) (sensory control, biomechanics)<br />
(ONR: Bio-Inspired Autonomous<br />
Systems), <strong>and</strong> Cornell (making increasingly<br />
complex machines) (Lipson).<br />
The modeling <strong>and</strong> underst<strong>and</strong>ing<br />
of unsteady hydrodynamics is also a<br />
challenge for underwater vehicle<br />
developers. While analogous to aerodynamic<br />
forces in some ways, hydrodynamics<br />
in unsteady, obstacle-laden<br />
environments is often more complex<br />
<strong>and</strong>, to date, difficult to model. In<br />
particular, controlled 6 degrees of freedom<br />
movement, particularly in unsteady<br />
conditions, is challenging for<br />
control, power, <strong>and</strong> propulsion systems.<br />
Underwater vehicles have a<br />
more challenging environment for<br />
controlled mobility than most UGVs<br />
due to this dynamic nature of their<br />
surroundings, <strong>and</strong> unless operating in<br />
open seas (blue water), they can face<br />
far more obstacles within tighter spaces<br />
than UAVs.<br />
Sensing <strong>and</strong> communication is another<br />
area where underwater systems<br />
have very different challenges. Biological<br />
systems have remarkable <strong>and</strong> innate<br />
abilities to detect obstacles or<br />
prey using a variety of sensory capabilities<br />
from echolocation (bottlenose<br />
dolphin biosonar is “probably the<br />
most sophisticated target location <strong>and</strong><br />
analysis system in existence”) (Fulton,<br />
2010) to lateral lines (in fish such as<br />
the tuna, water flow parallel to the motion<br />
deflects cilium in the lateral line<br />
which in turn defines water velocity)<br />
(Martiny et al., 2009), to electric fields<br />
(weakly electric black ghost knife fish<br />
generate an electric field that causes<br />
voltage perturbations due to the differences<br />
in electrical conductivity between<br />
an object <strong>and</strong> the water, thereby<br />
sensing its surroundings) (Martiny<br />
et al., 2009).<br />
26 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
However, unmanned systems can<br />
leverage only those sensory capabilities<br />
developed by the technology community<br />
at large. Herein lies the challenge,<br />
UAVs may have similar collision<br />
avoidance sensory needs, but the technologies<br />
for in-air sensing are currently<br />
more advanced <strong>and</strong>, due to relatively<br />
low signal attenuation in air, can effectively<br />
sense objects hundreds or<br />
thous<strong>and</strong>s of feet away (ignoring<br />
cloud cover <strong>and</strong> other aspects for sake<br />
of argument). The attenuation of<br />
various sensing signals in salt water is<br />
drastically more than in air, making<br />
technologies that many take for<br />
granted (WiFi <strong>and</strong> Bluetooth for example)<br />
impractical underwater. RF<br />
communications can work underwater<br />
but attenuation dem<strong>and</strong>s that low frequencies<br />
be used <strong>and</strong> the resulting<br />
trade-off is b<strong>and</strong>width (a reasonably<br />
sized acoustic modem can provide<br />
140 bps-15 kbps [www.benthos.com]<br />
where even limited WiFi [802.11 b for<br />
example] can provide 2 Mbps “in air”<br />
[Mitchell]). To achieve real-time<br />
or near real-time control, a UGV or<br />
UAV could use high-frequency,<br />
high-b<strong>and</strong>width RF communications.<br />
To date, this is a major challenge in underwater<br />
communications. Additionally,<br />
the change in medium<br />
(from air into <strong>and</strong> through water or<br />
vice versa) makes communication difficult<br />
due to refraction losses at the<br />
interface (losses are large for electromagnetic<br />
waves going from air into<br />
seawater <strong>and</strong> intrinsic wavelength differences<br />
make an underwater antenna<br />
for the same radio different than its<br />
in-air counterpart) (Butler, 2011).<br />
St<strong>and</strong>ard acoustic <strong>and</strong> light-based<br />
sensors are greatly affected by the attenuation<br />
as well. Additionally, their<br />
operation often dem<strong>and</strong>s smooth,<br />
controlled motion at relatively slow<br />
speeds. When coupling the hydrodynamic<br />
maneuverability challenges<br />
for AUVs with these sensing challenges,<br />
one can only marvel at the ability of<br />
fast-moving fish <strong>and</strong> marine mammals<br />
to detect <strong>and</strong> maneuver. For these animals,<br />
this detection <strong>and</strong> maneuver is a<br />
system level process where sensing<br />
provides exterioception <strong>and</strong> proprioception<br />
<strong>and</strong> enables the vorticity <strong>and</strong><br />
optimized flow control that is essential<br />
to maneuverability <strong>and</strong> fast swimming<br />
(Triantafyllou et al., 2002).<br />
The GPS is another technology<br />
taken for granted in the 21st century.<br />
However, current AUVs must surface<br />
to get GPS fixes because GPS will not<br />
work underwater; the signals from the<br />
satellites cannot sufficiently penetrate<br />
the water. This presents the AUV<br />
with a significant challenge in localization<br />
<strong>and</strong> navigation. Until these technologies<br />
advance in a cost effective<br />
manner for underwater applications,<br />
one of the most effective means for<br />
overcoming the challenge includes<br />
having the ability to surface quickly<br />
(<strong>and</strong> often covertly) where the manmade<br />
underwater system can then<br />
leverage in-air wireless <strong>and</strong> geopositioning<br />
technologies. This rapid<br />
surfacing can also benefit frombiologically<br />
inspired techniques, as the<br />
GhostSwimmer has implemented.<br />
Beyond Propulsion—<br />
Leveraging Biology<br />
to Advance the State<br />
of the Art<br />
Designing systems with bioinspired<br />
control systems can open avenues<br />
for efficiently performing, <strong>and</strong><br />
switching between, necessary behavior<br />
states. Advances in computing technology<br />
(increased memory <strong>and</strong> computational<br />
power in smaller, less<br />
expensive packages) now enables incorporation<br />
of newer <strong>and</strong> more advanced<br />
biologically inspired control<br />
systems. This includes the development<br />
of intelligent control through<br />
distributed control in systems containing<br />
their own “nervous systems,” structured<br />
like their biological analog that<br />
can then “learn” through adaptive<br />
techniques (Thomopoulos & Braught,<br />
1995).<br />
The study of biology itself can provide<br />
advances in the underst<strong>and</strong>ing of<br />
unsteady hydrodynamics as mentioned<br />
above. Biological systems<br />
achieve high efficiencies <strong>and</strong> maneuverability<br />
through the sensing, manipulation,<br />
<strong>and</strong> creation of optimized flow<br />
around them. This includes the manipulation<br />
of tip vortices at control<br />
<strong>and</strong> propulsive surfaces as well as the<br />
optimization of output motion to generate<br />
smooth <strong>and</strong> effective jets in the<br />
flow behind the body. The underst<strong>and</strong>ing<br />
of these phenomena is directly<br />
applicable to the creation of<br />
higher performance AUVs <strong>and</strong> even<br />
manned systems (Barrett, 1996).<br />
Boston Engineering’s ASG is currently<br />
applying many of these principles to<br />
develop improved control surfaces for<br />
the U.S. Navy Sea Systems Comm<strong>and</strong><br />
as well as advanced AUVs for ONR.<br />
Navigation is also an area that can<br />
benefit from biomimetics. Specific<br />
areas of interest in autonomy to the<br />
Navy include path planning, behavior<br />
development, localization, on-board<br />
mapping of environmental variability,<br />
<strong>and</strong> effective man-machine interfaces<br />
with a limited communication capability(Wernli,2001).Additionally,increasing<br />
uncertainty regarding Global<br />
Navigation Satellite Systems reliability<br />
has led unmanned systems developers<br />
to seek valuable alternatives. Integrating<br />
inertial systems with reference maps of<br />
Geophysical Fields of the Earth (GFE)<br />
is an area being explored by aerospace<br />
July/August 2011 Volume 45 Number 4 27
entities that offers promise through use<br />
of the recent advancements in embedded<br />
micro-processing, including<br />
memory devices’ capability <strong>and</strong> miniature<br />
size. GFEs, properties of the earth<br />
itself, are already well mapped in geographical<br />
system coordinates <strong>and</strong> can<br />
be considered a reliable navigation<br />
data source. Earth’s MagneticField<br />
(EMF) maps, models, <strong>and</strong> charts are<br />
currently in use for military <strong>and</strong> commercial<br />
entities (for directional information)<br />
<strong>and</strong> are available for at least<br />
98% of the earth’s surface (including<br />
water-covered areas) (Goldenberg,<br />
2006).<br />
Research <strong>and</strong> behavioral experiments<br />
have shown that various animals<br />
sense <strong>and</strong> use the earth’s magnetic field<br />
for navigation over both long <strong>and</strong> short<br />
distances (Johnsen & Lohmann,<br />
2005). Migratory animals, capable of<br />
sensing variations in geomagnetic<br />
fields, reference the earth’s mean field<br />
<strong>and</strong> its inclination at many points.<br />
Measurements of the field, including<br />
spatial variations <strong>and</strong> temporal evolutions,<br />
tabulated by the U.S. Geological<br />
Survey, allow the potential for geomagnetic<br />
field sensors to mimic animal<br />
behavior related to navigation (Zhai<br />
et al., 2007).<br />
Some animals, including certain<br />
birds, sea turtles, salam<strong>and</strong>ers, <strong>and</strong> lobsters<br />
can discriminate small differences<br />
in some of the earth’s magnetic features.<br />
They use positional information<br />
in the earth’s field in several different<br />
ways <strong>and</strong> some actually learn the magnetic<br />
topography of the areas they call<br />
home. Some have postulated that animals<br />
have two separate magnetosensory<br />
systems (Goldenberg, 2006). A<br />
compass alone is rarely sufficient to<br />
guide animals to specific destinations<br />
or along a long <strong>and</strong> complex migratory<br />
route due to currents <strong>and</strong> other errorinducing<br />
phenomena. Navigation<br />
must be enhanced by the ability to<br />
determine position relative to a destination<br />
(human travelers use a GPS),<br />
i.e., positional information inherent<br />
in the earth’s magneticfield provides<br />
a similar, although less precise, assessment<br />
of location (Goldenberg, 2006).<br />
Animals also use other cues; research<br />
has shown that bees navigate<br />
relative to the sun by using the sun as<br />
a fixed point <strong>and</strong> orienting themselves<br />
by maintaining a fixed angle between<br />
its line of flight <strong>and</strong> the line to the<br />
sun (www.physics.ohio-state.edu).<br />
Ocean waves combine with the earth’s<br />
magnetic field to serve as orientation<br />
cues for newly hatched turtles; while<br />
older turtles are following a map they<br />
learned that enables them to establish<br />
their position relative to some distant<br />
target (Lohmann et al., 2004). Regardless<br />
of surface currents or other<br />
oceanographic features, sockeye salmon<br />
use an internal “map sense” to navigate<br />
home after several migrations <strong>and</strong> induced<br />
errors. An olfactory “imprint”<br />
is made on smelts as they leave their<br />
home stream, but approaching their<br />
stream from open sea dem<strong>and</strong>s one<br />
other imprint. Fish are perceptive of<br />
the azimuth <strong>and</strong> altitude of the sun,<br />
but during overcast days, ferromagnetic<br />
mineral magnetite in their brain<br />
may function as a biological compass.<br />
Another means for a fish to sense the<br />
magnetic field is by merely moving<br />
through the water (like a wire moved<br />
across a magnetic field, electrical<br />
current occurs in the wire) (Gedney,<br />
1984). It is possible that AUVs<br />
could mimic these approaches.<br />
The electric field of ocean currents<br />
indicate to sharks their drift relative to<br />
the bottom or to deeper water layers.<br />
Electric current of an ocean stream invading<br />
a quiet bay may provide a shark<br />
with directional cues in familiar territory.<br />
They may also explore the fields<br />
by occasionally diving deeper or to the<br />
bottom <strong>and</strong> their orientating in uniform<br />
DC electric fields has been proven<br />
behaviorally. The electric sense operates<br />
in a passive mode, whereas magnetic<br />
field detection is an active mode, allowing<br />
them to simultaneously sense drift<br />
with ocean streams <strong>and</strong> magnetic headings.<br />
Based on this research, measured<br />
potentials, sensitivity <strong>and</strong> noise effects,<br />
<strong>and</strong> amplification are relatively well<br />
known (Kalmijn, 2000).<br />
The sun’s movement across the sky<br />
gives orientation signals that vary with<br />
time of day. Migratory birds gain information<br />
from the sun better than<br />
humans, detecting changing patterns<br />
of polarization. Using magnetism <strong>and</strong><br />
celestial rotation together can increase<br />
reliability (birds constantly cross check<br />
them <strong>and</strong> adjust) (www.teara.govt.nz).<br />
The sun’s position can be analyzed by<br />
looking at the polarization pattern of<br />
the sky arising from sunlight scattering<br />
(useful when the sun is obscured).<br />
Many insects use celestial polarization<br />
for compass orientation by using<br />
polarization-sensitive photoreceptors.<br />
Biological research has shown that<br />
some underwater animals use polarization<br />
of light for navigation, communication,<br />
<strong>and</strong> hunting. Studies have<br />
been performed using polarization of<br />
scattered light underwater to improve<br />
visibility. Polarization of light in water<br />
is caused by refraction through the surface,<br />
scattering light in water, refraction<br />
by polarizing objects, <strong>and</strong> emission<br />
from polarized light sources (Karpel &<br />
Schecher, 2011).<br />
The Present <strong>and</strong><br />
Future of Biomimetic<br />
Underwater Robotics<br />
Rather than relying on the future<br />
development of a novel <strong>and</strong><br />
28 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 3<br />
Boston Engineering’s GhostSwimmer PH I<br />
AUV in field testing.<br />
ground-breaking power source, Boston<br />
Engineering’sASG,alongwithitsteammate,<br />
Olin College Intelligent Vehicles<br />
Laboratory (Needham, MA), is taking<br />
inspiration from a comparably sized<br />
biological system to develop an AUV<br />
optimized for complex underwater<br />
environments.<br />
Thanks to funding from the ONR<br />
<strong>and</strong> a large internal R&D effort by<br />
both Boston Engineering <strong>and</strong> Olin<br />
College, this new generation of AUVs<br />
is emerging. By leveraging past research,<br />
the latest technologies, <strong>and</strong> innovative<br />
rapid prototyping techniques,<br />
the team developed an AUV that could<br />
autonomously swim like a fish in only<br />
6 months (Figure 3). The team is currently<br />
in Phase II with ONR <strong>and</strong> expects<br />
to showcase its next generation AUV in<br />
late 2011 to demonstrate advancements<br />
in efficiency <strong>and</strong> maneuverability while<br />
addressing <strong>and</strong> investigating the other<br />
challenges facing underwater vehicles<br />
as described above.<br />
While there have been many attempts<br />
to duplicate the swimming actions<br />
of a fish (Xianzhong DAI, 2003;<br />
Valdivia et al., 2006), a breakthrough<br />
achieved by the Boston Engineering<br />
research team has been in producing<br />
these movements efficiently within a<br />
tactically relevant vehicle. After proving<br />
that fish-like oscillating foil propulsion<br />
could indeed have higher<br />
propulsive efficiency <strong>and</strong> could be<br />
achieved by man-made methods, the<br />
next challenge is to properly control<br />
<strong>and</strong> coordinate the movements. The<br />
intendedsolutionforthiscontrolis<br />
itself bioinspired. The result is an<br />
AUV platform that could outperform<br />
conventional technologies deployed<br />
today in both endurance <strong>and</strong> mobility.<br />
This work holds the potential to provide<br />
a paradigm shift in underwater<br />
AUV capability <strong>and</strong> usefulness.<br />
The trend of mimicking biology<br />
in engineering certainly is not new,<br />
but through the efforts of the various<br />
parties working in the field of biologically<br />
inspired engineering <strong>and</strong> biological<br />
mechanics, a reduction in the<br />
number of limitations in enabling<br />
technologies <strong>and</strong> knowledge is being<br />
seen. Recent advances in the underst<strong>and</strong>ing<br />
of biological hydrodynamics,<br />
sensing, navigation techniques, <strong>and</strong><br />
control coupled with low-power highthroughput<br />
computing (including<br />
Quad Core Processors, www.intel.<br />
com), advances in material science<br />
(ionomers <strong>and</strong> self healing materials,<br />
www.bimat.org), <strong>and</strong> improved prototyping<br />
techniques (such as direct metal<br />
laser sintering, www.morristech.com)<br />
has enabled robotics engineers as<br />
never before. As industry <strong>and</strong> academia<br />
continue to make progress in demonstrating<br />
how effective biomimetic designs<br />
can be, the more developers of<br />
next generation tactical systems such<br />
as Boston Engineering’s Advanced<br />
Systems Group will be able to rise to<br />
the technical challenges presented by<br />
current adopters of unmanned technology,<br />
especially in the underwater<br />
space.<br />
Authors:<br />
Michael Rufo <strong>and</strong> Mark Smithers<br />
Advanced Systems Group,<br />
Boston Engineering Corporation<br />
411 Waverley Oaks Road,<br />
Waltham, MA 02452<br />
Emails: mrufo@boston-engineering.<br />
com; msmithers@boston-engineering.<br />
com<br />
References<br />
Azarsinal, F., Bose, N., & Seif, M.S. 2007.<br />
An underwater vehicle maneuvering simulation:<br />
Focus on turning maneuvers. J Ocean<br />
Technol. 2(1):54-73.<br />
Bar-Cohen, Y. 2006. <strong>Biomimetics</strong>: Biologically<br />
Inspired Technologies. Boca Raton,<br />
FL: CRC Press. 469 pp.<br />
Barrett, D.S. 1996. Propulsive efficiency of<br />
a flexible hull underwater vehicle. Ph.D.<br />
thesis, Massachusetts Institute of <strong>Technology</strong>.<br />
Bee Navigation. http://www.physics.ohiostate.edu/~wilkins/writing/Samples/shortmed/<br />
fiskemedium. Accessed June 20, 2011.<br />
Bird Migration. Navigating by the stars<br />
<strong>and</strong> sun. http://www.teara.govt.nz/<br />
EarthSeaAndSky/BirdsOfSeaAndShore/<br />
BirdMigration/7/en. Accessed June 20, 2011.<br />
Butler, L. 1987. Underwater Radio Communication.<br />
http://www.qsl.net/vk5br/<br />
UwaterComms.htm. Accessed June 20, 2011.<br />
Dickson, K.A., & Graham, J.B. 2004.<br />
Evolution <strong>and</strong> consequences of endothermy<br />
in fishes. Physiol Biochem Zool. 77(6):<br />
998-1018. doi: 10.1086/423743.<br />
Fuller, J. Top 10 bungled attempts at oneperson<br />
flight. http://science.howstuffworks.com/<br />
transport/flight/classic/ten-bungled-flightattempt4.htm.<br />
Accessed March 1, 2011.<br />
Fulton, J.T. 2010. Dolphin biosonar echolocation<br />
A case study. http://neuronresearch.net/<br />
July/August 2011 Volume 45 Number 4 29
hearing/pdf/Dolphin_biosonar_echolocation.<br />
pdf. Accessed June 20, 2011.<br />
Gedney, L. 1984. Do Salmon navigate by<br />
the earth’s magnetic field Article #691.<br />
http://www.gi.alaska.edu/ScienceForum/<br />
ASF6/691.html. Accessed June 20, 2011.<br />
Goldenberg, F. 2006. Geomagnetic navigation<br />
beyond the magnetic compass. Goodrich<br />
Corporation, Advanced Sensors Technical<br />
Center, 14300 Judicial Road, Burnsville,<br />
MN 55306. http://ieeexplore.ieee.org/xpl/<br />
freeabs_all.jsparnumber=1650662. Accessed<br />
June 20, 2011.<br />
Graham, J.B., & Dickson, K.A. 2004.<br />
Tuna comparative physiology. J Exp Biol.<br />
207(iii):4015-24. doi: 10.1242/jeb.01267.<br />
Huntsberger, T. 2001. Biologically inspired<br />
autonomous rover control autonomous<br />
robots. Auton Robot. 11:341-6.<br />
doi: 10.1023/A:1012467829785.<br />
Intel. Intel core quad processors. http://www.<br />
intel.com/products/processor/core2quad/.<br />
Accessed June 20, 2011.<br />
Johnsen, S., & Lohmann, K. 2005. The<br />
physics <strong>and</strong> neurobiology of magnetoreception.<br />
Nature Reviews Neuroscience. 6:703-12.<br />
doi: 10.1038/nrn1745.<br />
Kalmijn, J. 2000. Detection <strong>and</strong> processing<br />
of electromagnetic <strong>and</strong> near-field acoustic<br />
signals in elasmobranch fishes. Philos T R<br />
Soc Lond B. 355(1401):1135-41.<br />
doi: 10.1098/rstb.2000.0654.<br />
Karpel, N., & Schechner, Y. 2011. Portable<br />
polarimetric underwater imaging system with<br />
a linear response. Dept. of Electrical Eng.,<br />
Technion - Israel Institute of <strong>Technology</strong>,<br />
Haifa 32000, Israel. http://webee.technion.<br />
ac.il/~yoav/publications/polarimetric.pdf.<br />
Accessed June 20, 2011.<br />
Lipson, H. Associate Professor. Cornell<br />
University Mechanical <strong>and</strong> Aerospace Engineering.<br />
http://www.mae.cornell.edu/lipson/.<br />
Accessed June 20, 2011.<br />
Lohmann, K.J., Lohmann, C.M.F., Llewellyn,<br />
M., Ehrhart, L.M., Bagley, D.A., & Swing, T.<br />
2004. Animal behavior: Geomagnetic map<br />
used in sea-turtle navigation. Nature.<br />
428:909-10. doi: 10.1038/428909a.<br />
Magnuson, J.J. 1978. Locomotion by scombrid<br />
fishes: hydromechanics, morphology,<br />
<strong>and</strong> behavior. Fish Physiol. 7:239-313.<br />
doi: 10.1016/S1546-5098(08)60166-1.<br />
Martiny, N. 2009. Design of a lateral-line<br />
sensor for an autonomous underwater vehicle.<br />
In: 8th IFAC International Conference on<br />
Maneuvering <strong>and</strong> Control of <strong>Marine</strong> Craft,<br />
MCMC 2009. pp. 292-7. Sao Paulo, Brazil:<br />
IFAC.<br />
Mitchell, B. Wireless St<strong>and</strong>ards-802.11b<br />
802.11a 802.11g <strong>and</strong> 802.11n: The 802.11<br />
family explained. http://compnetworking.<br />
about.com/cs/wireless80211/a/aa80211st<strong>and</strong>ard.<br />
htm. Accessed June 20, 2011.<br />
Morris Technologies. Direct metal laser sintering<br />
(DMLS). http://www.morristech.com/<br />
dmls.asw. Accessed June 20, 2011.<br />
Office of Naval Research. Bio-inspired<br />
autonomous systems. http://www.onr.<br />
navy.mil/en/Media-Center/Fact-Sheets/<br />
Bio-Inspired-Autonomous-Systems.aspx.<br />
Accessed June 20, 2011.<br />
Paluska, D., & Herr, H. 2006. Series<br />
elasticity <strong>and</strong> actuator power output. In:<br />
Proceedings of the 2006 IEEE International<br />
Conference on Robotics <strong>and</strong> Automation.<br />
Orl<strong>and</strong>o, FL: IEEE.<br />
Teledyne benthos acoustic modems.<br />
http://www.benthos.com/acoustic-telesonarmodem-subsea-ATM885.asp.<br />
Accessed<br />
June 20, 2011.<br />
The Navy Unmanned Undersea Vehicle<br />
(UUV) Master Plan. November 9, 2004.<br />
Thomopoulos, S., & Braught, G. 1995. A<br />
biologically inspired mechanism of machine<br />
perception <strong>and</strong> intelligent control for multirobot<br />
coordination. Penn State University.<br />
Triantafyllou, M.S., Techet, A.H., Zhu, Q.,<br />
Beal, D.N., Hover, F.S., & Yue, D.K.P.<br />
2002. Vorticity control in fish-like propulsion<br />
<strong>and</strong> maneuvering. Integr Comp Biol.<br />
42(5):1026-31. doi: 10.1093/icb/42.5.1026.<br />
Triantafyllou, M.S., & Triantafyllou, G.S.<br />
1993. Propulsion through oscillating foils in<br />
nature <strong>and</strong> technology. J Fluids Struct.<br />
7:205-24. doi: 10.1006/jfls.1993.1012.<br />
Valdivia y Alvarado, P., & Youcef-Toumi, K.<br />
2006. Design of machines with compliant<br />
bodies for biomimetic locomotion in liquid<br />
environments. J Dyn Syst-T ASME.<br />
128(1,3):11.<br />
Wernli, R. 2001. Recent U.S. Underwater<br />
Vehicle Projects. San Diego, CA: Space <strong>and</strong><br />
Naval Warfare Systems Center.<br />
Wernli, R. 2011. Recent US Navy Underwater<br />
Vehicle Projects. http://www.nmri.go.<br />
jp/main/cooperation/ujnr/24ujnr_paper_us/<br />
Ocean_Engineering_<strong>and</strong>_Resource/OERD_<br />
Wernli.pdf. Accessed June 20, 2011.<br />
Xianzhong, D. 2003. Some new progress of<br />
advanced robotics in China. In: Report to JCF<br />
2003 of IARP, Xianzhong DAI Robotics<br />
Expert Group of National High-Tech R&D<br />
Programme, China. Madrid, Spain.<br />
Zhai, J. 2007. Geomagnetic sensor based on<br />
giant magnetoelectric effect. Appl Phys Lett.<br />
91(12):1-3. doi: 10.1063/1.2789391.<br />
30 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
PAPER<br />
Autonomous Robotic Fish as Mobile Sensor<br />
Platforms: Challenges <strong>and</strong> Potential Solutions<br />
AUTHOR<br />
Xiaobo Tan<br />
Smart Microsystems Laboratory,<br />
Department of Electrical <strong>and</strong><br />
Computer Engineering,<br />
Michigan State University<br />
Introduction<br />
ABSTRACT<br />
With advances in actuation <strong>and</strong> sensing materials <strong>and</strong> devices, there is a growing<br />
interest in developing underwater robots that propel <strong>and</strong> maneuver themselves<br />
as real fish do. Such robots, often known as robotic fish, could provide an engineering<br />
tool for underst<strong>and</strong>ing fish swimming. Equipped with communication<br />
capabilities <strong>and</strong> sensors, they could also serve as economical, dynamic samplers<br />
of aquatic environments. In this paper we discuss some of the major challenges in<br />
realizing adaptive, cost-effective, mobile sensor networks that are enabled by<br />
resource-constrained robotic fish. Such challenges include maneuvering in the<br />
presence of ambient disturbances, localization with adequate precision, sustained<br />
operation with minimal human interference, <strong>and</strong> cooperative control <strong>and</strong> sensing<br />
under communication constraints. We also present potential solutions <strong>and</strong> promising<br />
research directions for addressing these challenges, some of which are inspired<br />
by how fish solve similar problems.<br />
Keywords: robotic fish, adaptive sampling, mobile sensing platforms, aquatic<br />
sensor networks, water quality monitoring<br />
With 500 million years of evolution,<br />
fish <strong>and</strong> other aquatic animals<br />
areendowedwithavarietyofmorphological<br />
<strong>and</strong> structural features<br />
that enable them to move through<br />
water with speed, efficiency, <strong>and</strong> agility<br />
(Lauder & Drucker, 2004; Fish &<br />
Lauder, 2006). The remarkable feats<br />
in biological swimming have stimulated<br />
extensive theoretical, experimental,<br />
<strong>and</strong> computational research<br />
by biologists, mathematicians, <strong>and</strong><br />
engineers, in an effort to underst<strong>and</strong><br />
<strong>and</strong> mimic locomotion, maneuvering,<br />
<strong>and</strong> sensing mechanisms adopted by<br />
aquatic animals.<br />
Over the past two decades, there<br />
has also been significant interest in<br />
developing underwater robots that<br />
propel <strong>and</strong> maneuver themselves<br />
like real fish do (Triantafyllou &<br />
Triantafyllou, 1995; Kato, 2000;<br />
Anderson & Chhabra, 2002; Alvarado<br />
& Youcef-Toumi, 2006; Hu et al.,<br />
2006; Low, 2006; Epstein et al.,<br />
2006; Morgansen et al., 2007; Lauder<br />
et al., 2007; Chen et al., 2010; Aureli<br />
et al., 2010; Smithers, 2011). Often<br />
termed robotic fish, these robots provide<br />
an experimental platform for<br />
studying fish swimming <strong>and</strong> hold<br />
strong promise for a number of underwater<br />
applications. Instead of using<br />
propellers, robotic fish accomplish<br />
swimming by deforming the body<br />
<strong>and</strong>/or fin-like appendages, mostly<br />
functioning as caudal fins <strong>and</strong> sometimes<br />
as pectoral fins. Body deformation<br />
<strong>and</strong> fin movements are typically<br />
achieved with motors. On the other<br />
h<strong>and</strong>, advances in smart materials<br />
have been explored to actuate robotic<br />
fish in a noiseless <strong>and</strong> compact way<br />
(Paquette & Kim, 2004; Tangorra<br />
et al., 2007; Chen et al., 2010; Aureli<br />
et al., 2010). Robotic fish produce<br />
wake signatures similar to those of<br />
real fish <strong>and</strong> are thus less detectable<br />
than propeller-driven underwater vehicles,<br />
which is an important advantage<br />
in applications requiring stealth.<br />
Recent advances in computing,<br />
communication, electronics, <strong>and</strong><br />
materials have made it possible to<br />
create untethered robotic fish with<br />
onboard power, control, navigation,<br />
wireless communication, <strong>and</strong> sensing<br />
modules, which turns these robots<br />
into mobile sensing platforms in<br />
aquatic environments. Schools of robotic<br />
fish can form wireless sensor<br />
networks, which will have numerous<br />
promising applications, such as monitoring<br />
water quality, tracing oil spills,<br />
<strong>and</strong> patrolling harbors <strong>and</strong> coasts.<br />
Figure 1a shows a prototype of a robotic<br />
fish swimming in an inl<strong>and</strong><br />
lake. Figure 1b shows the close-up of<br />
another prototype, equipped with a<br />
dissolved oxygen (DO) sensor, global<br />
positioning system (GPS), <strong>and</strong> other<br />
electronic components, which has<br />
been developed for monitoring the<br />
DO level in aquafarms. Collected<br />
DO information will then be used to<br />
control the aerators to maintain a<br />
healthy environment for the aquatic<br />
animals on the farm.<br />
Autonomous robotic fish schools<br />
will provide a competitive alternative<br />
July/August 2011 Volume 45 Number 4 31
FIGURE 1<br />
Prototypes of autonomous robotic fish developed by the Smart Microsystems Laboratory at<br />
Michigan State University: (a) testing in a lake <strong>and</strong> (b) prototype for dynamically monitoring<br />
the DO level in aquafarms.<br />
to existing sensing technologies for<br />
aquatic <strong>and</strong> marine environments.<br />
Manual sampling, sometimes boat or<br />
ship-based, is still a common practice<br />
in environmental monitoring, which<br />
is labor-intensive with difficulty in<br />
capturing dynamic phenomena of<br />
interest. In-situ sensing with fixed or<br />
buoyedsensorsorverticalprofilers is<br />
another approach (Doherty et al.,<br />
1999; Reynolds-Fleming et al.,<br />
2002). However, these sensors have<br />
little freedom to move laterally, <strong>and</strong> it<br />
would require prohibitively many<br />
units for capturing distributed, spatially<br />
inhomogeneous information. The past<br />
decade has seen great progress in the use<br />
of robotic technology in aquatic sensing.<br />
Autonomous underwater vehicles<br />
(AUVs) (B<strong>and</strong>yopadhyay, 2005), for<br />
example, are being used for hydrographic<br />
survey, fishery operations, <strong>and</strong><br />
environmental monitoring (Hydroid,<br />
2009). Another highly successful technology<br />
is autonomous sea gliders,<br />
which has remarkable duration for<br />
continuous field operation because<br />
of highly energy-efficient design<br />
(Ericksen et al., 2001; Sherman et al.,<br />
2001; Webb et al., 2001; Rudnick<br />
et al., 2004). The downside for both<br />
AUVs <strong>and</strong> gliders is their cost, starting<br />
at US $50,000 per unit (not including<br />
the cost of sensors), prohibiting the<br />
deployment of many of them for<br />
observing with high spatial resolution,<br />
<strong>and</strong> excluding them from many applications<br />
(such as aquafarm monitoring)<br />
where cost is critical. The size (meters<br />
long) <strong>and</strong> weight (at the order of 50 kg)<br />
of these vehicles also make them cumbersome<br />
to h<strong>and</strong>le by a single person.<br />
Small autonomous robotic fish<br />
have the potential to address many of<br />
the aforementioned challenges. By a<br />
small robotic fish, we mean one that<br />
has length of 50 cm or less, displaces<br />
volume of up to 5 liters, <strong>and</strong> costs no<br />
more than US $5,000 (excluding that<br />
of aquatic sensors to be mounted). Its<br />
low cost, compact size, <strong>and</strong> light<br />
weight would make it affordable <strong>and</strong><br />
convenient to deploy these robots in<br />
groups for versatile applications <strong>and</strong><br />
various environments, such as ponds,<br />
lakes, rivers, <strong>and</strong> even oceans. Schools<br />
of robotic fish could form dynamic,<br />
adaptive sensor networks <strong>and</strong> provide<br />
distributed sensing coverage with desired<br />
spatiotemporal resolution.<br />
The realization of such a vision,<br />
however, is faced with a myriad of<br />
challenges. The size <strong>and</strong> cost considerations<br />
put stringent constraints on<br />
the robot’s locomotion, battery, computing,<br />
<strong>and</strong> communication capacities.<br />
The wide adoption of the<br />
robotic fish-based sensing technology<br />
will hinge on the robots’ ability to operate<br />
robustly in the unfriendly <strong>and</strong><br />
often unpredictable environment,<br />
with their limited onboard resources<br />
<strong>and</strong> with minimal human intervention.<br />
This poses challenges across a<br />
wide spectrum, ranging from locomotion<br />
<strong>and</strong> maneuvering mechanisms, to<br />
energy-efficient designs, localization<br />
<strong>and</strong> communication schemes, <strong>and</strong><br />
control <strong>and</strong> coordination strategies,<br />
to name a few. In this paper, we outline<br />
some of the most critical challenges<br />
<strong>and</strong> discuss potential approaches or<br />
opportunities in research <strong>and</strong> technology<br />
advancement for addressing the<br />
challenges.<br />
Maneuvering in<br />
Uncertain Environment<br />
As a sensor platform, the robotic<br />
fish often needs to survey a given<br />
path or hover over a particular region<br />
in the presence of ambient disturbances<br />
caused by wind, waves, currents,<br />
<strong>and</strong> turbulences. Regardless of<br />
its propulsion mechanism, however, a<br />
small robotic fish has limited actuation<br />
authority to counteract the disturbances.<br />
It is thus of great interest to<br />
be able to sense the flow <strong>and</strong> react in<br />
the most effective way under the actuation<br />
constraints. We can look to live<br />
fish for inspiration, because they deal<br />
with this very problem on a regular<br />
basis <strong>and</strong> have developed intricate<br />
sensing <strong>and</strong> actuation systems that<br />
offer us interesting insight.<br />
Artificial Lateral Line<br />
Most fish use the lateral line system<br />
as an important sensory organ to probe<br />
their environment (Coombs, 2001).<br />
A lateral line consists of arrays of so<br />
32 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
called neuromasts, each containing<br />
bundles of sensory hairs, encapsulated<br />
in a gelatinous structure called cupula.<br />
Under an impinging flow, the hairs are<br />
deflected, which elicits firing of the<br />
hair cell neurons <strong>and</strong> thus enables the<br />
animal to sense the flow field, perform<br />
hydrodynamic imaging, <strong>and</strong> identify<br />
new field objects of interest. The lateral<br />
line system plays an important role in<br />
various fish behaviors, including prey/<br />
predator detection, schooling, rheotaxis,<br />
courtship <strong>and</strong> communication.<br />
A lateral line-like sensory module<br />
or an artificial lateral line will be very<br />
useful for a robotic fish to improve its<br />
maneuverability. For example, with<br />
feedback from the lateral line, the<br />
robot could manipulate vortices in<br />
the flow with its actuated fins <strong>and</strong> exploit<br />
the ambient flow energy for locomotion<br />
(Beal et al., 2006) or perform<br />
station-keeping by responding appropriately<br />
to the sensed ambient flow.<br />
Artificial lateral line systems, where arrays<br />
of beam or hair-like structures are<br />
used to measure flow velocities, have<br />
been proposed based on various physical<br />
transduction principles, including<br />
hot wire anemometry (Yang et al.,<br />
2006), piezoresistivity (Yang et al.,<br />
2010), capacitive sensing (Dagamseh<br />
et al., 2010), <strong>and</strong> encapsulated interface<br />
bilayers (Sarles et al., 2011).<br />
Recently, we have exploited the intrinsic<br />
mechanosensory property of<br />
ionic polymer-metal composites<br />
(IPMCs) to construct artificial lateral<br />
lines (Abdulsadda & Tan, 2011;<br />
Abdulsadda et al., 2011). As illustrated<br />
in Figure 2, an IPMC consists of three<br />
layers, with an ion-exchange polymer<br />
membrane (e.g., Nafion) s<strong>and</strong>wiched<br />
by metal electrodes. Inside the polymer,<br />
(negatively charged) anions covalently<br />
fixed to polymer chains are balanced<br />
by mobile (positively charged) cations.<br />
An applied mechanical stimulus, such<br />
FIGURE 2<br />
Illustration of the IPMC sensing principle.<br />
as a flow impinging on the IPMC,<br />
redistributes the cations inside <strong>and</strong><br />
produces a detectable electrical signal<br />
(typically open-circuit voltage or<br />
short-circuit current) that is correlated<br />
with the mechanical or hydrodynamic<br />
stimulus (Chen et al., 2007). Conversely,<br />
an applied voltage across an<br />
IPMC leads to the transport of cations<br />
<strong>and</strong> accompanying solvent molecules,<br />
resulting in both differential swelling<br />
<strong>and</strong> electrostatic forces inside the<br />
FIGURE 3<br />
material, which cause the material to<br />
bend <strong>and</strong> hence the actuation effect<br />
(Shahinpoor & Kim, 2001). Figure 3a<br />
shows a prototype of an artificial lateral<br />
line consisting of four IPMC sensors.<br />
While the physical construction of<br />
robust <strong>and</strong> sensitive artificial lateral<br />
lines remains an active research area,<br />
it is of equal importance to make<br />
sense out of the data collected by the<br />
lateral line. Existing studies on biological<br />
<strong>and</strong> artificial lateral lines have<br />
Experimental results on localization of a dipole source with unknown location <strong>and</strong> vibration amplitude:<br />
(a) prototype of IPMC-based lateral line, consisting of four IPMC sensors, <strong>and</strong> (b) localization<br />
results along three different tracks, based on solving a model-based nonlinear estimation<br />
problem (Abdulsadda et al., 2011).<br />
July/August 2011 Volume 45 Number 4 33
mostly focused on the problem of<br />
localizing a vibrating sphere, known<br />
as a dipole, which is used to emulate periodic<br />
tail beating or other appendage<br />
movement of aquatic animals. Several<br />
approaches to signal processing have<br />
been reported, which include exploitation<br />
of the characteristic points (e.g.,<br />
zero-crossings, maxima, etc.) in the<br />
measured velocity profile (Dagamseh<br />
et al., 2010), matching of the measured<br />
data with preobtained templates<br />
(P<strong>and</strong>ya et al., 2006), beamforming<br />
techniques (Yang et al., 2010), <strong>and</strong><br />
artificial neural networks (Abdulsadda<br />
& Tan, 2011). We have further considered<br />
a source localization problem<br />
where both the source location <strong>and</strong><br />
its vibrating amplitude are unknown.<br />
The posed problem is interesting,<br />
since a source far away but with large<br />
vibration could produce a signal that<br />
has similar amplitude as a signal produced<br />
by a source nearby but with<br />
small vibration. By formulating <strong>and</strong><br />
solving a nonlinear estimation problem<br />
based on an analytical model for<br />
dipole-generated flow, we are able to<br />
resolve both the source location <strong>and</strong><br />
the vibration amplitude simultaneously<br />
(Abdulsadda et al., 2011). As shown in<br />
Figure 3b, experimental results on an<br />
IPMC-based lateral line prototype<br />
(Figure 3a) have confirmed the effectiveness<br />
of the model-based estimation<br />
approach.<br />
Other than the dipole source localization<br />
problem, there are a few<br />
interesting directions for the signal<br />
processing of artificial lateral lines.<br />
The first is the detection <strong>and</strong> localization<br />
of multiple, more sophisticated<br />
moving sources (including vortices).<br />
With the sources moving, the resultingflow<br />
is no longer at a steady state, <strong>and</strong><br />
the processing algorithm needs to<br />
localize the sources with minimal<br />
latency. Another major problem to<br />
consider is the information processing<br />
for a lateral line that is mounted on a<br />
robotic fish, where the motion of the<br />
robot itself <strong>and</strong> its fins adds significant<br />
“noise” to the lateral line signal. Biological<br />
fish deal with these problems<br />
effectively through biomechanical filtering<br />
for enhanced signal-to-noise<br />
ratio <strong>and</strong> through dynamic filtering<br />
in the central nervous system to remove<br />
the unwanted signal components<br />
(Coombs & Braun, 2003; Bodznick<br />
et al., 2003). For example, dynamic<br />
neural mechanisms have been identified<br />
for suppressing self-generated<br />
noise (Coombs & Braun, 2003). Such<br />
biological insight will prove valuable in<br />
devising the mechanical, electrical, <strong>and</strong><br />
digital filtering mechanisms for solving<br />
complex processing problems faced by<br />
artificial lateral lines.<br />
Bioinspired Fin<br />
Achieving high-maneuverability<br />
hinges on the ability to manipulate<br />
the fluid in a delicate manner. Fish<br />
often use their pectoral fins to perform<br />
sophisticated maneuvers (Drucker &<br />
Lauder, 2001, 2003). These maneuvers<br />
involve complex conformational<br />
changes of the fins, involving cupping,<br />
twisting, <strong>and</strong> bending motions.<br />
Robotic fish fins, on the other h<strong>and</strong>,<br />
often use rigid foils (Kato, 2000;<br />
Morgansen et al., 2007). Recently,<br />
advances in soft actuation materials,<br />
e.g., IPMCs, have led to the exploration<br />
of these materials as flexible propulsors<br />
(Paquette & Kim, 2004).<br />
However, the resulting robotic fins<br />
typically have simple deformation<br />
modes, e.g., bending only (Chen et al.,<br />
2010; Aureli et al., 2010), <strong>and</strong> fall<br />
short of emulating the complex deformation<br />
of biological fins.<br />
Underst<strong>and</strong>ing of the morphology<br />
<strong>and</strong> mechanics of fish fins has spurred<br />
effort on mimicking these features<br />
(typically at a higher level) in designing<br />
robotic fins (Lauder et al., 2007). In<br />
particular, the complex shape change<br />
of fish fins is enabled by multiple<br />
muscle-controlled, relatively rigid, bony<br />
fin rays that are connected via collagenous<br />
membrane (Lauder & Madden,<br />
2006). Coordinated movement of individual<br />
fin rays results in conformational<br />
changes of the fins desired in<br />
maneuvers. On the engineering side,<br />
by patterning electrodes of IPMC<br />
materials, one can expect to produce<br />
complex deformation by applying different<br />
voltage inputs to different electrode<br />
areas. The patterning can be<br />
achieved with masking during electroless<br />
plating or by selective removal of<br />
electrodes post-IPMC fabrication<br />
using laser or machining (Kim et al.,<br />
2011). Inspired by the pectoral fins<br />
of bluegill sunfish, we have developed<br />
a lithography-based monolithic fabrication<br />
process for creating IPMC actuators<br />
capable of sophisticated shape<br />
changes (Chen & Tan, 2010). As<br />
shown in Figure 4, the fabricated sampleconsistsofmultipleactiveIPMC<br />
regions, coupled through much thinner<br />
passive regions. By phasing the voltage<br />
inputs to different active regions, we<br />
can realize various deformation modes<br />
including bending, twisting, <strong>and</strong> cupping<br />
(Figure 5). For example, a peak-topeak<br />
twisting angle of 16° is achieved<br />
with actuation voltages of 3 V (Chen<br />
& Tan, 2010).<br />
While the progress made in biomimetic<br />
fins is encouraging, significant<br />
further advances in both material<br />
fabrication <strong>and</strong> fin control are needed,<br />
before robotic fish are capable of manipulating<br />
the flow in a manner close<br />
to what their biological counterparts<br />
do. In particular, the materials need<br />
to be improved so that they can produce<br />
much larger deformation with<br />
reasonable b<strong>and</strong>width (a few Hz). On<br />
34 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 4<br />
Monolithically fabricated IPMC sample inspired by fish fins: (a) top view <strong>and</strong> (b) SEM picture of the<br />
cross section, showing that the passive area is much thinner than the active area (Chen & Tan, 2010).<br />
the control side, we need to model <strong>and</strong><br />
underst<strong>and</strong> the deformation <strong>and</strong> its<br />
hydrodynamic consequences of a given<br />
input by combining observation of<br />
kinetic patterns of fish fin movement,<br />
nonlinear elasticity modeling, computational<br />
fluid dynamics modeling,<br />
<strong>and</strong> experimental flow measurements<br />
using digital particle image velocimetry.<br />
Energy-Efficient<br />
Sustained Operation<br />
For the robotic fish-based sensing<br />
technology to gain widespread adoption,<br />
these robots will have to be able<br />
FIGURE 5<br />
to work continuously in the field<br />
with minimal human intervention. In<br />
particular, they need to operate for at<br />
least weeks, if not for months, before<br />
returning for battery recharge <strong>and</strong><br />
other manual maintenance. Power is<br />
arguably the most crucial factor that<br />
limits the operational time. While<br />
fuel represents a potential energy<br />
source with high power <strong>and</strong> energy<br />
density, it is unclear when fuel-based<br />
propulsion will become feasible for<br />
small underwater robots. Therefore,<br />
battery is expected to be the primary<br />
power source for robotic fish, for at<br />
least the next 5–10 years.<br />
Examples of deformation modes demonstrated by the fabricated IPMC fin: (a) bending <strong>and</strong><br />
(b) twisting.<br />
Thereareanumberofwaysone<br />
can potentially extend the run time<br />
of battery-powered robotic fish. For<br />
example, many onboard devices can<br />
be put to the sleep mode to save energy,<br />
when they are not active. Photovoltaic<br />
films can be mounted on the<br />
robot to harvest solar energy <strong>and</strong> replenish<br />
the battery, when the robot is<br />
on the water surface. Wave energy<br />
could be another source to tap into,<br />
but how to harvest it on an untethered<br />
<strong>and</strong> often goal-oriented robotic fish remains<br />
a challenge.<br />
Whilealltheaforementionedapproaches<br />
could stretch the mileage<br />
per battery charge to some extent,<br />
they are not game-changers. Design<br />
of energy-efficient locomotion mechanisms<br />
will be critical in realizing<br />
long-duration field operation, since<br />
locomotion is the biggest source of<br />
energy expenditure for autonomous<br />
robotic fish. To this end, we are currently<br />
developing a novel class of underwater<br />
robots, called gliding robotic<br />
fish (Figure 6a). Such a robot will represent<br />
a hybrid of underwater glider<br />
<strong>and</strong> robotic fish; for example, it will<br />
have wings for gliding <strong>and</strong> fins for maneuvering<br />
<strong>and</strong> assistive propulsion.<br />
Consequently, a gliding robotic fish<br />
is expected to possess both high energy<br />
efficiency <strong>and</strong> great maneuverability.<br />
Figure 6b further illustrates the<br />
gliding principle <strong>and</strong> why a gliding robotic<br />
fish will be energy-efficient.<br />
Under the combined influence of gravity<br />
<strong>and</strong> buoyancy, the body will experience<br />
vertical (up or down) motion.<br />
When the glider is properly pitched,<br />
the lift generated during buoyancyinduced<br />
vertical motion will enable<br />
horizontal travel. Through the control<br />
of pitch direction <strong>and</strong> buoyancy, one<br />
can switch between the descent/ascent<br />
gliding motion, resulting in a sawtoothshaped<br />
trajectory. Since buoyancy<br />
July/August 2011 Volume 45 Number 4 35
FIGURE 6<br />
Energy-efficient gliding robotic fish: (a) the concept of a gliding robotic fish with a hydrodynamic<br />
gliding body <strong>and</strong> a caudal fin <strong>and</strong> (b) illustration of the gliding principle.<br />
control <strong>and</strong> pitch control are the major<br />
sources of energy expenditure <strong>and</strong> take<br />
placeonlyduringascent/descent<br />
switching, the motion is very energyefficient,<br />
especially if the dive depth<br />
is relatively large.<br />
Communication <strong>and</strong><br />
Localization<br />
Robotic fish need to communicate<br />
with a base station to receive comm<strong>and</strong>s<br />
<strong>and</strong> send back the collected environmental<br />
information. They also<br />
need to communicate with each<br />
other for information relay <strong>and</strong> motion<br />
coordination. Underwater communication,<br />
however, is particularly challenging<br />
for small robotic fish that<br />
have stringent power <strong>and</strong> size constraints.<br />
Radio frequency (RF) signals<br />
attenuate quickly in water, severely<br />
limiting the achievable communication<br />
range <strong>and</strong> data rate. Light communication<br />
is possible (Verzijlenberg<br />
& Jenkin, 2010), but again the range<br />
<strong>and</strong> data rate are very limited <strong>and</strong> it<br />
does not work in a turbid environment.<br />
Acoustic <strong>and</strong> sonar communication<br />
underwater has been studied for<br />
many decades <strong>and</strong> was recently explored<br />
for communication among robotic<br />
fish (Science Daily, 2008).<br />
However, the associated power <strong>and</strong><br />
hardware required to achieve reasonably<br />
large communication distance<br />
<strong>and</strong> data rate are typically not affordable<br />
by small robotic fish. For these<br />
reasons, the most viable solution<br />
wouldbetocommunicatewhenthe<br />
robot surfaces, in which case lowpower,<br />
low-cost RF communication<br />
protocols such as ZigBee can be readily<br />
used. Unlike the Bluetooth protocol,<br />
which is intended for eliminating cables<br />
between electronic devices, the ZigBee<br />
protocol is built on top of the IEEE<br />
802.15.4 st<strong>and</strong>ard <strong>and</strong> it targets specifically<br />
wireless sensor network applications.<br />
For wider range communication,<br />
cellular networks could also be employed<br />
if such networks are available.<br />
Limiting the communication to the<br />
water surface entails additional challenges<br />
in robotic fish coordination,<br />
control, <strong>and</strong> networking. For effective<br />
FIGURE 7<br />
networking, we need to have a sufficient<br />
number of nodes on the surface.<br />
This can be achieved through joint<br />
motion planning <strong>and</strong> control. For example,<br />
we can hold robotic fish on<br />
the surface until a network with adequate<br />
density <strong>and</strong> coverage is formed<br />
<strong>and</strong> completes data transmission.<br />
Localization is another challenge in<br />
robotic fish-based sensor networks.<br />
For small robotic fish, having onboard<br />
localization capability is essential for<br />
successful navigation of the robot <strong>and</strong><br />
for effective coordination of robotic<br />
fish networks. Accurate localization is<br />
also critical for tagging the sensed information<br />
so that the data collected<br />
by robotic fish are associated correctly<br />
to the physical location in water. While<br />
the GPS is readily available <strong>and</strong> does<br />
not take up much space, its typical precision<br />
of 5–10 m is inadequate for<br />
many applications of robotic fish due<br />
to their small size <strong>and</strong> relatively low<br />
speeds (50 cm/s or less). In addition,<br />
theGPSmaytakeafewminutesto<br />
lock satellites every time the robot<br />
emerges from underwater, which severely<br />
limits the networking <strong>and</strong> control<br />
performance. More agile <strong>and</strong> precise<br />
localization technology is needed.<br />
We have developed an efficient localization<br />
scheme for small robotic fish<br />
(Shatara & Tan, 2010). As illustrated<br />
in Figure 7a, the scheme is based on<br />
Underwater acoustic ranging-based localization: (a) schematic of the ranging protocol <strong>and</strong><br />
(b) localization performance in a pool test (Shatara & Tan, 2010).<br />
36 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
acoustic ranging, which measures the<br />
time it takes an acoustic signal to travel<br />
from one node to the other. For example,<br />
node 1 simultaneously sends an<br />
RF packet <strong>and</strong> an acoustic pulse to<br />
node 2. When node 2 starts its onboard<br />
timer when it receives the RF<br />
packet <strong>and</strong> then stops its timer when<br />
it detects the acoustic pulse. Since the<br />
RF signal travels much faster than<br />
the acoustic signal, we can estimate<br />
thedistancebetweenthetwonodes<br />
based on the timer reading. The<br />
scheme involves simple hardware, a<br />
buzzer <strong>and</strong> a microphone, for each<br />
node. A sliding discrete Fourier transform<br />
algorithm, implemented on a<br />
digital signal controller, is employed<br />
for the detection of arrival of the acoustic<br />
signal. Figure 7b shows the results<br />
from experiments in a swimming<br />
pool, where a small robotic fish was<br />
towed across the deep side of the<br />
pool (about 13 m long) while its distances<br />
to the two beacon nodes<br />
mounted on the pool wall were measured<br />
through acoustic ranging. The<br />
resulting localization error was less<br />
than 1 m for the entire tested range,<br />
which was a significant improvement<br />
over the precision of a commercial<br />
GPS.<br />
Note that the above localization<br />
scheme works only when the robot<br />
surfaces, since it involves RF communication.<br />
The location of a robot when<br />
it is underwater can be inferred using<br />
dead reckoning. The scheme in<br />
Figure 7 does require beacon nodes<br />
(whose locations are known) to obtain<br />
the absolute location of a node. In the<br />
absence of such beacon nodes, the<br />
scheme can be used to get relative locations<br />
among nodes, which is of interest<br />
in coordinating schools of robotic fish.<br />
There are many other in-air localization<br />
schemes for wireless sensor networks,<br />
e.g., ranging based on received<br />
signal strength. While these schemes<br />
can be adapted for robotic fish-based<br />
aquatic networks, care must be taken<br />
to address the challenges associated<br />
with noises, disturbances, <strong>and</strong> signal<br />
attenuation at the air/water interface.<br />
Autonomous Control<br />
<strong>and</strong> Coordination<br />
With onboard communication,<br />
navigation, control, <strong>and</strong> sensing devices,<br />
robotic fish are desired to operate<br />
autonomously, as individuals <strong>and</strong> as<br />
schools, to carry out envisioned monitoring<br />
tasks. A few challenges arise in<br />
the control <strong>and</strong> coordination of these<br />
robots. A robotic fish needs to h<strong>and</strong>le<br />
multiple functions subject to environmental<br />
uncertainties <strong>and</strong> resource constraints.<br />
In particular, the functions<br />
could include sampling the environment,<br />
processing <strong>and</strong> transmitting<br />
the measured data, maintaining network<br />
connectivity, <strong>and</strong> controlling<br />
its motion. There are various uncertainties<br />
that interfere with these functions,<br />
examples of which include<br />
motion perturbations due to waves<br />
<strong>and</strong> turbulences, imperfect sensor<br />
measurements, <strong>and</strong> localization error<br />
<strong>and</strong> communication packet drops.<br />
Furthermore, all of these functions<br />
compete for limited onboard computing<br />
<strong>and</strong> power resources. This is a classic<br />
multi-objective, multi-constraint<br />
optimization problem, <strong>and</strong> it dem<strong>and</strong>s<br />
asystematicapproachtothejoint<br />
consideration of control, networking,<br />
<strong>and</strong> sensor fusion. Evolutionary algorithms<br />
(Deb, 2001), which codify<br />
basic principles of genetic evolution,<br />
can offer a promising solution to this<br />
multi-objective optimization problem.<br />
Another challenge lies in coordinating<br />
a school of robotic fish. It is intriguing<br />
to deploy groups of robotic<br />
fish that cooperatively perform sensing<br />
tasks. In that case, it is often desirable<br />
not to use centralized control, because<br />
the centralized paradigm would entail<br />
prohibitive cost in communication,<br />
<strong>and</strong> it would paralyze the whole network<br />
if the comm<strong>and</strong> node fails.<br />
Therefore, individual robotic fish are<br />
expected to communicate only with<br />
their local neighbors <strong>and</strong> make decisions<br />
in a distributed manner. Animals<br />
including fish often exhibit coordinated<br />
collective movement facilitated<br />
by only local interactions, which has<br />
inspired great interest from the controls<br />
community in analyzing <strong>and</strong> synthesizing<br />
control laws for groups of<br />
unmanned vehicles. Significant progress<br />
has been made in this area, even<br />
with some demonstrated success in<br />
adaptive sampling using underwater<br />
gliders (Leonard et al., 2007).<br />
While these accomplishments can<br />
provide a sound starting point for the<br />
control <strong>and</strong> coordination of robotic<br />
fish schools, we need to recognize<br />
many new <strong>and</strong> subtle difficulties<br />
faced by the latter. For example, a robotic<br />
fish can only communicate with<br />
its peers when it surfaces, which renders<br />
communication <strong>and</strong> feedback<br />
intermittent <strong>and</strong> asynchronous. This<br />
again points to the need to jointly consider<br />
control, communication, <strong>and</strong><br />
networking issues.<br />
Other Challenges<br />
As sensor platforms, the potential<br />
of robotic fish in environmental sensing<br />
will be ultimately limited by the<br />
availability of versatile sensors that are<br />
compact <strong>and</strong> easy to interface with.<br />
Most commercial sensors available<br />
today are not amenable to integration<br />
into small robotic fish, since sensor<br />
manufacturers have mostly been targeting<br />
h<strong>and</strong>held, fixed, or buoyed<br />
July/August 2011 Volume 45 Number 4 37
sensors where miniaturization is not<br />
critical. It is expected that, with the development<br />
of robotic fish <strong>and</strong> wireless<br />
networking technologies, manufacturers<br />
will see the growth opportunities<br />
in robotic fish-enabled aquatic<br />
sensing <strong>and</strong> start investing in the development<br />
of compact, economical, <strong>and</strong><br />
robust aquatic sensors.<br />
There are other engineering challenge,<br />
one example of which is biofouling,<br />
where microorganisms <strong>and</strong><br />
other organisms accumulate on the<br />
surface of robotic fish <strong>and</strong> their sensors,<br />
degrading movement <strong>and</strong> sensing<br />
performance. Periodic cleaning is<br />
an option; thanks to the mobility of<br />
robotic fish, access to these robots is<br />
relatively easy. Another possibility is<br />
to apply anti-fouling coatings.<br />
Conclusion<br />
In this paper, we have explored the<br />
potential of small robotic fish as<br />
mobile sensor platforms for aquatic<br />
<strong>and</strong> marine environments. Realization<br />
of this vision poses a rich set of challenges<br />
across a wide spectrum of areas,<br />
such as actuation/sensing materials,<br />
mechanism design, communication,<br />
control, <strong>and</strong> packaging. We have reviewed<br />
some of the major challenges<br />
<strong>and</strong> discussed possible routes to overcome<br />
them. The list of challenges outlined<br />
in this paper is by no means<br />
exhaustive, but even partial success<br />
in addressing them could have farreaching<br />
impact on aquatic environmental<br />
monitoring <strong>and</strong> other<br />
engineering applications.<br />
Acknowledgment<br />
This work was supported in part by<br />
the Office of Naval Research under<br />
grant N000140810640 (program<br />
manager Dr. T. McKenna) <strong>and</strong> the<br />
National Science Foundation under<br />
grants ECCS 0547131, CCF<br />
0820220, EEC 0908810, IIS<br />
0916720, DBI 0939454, ECCS<br />
1050236, <strong>and</strong> ECCS 1029683. The<br />
author gratefully acknowledges the<br />
contributions of many former <strong>and</strong><br />
current members of the Smart Microsystems<br />
Laboratory at Michigan State<br />
University, for the results <strong>and</strong> ideas<br />
presented in this paper.<br />
Author:<br />
Xiaobo Tan<br />
Smart Microsystems Laboratory<br />
Department of Electrical <strong>and</strong><br />
Computer Engineering<br />
Michigan State University,<br />
East Lansing, MI 48824<br />
Email: xbtan@egr.msu.edu<br />
References<br />
Abdulsadda, A.T., & Tan, X. 2011.<br />
Underwater source localization using an<br />
IPMC-based artificial lateral line. In:<br />
Proceedings of the 2011 IEEE International<br />
Conference on Robotics <strong>and</strong> Automation,<br />
to appear. 2719-24. Shanghai, China: IEEE.<br />
Abdulsadda, A.T., Zhang, F., & Tan, X.<br />
2011. Localization of source with unknown<br />
amplitude using IPMC sensor arrays. In:<br />
Proceedings of Electroactive Polymer Actuators<br />
<strong>and</strong> Devices (EAPAD) XIII. Proc. SPIE, eds.<br />
Bar-Cohen, Y., & Carpi, F., 797627(1-11).<br />
San Diego, CA: SPIE or <strong>Society</strong> of Photo-<br />
Optical Instrumentation Engineers.<br />
Alvarado, P.V., & Youcef-Toumi, K. 2006.<br />
Design of machines with compliant bodies<br />
for biomimetic locomotion in liquid environments.<br />
Trans ASME J Dyn Syst Meas<br />
Control. 128:3-13. doi: 10.1115/1.2168476.<br />
Anderson, J.M., & Chhabra, N.K. 2002.<br />
Maneuvering <strong>and</strong> stability performance of a<br />
robotic tuna. Integr Comp Biol. 42:118-126.<br />
doi: 10.1093/icb/42.1.118.<br />
Aureli, M., Kopman, V., & Porfiri, M. 2010.<br />
Free-locomotion of underwater vehicles<br />
actuated by ionic polymer metal composites.<br />
IEEE/ASME Trans Mech. 15(4):603-14.<br />
doi: 10.1109/TMECH.2009.2030887.<br />
B<strong>and</strong>yopadhyay, P.R. 2005. Trends in<br />
biorobotic autonomous undersea vehicles.<br />
IEEE J Oceanic Eng. 30:109-39.<br />
doi: 10.1109/JOE.2005.843748.<br />
Beal, D.N., Hover, F.S., Triantafyllou,<br />
M.S., Liao, J.C., & Lauder, G.V. 2006.<br />
Passive propulsion in vortex wakes.<br />
J Fluid Mech. 549:385-402. doi: 10.1017/<br />
S0022112005007925.<br />
Bodznick, D., Montgomery, J., & Tricas,<br />
T.C. 2003. Electroreception: Extracting<br />
behaviorally important signals from noise.<br />
In: Sensory Processing in Aquatic Environments,<br />
eds. Collin, S.P., & Marshall, N.J.,<br />
389-403. New York: Springer-Verlag.<br />
Chen, Z., Shatara, S., & Tan, X. 2010.<br />
Modeling of biomimetic robotic fish propelled<br />
by an ionic polymer-metal composite<br />
caudal fin. IEEE/ASME Trans Mech. 15(3):<br />
448-459. doi: 10.1109/TMECH.2009.<br />
2027812.<br />
Chen, Z., & Tan, X. 2010. Monolithic<br />
fabrication of ionic polymer-metal composite<br />
actuators capable of complex deformation.<br />
Sensor Actuat A: Phys. 157:246-57.<br />
doi: 10.1016/j.sna.2009.11.024.<br />
Chen, Z, Tan, X., Will, A., & Ziel, C. 2007.<br />
A dynamic model for ionic polymer-metal<br />
composite sensors. Smart Mater Struct. 16:<br />
1477-88. doi: 10.1088/0964-1726/16/4/063.<br />
Coombs, S. 2001. Smart skins: Information<br />
processing by lateral line flow sensors.<br />
Auton Robot. 11:255-61. doi: 10.1023/<br />
A:1012491007495.<br />
Coombs, S., & Braun, C.B. 2003. Information<br />
processing by the lateral line system.<br />
In: Sensory Processing in Aquatic Environments,<br />
eds. Collin, S.P., & Marshall, N.J.,<br />
122-38. New York: Springer-Verlag.<br />
Dagamseh, A., Lammerink, T., Kolster, M.,<br />
Bruinink, C., Wiegerink, R., & Krijnen, G.<br />
38 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
2010. Dipole-source localization using biomimetic<br />
flowsensor arrays positioned as<br />
lateral-line system. Sensor Actuat A. 162(2):<br />
355-60. doi: 10.1016/j.sna.2010.02.016.<br />
Deb, K. 2001. Multi-objective Optimization<br />
using Evolutionary Algorithms. Chichester,<br />
UK: John Wiley & Sons.<br />
Doherty, K.W., Frye, D.E., Liberatore,<br />
S.P., & Toole, J.M. 1999. A moored profiling<br />
instrument. J Atmos Ocean Technol.<br />
16:1816-29. 2.0.CO;2"target=_blankdoi:-<br />
10.1175/1520-0426(1999)0162.0.CO;2.<br />
Drucker, E.G., & Lauder, G.V. 2001.<br />
Wake dynamics <strong>and</strong> fluid forces of turning<br />
maneuvers in sunfish. J Exp Biol. 204:431-42.<br />
Drucker, E.G., & Lauder, G.V. 2003.<br />
Function of pectoral fins in rainbow trout:<br />
behavioral repertoire <strong>and</strong> hydrodynamic<br />
forces. J Exp Biol. 206:813-26. doi: 10.1242/<br />
jeb.00139.<br />
Epstein, M., Colgate, J.E., & MacIver,<br />
M.A. 2006. Generating thrust with a<br />
biologically-inspired robotic ribbon fin.<br />
In: Proceedings of the 2006 IEEE/RSJ<br />
International Conference on Intelligent<br />
Robots <strong>and</strong> Systems. 2412-17. Beijing, China:<br />
IEEE. doi: 10.1109/IROS.2006.281681.<br />
Ericksen, C.C., Osse, T.J., Light, R.D.,<br />
Wen, T., Lehman, T.W., Sabin, P.L., …<br />
Chiodi, A.M. 2001. Seaglider: A long-range<br />
autonomous underwater vehicle for oceanographic<br />
research. IEEE J Oceanic Eng. 26:<br />
424-36. doi: 10.1109/48.972073.<br />
Fish, F.E., & Lauder, G.V. 2006. Passive<br />
<strong>and</strong> active flow control by swimming fishes<br />
<strong>and</strong> mammals. Annu Rev Fluid Mech.<br />
38:193-224. doi: 10.1146/annurev.<br />
fluid.38.050304.092201.<br />
Hu, H., Liu, J., Dukes, I., & Francis, G.<br />
2006. Design of 3D swim patterns for<br />
autonomous robotic fish. In: Proc. 2006<br />
IEEE/RSJ Int. Conf. Intell. Robots Syst.<br />
2406-11. Beijing, China: IEEE.<br />
Hydroid, LLC. 2009. REMUS: Autonomous<br />
underwater vehicles - REMUS AUVs.<br />
Available at http://www.km.kongsberg.<br />
com/ks/web/nokb0240.nsf/AllWeb/<br />
D5682F98CBFBC05AC1257497002976E4<br />
OpenDocument/ (accessed 5 July 2011).<br />
Kato, N. 2000. Control performance in<br />
the horizontal plane of a fish robot with<br />
mechanical pectoral fins. IEEE J Oceanic<br />
Eng. 25(1):121-9. doi: 10.1109/48.820744.<br />
Kim, K.J., Pugal, D., & Leang, K.K. 2011.<br />
A twistable ionic polymer-metal composite<br />
artificial muscle for marine applications.<br />
Mar Technol Soc J. This issue.<br />
Lauder, G.V., Anderson, E.J., Tangorra, J.,<br />
& Madden, P.G.A. 2007. Fish biorobotics:<br />
Kinematics <strong>and</strong> hydrodynamics of selfpropulsion.<br />
J Exp Biol. 210:2767-80.<br />
doi: 10.1242/jeb.000265.<br />
Lauder, G.V., & Drucker, E.G. 2004.<br />
Morphology <strong>and</strong> experimental hydrodynamics<br />
of fish fin control surfaces. IEEE J Oceanic<br />
Eng. 29(3):556-71. doi: 10.1109/JOE.<br />
2004.833219.<br />
Lauder, G.V., & Madden, P.G.A. 2006.<br />
Learning from fish: Kinematics <strong>and</strong> experimental<br />
hydrodynamics for roboticists. Int<br />
J Automat Comput. 4:325-35. doi: 10.1007/<br />
s11633-006-0325-0.<br />
Leonard, N.E., Paley, D.A., Lekien, F.,<br />
Sepulchre, R., Fratantoni, D.M., & Davis,<br />
R.E. 2007. Collective motion, sensor<br />
networks, <strong>and</strong> ocean sampling. Proc IEEE.<br />
95(1):48-74. doi: 10.1109/JPROC.2006.<br />
887295.<br />
Low, K.H. 2006. Locomotion <strong>and</strong> depth control<br />
of robotic fish with modular undulating<br />
fins. Int J Automat Comput. 4:348-57.<br />
doi: 10.1007/s11633-006-0348-6.<br />
Morgansen, K.A., Triplett, B.I., & Klein,<br />
D.J. 2007. Geometric methods for modeling<br />
<strong>and</strong> control of free-swimming fin-actuated<br />
underwater vehicles. IEEE T Robot. 23(6):<br />
1184-99. doi: 10.1109/LED.2007.911625.<br />
P<strong>and</strong>ya, S., Yang, Y., Jones, D.L., Engel, J.,<br />
& Liu, C. 2006. Multisensor processing<br />
algorithms for underwater dipole localization<br />
<strong>and</strong> tracking using MEMS artificial<br />
lateral-line sensors. EURASIP J Appl Signal<br />
Processing. 2006:76593. doi: 10.1155/<br />
ASP/2006/76593.<br />
Paquette, J.W., & Kim, K.J. 2004. Ionomeric<br />
electroactive polymer artificial muscle for<br />
naval applications. IEEE J Oceanic Eng. 29(3):<br />
729-37. doi: 10.1109/JOE.2004.833132.<br />
Reynolds-Fleming, J.V., Fleming, J.G., &<br />
Luettich, R.A. 2002. Portable autonomous<br />
vertical profiler for estuarine applications.<br />
Estuaries. 25:142-7. doi: 10.1007/BF02696058.<br />
Rudnick, D.L., Eriksen, C.C., Fratantoni,<br />
D.M., & Perry, M.J. 2004. Underwater gliders<br />
for ocean research. Mar Technol Soc J. 38:<br />
48-59. doi: 10.4031/002533204787522703.<br />
Sarles, S.A., & Leo, D.J. 2011. Hair cell<br />
sensing with encapsulated interface bilayers.<br />
In: Proceedings of Bioinspiration, <strong>Biomimetics</strong>,<br />
<strong>and</strong> Bioreplication. Proc. SPIE, eds. Martin-<br />
Palma, R.J., & Lakhtakia, A., 797509(1-11).<br />
San Diego, CA: SPIE.<br />
Science Daily. 2008. School of robofish communicate<br />
with each other in underwater robot<br />
teams. Science Daily. http://www.sciencedaily.<br />
com/releases/2008/06/080606105454.htm<br />
(accessed 5 July 2011).<br />
Shahinpoor, M., & Kim, K. 2001. Ionic<br />
polymer-metal composites: I. Fundamentals.<br />
Smart Mater Struct. 10:819-33. doi: 10.1088/<br />
0964-1726/10/4/327.<br />
Shatara, S., & Tan, X. 2010. An efficient,<br />
time-of-flight-based underwater acoustic<br />
ranging system for small robotic fish. IEEE<br />
J Oceanic Eng. 35(4):837-46. doi: 10.1109/<br />
JOE.2010.2060810.<br />
Sherman, J., Davis, R.E., Owens, W.B., &<br />
Valdes, J. 2001. The autonomous underwater<br />
glider “spray”. IEEE J Oceanic Eng.<br />
26:437-46. doi: 10.1109/48.972076.<br />
Smithers, M. 2011. Designing future<br />
autonomous underwater vehicles. Defense<br />
Tech Briefs. Available at http://www.<br />
defensetechbriefs.com/component/content/<br />
article/9075 (accessed 5 July 2011).<br />
Tangorra, J., Anquetil, P., Fofonoff, T.,<br />
Chen, A., Del Zio, M., & Hunter, I. 2007.<br />
July/August 2011 Volume 45 Number 4 39
The application of conducting polymers to a<br />
biorobotic fin propulsor. Bioinspir Biomim.<br />
2:S6-17. doi: 10.1088/1748-3182/2/2/S02.<br />
Triantafyllou, M.S., & Triantafyllou, G.S.<br />
1995. An efficient swimming machine.<br />
Sci Amer. 272:64-71. doi: 10.1038/<br />
scientificamerican0395-64.<br />
Verzijlenberg, B., & Jenkin, M. 2010.<br />
Swimming with robots: Human robot<br />
communication at depth. In: Proceedings of<br />
the 2010 IEEE/RSJ International Conference<br />
on Intelligent Robots <strong>and</strong> Systems.<br />
4023-28. Taipei, Taiwan: IEEE. doi: 10.1109/<br />
IROS.2010.5652751.<br />
Webb, D.C., Simonetti, P.J., & Jones, C.P.<br />
2001. “SLOCUM”: An underwater glider<br />
propelled by environmental energy. IEEE J<br />
Oceanic Eng. 26:447-52. doi: 10.1109/<br />
48.972077.<br />
Yang, Y., Chen, J., Engel, J., P<strong>and</strong>ya, S.,<br />
Chen, N., Tucker, C., … Liu, C. 2006.<br />
Distant touch hydrodynamic imaging with<br />
an artificial lateral line. Proc Nat Acad<br />
Sci. 103(50):18891-5. doi: 10.1073/<br />
pnas.0609274103.<br />
Yang, Y., Nguyen, N., Chen, N., Lockwood,<br />
M., Tucker, C., Hu, H., … Jones, D.L. 2010.<br />
Artificial lateral line with biomimetic neuromasts<br />
to emulate fish sensing. Bioinspir<br />
Biomim. 5:016 001.<br />
40 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
PAPER<br />
Robotic Models for Studying Undulatory<br />
Locomotion in Fishes<br />
AUTHORS<br />
George V. Lauder<br />
Jeanette Lim<br />
Ryan Shelton<br />
Museum of Comparative Zoology,<br />
Harvard University<br />
Chuck Witt<br />
Erik Anderson<br />
Department of Mechanical<br />
Engineering, Grove City College<br />
James L. Tangorra<br />
Department of Mechanical<br />
Engineering, Drexel University<br />
Introduction<br />
Fish moving through the water are<br />
capable of using a variety of locomotor<br />
modes. Some species swim nearly exclusively<br />
using their fins <strong>and</strong> generate<br />
propulsive <strong>and</strong> maneuvering forces<br />
using midline fins (dorsal <strong>and</strong> anal)<br />
or paired fins (pectoral <strong>and</strong> pelvic)<br />
(Drucker & Lauder, 1999, 2001;<br />
Hove et al., 2001; St<strong>and</strong>en, 2008;<br />
St<strong>and</strong>en & Lauder, 2007). Other<br />
species generate thrust primarily by<br />
activating body musculature to bend<br />
the body <strong>and</strong> generate waves passing<br />
from the head toward the tail (Gillis,<br />
1996; Jayne & Lauder, 1994, 1995c;<br />
Lauder & Tytell, 2006; Rome et al.,<br />
1993). The caudal or tail fin is generally<br />
considered as an extension of the<br />
body in most analyses of fish locomotion,<br />
but the tail also possesses a<br />
substantial array of intrinsic muscles<br />
that appear to stiffen the tail during<br />
steady forward swimming <strong>and</strong> generate<br />
a variety of complex tail conformations<br />
during maneuvering (Flammang<br />
ABSTRACT<br />
Many fish swim using body undulations to generate thrust <strong>and</strong> maneuver in<br />
three dimensions. The pattern of body bending during steady rectilinear locomotion<br />
has similar general characteristics in many fishes <strong>and</strong> involves a wave of increasing<br />
amplitude passing from the head region toward the tail. While great progress has<br />
been made in underst<strong>and</strong>ing the mechanics of undulatory propulsion in fishes, the<br />
inability to control <strong>and</strong> precisely alter individual parameters such as oscillation frequency,<br />
body shape, <strong>and</strong> body stiffness, <strong>and</strong> the difficulty of measuring forces on<br />
freely swimming fishes have greatly hampered our ability to underst<strong>and</strong> the fundamental<br />
mechanics of the undulatory mode of locomotion in aquatic systems. In this<br />
paper, we present the use of a robotic flapping foil apparatus that allows these parameters<br />
to be individually altered <strong>and</strong> forces measured on self-propelling flapping<br />
flexible foils that produce a wave-like motion very similar to that of freely swimming<br />
fishes. We use this robotic device to explore the effects of changing swimming<br />
speed, foil length, <strong>and</strong> foil-trailing edge shape on locomotor hydrodynamics, the<br />
cost of transport, <strong>and</strong> the shape of the undulating foil during locomotion. We<br />
also examine the passive swimming capabilities of a freshly dead fish body. Finally,<br />
we model fin-fin interactions in fishes using dual-flapping foils <strong>and</strong> show that thrust<br />
can be enhanced under correct conditions of foil phasing <strong>and</strong> spacing as a result of<br />
the downstream foil making use of vortical energy released by the upstream foil.<br />
Keywords: fish, robot, swimming, biomechanics<br />
& Lauder, 2008; Flammang &<br />
Lauder, 2009).<br />
Despite the large number of studies<br />
of fish undulatory locomotion<br />
over the last 20 years, there are still<br />
many unanswered questions about<br />
how the pattern of body deformation<br />
is generated, the effect of body stiffness<br />
on locomotor performance, what<br />
effect different tail shapes have on locomotor<br />
function, <strong>and</strong> how hydrodynamic<br />
interactions among different<br />
fins might influence the generation<br />
of swimming forces. Studies of freely<br />
swimming fishes have contributed<br />
enormously to our underst<strong>and</strong>ing of<br />
the mechanics of aquatic locomotion,<br />
but such an approach is necessarily<br />
limited to the behaviors voluntarily<br />
executed by living fishes. And measuring<br />
locomotor forces on freely<br />
swimming fishes is a challenging proposition.<br />
Furthermore, a wide array of<br />
interesting experimental manipulations,<br />
including changing the flexural<br />
stiffness of the body, altering the<br />
shape of the tail, <strong>and</strong> changing the<br />
spacing between fins, are clearly impossible<br />
to conduct in living animals.<br />
Robotics offers a complementary<br />
approach to studies of living fishes by<br />
allowing manipulation of variety of<br />
parameters such as flexural stiffness,<br />
aspect ratio, tail shape, <strong>and</strong> spacing between<br />
adjacent fins. A simple robotic<br />
flapping foil device can be used to generate<br />
undulatory locomotion in flexible<br />
fish-like materials, <strong>and</strong> forces can<br />
July/August 2011 Volume 45 Number 4 41
e measured <strong>and</strong> the effect of changes<br />
in body length, stiffness <strong>and</strong> tail shape<br />
can be quantified.<br />
The focus of this paper will be on<br />
undulatory locomotion in the water<br />
<strong>and</strong> the use of a robotic flapping apparatus<br />
to produce swimming in both<br />
flexible plastic foils <strong>and</strong> a passive<br />
freshly dead fish body. After a brief<br />
overview of undulatory locomotion<br />
in fishes, we discuss the design of a robotic<br />
controller for flapping flexible<br />
foils that allows measurement of selfpropelled<br />
speeds (SPS), forces, <strong>and</strong><br />
torques <strong>and</strong> the cost of transport associated<br />
with foils of different shapes <strong>and</strong><br />
stiffnesses. In addition, we address<br />
an important technical issue in studies<br />
of aquatic propulsion: the effect of<br />
swimming at non-SPS on body waveform,<br />
patterns of force production <strong>and</strong><br />
wake flow patterns. Finally, we present<br />
data on the swimming performance<br />
of passive fish bodies <strong>and</strong> discuss the<br />
future for studies of robotically controlled<br />
undulatory locomotion. Our<br />
overall aim is to introduce a number<br />
of case studies with data that show<br />
the utility of this approach for studying<br />
underwater propulsion <strong>and</strong> to present<br />
an overview of how such studies can<br />
provide new ideas <strong>and</strong> tests for current<br />
views of how fish swim.<br />
Overview of Undulatory<br />
Propulsion in Fishes<br />
When fish swim using their bodies<br />
<strong>and</strong> tail fin as the primary thrust generators,<br />
they pass a wave of bending<br />
down the body that increases in amplitude<br />
from the head toward the tail<br />
(Donley & Dickson, 2000; Gillis,<br />
1996; Jayne & Lauder, 1995a, 1995b;<br />
Liao, 2002; Long et al., 1994). This<br />
bending wave is produced by a wave<br />
of muscular activity that also moves<br />
posteriorly <strong>and</strong> is created by spinal<br />
cord <strong>and</strong> hindbrain pattern generators<br />
(Bone et al., 1978; Fetcho & Svoboda,<br />
1993; Fetcho, 1986; Shadwick &<br />
Gemballa, 2006). Muscular power<br />
to generate thrust is produced primarily<br />
by the segmented myotomal<br />
musculature in the posterior region of<br />
fish, especially during slow swimming<br />
(Johnson et al., 1994; Rome et al.,<br />
1993; Syme, 2006), <strong>and</strong> as speed increases<br />
more anterior body muscles<br />
are recruited to power locomotion.<br />
The division of the segmented body<br />
musculature into superficial “red” fibers<br />
<strong>and</strong> deeper <strong>and</strong> more complexly<br />
arranged “white” fibers also plays an<br />
important role in underst<strong>and</strong>ing the<br />
multiple “gaits” used by fishes; the relative<br />
roles of these two types of muscle<br />
fibers can change dramatically as fish<br />
change swimming speed <strong>and</strong> execute<br />
rapid locomotor behaviors such as<br />
fast-start escapes (Jayne & Lauder,<br />
1993; Tytell & Lauder, 2002).<br />
When fish swim slowly using body<br />
undulations, oscillation of the front<br />
third or so of the body is quite small<br />
even in species as diverse as eels, trout,<br />
<strong>and</strong> tuna (Donley & Dickson, 2000;<br />
Gillis, 1998; Lauder & Tytell, 2006),<br />
<strong>and</strong> a primary role of this reduced oscillation<br />
appears to be drag reduction<br />
by minimizing the frontal area of the<br />
fish that encounters oncoming flow.<br />
As swimming speed increases, the<br />
front region of fish shows increasingly<br />
large side-to-side oscillations, which is<br />
in part a reflection of the recruitment<br />
of body musculature in this more anterior<br />
region of the fish.<br />
Even when fish swim by undulatory<br />
propulsion using the production<br />
of traveling waves, other fins often<br />
are used too, <strong>and</strong> it is incorrect to suggest<br />
that fish locomotor modes represent<br />
completely distinct patterns of<br />
motion. For example, when trout or<br />
bluegill sunfish swim using body undulations,<br />
they are also actively using<br />
their dorsal <strong>and</strong> anal fins, which play<br />
a key role in balancing roll torques<br />
<strong>and</strong> generating thrust (Drucker &<br />
Lauder, 2001, 2005; St<strong>and</strong>en &<br />
Lauder, 2005, 2007). In addition,<br />
the pelvic fins play an important role<br />
in controlling body stability during<br />
locomotion in fishes (Harris, 1936,<br />
1938; St<strong>and</strong>en, 2008, 2010). Fishes<br />
also vary in tail shape, <strong>and</strong> the distinction<br />
between the externally symmetrical<br />
(homocercal) tail of teleost fishes<br />
<strong>and</strong> the asymmetrical tail in sharks<br />
<strong>and</strong> fish such as sturgeon (Lauder,<br />
1989, 2000) is well known. The heterocercal<br />
tail shape induces torques around<br />
the body center of mass that requires<br />
compensatory changes in body position<br />
to allow steady horizontal swimming<br />
(Liao & Lauder, 2000; Wilga<br />
& Lauder, 2002, 2004b).<br />
Although considerable progress has<br />
been made in studies of fish locomotion<br />
by investigating the kinematics,<br />
muscle activity, <strong>and</strong> hydrodynamics<br />
of live fishes swimming steadily <strong>and</strong><br />
maneuvering, there are many limits<br />
to studies of this kind. Perhaps the<br />
two greatest limitations to research<br />
on live fishes are (1) the considerable<br />
difficulty in measuring locomotor<br />
forces <strong>and</strong> torques produced by the<br />
bending body as fish swim freely <strong>and</strong><br />
(2) the inability to manipulate key<br />
variables such as body length, aspect<br />
ratio, tail shape, <strong>and</strong> stiffness. Without<br />
an ability to alter such key components<br />
that govern locomotor dynamics, we<br />
will be limited in our underst<strong>and</strong>ing<br />
of the factors that influence undulatory<br />
propulsion in the water.<br />
The purpose of this paper is to present<br />
a number of new case studies<br />
using a robotic flapping device for<br />
generating undulatory locomotion in<br />
engineered materials as a means of<br />
42 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
etter underst<strong>and</strong>ing how fish swim<br />
<strong>and</strong> the dynamics of undulatory propulsion.<br />
We discuss different examples<br />
to show the utility of this approach <strong>and</strong><br />
how use of simple flexible foils informs<br />
studies of fish locomotion <strong>and</strong> points<br />
to new avenues of research.<br />
Simple Robotic Models of<br />
Fish Undulatory Locomotion<br />
We have designed a robotic apparatus<br />
that produces controlled heave <strong>and</strong><br />
pitch motions of flapping foils. The<br />
most important features of this device<br />
are (1) that it can be set up to be selfpropelling<br />
(allowing locomotion by<br />
flapping foils at their natural swimming<br />
speed <strong>and</strong> not only at imposed<br />
speeds) <strong>and</strong> (2) that forces <strong>and</strong> torques<br />
can be measured on flapping foils<br />
during self-propelled swimming so<br />
that within-cycle patterns of force <strong>and</strong><br />
torque oscillation can be compared<br />
among foils with different shapes<br />
<strong>and</strong> stiffnesses at different swimming<br />
speeds. This apparatus was designed<br />
with two sets of flapping foils in series<br />
in order to be able to model, with foils,<br />
the interactions that can occur between<br />
fins of fishes that are arranged<br />
in series such as the dorsal <strong>and</strong> anal<br />
fins <strong>and</strong> the tail fin (discussed further<br />
in the section on fin-fin interactions<br />
below).<br />
A general description of the first<br />
generation of this apparatus is presented<br />
in Lauder et al. (2007). Briefly,<br />
heave <strong>and</strong> pitch motors <strong>and</strong> rotary<br />
encoders that allow readouts of foil position<br />
are mounted on a carriage above<br />
a recirculating flow tank. This carriage<br />
is supported on low friction air bearings,<br />
which allow the carriage to move<br />
in response to foil thrust <strong>and</strong> drag forces<br />
generated during flapping motions.<br />
The second generation version of this<br />
apparatus has an ATI Nano-17 sixaxis<br />
force/torque sensor mounted on<br />
the shaft supporting the foils (Figure<br />
1A). This permits measurement<br />
of three forces <strong>and</strong> three torques during<br />
self-propulsion at sample rates that<br />
allow quantification of within-cycle<br />
patterns of force production even during<br />
self-propulsion. In addition, a linear<br />
encoder mounted on the carriage<br />
allows a readout of carriage position,<br />
FIGURE 1<br />
<strong>and</strong> this is used by a Labview program<br />
to calculate SPS from data generated<br />
by a series of swimming tests at a<br />
range of speeds. Synchronizing signals<br />
from the LabView program controlling<br />
the heave <strong>and</strong> pitch motors are<br />
used to trigger data acquisition from<br />
theATIsensor<strong>and</strong>alsototrigger<br />
image acquisition from three synchronized<br />
Photron high-speed video<br />
Images of a variety of flexible foils <strong>and</strong> freshly dead trout (Oncorhynchus mykiss) attached to a<br />
robotic flapping foil apparatus (see Lauder et al., 2007, for details on the basic design). (A) Flexible<br />
foil actuated at its leading edge in heave <strong>and</strong> pitch, suspended in a recirculating flow tank. The red<br />
arrow points to an ATI Nano-17 6-axis force/torque transducer on the foil shaft. (B, C) Flexible foils<br />
with different trailing edge shapes are used to study the effect of tail shape on swimming performance.<br />
(D, E, F) Images of a trout held behind the head to allow imposition of heave <strong>and</strong> pitch motions to<br />
study passive body properties. (E) An image from a trout self-propelling under an imposed heave<br />
motion; the body waveform produced is very similar to that generated during swimming.<br />
July/August 2011 Volume 45 Number 4 43
cameras. Images of the flapping foils to<br />
quantify both foil motion <strong>and</strong> hydrodynamic<br />
flow patterns are thus<br />
synchronized with force <strong>and</strong> torque<br />
measurements on the foil <strong>and</strong> with<br />
the imposed heave <strong>and</strong> pitch motions<br />
on the foil.<br />
The overall goals of using a flapping<br />
foil robotic device are to simplify<br />
<strong>and</strong> control as much as possible patterns<br />
of motion imposed on flexible<br />
<strong>and</strong> rigid foils that swim through the<br />
water <strong>and</strong> to allow direct testing of<br />
the effect on propulsion of a variety<br />
of key factors relevant to underst<strong>and</strong>ing<br />
fish locomotion: tail shape (Figures<br />
1B <strong>and</strong> 1C), foil flexibility, <strong>and</strong><br />
interactions among fins. This robotic<br />
apparatus can also be used to examine<br />
the passive swimming capabilities of a<br />
freshly dead fish body (Figures 1D, 1E,<br />
<strong>and</strong> 1F ), <strong>and</strong> below we present data on<br />
the ability of fish bodies to passively<br />
propel <strong>and</strong> generate propulsive waveforms<br />
under imposed motions. Foils<br />
of various kinds are attached to a stainless<br />
steel s<strong>and</strong>wich bar (to hold the<br />
leading edge of flexible materials; Figures<br />
1A, 1B, <strong>and</strong> 1C) to a solid 8-mm<br />
shaft (for rigid NACA 0012 foils; see<br />
Lauder et al., 2007, <strong>and</strong> data shown<br />
in Figure 11), <strong>and</strong> flexible fish bodies<br />
aremountedinaholderthatisattached<br />
behind the head (Figures 1D<br />
<strong>and</strong> 1F). Holding systems are designed<br />
not to flex in response to imposed<br />
heave <strong>and</strong> pitch motions <strong>and</strong> to allow<br />
forces <strong>and</strong> torques generated by swimming<br />
foils to be transmitted to the<br />
force/torque sensor on the shaft. Holding<br />
systems such as the s<strong>and</strong>wich bar<br />
system do have their own drag, <strong>and</strong><br />
this drag is time-dependent due to<br />
the heaving <strong>and</strong> pitching motion of<br />
the holding system as the foils selfpropel.<br />
It is thus not possible to give<br />
asinglevalueforthedragofthefoil<br />
holding system. Since all foil comparisons<br />
within a single experimental type<br />
used the same holding system <strong>and</strong> were<br />
treated identically, we do not present<br />
data on the performance of the holding<br />
apparatus alone.<br />
Sample data from a self-propelled<br />
flapping foil actuated in heave only<br />
are shown in Figure 2. Monitoring<br />
foil shaft position <strong>and</strong> measuring forces<br />
in the X (upstream-downstream) <strong>and</strong><br />
Y (side to side) directions allows calculation<br />
of the heave velocity <strong>and</strong> other<br />
derived quantities such as the instantaneous<br />
power required by the foil to<br />
swim <strong>and</strong> the coefficients of thrust<br />
<strong>and</strong> power. The cost of transport is calculated<br />
by measuring the foil mass <strong>and</strong><br />
dividing the cost/meter by this value<br />
for each foil (see Table 1). Typical<br />
peak thrust coefficients for highly flexible<br />
foils (flexural stiffness in the range<br />
of 10 −4 to 10 −6 Nm 2 ) are in the range<br />
of±0.2,lowerthantypicalforrigid<br />
foils, but these flexible foils nonetheless<br />
are capable of self-propulsion<br />
at speeds of 10-30 cm s −1 .<br />
FIGURE 2<br />
A key technical issue arises in studies<br />
of flapping foil propulsion that<br />
hope to imitate the self-propelled condition<br />
achieved by swimming fishes:<br />
foils that are not self-propelling may<br />
exhibit patterns of thrust oscillation<br />
that are not centered around zero.<br />
Any foil that is truly self-propelling<br />
(<strong>and</strong> not being dragged through the<br />
wateratspeedsslowerorfasterthan<br />
it would naturally move) should generate<br />
thrust in an oscillatory pattern, <strong>and</strong><br />
thrust integrated over a single flapping<br />
cycle should equal zero. Graphs in the<br />
literature of thrust coefficients during<br />
foil-based locomotion that are not<br />
centered around zero indicate that<br />
the foil was being towed above or<br />
below the SPS <strong>and</strong> do not reflect the<br />
self-propelled condition. Data from<br />
recordings of foil forces generated<br />
during self-propulsion <strong>and</strong> at speeds<br />
below <strong>and</strong> above the SPS are shown<br />
in Figure 3. During self-propelled<br />
swimming, the thrust coefficient has<br />
a mean of zero over each flapping<br />
Sample data from a self-propelling flexible plastic foil (flexural stiffness = 9.2 × 10 −5 Nm 2 ) actuated<br />
in heave at the leading edge at 2-Hz frequency. Heave position is monitored by rotary encoders,<br />
<strong>and</strong> X <strong>and</strong>Y forces are measured by a force transducer mounted on the foil shaft. From the data on<br />
foil position, force, <strong>and</strong> velocity, we make calculations of the instantaneous power, <strong>and</strong> dimensionless<br />
thrust <strong>and</strong> power coefficients. The dashed lines indicate zero for each trace.<br />
44 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
TABLE 1<br />
Locomotor properties of flexible plastic foils of three lengths while self-propelling.<br />
Foil length SPS (m/s) Reynolds number Strouhal number Work/cycle (mJ) Cost/meter (mJ/m) Cost of transport (mJ/g/m)<br />
20 0.105 21,114 0.831 2.248 42.59 106.47<br />
25 0.101 25,250 0.775 2.257 44.69 89.39<br />
35 0.091 31,850 0.349 2.233 49.08 70.11<br />
The three foils were made of the same material with a flexural stiffness of 3.1 × 10 −6 Nm 2 . These highly flexible foils propel at relatively high Strouhal numbers at<br />
shorter lengths.<br />
Reynolds <strong>and</strong> Strouhal numbers are dimensionless.<br />
Foils were actuated at their leading edge with ±1 cm heave, no pitch, at 2 Hz.<br />
Foil span was 6.8 cm for all three foils.<br />
St<strong>and</strong>ard errors for all parameters ranged from 0.3% to 1.5% of the mean values in each column.<br />
cycle. When foils are forced to swim<br />
above their SPS, the thrust coefficient<br />
curves shift above the zero baseline,<br />
<strong>and</strong> when forced swimming occurs at<br />
speeds below the SPS, data are shifted<br />
below the baseline.<br />
Change in foil swimming speed<br />
above <strong>and</strong> below the SPS can also<br />
have dramatic effects on the kinematics<br />
of the foil (Lauder et al., 2011) <strong>and</strong><br />
on the hydrodynamic wakes displayed<br />
by swimming foils. Figure 4 shows<br />
FIGURE 3<br />
the shape observed during swimming<br />
of a flexible foil (flexural stiffness,<br />
3.1 × 10 −6 Nm 2 )<strong>and</strong>thehydrodynamic<br />
wakes that result from swimming<br />
at, below, <strong>and</strong> above the average<br />
SPS. During swimming at an imposed<br />
speed below the SPS, the trailing edge<br />
ofthefoilhasalargeamplitude<strong>and</strong><br />
irregular motion that produces a wide<br />
bifurcating wake with separate momentum<br />
jets to each side (Figure 4A).<br />
At the SPS where the foil is allowed to<br />
Graph showing the dimensionless thrust coefficient versus time for a flexible plastic foil (flexural<br />
stiffness = 9.2 × 10 −5 Nm 2 ), 20 cm long, 6.8 cm high, actuated in heave ±1 cm at the leading<br />
edge at 2-Hz frequency. The red curve shows data for the self-propelled condition (18.8 cm/s),<br />
during which the thrust coefficient integrated over a single flapping cycle equals zero. Note that<br />
when experiments are done under non-self-propelling conditions (green <strong>and</strong> blue curves) the<br />
plots shift up or down so that the integrated coefficient over a flapping cycle is no longer zero.<br />
Data shown have been digitally filtered with a b<strong>and</strong>pass filter. Strouhal number for this experiment<br />
= 0.3 at the SPS (red curve). (Color versions of figures available online at: http://www.<br />
ingentaconnect.com/content/mts/mtsj/2011/00000045/00000004.)<br />
swim freely with no imposed constraints<br />
on speed, the foil bends into<br />
a regular ribbon-like pattern with<br />
a fish-likebodywake(seeNauen&<br />
Lauder, 2002a, 2002b) <strong>and</strong> alternating<br />
centers of vorticity with a fluid jet that<br />
me<strong>and</strong>ers in between these vortical<br />
centers (Figure 4B). Above the SPS<br />
where the external free-stream flow<br />
is increased above SPS <strong>and</strong> the foil is<br />
forced to swim against the increased<br />
flow, the foil shape exhibits large<br />
amplitude wave-like motion <strong>and</strong> a<br />
substantial drag-like wake with fluid<br />
velocities below that of the free stream<br />
in the wake (Figure 4C). These changing<br />
patterns of foil kinematics <strong>and</strong><br />
hydrodynamics during swimming<br />
that are not under conditions of selfpropulsion<br />
are most easily seen in highly<br />
flexible materials (flexural stiffnesses in<br />
the range of 10 −3 to 10 −6 Nm 2 ) where<br />
the fluid-structure interaction is most<br />
evident visually. These flexible materials<br />
which have flexural stiffnesses similar<br />
to those of fish (McHenry et al., 1995)<br />
show fish-like propulsion <strong>and</strong> deform<br />
into wave-like patterns with a fish-like<br />
wake (Figure 4B) when allowed to selfpropel<br />
in a low-friction system.<br />
The structure of the wake behind<br />
flapping flexible foils has a significant<br />
three-dimensional component due<br />
to the finite chord <strong>and</strong> span, <strong>and</strong> we<br />
July/August 2011 Volume 45 Number 4 45
FIGURE 4<br />
Hydrodynamics of propulsion in a flexible foil (flexural stiffness = 3.1 × 10 −6 Nm 2 ) swimming<br />
below, at, <strong>and</strong> above its SPS (8.55 cm/s). The foil was actuated at the leading edge with amplitude<br />
±0.5 cm, no pitch, at 3Hz, <strong>and</strong> is 20 cm long; Strouhal number at the SPS is 0.83 (see Table 1). The<br />
bottom margin of the foil is marked in white. Yellow arrows indicate water velocity; red color indicates<br />
counterclockwise vorticity; blue color indicates clockwise vorticity. The panels on the left<br />
show the whole foil swimming, while the matched panels on the right show a close-in view of the<br />
wake structure. The effect of swimming at a non SPS is dramatic, both on foil shape <strong>and</strong> on wake<br />
structure. In (A) the foil swam at an imposed speed of 4.75 cm/s, <strong>and</strong> in (C) the foil swam at an<br />
imposed speed of 25.7 cm/s.<br />
quantified this aspect of foil locomotor<br />
dynamics using the volumetric flow<br />
visualization system described by<br />
Flammang et al. (2011a, 2011b) <strong>and</strong><br />
Troolin <strong>and</strong> Longmire (2010). Figure<br />
5 shows how each of the centers<br />
of vorticity in the ribbon-like pattern<br />
shown in Figure 4B actually represent<br />
vortical columns on each side of the<br />
foil, which connect to each other above<br />
<strong>and</strong> below the foil. These inter-column<br />
connections occur both to upstream<br />
<strong>and</strong> downstream columns on the<br />
same side <strong>and</strong> also across the foil to<br />
vortical columns on the opposite side<br />
(Figure 5B). To date no other studies<br />
have provided three-dimensional<br />
wake snapshots during self-propulsion<br />
in highly flexible foils, but such studies<br />
in the future may reveal interesting<br />
patterns of wake interaction among<br />
components of the foil <strong>and</strong> may contribute<br />
to explaining the dynamics of<br />
propulsion in the flexing bodies.<br />
These results also suggest that at<br />
least some of the diversity of fish<br />
wakes reported in the literature results<br />
from situations in which the fishes<br />
were not swimming steadily, either<br />
in still water or against imposed<br />
flows, as wakes that look like those<br />
in Figures 4A <strong>and</strong> 4C are frequently<br />
presented. Fish commonly accelerate<br />
<strong>and</strong> execute small maneuvers during<br />
locomotion <strong>and</strong> considerable effort is<br />
needed to ensure that kinematics <strong>and</strong><br />
wake flow patterns are taken at moments<br />
when the fish is swimming<br />
steadily. Even small accelerations can<br />
substantially change fish wake flows<br />
(Tytell, 2004), <strong>and</strong> data from the<br />
flapping foil robot here illustrate that<br />
these kinematic <strong>and</strong> hydrodynamic alterations<br />
can be reproduced with flexible<br />
foils.<br />
One of the most straightforward<br />
questions that could be asked about<br />
undulatory propulsion <strong>and</strong> one that<br />
is difficult to study with live fishes is<br />
the effect of changing length alone on<br />
locomotor performance. We clearly cannot<br />
alter fish length experimentally<br />
<strong>and</strong> expect reasonable swimming performance,<br />
<strong>and</strong> comparing fish of<br />
different lengths (while useful for studies<br />
of scaling) does not account for the<br />
many other changes in the musculature<br />
<strong>and</strong> skeleton that occur as fish<br />
grow. Table 1 shows data obtained<br />
from three flexible foils of different<br />
46 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 5<br />
Volumetric flow visualization using the V3V technique (see Flammang et al., 2011a, 2011b) to<br />
image the 3D wake structure behind the flexible foil shown in Figure 4B. The trailing edge of<br />
the foil is located at the −60 mm position on the x-axis, <strong>and</strong> the fluid structures shown are all<br />
in the wake of the foil. This foil was self-propelling <strong>and</strong> was actuated using the same parameters<br />
shown for Figure 4B. (A) The flexible foil achieved a ribbon-like shape <strong>and</strong> columns of vorticity<br />
extend vertically on either side of the foil. Vorticity is isosurfaced at a value of 3.1 (blue surface),<br />
<strong>and</strong> a horizontal slice through the wake is shown to correspond to that shown using 2D piv in<br />
Figure 4B (green plane with velocity vectors). (B) The vortical columns on each side of the<br />
ribbon-like foil motion connect to each other across the top <strong>and</strong> bottom of the same side (yellow<br />
arrows) <strong>and</strong> opposite sides (red arrows).<br />
The work per cycle stays nearly constant,<br />
as does the cost per meter<br />
(Table 1), but the cost of transport<br />
decreases substantially as the increased<br />
mass of the longer foils does not result<br />
in increased energy requirements for<br />
propulsion. Foils such as these that<br />
are composed of a very flexible material<br />
self-propel at shorter lengths with a<br />
Strouhal number that is quite large<br />
relative to most self-propelling bodies.<br />
At longer lengths, the Strouhal number<br />
approaches that of many swimming<br />
fishes or rigid foils (Table 1).<br />
Figure 6 shows changes in shape<br />
of self-propelling foils made of the<br />
same material (flexural stiffness, 3.1 ×<br />
10 −6 Nm 2 ) that occur due to change<br />
FIGURE 6<br />
Graphs to show the effect of length on the<br />
shape (amplitude envelope) of a flexible foil<br />
(flexural stiffness = 3.1 × 10 −6 Nm 2 )swimming<br />
at its SPS. This foil was 6.85 cm in<br />
chord, <strong>and</strong> of varying length, given in the<br />
color-coded legend. The leading edge was actuated<br />
at 2 Hz <strong>and</strong> with amplitude of ±1 cm.<br />
Each plot shows the shape of the foil during<br />
self-propulsion as indicated by the peak-topeak<br />
amplitude of the sideways flapping motion.<br />
(A, B) Foil shapes as the absolute distance<br />
along the foil <strong>and</strong> as percentage of the total foil<br />
length, respectively.<br />
lengths made of the same material <strong>and</strong><br />
actuated in heave only at the leading<br />
edge. Altering the length of this flexible<br />
swimming foil from 20 to 35 cm produces<br />
only minor changes in the SPS<br />
(from10toabout9cms −1 )<strong>and</strong><br />
hence in Reynolds number but dramatically<br />
lowers the Strouhal number<br />
(from0.83to0.34;Table1)dueto<br />
changes in foil trailing edge amplitude.<br />
July/August 2011 Volume 45 Number 4 47
in length. Longer foils show peak excursions<br />
at the same locations as shorter<br />
foils (Figure 6A) but amplitudes that<br />
are lower. Each of the foils of this material<br />
generates a ribbon-like shape<br />
when self-propelling with a consistent<br />
wave-like pattern down its length (see<br />
Figure 4B). When the amplitude of<br />
each foil as a percentage of foil length<br />
is plotted (Figure 6B), changes in<br />
amplitude are more evident with<br />
side-to-side excursion amplitudes decreasing<br />
as length increases while the<br />
wave-like pattern is retained. The<br />
shortest foils have high amplitudes<br />
near the trailing edge <strong>and</strong> an amplitude<br />
envelope that grows along the foil<br />
while the longer foils display a tapering<br />
amplitude envelope (compare blue <strong>and</strong><br />
black curves in Figure 6B). These data<br />
show that length alone can have significant<br />
effects on locomotor efficiency<br />
<strong>and</strong> foil kinematics <strong>and</strong> that, all other<br />
things held constant, foil length increases<br />
on the order of 100% can provide<br />
reduced costs of transport.<br />
The Effect of Trailing<br />
Edge Shape on<br />
Swimming Performance<br />
The shape of the trailing tail edge<br />
of swimming fishes has been the subject<br />
of considerable discussion in the<br />
literature, with various authors considering<br />
the advantages or disadvantages<br />
of fish tails with symmetrical,<br />
asymmetrical, or forked shapes (Affleck,<br />
1950; Aleev, 1969; Lauder, 1989;<br />
Plaut, 2000; Thomson, 1971, 1976;<br />
Wilga & Lauder, 2004a). One advantage<br />
of a robotic flapping foil approach<br />
is that the trailing edge of a flexible<br />
flapping foil can be altered <strong>and</strong> a variety<br />
of configurations constructed that<br />
allow the effect of trailing edge shape<br />
alone on locomotor performance to<br />
FIGURE 7<br />
Propulsion by flexible foils of different shapes (material flexural stiffness = 3.1 × 10 −4 Nm 2 ) swimming<br />
at their SPS. Each foil was actuated at ±1 cm heave at 2 Hz. Error bars are ±2 SE. Strouhal<br />
numbers for these experiments range from 0.2 to 0.3. Foils were constructed to be of differing<br />
shapes <strong>and</strong> areas as follows: foil 1= square trailing edge, area = 131.2 cm 2 , dimensions = 6.85 cm ×<br />
19.15 cm; foil 2 = angled trailing edge, same length as foil 1, area = 107.71 cm 2 ; foil 3 = angled<br />
trailing edge, same area as foil 1; foil 4 = angled trailing edge: same length <strong>and</strong> area as foil 1; foil 5 =<br />
forked trailing edge: same area as foil 1.<br />
be investigated. We designed five different<br />
flexible foils (material flexural<br />
stiffness = 3.1 × 10 −4 Nm 2 ) that control<br />
for total length <strong>and</strong> foil area <strong>and</strong><br />
allow different tail shapes to be compared<br />
for the effect of these changes<br />
on swimming speed (Figure 7). Foil 1<br />
corresponds to a highly abstracted<br />
“trout-like” body shape with a mostly<br />
vertical trailing tail edge, while foils 3<br />
<strong>and</strong> 4 present a shark-like tail trailing<br />
edge shape. Many fish have forked<br />
tails, <strong>and</strong> this shape is represented by<br />
foil 5.<br />
The fastest swimming foil (Figure<br />
7, foil 4) is the one with the<br />
most area near the axis of actuation,<br />
even though it has the same area as<br />
several of the other foils. This result<br />
corresponds with our previous results<br />
showing that foils with higher aspect<br />
ratios <strong>and</strong> hence more material near<br />
the actuator swim significantly faster<br />
(Lauderetal.,2011).Interestingly,<br />
the foil with the angled trailing edge<br />
<strong>and</strong> the shark tail shape (Figure 7,<br />
foil 3) swims significantly faster (P <<br />
0.003) than a foil with the same<br />
area but a straight trailing edge (Figure<br />
7, foil 1). The foil shape with<br />
the significantly lowest swimming<br />
speed was the notched shape (Figure<br />
7, foil 5) even though it possesses<br />
the same area as foils 1 <strong>and</strong> 3. The reduced<br />
performance of the notched<br />
shape may be due to bending of the<br />
upper <strong>and</strong> lower “lobes” of the tail<br />
during the flapping motion, <strong>and</strong> kinematic<br />
data obtained for these foils<br />
does show that each lobe of this shape<br />
twists during the flapping cycle. Twisting<br />
of the tail could be reduced by introducing<br />
stiffening elements, <strong>and</strong> this<br />
may be one reason that fishes with<br />
high-performance tail shapes such as<br />
tuna possess substantial stiffening of<br />
both the upper <strong>and</strong> lower tail lobes<br />
(Fierstine & Walters, 1968; Westneat<br />
& Wainwright, 2001).<br />
Although the angled foil shape<br />
swims significantly faster than a foil<br />
of the same area with a straight trailing<br />
edge (Figure 7: compare foils 1 <strong>and</strong> 3),<br />
more power is required for this foil to<br />
swim at this (self-propelled) speed<br />
(Figure 8A). The foils were made of<br />
48 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 8<br />
Graphs of power consumed (A) <strong>and</strong> the cost of<br />
transport (B) of foils 1 <strong>and</strong> 3 (see Figure 7) selfpropelling.<br />
Both foils have the same surface<br />
area (131.2 cm 2 ) <strong>and</strong> were actuated with<br />
±1 cm heave at 2 Hz. Power values are significantly<br />
different at P =0.003.Costoftransport<br />
values are not significantly different at P =0.07.<br />
thesamematerial<strong>and</strong>havethesame<br />
area, so the cost of transport can be<br />
calculated in mJ/m. Figure 8B shows<br />
that there is no significant difference<br />
(P = 0.07) between the foils in cost of<br />
transport, although there is a trend in<br />
the data toward the angled foil edge<br />
costing less per meter resulting in the<br />
marginally non-significant difference.<br />
Further experiments on both foils<br />
will be needed to extend this result<br />
<strong>and</strong> to determine if in fact there is a<br />
small difference in cost of transport between<br />
these two foil types.<br />
fishes that minimizes the complexity<br />
of the foil <strong>and</strong> maintains constant material<br />
properties along the foil length,<br />
fish bodies are clearly different in<br />
showing changing material properties<br />
from head to tail (McHenry et al.,<br />
1995). To better underst<strong>and</strong> the locomotor<br />
properties of the passive fish<br />
body alone, we used freshly dead rainbow<br />
trout (Oncorhynchus mykiss) <strong>and</strong><br />
attached the body to the robotic flapping<br />
foil apparatus (Figures 1D, 1E,<br />
<strong>and</strong> 1F). We took care to ensure<br />
that rigor mortis had not set in during<br />
the experiments <strong>and</strong> to ensure that<br />
the body had thus not stiffened during<br />
the time the flapping trials were<br />
conducted. By actuating the passive<br />
trout body in heave <strong>and</strong> pitch in various<br />
combinations just behind the<br />
head,wewereabletoconstructa<br />
swimming performance surface for<br />
the passive fish body (Figure 9).<br />
Body waveforms that are remarkably<br />
like those occurring in live trout<br />
were generated by the passive fish<br />
bodies. These data show that heave<br />
amplitude overall has the greater effect<br />
on swimming performance than<br />
FIGURE 9<br />
changes in pitch alone. At any given<br />
heave value increases in pitch further<br />
increase swimming speed, but at<br />
higher heave amplitudes near ±2 cm,<br />
changes in pitch only produce modest<br />
increases in SPS (Figure 9). These<br />
data provide an interesting comparison<br />
to our previous results showing<br />
the effects of heave <strong>and</strong> pitch actuation<br />
on flexible foil propulsion (Lauder<br />
et al., 2011). In that study adding<br />
pitch motions to baseline heave actuation<br />
for foils of varying flexural stiffness<br />
did not produce significant<br />
increases in foil SPS but did allow<br />
stiffer foils to maintain a relatively<br />
high swimming speed that would<br />
have declined with heave actuation<br />
only.<br />
As driving frequency increases, the<br />
tail beat amplitude of the passive flapping<br />
trout body remains relatively<br />
constant from 0.1 to 2.0 Hz before<br />
increasing steadily to a peak at<br />
3.5 Hz (Figure 10). These tail beat<br />
amplitude values are very similar to<br />
those observed in live trout swimming<br />
under a variety of locomotor conditions<br />
(Liao et al., 2003a; Webb,<br />
Performance surface of a freshly dead trout (Oncorhynchus mykiss) attached to a robotic controller<br />
(see Figures 1D, 1E, 1F) driven at 2 Hz at a variety of heave <strong>and</strong> pitch amplitudes. This trout was<br />
25.3 cm in total length. The graph shows how SPS of the passive trout body varies with different<br />
pitch <strong>and</strong> heave actuation parameters.<br />
Undulatory Locomotion<br />
of a Passive Fish Body<br />
Although flapping flexible foils<br />
provide a reasonable <strong>and</strong> simple<br />
model for undulatory propulsion in<br />
July/August 2011 Volume 45 Number 4 49
FIGURE 10<br />
Graph of tail-tip amplitude versus heave actuation frequency for a freshly dead trout (Oncorhynchus<br />
mykiss) attached to a robotic controller driving the passive body at a variety of frequencies. This trout<br />
was 25.3 cm in total length <strong>and</strong> was actuated with a constant heave (±2 cm) <strong>and</strong> pitch (±5°). Error<br />
bars are ±1 SE of the mean.<br />
1971; Webb et al., 1984) <strong>and</strong> increases<br />
in tail beat amplitude of the<br />
magnitude shown here for the passive<br />
trout body are similar to those observed<br />
previously as fish increase<br />
swimming speed <strong>and</strong> alter both frequency<br />
(primarily) <strong>and</strong> amplitude<br />
(to a lesser extent) of the tail beat.<br />
Although during routine undulatory<br />
swimming fish bodies are not<br />
passive <strong>and</strong> are certainly stiffened by<br />
body musculature as fish swim, for<br />
all but the fastest swimming speeds<br />
locomotion is powered by red muscle<br />
fibers, which can make up a rather<br />
small percentage of body mass<br />
(around 1.5% in largemouth bass;<br />
see Johnson et al., 1994). The bulk<br />
of the locomotor musculature is composed<br />
of white fibers that are not activated<br />
until the fastest swimming<br />
speeds are needed (Jayne & Lauder,<br />
1994, 1995a). Thus, the red muscle<br />
fibers can be considered as acting to<br />
bend a mostly passive fish body composed<br />
of white muscle fibers <strong>and</strong> associated<br />
skeletal tissues that may not be<br />
much different in flexural stiffness<br />
from the passive bodies studied here.<br />
Also, data on the propulsion of passive<br />
fish bodies are relevant to fishes<br />
swimming in turbulent flows. Trout<br />
swimming in a vortex street have been<br />
shown to greatly alter body kinematics<br />
<strong>and</strong> to utilize vortical energy shed from<br />
objects in the flow (Liao, 2004; Liao<br />
et al., 2003b). The amplitude of the<br />
center of mass oscillation by trout<br />
swimming in the Karman gait is up to<br />
±2 cm for a 10-cm-long trout, which<br />
corresponds on a length-specific basis<br />
to the middle region of the performance<br />
surface in Figure 9 (Liao et al., 2003a).<br />
By greatly reducing muscle activity<br />
when trout enter a vortex street, the<br />
body becomes largely passive <strong>and</strong> deforms<br />
in response to the oncoming<br />
flows. Trouts are able to maintain<br />
their position in such flows entirely passively<br />
<strong>and</strong> allow their bodies to extract<br />
energy from the oncoming flow <strong>and</strong><br />
generate thrust. This phenomenon<br />
was further demonstrated in a study<br />
using passive foils <strong>and</strong> freshly dead<br />
fish bodies to show that the passive<br />
trout body alone in a vortex wake can<br />
generate sufficient thrust to maintain<br />
position in flow (Beal et al., 2006).<br />
Fin-Fin Hydrodynamic<br />
Interactions<br />
One of the most intriguing aspects<br />
of fish functional design is the arrangement<br />
of two fins in series that could<br />
allow enhanced locomotor efficiency<br />
through hydrodynamic interactions<br />
between the fins. For example, the<br />
dorsal fin<strong>and</strong>theanalfininmostfishes<br />
are located upstream of the caudal fin,<br />
<strong>and</strong> the wakes shed from these fins<br />
could interact with the tail fin during<br />
locomotion (Drucker & Lauder, 2001,<br />
2005; Tytell, 2006). Undulatory locomotion<br />
using the body also causes the<br />
attached median fins to oscillate from<br />
side to side, <strong>and</strong> in the fish species studied<br />
so far these fins have been shown also<br />
to be actively oscillated by intrinsic fin<br />
musculature (Jayne et al., 1996) <strong>and</strong><br />
thus can generate thrust on their own.<br />
The possibility of fin-fin hydrodynamic<br />
interactions during locomotion<br />
has been explored in a number of<br />
experimental papers on living fishes<br />
(Drucker & Lauder, 2001, 2005;<br />
St<strong>and</strong>en, 2008; St<strong>and</strong>en & Lauder,<br />
2007; Tytell, 2006; Webb & Keyes,<br />
1981) as well as using computational<br />
approaches (Akhtar et al., 2007; Weihs<br />
et al., 2006). Akhtar et al. (2007) used<br />
data on the kinematics of the bluegill<br />
dorsal fin <strong>and</strong> tail from Drucker <strong>and</strong><br />
Lauder (2001) <strong>and</strong> performed a twodimensional<br />
computational analysis of<br />
the effect of having the fins in series<br />
<strong>and</strong> found that vortex shedding from<br />
the dorsal fin can increase thrust of the<br />
tail <strong>and</strong> that the amount of this thrust<br />
increase depends on the phasing of<br />
dorsal fin <strong>and</strong>tailmotion.<br />
In order to experimentally evaluate<br />
hydrodynamic fin-fin interactions<br />
using an apparatus in which phasing<br />
<strong>and</strong> the distance between fins can be<br />
experimentally manipulated, we used<br />
our robotic flapping foil apparatus<br />
in the dual-foil configuration with<br />
50 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
upstream <strong>and</strong> downstream foils. Foils<br />
were rigid aluminum plates in a<br />
NACA 0012 airfoil shape, <strong>and</strong> each<br />
foil could be moved in heave <strong>and</strong><br />
pitch, <strong>and</strong> both foils were driven from<br />
a common carriage mounted on air<br />
bearings as described above. Foils were<br />
moved according to the parameters<br />
measured for bluegill sunfish dorsal<br />
<strong>and</strong> caudal fins (Drucker & Lauder,<br />
2001) <strong>and</strong> used by Akhtar et al.<br />
(2007) for their computational study<br />
of this problem: the upstream foil was<br />
moved with heave amplitude of<br />
2.5 cm, the downstream foil at<br />
3.5 cm heave amplitude. Pitch amplitude<br />
for both foils was ±20°, <strong>and</strong> the<br />
frequency for both foils was 1.7 Hz,<br />
corresponding to the frequency of fin<br />
flapping in the bluegill sunfish model<br />
case (Drucker & Lauder, 2001).<br />
Increases in SPS as a result of<br />
changing foil phase <strong>and</strong> distance reflect<br />
thrust enhancement beyond the<br />
thrust achieved by the two foils operating<br />
separately (assessed by offsetting<br />
the foils from each other so that the<br />
downstream foil was no longer in<br />
the wake of the upstream foil). A general<br />
description of the dual-foil configuration<br />
<strong>and</strong> images of flows<br />
around <strong>and</strong> between the foils are presented<br />
in Lauder et al. (2007). Data<br />
for SPS were plotted over a range of<br />
phase differences between the upstream<br />
<strong>and</strong> downstream foils <strong>and</strong><br />
polynomial fits to these data points<br />
were used to determine the change<br />
in SPS with phase (e.g., Figure 11).<br />
Here we present the results of experiments<br />
measuring the SPS of dual<br />
foil flapping. Figure 11a shows the effect<br />
of changing the phase of sinusoidal<br />
motion of the downstream<br />
foil relative to the upstream foil on<br />
SPS for three different spacings between<br />
the foils. In each case, there is<br />
a clear peak in swimming speed, <strong>and</strong><br />
FIGURE 11<br />
Robotic model of fish fin hydrodynamic interactions. Dual NACA 0012 flapping foils in series are<br />
driven by a robotic flapping foil apparatus. Details of this device <strong>and</strong> of the experimental setup are<br />
given in Lauder et al. (2007). (A) SPS plotted against the phase difference between the upstream<br />
<strong>and</strong> downstream foils. Foils were arranged with 0 side-to-side offset (in cm) <strong>and</strong> are thus moving<br />
in line with each other with the downstream foil at varying chord length separations from the<br />
upstream foil: 0.5, 1, <strong>and</strong> 2 chord lengths separation; foil chord length was 6.85 cm. Further<br />
experimental details are provided in the text. (B) Effect of changing the foil offsets (in cm) so<br />
that the downstream foil is not moving in the direct wake of the upstream foil. The one chord length<br />
spacing with zero offset curve (dark blue) is the same as in panel (A) <strong>and</strong> is shown for reference.<br />
The other three plots show the effect of a 3.5 cm offset of the midline motion of the downstream<br />
foil relative to the upstream foil, removing it from the upstream foil wake.<br />
as the distance between the foils is increased<br />
the peak swimming speed<br />
shifts toward a larger phase lag of<br />
the downstream foil. Interestingly,<br />
the maximal swimming speed of the<br />
two foils together does not change significantly<br />
as the distance between foils<br />
changes, <strong>and</strong> alterations in phasing<br />
between the foils can thus be used<br />
to compensate for changes in spacing.<br />
For each interfoil spacing, however,<br />
there are phase relationships that significantly<br />
reduce swimming performance,<br />
indicating that thrust of the<br />
entire two-foil system is sensitive to<br />
phase relationships between two foils<br />
flapping. These data correspond very<br />
well to the computational results of<br />
Akhtar et al. (2007), which showed<br />
peak thrust enhancement at a phase<br />
July/August 2011 Volume 45 Number 4 51
of about 40°, very similar to our data<br />
(Figure11A,bluecurve)wherethe<br />
plateau around the SPS peak includes<br />
the 40° value.<br />
Figure 11B illustrates the changes<br />
in propulsion that result from moving<br />
the downstream foil to the side by an<br />
offset of 3.5 cm to move it out of the<br />
wake of the upstream foil. This effectively<br />
creates a t<strong>and</strong>em foil configuration<br />
in which fluid dynamic<br />
interactions between the foils are<br />
minimized, although not completely<br />
eliminated. Comparison of the three<br />
offset curves to the zero offset curve<br />
shows that offsetting the foils reduces<br />
both the SPS <strong>and</strong> the effect of foil<br />
phasing. Offset plots show reduced<br />
effects of foil phasing (with lower<br />
maxima <strong>and</strong> higher minima) <strong>and</strong> less<br />
distinct overall peaks in SPS, showing<br />
that the downstream foil was not able<br />
to improve swimming speed of the<br />
two foils together when removed<br />
from the upstream foil wake.<br />
Computational work <strong>and</strong> preliminary<br />
flow visualization (Lauder et al.,<br />
2007) data indicate that the behavior<br />
of t<strong>and</strong>em foils can be explained in<br />
terms of how the downstream foil interacts<br />
with fluid structures generated by<br />
the upstream foil. The shifting of the<br />
peak SPS with increased spacing of<br />
the t<strong>and</strong>em foils can be explained by<br />
the fact that, for larger spacings, the convection<br />
of those structures to the downstream<br />
foil takes longer. That time, tc,is<br />
simply the spacing between the foils, s,<br />
divided by the convection velocity, Uc.<br />
The increase in phase lag, Δϕ, needed<br />
so that the downstream foil meets the<br />
fluid structures at the right time is simply<br />
ðS 2 S 1 Þ<br />
Δϕ ¼ 2πf<br />
U c<br />
where f is the frequency of flapping<br />
<strong>and</strong> s 2 <strong>and</strong> s 1 are two different spacings.<br />
If we assume that Uc is close to the<br />
SPS, this equation estimates the Δϕ<br />
between the peak SPS well for the<br />
three spacings we used. The equation<br />
predicts Δϕ = 0.648 radians, or 37°,<br />
for a spacing difference of 0.5 chord<br />
lengths, <strong>and</strong> 74° for a difference of<br />
one chord length. The Δϕ from our<br />
data are approximately 30°<strong>and</strong> 75° for<br />
the corresponding spacing differences.<br />
The nearness of the foils <strong>and</strong>/or higher<br />
convection velocities at the smaller<br />
spacings may explain the lower than<br />
predicted Δϕ for the peaks in SPS in<br />
these cases. Overall, the agreement is<br />
very good considering that the peaks<br />
are somewhat broad <strong>and</strong> the equation<br />
for Δϕ is a simplification of the fluid<br />
dynamics.<br />
The Future of<br />
Undulatory Biorobotics<br />
In this paper, we use a robotic tool<br />
for investigating a variety of phenomena<br />
relating to undulatory propulsion<br />
in fishes <strong>and</strong> present experimental<br />
data that would be difficult if not impossible<br />
to obtain from studying live<br />
animals. The promise of robotic models<br />
for studying the biomechanics of<br />
locomotion in fishes has just begun<br />
to be realized (Curet et al., 2011;<br />
Long et al., 2006, 2010; Tangorra<br />
et al., 2010, 2011), <strong>and</strong> fundamental<br />
questions relating to the mechanics of<br />
undulatory propulsion remain to be<br />
addressed. In particular, key unresolved<br />
issues are the extent to which<br />
changes in body stiffness during propulsion<br />
affect locomotor performance<br />
(see Long & Nipper, 1996) <strong>and</strong> how<br />
active modulation of stiffness during<br />
an undulatory cycle <strong>and</strong> across<br />
changes in swimming speed are<br />
achieved <strong>and</strong> affect propulsive speed<br />
<strong>and</strong> efficiency.<br />
To address these questions, a new<br />
generation of robotic undulatory devices<br />
will be needed that allow for<br />
controlled modulation of body stiffness<br />
<strong>and</strong> the phasing of stiffness<br />
changes with undulatory cycles of<br />
compression <strong>and</strong> tension on the<br />
bending fish or foil body. An additional<br />
arena that is key to making<br />
progress in underst<strong>and</strong>ing undulatory<br />
mechanics is the ability to perturb the<br />
locomotor system to assess how stiffness<br />
of the body relates to the ability<br />
to recover from perturbations. There<br />
have been very few studies of perturbations<br />
of undulatory locomotor systems<br />
(see Webb, 2004), <strong>and</strong> yet fishes<br />
oftenswiminchallenginghydrodynamic<br />
environments in which they<br />
are forced to recover from impulsive<br />
challenges to their undulatory<br />
pattern.<br />
Acknowledgments<br />
This work was supported the Office<br />
of Naval Research grant N00014-<br />
09-1-0352 on fin neuromechanics<br />
monitored by Dr. Thomas McKenna<br />
<strong>and</strong> by the National Science Foundation<br />
grant EFRI-0938043. We<br />
thank the members of the Lauder<br />
<strong>and</strong> Tangorra labs for many helpful<br />
discussions on fish fins <strong>and</strong> flexible<br />
flapping foil propulsion <strong>and</strong> Nate<br />
Jackson for his assistance with the<br />
dual-flapping foil experiments. Many<br />
thanks to Brooke Flammang, Tyson<br />
Str<strong>and</strong>, <strong>and</strong> Dan Troolin for assistance<br />
with the V3V volumetric flow<br />
imaging experiments on the undulating<br />
plastic foil.<br />
Lead Author:<br />
George V. Lauder<br />
The Museum of<br />
Comparative Zoology<br />
52 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
26 Oxford Street, Harvard University<br />
Cambridge, MA 02138<br />
Email: glauder@oeb.harvard.edu<br />
References<br />
Affleck, R.J. 1950. Some points in the<br />
function, development, <strong>and</strong> evolution of the<br />
tail in fishes. Proc Zool Soc Lond. 120:349-68.<br />
doi: 10.1111/j.1096-3642.1950.tb00954.x.<br />
Akhtar, I., Mittal, R., Lauder, G.V., &<br />
Drucker, E. 2007. Hydrodynamics of a<br />
biologically inspired t<strong>and</strong>em flapping foil<br />
configuration. Theor Comp Fluid Dyn.<br />
21:155-70. doi: 10.1007/s00162-007-0045-2.<br />
Aleev, Y.G. 1969. Function <strong>and</strong> Gross<br />
Morphology in Fish. Jerusalem: Keter Press.<br />
Beal, D.N., Hover, F.S., Triantafyllou,<br />
M.S., Liao, J., & Lauder, G.V. 2006.<br />
Passive propulsion in vortex wakes.<br />
J Fluid Mech. 549:385-402. doi: 10.1017/<br />
S0022112005007925.<br />
Bone, Q., Kiceniuk, J., & Jones, D.R. 1978.<br />
On the role of the different fibre types in fish<br />
myotomes at intermediate swimming speeds.<br />
Fish Bull. 76:691-9.<br />
Curet, O.M., Patankar, N.A., Lauder, G.V.,<br />
& MacIver, M.A. 2011. Aquatic manoeuvering<br />
with counter-propagating waves: A novel<br />
locomotive strategy. J R Soc Interface.<br />
8:1041-50. doi: 10.1098/rsif.2010.0493.<br />
Donley, J., & Dickson, K.A. 2000. Swimming<br />
kinematics of juvenile Kawakawa tuna<br />
(Euthynnus affinis) <strong>and</strong> chub mackerel (Scomber<br />
japonicus). J Exp Biol. 203:3103-16.<br />
Drucker, E.G., & Lauder, G.V. 1999.<br />
Locomotor forces on a swimming fish: Threedimensional<br />
vortex wake dynamics quantified<br />
using digital particle image velocimetry.<br />
J Exp Biol. 202:2393-412.<br />
Drucker, E.G., & Lauder, G.V. 2001.<br />
Locomotor function of the dorsal fin in teleost<br />
fishes: Experimental analysis of wake forces<br />
in sunfish. J Exp Biol. 204:2943-58.<br />
Drucker, E.G., & Lauder, G.V. 2005.<br />
Locomotor function of the dorsal fin in<br />
rainbow trout: Kinematic patterns <strong>and</strong> hydrodynamic<br />
forces. J Exp Biol. 208:4479-94.<br />
doi: 10.1242/jeb.01922.<br />
Fetcho, J., & Svoboda, K.R. 1993. Fictive<br />
swimming elicited by electrical stimulation<br />
of the midbrain in goldfish. J Neurophysiol.<br />
70:764-80.<br />
Fetcho, J.R. 1986. The organization of the<br />
motoneurons innervating the axial musculature<br />
of vertebrates: I. Goldfish (Carassius<br />
auratus) <strong>and</strong> mudpuppies (Necturus maculosus).<br />
J Comp Neurol. 249:521-50. doi: 10.1002/<br />
cne.902490408.<br />
Fierstine, H.L., & Walters, V. 1968. Studies<br />
in locomotion <strong>and</strong> anatomy of scombroid<br />
fishes. Mem Southern Calif Acad Sci. 6:1-31.<br />
Flammang, B.E., & Lauder, G.V. 2008.<br />
Speed-dependent intrinsic caudal fin muscle<br />
recruitment during steady swimming in<br />
bluegill sunfish, Lepomis macrochirus. J Exp<br />
Biol. 211:587-98. doi: 10.1242/jeb.012096.<br />
Flammang, B.E., & Lauder, G.V. 2009.<br />
Caudal fin shape modulation <strong>and</strong> control<br />
during acceleration, braking <strong>and</strong> backing<br />
maneuvers in bluegill sunfish, Lepomis<br />
macrochirus. J Exp Biol. 212:277-86.<br />
doi: 10.1242/jeb.021360.<br />
Flammang, B.E., Lauder, G.V., Troolin,<br />
D.R., & Str<strong>and</strong>, T. 2011a. Volumetric<br />
imaging of shark tail hydrodynamics reveals<br />
a three-dimensional dual-ring vortex wake<br />
structure. P Roy Soc B. 278. doi: 10.1098/<br />
rspb.2011.0489.<br />
Flammang, B.E., Lauder, G.V., Troolin,<br />
D.R., & Str<strong>and</strong>, T.E. 2011b. Volumetric<br />
imaging of fish locomotion. Biol Lett.<br />
doi: 10.1098/rsbl.2011.0282.<br />
Gillis, G.B. 1996. Undulatory locomotion<br />
in elongate aquatic vertebrates: Anguilliform<br />
swimming since Sir. James Gray. Am Zool.<br />
36:656-65.<br />
Gillis, G.B. 1998. Environmental effects on<br />
undulatory locomotion in the American eel<br />
Anguilla rostrata: Kinematics in water <strong>and</strong><br />
on l<strong>and</strong>. J Exp Biol. 201:949-61.<br />
Harris, J.E. 1936. The role of the fins in the<br />
equilibrium of the swimming fish: I. Wind<br />
tunnel tests on a model of Mustelus canis<br />
(Mitchell). J Exp Biol. 13:476-93.<br />
Harris, J.E. 1938. The role of the fins in the<br />
equilibrium of the swimming fish: II. The role<br />
of the pelvic fins. J Exp Biol. 16:32-47.<br />
Hove, J.R., O’Bryan, L.M., Gordon, M.S.,<br />
Webb, P.W., & Weihs, D. 2001. Boxfishes<br />
(Teleostei: Ostraciidae) as a model system for<br />
fishes swimming with many fins: Kinematics.<br />
J Exp Biol. 204:1459-71.<br />
Jayne, B.C., & Lauder, G.V. 1993. Red <strong>and</strong><br />
white muscle activity <strong>and</strong> kinematics of the<br />
escape response of the bluegill sunfish during<br />
swimming. J Comp Physiol A. 173:495-508.<br />
doi: 10.1007/BF00193522.<br />
Jayne, B.C., & Lauder, G.V. 1994. How<br />
swimming fish use slow <strong>and</strong> fast muscle fibers:<br />
Implications for models of vertebrate muscle<br />
recruitment. J Comp Physiol A. 175:123-31.<br />
doi: 10.1007/BF00217443.<br />
Jayne, B.C., & Lauder, G.V. 1995a. Are<br />
muscle fibers within fish myotomes activated<br />
synchronously Patterns of recruitment within<br />
deep myomeric musculature during swimming<br />
in largemouth bass. J Exp Biol.<br />
198:805-15.<br />
Jayne, B.C., & Lauder, G.V. 1995b. Red<br />
muscle motor patterns during steady swimming<br />
in largemouth bass: Effects of speed<br />
<strong>and</strong> correlations with axial kinematics.<br />
J Exp Biol. 198:1575-87.<br />
Jayne, B.C., & Lauder, G.V. 1995c.<br />
Speed effects on midline kinematics during<br />
steady undulatory swimming of largemouth<br />
bass, Micropterus salmoides. JExpBiol.<br />
198:585-602.<br />
Jayne, B.C., Lozada, A., & Lauder, G.V.<br />
1996. Function of the dorsal fin in bluegill<br />
sunfish: Motor patterns during four locomotor<br />
behaviors. J Morphol. 228:307-26.<br />
doi:10.1002/(SICI)1097-4687(199606)<br />
228:33.0.CO;2-Z.<br />
Johnson, T.P., Syme, D.A., Jayne, B.C.,<br />
Lauder, G.V., & Bennett, A.F. 1994. Modeling<br />
red muscle power output during steady <strong>and</strong> unsteady<br />
swimming in largemouth bass (Micropterus<br />
salmoides). Am J Physiol. 267:R481-8.<br />
July/August 2011 Volume 45 Number 4 53
Lauder, G., Madden, P.G.A., Tangorra, J.,<br />
Anderson, E., & Baker, T.V. 2011.<br />
Bioinspiration from fish for smart material<br />
design <strong>and</strong> function. Smart Mater Struct.<br />
in press.<br />
Lauder, G.V. 1989. Caudal fin locomotion<br />
in ray-finned fishes: Historical <strong>and</strong> functional<br />
analyses. Am Zool. 29:85-102.<br />
Lauder, G.V. 2000. Function of the caudal<br />
fin during locomotion in fishes: Kinematics,<br />
flow visualization, <strong>and</strong> evolutionary patterns.<br />
Am Zool. 40:101-22. doi: 10.1668/0003-<br />
1569(2000)040[0101:FOTCFD]2.0.CO;2.<br />
Lauder, G.V., Anderson, E.J., Tangorra, J.,<br />
& Madden, P.G.A. 2007. Fish biorobotics:<br />
Kinematics <strong>and</strong> hydrodynamics of selfpropulsion.<br />
J Exp Biol. 210:2767-80.<br />
doi: 10.1242/jeb.000265.<br />
Lauder, G.V., & Tytell, E.D. 2006. Hydrodynamics<br />
of undulatory propulsion. In:<br />
Fish Biomechanics. Volume 23 in Fish<br />
Physiology, eds. Shadwick, R.E., & Lauder,<br />
G.V., 425-68. San Diego: Academic Press.<br />
Liao, J. 2002. Swimming in needlefish<br />
(Belonidae): Anguilliform locomotion with<br />
fins. J Exp Biol. 205:2875-84.<br />
Liao, J. 2004. Neuromuscular control of trout<br />
swimming in a vortex street: Implications for<br />
energy economy during the Karman gait.<br />
J Exp Biol. 207:3495-506. doi: 10.1242/<br />
jeb.01125.<br />
Liao, J., Beal, D.N., Lauder, G.V., &<br />
Triantafyllou, M.S. 2003a. The Kármán gait:<br />
Novel body kinematics of rainbow trout<br />
swimming in a vortex street. J Exp Biol.<br />
206:1059-73. doi: 10.1242/jeb.00209.<br />
Liao, J., & Lauder, G.V. 2000. Function<br />
of the heterocercal tail in white sturgeon: Flow<br />
visualization during steady swimming <strong>and</strong><br />
vertical maneuvering. J Exp Biol. 203:3585-94.<br />
Liao, J.C., Beal, D.N., Lauder, G.V., &<br />
Triantafyllou, M.S. 2003b. Fish exploiting<br />
vortices decrease muscle activity. Science.<br />
302:1566-9. doi: 10.1126/science.1088295.<br />
Long, J.H., Jr., Koob, T.J., Irving, K.,<br />
Combie, K., Engel, V., Livingston, N., …<br />
Schumacher, J. 2006. Biomimetic evolutionary<br />
analysis: Testing the adaptive value of<br />
vertebrate tail stiffness in autonomous swimming<br />
robots. J Exp Biol. 209:4732-46.<br />
doi: 10.1242/jeb.02559.<br />
Long, J.H., McHenry, M.J., & Boetticher,<br />
N.C. 1994. Undulatory swimming: How<br />
traveling waves are produced <strong>and</strong> modulated<br />
in sunfish (Lepomis gibbosus). J Exp Biol.<br />
192:129-45.<br />
Long, J.H., & Nipper, K.S. 1996. The<br />
importance of body stiffness in undulatory<br />
propulsion. Am Zool. 36:678-694.<br />
Long, J.H., Porter, M.E., Root, R.G., &<br />
Liew, C.W. 2010. Go reconfigure: How<br />
fish change shape as they swim <strong>and</strong> evolve.<br />
Integr Comp Biol. 50:1120-39. doi: 10.1093/<br />
icb/icq066.<br />
McHenry, M.J., Pell, C.A., & Long, J.A.<br />
1995. Mechanical control of swimming speed:<br />
Stiffness <strong>and</strong> axial wave form in undulating<br />
fish models. J Exp Biol. 198:2293-305.<br />
Nauen, J.C., & Lauder, G.V. 2002a. Hydrodynamics<br />
of caudal fin locomotion by chub<br />
mackerel, Scomber japonicus (Scombridae).<br />
J Exp Biol. 205:1709-24.<br />
Nauen, J.C., & Lauder, G.V. 2002b.<br />
Quantification of the wake of rainbow trout<br />
(Oncorhynchus mykiss) using three-dimensional<br />
stereoscopic digital particle image velocimetry.<br />
J Exp Biol. 205:3271-9.<br />
Plaut, I. 2000. Effects of fin size on swimming<br />
performance, swimming behavior <strong>and</strong><br />
routine activity of zebrafish Danio rerio.<br />
J Exp Biol. 203:813-20.<br />
Rome, L.C., Swank, D., & Corda, D. 1993.<br />
How fish power swimming. Science. 261:<br />
340-3. doi: 10.1126/science.8332898.<br />
Shadwick, R., & Gemballa, S. 2006.<br />
Structure, kinematics, <strong>and</strong> muscle dynamics in<br />
undulatory swimming. In: Fish Biomechanics.<br />
Volume 23 in Fish Physiology, eds. Shadwick,<br />
R.E., & Lauder, G.V., 241-280. San Diego:<br />
Academic Press.<br />
St<strong>and</strong>en, E.M. 2008. Pelvic fin locomotor<br />
function in fishes: Three-dimensional kinematics<br />
in rainbow trout (Oncorhynchus mykiss).<br />
J Exp Biol. 211:2931-42. doi: 10.1242/<br />
jeb.018572.<br />
St<strong>and</strong>en, E.M. 2010. Muscle activity <strong>and</strong><br />
hydrodynamic function of pelvic fins in<br />
trout (Oncorhynchus mykiss). J Exp Biol.<br />
213:831-41. doi: 10.1242/jeb.033084.<br />
St<strong>and</strong>en, E.M., & Lauder, G.V. 2005. Dorsal<br />
<strong>and</strong> anal fin function in bluegill sunfish<br />
(Lepomis macrochirus): Three-dimensional<br />
kinematics during propulsion <strong>and</strong> maneuvering.<br />
J Exp Biol. 205:2753-63. doi: 10.1242/<br />
jeb.01706.<br />
St<strong>and</strong>en, E.M., & Lauder, G.V. 2007.<br />
Hydrodynamic function of dorsal <strong>and</strong> anal<br />
fins in brook trout (Salvelinus fontinalis).<br />
J Exp Biol. 210:325-39. doi: 10.1242/<br />
jeb.02661.<br />
Syme, D.A. 2006. Functional properties of<br />
skeletal muscle. In: Fish Biomechanics.<br />
Volume 23 in Fish Physiology, eds. Shadwick,<br />
R.E., & Lauder, G.V., 179-240. San Diego:<br />
Academic Press.<br />
Tangorra, J., Phelan, C., Esposito, C., &<br />
Lauder, G. 2011. Biorobotic models of highly<br />
deformable fins for studies of the mechanics<br />
<strong>and</strong> control of fin forces. Integr Comp Biol.<br />
in press. doi: 10.1093/icb/icr036.<br />
Tangorra, J.L., Lauder, G.V., Hunter, I.,<br />
Mittal, R., Madden, P.G., & Bozkurttas, M.<br />
2010. The effect of fin ray flexural rigidity<br />
on the propulsive forces generated<br />
by a biorobotic fish pectoral fin. J Exp<br />
Biol. 213:4043-54. doi: 10.1242/<br />
jeb.048017.<br />
Thomson, K.S. 1971. The adaptation <strong>and</strong><br />
evolution of early fishes. Q Rev Biol. 46:<br />
139-66. doi: 10.1086/406831.<br />
Thomson, K.S. 1976. On the heterocercal<br />
tail in sharks. Paleobiology. 2:19-38.<br />
Troolin, D., & Longmire, E. 2010. Volumetric<br />
velocity measurements of vortex rings from<br />
inclined exits. Exp Fluids. 48:409-20.<br />
doi: 10.1007/s00348-009-0745-z.<br />
Tytell, E.D. 2004. Kinematics <strong>and</strong> hydrodynamics<br />
of linear acceleration in eels,<br />
54 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
Anguilla rostrata. P Roy Soc Lond B.<br />
271:2535-40. doi: 10.1098/rspb.2004.2901.<br />
Tytell, E.D. 2006. Median fin function<br />
in bluegill sunfish, Lepomis macrochirus:<br />
Streamwise vortex structure during steady<br />
swimming. J Exp Biol. 209:1516-34.<br />
doi: 10.1242/jeb.02154.<br />
Wilga, C.D., & Lauder, G.V. 2004b.<br />
Hydrodynamic function of the shark’s tail.<br />
Nature. 430:850. doi: 10.1038/430850a.<br />
Tytell, E.D., & Lauder, G.V. 2002. The<br />
c-start escape response of Polypterus senegalus:<br />
Bilateral muscle activity <strong>and</strong> variation during<br />
stage 1 <strong>and</strong> 2. J Exp Biol. 205:2591-603.<br />
Webb, P.W. 1971. The swimming energetics<br />
of trout: I. Thrust <strong>and</strong> power output at<br />
cruising speeds. J Exp Biol. 55:489-520.<br />
Webb, P.W. 2004. Response latencies to<br />
postural disturbances in three species of<br />
teleostean fishes. J Exp Biol. 207:955-61.<br />
doi: 10.1242/jeb.00854.<br />
Webb, P.W., & Keyes, R.S. 1981. Division<br />
of labor between median fins in swimming<br />
dolphin (Pisces: Coryphaenidae). Copeia.<br />
1981:901-04. doi: 10.2307/1444198.<br />
Webb, P.W., Kostecki, P.T., & Stevens,<br />
E.D. 1984. The effect of size <strong>and</strong> swimming<br />
speed on the locomotor kinematics of rainbow<br />
trout. J Exp Biol. 109:77-95.<br />
Weihs, D., Ringel, M., & Victor, M. 2006.<br />
Aerodynamic interactions between adjacent<br />
slender bodies. AIAA J. 44:481-4.<br />
doi: 10.2514/1.18902.<br />
Westneat, M., & Wainwright, S.A. 2001.<br />
Mechanical design for swimming: Muscle,<br />
tendon, <strong>and</strong> bone. In: Tuna: Physiology,<br />
Ecology, <strong>and</strong> Evolution, eds. Block, B., &<br />
Stevens, E.D., 271-311. San Diego: Academic<br />
Press. doi: 10.1016/S1546-5098(01)19008-4.<br />
Wilga, C.D., & Lauder, G.V. 2002. Function<br />
of the heterocercal tail in sharks: Quantitative<br />
wake dynamics during steady horizontal<br />
swimming <strong>and</strong> vertical maneuvering.<br />
J Exp Biol. 205:2365-74.<br />
Wilga, C.D., & Lauder, G.V. 2004a.<br />
Biomechanics of locomotion in sharks, rays<br />
<strong>and</strong> chimeras. In: Biology of Sharks <strong>and</strong> Their<br />
Relatives, eds. Carrier, J.C., Musick, J.A.,<br />
& Heithaus, M.R., 139-64. Boca Raton:<br />
CRC Press.<br />
July/August 2011 Volume 45 Number 4 55
PAPER<br />
Thrust Production in Highly Flexible Pectoral<br />
Fins: A Computational Dissection<br />
AUTHORS<br />
Srinivas Ramakrishnan 1<br />
ANSYS, Inc.<br />
Meliha Bozkurttas<br />
Franklin W. Olin<br />
College of Engineering<br />
Rajat Mittal<br />
Department of Mechanical Engineering,<br />
Johns Hopkins University<br />
George V. Lauder<br />
Department of Organismic<br />
<strong>and</strong> Evolutionary Biology,<br />
Harvard University<br />
Introduction<br />
Robust design based on natural<br />
systems is a significant engineering<br />
challenge. Evolution-based design is<br />
inherently a multi-objective optimization<br />
problem. Natural selection puts<br />
pressure on organisms to produce<br />
locomotion abilities that balance competing<br />
requirements of speed, efficiency,<br />
<strong>and</strong> effectiveness. The goals<br />
of ongoing research efforts are to elucidate<br />
the competing requirements that<br />
have enabled the evolution of highly<br />
maneuverable propulsion/locomotion<br />
at low speeds. Prominent natural systems<br />
of interest are flapping flight in<br />
air <strong>and</strong> aquatic locomotion <strong>and</strong> a common<br />
feature among these systems is<br />
the presence of highly compliant control<br />
surfaces. Organisms that employ<br />
these models of locomotion appear to<br />
exploit the flexibility of their wings/<br />
1 Work presented here was done while the author<br />
was a postdoctoral scientist at The George Washington<br />
University prior to joining Ansys Inc.<br />
ABSTRACT<br />
Bluegill sunfish pectoral fins represent a remarkable success in evolutionary<br />
terms as a means of propulsion in challenging environments. Attempts to mimic<br />
their design in the context of autonomous underwater vehicles have overwhelmingly<br />
relied on the analysis of steady swimming. Experimental observations of<br />
maneuvers reveal that the kinematics of fin <strong>and</strong> wake dynamics exhibit characteristics<br />
that are distinctly different from steady swimming. We present a computational<br />
analysis that compares, qualitatively <strong>and</strong> quantitatively, the wake hydrodynamics<br />
<strong>and</strong> performance of the bluegill sunfish pectoral fin for two modes of swimming:<br />
steady swimming <strong>and</strong> a yaw turn maneuver. It is in this context that we comment on<br />
the role that flexibility plays in the success of the pectoral fin as a versatile propulsor.<br />
Specifically, we assess the performance of the fin by conducting a “virtual dissection”<br />
where only a portion of fin is retained. Approximately 90% of peak thrust for<br />
steady swimming is recovered using only the dorsal half. This figure drops to 70%<br />
for the yaw turn maneuver. Our findings suggest that designs based on fin analysis<br />
that account for various locomotion modes can lead to more robust performance<br />
than those based solely on steady swimming.<br />
Keywords: computational fluid dynamics (CFD), immersed boundary methods<br />
(IBM), bluegill sunfish, biological locomotion<br />
fins to achieve high maneuverability<br />
at low speeds. This paper presents the<br />
analysis of one such control surface:<br />
the bluegill sunfish pectoral fin.<br />
Atypicalsunfish pectoral fin<br />
consists of 14 fin raysasshownin<br />
Figure 1. We see the fin raysnumbered<br />
sequentially starting from the<br />
dorsal edge (ray 1) to the ventral edge<br />
(ray 14). These rays support an asymmetric<br />
planform shape for the pectoral<br />
fin. Figure 3 shows different frames of<br />
the sunfish executing a maneuver from<br />
a ventral view. The motion of the pectoral<br />
fin <strong>and</strong> body are captured using<br />
multiple high-speed video cameras<br />
simultaneously operating at 250 or<br />
more frames per second with a 1024 ×<br />
1024 resolution (Lauder et al., 2006).<br />
The wing surface is digitized at about<br />
300 spatial locations at several points<br />
during the fin cycle.Thus,thekinematics<br />
of the fin motionisacquired<br />
for the simulation. The collaboration<br />
with experimentalists (biologists <strong>and</strong><br />
FIGURE 1<br />
Bluegill sunfish pectoral fin consists of<br />
14 rays, which form the full planform. The<br />
dissected planform is interpolated from rays<br />
1–8 to investigate the flow <strong>and</strong> performance.<br />
56 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 2<br />
Bioinspired design paradigm.<br />
engineers), through a multi-disciplinary<br />
effort (Lauder et al., 2006; Mittal et al.,<br />
2006), has enabled high-fidelity data<br />
to be used in the computational analysis<br />
(see Figure 2).<br />
It is clear from looking at the fin<br />
motion during the maneuver (see Figure<br />
3) that the kinematics of the fin<br />
involves both deformation <strong>and</strong> translation.<br />
This poses severe challenges for<br />
traditional body-fitted computational<br />
methods. Here, the immersed boundary<br />
method, with its ability to h<strong>and</strong>le<br />
complex deforming structures, enables<br />
us to undertake high-fidelity computational<br />
fluid dynamics (CFD) analysis of<br />
the pectoral fin hydrodynamics(see<br />
Computational Methodology). It has<br />
been used to gain valuable insight into<br />
pectoral fin hydrodynamics in steady<br />
swimming (Bozkurttas et al., 2009;<br />
Dong et al., 2010). The experimentally<br />
obtained steady swimming kinematics<br />
was analyzed, <strong>and</strong> an efficient reconstruction<br />
of the kinematics using proper<br />
orthogonal decomposition (POD)<br />
was obtained. The POD modes using<br />
a combination of the first three modes<br />
(hereafterreferredtoasMode1+2+3)<br />
were successful in reproducing two<br />
thirds of the full fin kinematics. More<br />
significantly, this combination of<br />
modes was found to retain 92%<br />
of the thrust produced using the actual<br />
kinematics (Bozkurttas, 2007;<br />
Bozkurttas et al., 2009). Further, detailed<br />
analysis of the pressure distribution<br />
over the full fin surface (rays<br />
1-14) during steady swimming also revealed<br />
that most of the thrust was produced<br />
by the dorsal part mainly around<br />
the spanwise tip region (Bozkurttas<br />
et al., 2009; Dong et al., 2010) (see Figure<br />
5). Since different sections of the<br />
pectoral fin trace different trajectories<br />
during a fin stroke, the contribution<br />
of each region of the fin to its overall<br />
performance may not be uniform. Naturally,<br />
this leads us to the central theme<br />
of this paper, the idea of examining the<br />
thrust production of different sections<br />
of the fin. The goal is to enable a virtual<br />
“dissection” or “ablation” of the pectoral<br />
fin dynamics <strong>and</strong> the effect of this<br />
ablation on the fin performance. It is<br />
expected that this will yield useful insight<br />
into the hydrodynamic function<br />
of the fin in various swimming modes.<br />
FIGURE 3<br />
A bluegill sunfish during a maneuver: ventral (bottom) view. Images are frames from a highspeed<br />
video. Note the differential motion of the left <strong>and</strong> right side fins. Top row: t/T =0,t/T =<br />
0.23, t/T = 0.30. Bottom row: t/T = 0.46, t/T = 0.70, t/T = 0.84.<br />
Computational<br />
Methodology<br />
We present a brief description of<br />
the Cartesian grid-based immersed<br />
boundary method for moving boundaries<br />
starting with the governing<br />
equations. The three-dimensional<br />
unsteady, viscous incompressible<br />
Navier-Stokes equations are given as<br />
∂u i<br />
¼ 0<br />
∂x i<br />
∂u i<br />
∂t þ ∂ u <br />
iu j<br />
¼ 1 ∂p<br />
þ ν ∂ <br />
∂u i<br />
∂x j ρ ∂x i ∂x j ∂x j<br />
ð1Þ<br />
where i; j =1,2,3,u i are the velocity<br />
component, p is the pressure, <strong>and</strong> ρ<br />
July/August 2011 Volume 45 Number 4 57
<strong>and</strong> ν are the fluid density <strong>and</strong> kinematicviscosity.Wehaveemployeda<br />
conventional notation where repeated<br />
indices imply summation.<br />
1. Numerical Method<br />
The Navier-Stokes equations<br />
(Eq. 1) are discretized using a cellcentered,<br />
collocated (non-staggered)<br />
arrangement of the primitive<br />
variables (u i , p). In addition to the<br />
cell-centered velocities (u i ), the<br />
face-centered velocities, U i ,are<br />
computed. A second-order Adams-<br />
Bashforth scheme is employed for<br />
the convective terms while the diffusion<br />
terms are discretized using<br />
an implicit Crank-Nicolson scheme<br />
whicheliminatestheviscousstability<br />
constraint. The spatial derivatives<br />
are computed using a<br />
second-order accurate central difference<br />
scheme. The equations<br />
are integrated in time using the<br />
fractional step method (Chorin,<br />
1967). In the first sub-step of this<br />
method, a modified momentum<br />
equation is solved <strong>and</strong> an intermediate<br />
velocity u* obtained. The second<br />
sub-step requires the solution<br />
of the pressure correction equation<br />
which is solved with the constraint<br />
that the final velocity u i<br />
n+1<br />
be divergence-free. This gives a<br />
Poisson equation for the pressure<br />
correction <strong>and</strong> a Neumann boundary<br />
condition imposed on this pressure<br />
correction at all boundaries.<br />
This Poisson equation is solved<br />
with a highly efficient geometric<br />
multigridmethodwhichemploys<br />
a Gauss-Siedel line-SOR smoother.<br />
Once the pressure correction is obtained,<br />
the pressure <strong>and</strong> velocity are<br />
updated (see Dong et al., 2006 <strong>and</strong><br />
Mittal et al., 2008, for additional<br />
details). These separately updated<br />
face velocities satisfy discrete mass<br />
conservation to machine accuracy<br />
<strong>and</strong> use of these velocities in estimating<br />
the non-linear convective flux<br />
leads to a more accurate <strong>and</strong> robust<br />
solution procedure. The advantage<br />
of separately computing the facecentered<br />
velocities was initially proposed<br />
by Zang et al. (1994) <strong>and</strong><br />
discussed in the context of the<br />
Cartesian grid methods in Ye et al.<br />
(1999) <strong>and</strong> Mittal et al. (2008).<br />
2. Immersed Boundary Treatment<br />
The immersed boundary method<br />
used here employs a multidimensional<br />
ghost cell methodology<br />
to impose the boundary conditions<br />
on the immersed boundary. The<br />
current solver is designed from the<br />
start for fast, efficient, <strong>and</strong> accurate<br />
solution of flows with complex<br />
three-dimensional, moving boundaries.<br />
Also, the current method is<br />
a “sharp“ interface method in that<br />
the boundary conditions on the<br />
immersed boundary are imposed<br />
at the precise location of the immersed<br />
body, <strong>and</strong> there is no spurious<br />
spreading of boundary forcing<br />
into the fluid as what usually occurs<br />
with diffuse interface methods<br />
(Mittal & Iaccarino, 2005).<br />
3. Geometric Representation<br />
of Immersed Boundary<br />
The current method is designed to<br />
simulate flows over arbitrarily complex<br />
2D <strong>and</strong> 3D immersed stationary<br />
<strong>and</strong> moving boundaries <strong>and</strong> the<br />
approach chosen to represent the<br />
boundary surface should be flexible<br />
enough so as not to limit the type<br />
of geometries that can be h<strong>and</strong>led.<br />
A number of different approaches<br />
are available for representing the<br />
surface of the immersed boundary,<br />
including level sets (Osher &<br />
Sethian, 1988; Tran & Udaykumar,<br />
2004), <strong>and</strong> unstructured surface<br />
grids. In the current solver, we<br />
choose to represent the surface of<br />
the immersed boundary by an unstructured<br />
mesh with triangular elements.<br />
This approach is very well<br />
suited for the wide variety of engineering<br />
<strong>and</strong> biological configurations<br />
that are of interest to us <strong>and</strong><br />
is compatible with the immersed<br />
boundary methodology used in<br />
the current solver.<br />
4. Boundary Motion<br />
Boundary motion can be included<br />
into immersed boundary formulation<br />
with relative ease. In advancing<br />
the field equations from time level<br />
n to n + 1 in the case of a moving<br />
boundary, the first step is to move<br />
from its current location to the<br />
new location. This is accomplished<br />
by moving the nodes of the surface<br />
triangles with a known velocity.<br />
Thus, we employ the following<br />
equation to update the coordinates<br />
(X i ) of the surface element vertices,<br />
X nþ1<br />
i<br />
Δt<br />
X n<br />
i<br />
¼ V nþ1<br />
i<br />
ð2Þ<br />
where V i is the vertex velocity. The<br />
vertex velocity can either be prescribed<br />
or it can be computed<br />
from a dynamical equation if the<br />
body motion is coupled to the<br />
fluid. The next step is to determine<br />
the ghost cells for this new immersed<br />
boundary location <strong>and</strong><br />
recompute interpolation weights<br />
associated with the ghost point<br />
methodology. Subsequently, the<br />
flow equations, which are written<br />
in Eulerian form, are advanced<br />
in time. The general framework<br />
described above can, therefore, be<br />
considered as Eulerian-Lagrangian,<br />
wherein the immersed boundaries<br />
are explicitly tracked as surfaces in<br />
a Lagrangian mode, while the flow<br />
computations are performed on a<br />
fixed Eulerian mesh. Additional<br />
58 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
details regarding the current immersed<br />
boundary methodology<br />
may be found in Mittal et al. (2008).<br />
Computational Setup<br />
All simulations are conducted in<br />
a rectangular computational domain.<br />
The boundary conditions on the<br />
bounding box of the domain are freestream<br />
on the left (x direction), outflow<br />
on the right while the remaining<br />
boundaries (top <strong>and</strong> bottom ( y direction)<br />
<strong>and</strong> front <strong>and</strong> back (z direction))<br />
employ slip boundary conditions (see<br />
Figure 4). The fin surface <strong>and</strong> fish<br />
body are considered as no-slip boundaries.<br />
The fins are treated as deforming<br />
membranes while the body, where applicable,<br />
is treated as rigid body undergoing<br />
general motion. The Reynolds<br />
number in the present work is defined<br />
as Re = UL s /ν where U, L s ,<strong>and</strong>ν are<br />
the swimming velocity, spanwise fin<br />
length, <strong>and</strong> the kinematic viscosity of<br />
FIGURE 4<br />
water (ν =1.007×10 −6 m 2 s −1 at room<br />
temperature), respectively.<br />
Based on a swimming speed of<br />
1.1 body length per second, the<br />
Reynolds number for the steady swimming<br />
is 6300. However, a comparison<br />
of the force coefficients obtained<br />
at Re = 1440 with those at the experimental<br />
Reynolds number appear to be<br />
in good agreement both quantitatively<br />
<strong>and</strong> qualitatively (Bozkurttas, 2007).<br />
So, for computational expediency, we<br />
use the lower Reynolds number in the<br />
steady swimming analysis (Dong et al.,<br />
2010). As mentioned earlier, low dimensional<br />
model performance analyses<br />
have shown that Mode 1 + 2 + 3<br />
gait that accounts for 67% of the fin<br />
motion still produces 92% of the<br />
thrust (Bozkurttas et al., 2009). Therefore,<br />
in lieu of the experimentally<br />
extracted fin kinematics, this simplified<br />
model has been used here. The<br />
grid size in these simulations is 153 ×<br />
161 × 97, which is about 2.35 million<br />
Cartesian grid (4.8 million grid points) <strong>and</strong> unstructured mesh employed for yaw maneuver:<br />
(a) x-y plane section, (b) x-z plane section, (c) y-z plane section (strongside fin ontheleft<br />
<strong>and</strong> weakside on right of the body), <strong>and</strong> (d) unstructured surface mesh (pectoral fin only, number<br />
of nodes = 10,000, number of elements = 19,602).<br />
grid points. A domain size of 3.8L s ×<br />
4.5L s ×1.8L s is selected where L s is<br />
the span wise size of the fin. Comprehensive<br />
studies have been carried out<br />
to assess the effect of the grid resolution<br />
<strong>and</strong> domain size on the salient<br />
features of the flow <strong>and</strong> also to demonstrate<br />
the accuracy of the selected grid<br />
(Bozkurttas, 2007).<br />
The Reynolds number for the turning<br />
maneuver based on a freestream<br />
velocity of 0.5 body lengths per second<br />
is approximately 3500. The domain<br />
size employed for the maneuver is<br />
7.5L s ×5L s ×5L s . The pectoral fins<br />
<strong>and</strong> an idealized body, immersed in<br />
the computational grid, are shown in<br />
Figure 4. The nominal grid size used<br />
in the current simulation is 241 ×<br />
145 × 145 (see Figure 4). Finally, the<br />
domain size for the maneuver with<br />
just the strongside (outside) fin is<br />
4L s ×4L s ×4L s with a non-uniform<br />
grid using 128 points in all three<br />
dimensions.<br />
We note in passing that all the steady<br />
swimming cases <strong>and</strong> ablated fin simulations<br />
(for the maneuver) do not include<br />
the fish body. This is reasonable<br />
sincewehaveobservedthatthedifference<br />
in the thrust coefficients with<br />
<strong>and</strong> without the body is minimal. As<br />
we shall see shortly, the wake dynamics<br />
for both steady swimming <strong>and</strong> maneuver<br />
are dominated by vortex structures<br />
generated far from the fish body (see<br />
Figures 5 <strong>and</strong> 8). Thus, the interaction<br />
between the body <strong>and</strong> the fin hydrodynamics<br />
is minimal.<br />
The performance of the fin is evaluated<br />
using the computed force coefficients<br />
which are defined as,<br />
C T ¼<br />
2F x<br />
ρU∞ 2A fin<br />
; C L ¼ 2F y<br />
ρU∞ 2A ;<br />
fin<br />
C Z ¼<br />
2F z<br />
ρU 2 ∞ A fin<br />
ð3Þ<br />
July/August 2011 Volume 45 Number 4 59
FIGURE 5<br />
The anatomy of the principal vortex dynamics<br />
involved in steady swimming.<br />
where F x , F y <strong>and</strong> F z are the forces respectively<br />
in the streamwise (drag/<br />
thrust), vertical (lift), <strong>and</strong> spanwise<br />
(lateral) directions, A fin is the nominal<br />
fin area,<strong>and</strong>ρ is the density of the<br />
fluid. U ∞ is the forward swimming<br />
velocity. The force components are<br />
calculated by directly integrating the<br />
computed pressure <strong>and</strong> shear stress<br />
on the fin surface.<br />
Results<br />
Steady Swimming<br />
A snapshot of the vortex dynamics<br />
at the end of a steady swimming fin<br />
beatisshowninFigure5.Notethe<br />
FIGURE 6<br />
complex interaction of among vortices<br />
generated by the path traversed by the<br />
fin tip during a fin beat. Clearly, both<br />
adduction <strong>and</strong> abduction appear to<br />
produce distinct vortex structures.<br />
This is in stark contrast with a simple<br />
ring vortex created during the maneuver<br />
(see Figure 8). The time variations<br />
of the force coefficients (C T , C L <strong>and</strong><br />
C Z ) for three fin planforms are plotted<br />
in Figure 6. Note the presence of two<br />
distinct <strong>and</strong> comparable peaks corresponding<br />
to the adduction <strong>and</strong> abduction<br />
phases. This force signature<br />
bears the trademark of efficiency where<br />
the fin sustains net forward thrust<br />
throughout its fin beat. Clearly, the<br />
chordwise <strong>and</strong> spanwise compliance<br />
of the fin allows the simultaneous formation<br />
<strong>and</strong> persistence of two distinct<br />
vortex structures within a single fin<br />
beat. A rigid planform would lead to<br />
a more restrictive envelope for the fin<br />
tip path resulting in vortex dynamics<br />
that have stronger interactions detrimental<br />
to sustained net thrust production<br />
(see Akhtar et al., 2007).<br />
We now construct two different<br />
ablated fin models: one that contains<br />
only the rays 1-4 <strong>and</strong> one that contains<br />
rays 1-8 (see Figure 1). The motion of<br />
these dissected fins is precisely the<br />
same as that for the full fin <strong>and</strong>we<br />
carry out flow simulations for both of<br />
these cases. Examining the results from<br />
our virtual dissection, we notice that<br />
the dorsal half of the fin (rays 1-8) captures<br />
the two main peaks of the thrust<br />
<strong>and</strong> preserves 90% of the thrust production<br />
of the full fin planform. Consequently,<br />
the ventral contribution of<br />
the fin, represented by rays 9-14 in<br />
Figure 1, to the thrust production is<br />
found to be insignificant. Also, the<br />
planform interpolated from rays 1-4<br />
has a similar trend in thrust variation<br />
during the entire fin-beat cycle albeit<br />
with smaller amplitudes. Interestingly,<br />
it has two main peaks <strong>and</strong> even the two<br />
local peaks in the abduction phase as in<br />
the full fin case. This further reinforces<br />
the notion that the dorsal leading edge<br />
of the bluegill’s pectoral fin dominates<br />
the overall performance during steady<br />
swimming propulsion. This planform<br />
produces almost 40% of the thrust<br />
produced by the fish fin while undergoing<br />
Mode 1 + 2 + 3 gait. Finally,<br />
we observe similar tendencies for lift<br />
<strong>and</strong> spanwise force coefficients for the<br />
three planforms, except the case with<br />
just rays 1-4 where the values show attenuation.<br />
The key observation here is<br />
that the dorsal half of the pectoral fin<br />
Comparison of time variation of force coefficients for three different fin planforms (rays 1–4, rays 1–8, full planform) at Mode 1+2+3gait:(a)<br />
streamwise force, (b) vertical force, <strong>and</strong> (c) lateral force.<br />
60 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
(rays 1-8) is responsible for producing<br />
a majority of the thrust. These results<br />
bring into question the need for the<br />
ventral portion of the fin. We explore<br />
this in detail as we consider the case of<br />
the yaw turn maneuver.<br />
FIGURE 7<br />
Comparison of time variation of force coefficients: (a) strongside <strong>and</strong> (b) weakside.<br />
Yaw Turn Maneuver<br />
The evolution of wake structure<br />
from the strongside fin, that drives<br />
the maneuver, is shown from two vantage<br />
points: lateral (Figure 8 (a,c,e))<br />
<strong>and</strong>dorsal(Figure8(b,d,f)).The<br />
well-defined vortex ring formed during<br />
the outstroke (abduction) produces a<br />
lateral jet oriented normal to the fish<br />
body (see Figure 9). This type of vortex<br />
ring <strong>and</strong> associated lateral jet shown in<br />
Figures 8 <strong>and</strong> 9 have also been observed<br />
in experimental visualization<br />
(Drucker & Lauder, 2001). The peak<br />
lateral velocity is found to be greater<br />
than three times the freestream velocity.<br />
Consequently, the lateral forces<br />
developed are several times that observed<br />
in forward thrust for the steady<br />
swimming case (Bozkurttas, 2007).<br />
Preliminary estimates for stroke-averaged<br />
force coefficients ratio between<br />
lateral force in maneuvering ( ― C Z =<br />
6.1) to steady swimming thrust ( ― C T =<br />
1.29) is approximately 4 (( ‐ ) denotes<br />
averageoverstroke).Thisfactorisin<br />
reasonable agreement with the forces<br />
measured experimentally (Drucker &<br />
Lauder, 2001).<br />
Returning to Figure 7(a), we note<br />
that the C Z peak is reached between<br />
t/T = 0.15 <strong>and</strong> t/T = 0.3. Shortly thereafter,<br />
the C T peak occurs between<br />
t/T = 0.3 <strong>and</strong> t/T = 0.4. As expected,<br />
the first priority in the maneuver is<br />
to evade the stimulus (an obstacle or<br />
predator in the wild) by quickly generating<br />
a strong lateral force (maximum<br />
occurs at t/T = 0.2). Thereafter, the<br />
drag force developed in the streamwise<br />
direction is likely used to modulate the<br />
direction of the resultant force as the<br />
sunfish turns away from the stimulus.<br />
The evolving vortex ring, clearly seen<br />
in Figures 8(d) <strong>and</strong> 8(f ), continues to<br />
be oriented nearly parallel to the fish<br />
body. Consequently, the lateral jet orientation<br />
ensures that the maximum<br />
lateral force continues to act normal<br />
to the fish body for the duration of<br />
the maneuver. Here, the inherent flexibility<br />
of the pectoral fin structure <strong>and</strong><br />
the ability to continuously alter planform<br />
area is likely to be very useful.<br />
Finally, an examination of the force<br />
histories for the dissected fin reveals<br />
that the peak lateral thrust developed<br />
by the dorsal part (rays 1-8) is approximately<br />
70% of the total as opposed to<br />
30% for the ventral (rays 8-14) portion<br />
(see Figure 10 <strong>and</strong> Figure 11c). The<br />
streamwise drag is slightly more comparable,<br />
although the dorsal part peak is<br />
higher (see Figure 11a). Overall, while<br />
the dorsal portion contributes to the majority<br />
of lateral force production, the ratio<br />
of dorsal to ventral contribution appears<br />
to be more equitable than the steady<br />
swimming case.<br />
Conclusions<br />
A comparative analysis of the pectoral<br />
fin performance in steady swimming<br />
<strong>and</strong> yaw turn maneuver reveals<br />
that the dorsal part of the pectoral fin<br />
is responsible for the majority of force<br />
production. The chordwise <strong>and</strong> spanwise<br />
flexibility of the pectoral fin <strong>and</strong><br />
its ability to have them function either<br />
in concert or independently seems to<br />
enable the bluegill sunfish to achieve<br />
a variety of maneuvers. The virtual<br />
dissection reveals a significant loss of<br />
performance with maneuvering with<br />
respect to peak lateral thrust when<br />
the ventral portion is removed. Thus,<br />
a fin design using just the dorsal portion<br />
of the pectoral fin might perform<br />
as well as the full fin insteadyswimming<br />
but will not retain the same<br />
maneuverability. Hence, any effective<br />
design based on the pectoral fin that<br />
aims to preserve its performance over<br />
all locomotion mode needs to retain<br />
a greater portion of the fin thanthat<br />
suggested by steady swimming alone.<br />
The pectoral fins of fishes display a<br />
diversity of shapes (e.g., Drucker &<br />
Lauder, 2002; Thorsen & Westneat,<br />
2005), <strong>and</strong> although some general<br />
conclusions about correlations of fin<br />
shape with fish ecology have been possible<br />
(see Wainwright et al., 2002),<br />
there are very few data on functional<br />
regionalization of pectoral fins <strong>and</strong> on<br />
therolethatdifferentfin rays within<br />
July/August 2011 Volume 45 Number 4 61
FIGURE 8<br />
Formation of the vortex ring due to the strongside pectoral fin motion: (a), (c), <strong>and</strong> (e) are lateral views at t/T = 0.22, t/T = 0.49, <strong>and</strong> t/T = 0.66,<br />
respectively; (b), (d), <strong>and</strong> (f) are the corresponding dorsal views at t/T = 0.22, t/T = 0.49, <strong>and</strong> t/T = 0.66, respectively.<br />
62 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 9<br />
The strongside lateral jet associated with the vortex structures in Figure 8 (c) at t/T = 0.49.<br />
FIGURE 10<br />
Formation of the vortex ring due to the strongside pectoral fin motion: (a) full, (b) dorsal, <strong>and</strong><br />
(c) ventral portion of the fin sections.<br />
the pectoral fin might play in controlling<br />
locomotor performance. Taft et al.<br />
(2008) discussed functional regionalization<br />
during steady swimming in<br />
sculpin, but the role that different fin<br />
rays play during maneuvering behaviors<br />
has not previously been analyzed.<br />
The results presented here suggest that<br />
the ventral region of the fin playsan<br />
important role in modulating maneuvering<br />
forces, <strong>and</strong> future studies on the<br />
diversity of fish pectoral fin shapes<br />
could focus on the surface area <strong>and</strong><br />
mechanical properties of this region<br />
of the fin incorrelationwithmaneuvering<br />
performance. No data are currently<br />
available that would permit<br />
even general conclusions about the diversification<br />
of pectoral fin structure in<br />
relation to maneuvering capability,<br />
<strong>and</strong> this represents a new <strong>and</strong> very interesting<br />
direction for future work that<br />
integrates approaches from biomechanics<br />
<strong>and</strong> fluid dynamics with behavioral<br />
<strong>and</strong> ecological studies of fish<br />
locomotion.<br />
Acknowledgments<br />
This work was done while the first<br />
three authors were at The George<br />
FIGURE 11<br />
Comparison of forces produced on the dorsal <strong>and</strong> ventral halves of the strongside fin with respect to the full fin: (a) streamwise force, (b) vertical<br />
force, <strong>and</strong> (c) lateral force.<br />
July/August 2011 Volume 45 Number 4 63
Washington University, <strong>and</strong> the work<br />
wassupportedunderONR-MURI<br />
grant N00014-03-1-0897 monitored<br />
by Dr. Thomas McKenna.<br />
Lead Author:<br />
Srinivas Ramakrishnan<br />
ANSYS, Inc.<br />
10 Cavendish Court,<br />
Lebanon, NH 03766<br />
Email: srinivas.ramakrishnan@<br />
ansys.com<br />
References<br />
Akhtar, I., Mittal, R., Lauder, G.V., &<br />
Drucker, E.G. 2007. Hydrodynamics of a<br />
biologically inspired t<strong>and</strong>em flapping foil<br />
configuration. Theor Comp Fluid Dyn.<br />
21:155-70. doi: 10.1007/s00162-007-0045-2.<br />
Bozkurttas, M. 2007. Hydrodynamic performance<br />
of fish pectoral fins with application to<br />
autonomous underwater vehicles. Ph.D. Thesis,<br />
The George Washington University. p. 9.<br />
Bozkurttas, M., Mittal, R., Dong, H.,<br />
Lauder, G.V., & Madden, P. 2009. A lowdimensional<br />
models <strong>and</strong> performance scaling<br />
of a highly deformable fish pectoral fin.<br />
J Fluid Mech. 631:311-42. doi: 10.1017/<br />
S0022112009007046.<br />
Chorin, A.J. 1967. A numerical method<br />
for solving incompressible viscous flow<br />
problems. J Comput Phys. 2:1-9.<br />
doi: 10.1016/0021-9991(67)90037-X.<br />
Dong, H., Bozkurttas, M., Mittal, R., Lauder,<br />
G.V., & Madden, P. 2010. A Computational<br />
modelling <strong>and</strong> analysis of the hydrodynamics<br />
of a highly deformable fish pectoral fin.<br />
J Fluid Mech. 645:345-73. doi: 10.1017/<br />
S0022112009992941.<br />
Dong, H., Mittal, R., & Najjar, F.M. 2006.<br />
Wake topology <strong>and</strong> hydrodynamic performance<br />
of low-aspect-ratio flapping foils.<br />
J Fluid Mech. 566:309-43. doi: 10.1017/<br />
S002211200600190X.<br />
Drucker, E.G., & Lauder, G.V. 2001. Wake<br />
dynamics <strong>and</strong> fluid forces of turning maneuvers<br />
in sunfish. J Exp Biol. 204:431-42.<br />
Drucker, E.G., & Lauder, G.V. 2002. Experimental<br />
hydrodynamics of fish locomotion:<br />
Functional insights from wake visualization.<br />
Integr Comp Biol. 42:243-57. doi: 10.1093/<br />
icb/42.2.243.<br />
Lauder, G.V., Madden, P.G.A., Mittal, R.,<br />
Dong, H., & Bozkurttas, M. 2006. Locomotion<br />
with flexible propulsors: I. Experimental<br />
analysis of pectoral fin swimming in sunfish.<br />
Bioinspir Biomim. 1:35-41. doi: 10.1088/<br />
1748-3182/1/4/S04.<br />
Mittal, R., Dong, H., Bozkurttas, M., &<br />
Lauder, G.V. 2006. Locomotion with flexible<br />
propulsors: II. Computational modeling of<br />
pectoral fin swimming in sunfish. Bioinspir<br />
Biomim. 1:25-34. doi: 10.1088/1748-3182/<br />
1/4/S05.<br />
Mittal, R., Dong, H., Bozkurttas, M., Najjar,<br />
F.M., Vargas, A., & Von Loebbecke, A. 2008.<br />
A versatile sharp interface boundary method<br />
for incompressible flows with complex boundaries.<br />
J Comput Phys. 227:1-9. doi: 10.1016/<br />
j.jcp.2008.01.028.<br />
Mittal, R., & Iaccarino, G. 2005. Immersed<br />
boundary methods. Annu Rev Fluid Mech.<br />
37:239-61. doi: 10.1146/annurev.fluid.37.<br />
061903.175743.<br />
Osher, S., & Sethian, J.A. 1988. Fronts<br />
propagating with curvature-dependent.<br />
J Comput Phys. 79:12-49. doi: 10.1016/<br />
0021-9991(88)90002-2.<br />
Taft, N., Lauder, G.V., & Madden, P.G.<br />
2008. Functional regionalization of the<br />
pectoral fin of the benthic longhorn sculpin<br />
during station holding <strong>and</strong> swimming.<br />
J Zool Lond. 276:159-67. doi: 10.1111/<br />
j.1469-7998.2008.00472.x.<br />
Thorsen, D.H., & Westneat, M. 2005.<br />
Diversity of pectoral fin structure <strong>and</strong> function<br />
in fishes with labriform propulsion. J Morphol.<br />
263:133-50. doi: 10.1002/jmor.10173.<br />
Tran, L.B., & Udaykumar, H.S. 2004. A<br />
particle-level set-based sharp interface cartesian<br />
grid method for impact, penetration, <strong>and</strong><br />
void collapse. J Comput Phys. 193:469-510.<br />
doi: 10.1016/j.jcp.2003.07.023.<br />
Wainwright, P., Bellwood, D.R., & Westneat,<br />
M. 2002. Ecomorphology of locomotion in<br />
labrid fishes. Environ Biol Fish. 65:47-62.<br />
doi: 10.1023/A:1019671131001.<br />
Ye, T., Mittal, R., Udaykumar, H.S., &<br />
Shyy, W. 1999. An accurate Cartesian grid<br />
method for simulation of viscous incompressible<br />
flows with complex immersed boundaries.<br />
J Comput Phys. 156:209-40. doi: 10.1006/<br />
jcph.1999.6356.<br />
Zang, Y., Streett, R.L., & Koseff, J.R. 1994.<br />
A non-staggered fractional step method for<br />
time-dependent incompressible Navier-<br />
Stokes equations in curvilinear coordinates.<br />
J Comput Phys. 114:18-33. doi: 10.1006/<br />
jcph.1994.1146.<br />
64 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
PAPER<br />
Learning From the Fins of Ray-Finned<br />
Fish for the Propulsors of Unmanned<br />
Undersea Vehicles<br />
AUTHORS<br />
James L. Tangorra<br />
Department of Mechanical<br />
Engineering, Drexel University<br />
Timo Gericke<br />
George V. Lauder<br />
Museum of Comparative Zoology,<br />
Harvard University<br />
Introduction<br />
Military <strong>and</strong> civilian studies have<br />
identified that two of the most significant<br />
technological obstacles to deploying<br />
unmanned undersea vehicles<br />
(UUVs) are energy <strong>and</strong> autonomy<br />
(Nicholson & Healey, 2008; Office<br />
of the Secretary of Defense, 2009).<br />
The energy <strong>and</strong> the rate it is used<br />
(power) limit the duration <strong>and</strong> distance<br />
of operations <strong>and</strong> bound the<br />
type of activity that can occur even<br />
for short periods. Autonomy defines<br />
the degree to which humans must supervise<br />
UUV operations <strong>and</strong> provides<br />
UUVs with the ability to react to<br />
external stimuli without human intervention.<br />
Among the enabling technologies<br />
that are critical for solving the<br />
challenges associated with energy <strong>and</strong><br />
autonomyaremoreeffectivepropulsors<br />
(Office of the Secretary Defense,<br />
2009). Propulsors are required to provide<br />
increased maneuverability, stealth,<br />
<strong>and</strong> endurance for the widespread range<br />
of missions envisioned for UUVs, from<br />
long duration sensing in the open ocean<br />
to mine countermeasures in very shallow,<br />
high-energy water.<br />
ABSTRACT<br />
Advanced propulsors are required to help unmanned undersea vehicles (UUVs)<br />
overcome major challenges associated with energy <strong>and</strong> autonomy. The fins of rayfinned<br />
fish provide an excellent model from which to develop propulsors that can<br />
create forces efficiently <strong>and</strong> drive a wide range of behaviors, from hover to lowspeed<br />
maneuvers to high-speed travel. Although much is known about the mechanics<br />
of fins, little is known about the fin’s sensorimotor systems or how fins<br />
are regulated in response to external disturbances. This information is crucial for<br />
implementing propulsive <strong>and</strong> control systems that exploit the same phenomena as<br />
the biological fins for efficiency, effectiveness, <strong>and</strong> autonomous regulation. Experiments<br />
were conducted to evaluate the in vivo response of the sunfish <strong>and</strong> its pectoral<br />
fins to vortex perturbations applied directly to the fish <strong>and</strong> to the fins. The fish<br />
<strong>and</strong> the fins responded actively to perturbations that disturbed the motion of the fish<br />
body. Surprisingly, perturbations that deformed the fins extensively did not cause a<br />
reaction from either the fins or the body. These results indicate that the response of the<br />
pectoral fins to large deformations is not reflexive <strong>and</strong> that fin motions are regulated<br />
when it is necessary to correct for disturbances to the motion of the fish. The results<br />
also demonstrate a benefit of compliance in propulsors, in that external perturbations<br />
can disturb the fins without having its impact be transferred to the fish body.<br />
Keywords: biorobotics, flapping fins, vortex pertubations, sensory-based control<br />
Fish are important biological models<br />
from which to learn methods of<br />
propulsion that are effective <strong>and</strong> efficient<br />
over a wide range of operating<br />
conditions. Bony fish, such as the<br />
bluegill sunfish (Lepomis macrochirus)<br />
<strong>and</strong> the swordfish (Xiphias gladius),<br />
are able to hover, swim <strong>and</strong> maneuver<br />
at low speeds, manipulate the orientation<br />
of their bodies, conduct acrobatics<br />
to escape or to attack prey, <strong>and</strong>, especially<br />
for the swordfish, sustain high<br />
swimming speeds. These behaviors<br />
can be accomplished in smooth water<br />
<strong>and</strong> in high-energy flows <strong>and</strong> relate<br />
directly to the behaviors desired for<br />
UUVs. The remarkable swimming<br />
abilities of these fish are due, in large<br />
part, to the fish having multiple, highly<br />
actuated, flexible fins that are able to<br />
create <strong>and</strong> to modulate large-magnitude<br />
forces.<br />
A great deal is known about the<br />
mechanisms that contribute to the<br />
production of hydrodynamic forces<br />
by flapping the fins <strong>and</strong> the fish<br />
body. Forces are created through the<br />
dynamic interaction of the fins, the<br />
body, <strong>and</strong> the fluid, which results in<br />
energy being added to, or taken from,<br />
the fluid. A review of seminal work<br />
that explains the way in which marine<br />
animals control vorticity is presented<br />
in Triantafyllou et al. (2002) <strong>and</strong><br />
July/August 2011 Volume 45 Number 4 65
Zhu et al. (2002). Numerical <strong>and</strong><br />
experimental studies of flexible fins<br />
with two-dimensional kinematics<br />
(heaving <strong>and</strong> pitching) include, but<br />
are in no way limited to, studies<br />
of McHenry (1995), Liu <strong>and</strong> Bose<br />
(1997), Prempraneerach et al. (2003),<br />
Triantafyllou et al. (2005), Fish et al.<br />
(2006), Lauder et al. (2006), Mittal<br />
et al. (2006), Lauder <strong>and</strong> Madden<br />
(2007), <strong>and</strong> Zhu <strong>and</strong> Shoele (2008).<br />
Recent studies that considered deformable<br />
fins with complex kinematics<br />
are presented in, for example, Shoele<br />
<strong>and</strong> Zhu (2009), Dong et al. (2010),<br />
<strong>and</strong> Tangorra et al. (2010).<br />
In contrast to our underst<strong>and</strong>ing<br />
of the mechanics of fins <strong>and</strong> of hydrodynamic<br />
forces, little is known about<br />
how fishes sense their interaction<br />
with the water <strong>and</strong> use sensory information<br />
to regulate the fins. Knowledge<br />
of fin sensorimotor control is critical<br />
if engineered systems are to take full<br />
advantage of the mechanisms used by<br />
fins to create forces efficiently <strong>and</strong> to<br />
react to changes in the environment.<br />
The focus of this paper will be on<br />
the pectoral fins of sunfish <strong>and</strong>, in particular,<br />
on how <strong>and</strong> when sunfish alter<br />
the use of the pectoral fins in response<br />
to external perturbations. We begin<br />
with an overview of pectoral fin swimming<br />
in sunfish <strong>and</strong> briefly present robotic<br />
fins that produce <strong>and</strong> modulate<br />
forces like the biological fins. A series<br />
of experiments where the biological<br />
fin is perturbed during steady swimming<br />
is then presented. These experiments<br />
address the response of the fins<br />
in the context of using the fins to control<br />
<strong>and</strong> stabilize the fish body.<br />
Ray-Finned Fish <strong>and</strong> Robots<br />
Sunfish Swimming<br />
The ability of the sunfish to control<br />
the magnitude <strong>and</strong> direction of its propulsive<br />
forces is due to its ability to<br />
modulate the kinematics, coordination,<br />
<strong>and</strong> mechanical properties of<br />
its fins <strong>and</strong> muscular tail (Tangorra<br />
et al., 2010, 2011). Hydrodynamic<br />
forces are created through an exchange<br />
of energy between the propulsive surfaces<br />
<strong>and</strong> the surrounding fluid. As the<br />
fish moves through the water, vortices<br />
develop along the body <strong>and</strong> fins, the<br />
propulsive structures bend <strong>and</strong> store<br />
energy, <strong>and</strong> the vortices are shed into<br />
the flow along with directed jets<br />
(Triantafyllou et al., 2002; Dong<br />
et al., 2010). The complex motions<br />
that cause this exchange of energy are<br />
the result of driven motions of the<br />
fin rays <strong>and</strong> a dynamic interaction of<br />
the deformable fin surfaces with the<br />
water. The forces created by fins are,<br />
therefore, modulated through changes<br />
to the kinematics of the fin <strong>and</strong> active<br />
adjustments of fin’s mechanical properties<br />
(Lauder et al., 2006; Mittal et al.,<br />
2006; Akhtar et al., 2007; Tangorra<br />
et al., 2010). The changes may be subtle,<br />
as in steady swimming where the<br />
stiffness of the fin rays is gradually increased<br />
with speed, but where the motions<br />
of the fins are approximately the<br />
same. Or the changes may be obvious,<br />
as when the fish interrupts a cyclic<br />
swimming pattern <strong>and</strong> uses a stiff, impulsive<br />
fin motion to slow the fish <strong>and</strong><br />
turn it away from an obstacle (Gottlieb<br />
et al., 2010).<br />
Ray-Finned Robotic Systems<br />
Robotic fins(Figure1)havebeen<br />
developed that produce motions,<br />
forces, <strong>and</strong> flows like the biological<br />
fins (Tangorra, Davidson et al., 2007;<br />
Phelan, Tangorra et al., 2010; Tangorra,<br />
Lauder et al., 2010). These fins were<br />
designed originally as physical models<br />
with which to conduct experimental<br />
studies that would have been difficult<br />
to conduct with the living fish<br />
FIGURE 1<br />
Biorobotic models of the sunfish pectoral fin<br />
(A) <strong>and</strong> caudal fin (B). The pectoral fin is instrumented<br />
with strain gages along the fin<br />
rays <strong>and</strong> pressure sensors along the body<br />
plate in order to model distributed sensing in<br />
the sunfish. Modified versions of the robotic<br />
pectoral <strong>and</strong> caudal fins,aswellasdorsal<strong>and</strong><br />
anal fins, are implemented on a fish robot (C).<br />
The fish robot can swim freely or be attached<br />
to a rigid mast (shown) so that forces can be<br />
measured. The grooves in the side of the fish<br />
body are used for pressure lines <strong>and</strong> ports.<br />
(Tangorra, Phelan et al., 2011). The<br />
fins are comprised of fin rays, each<br />
with multiple actuated degrees of freedom<br />
(DOF), within a thin, flexible<br />
webbing. The geometries of the fin<br />
rays were defined so that the stiffness<br />
of the robotic fin was proportional to<br />
that of the biological fin across the<br />
fin’s chord <strong>and</strong> span. The architecture<br />
of the robotic fin provides a great degree<br />
of control over the fin’s motions<br />
<strong>and</strong> mechanical properties, which enables<br />
the magnitude <strong>and</strong> direction of<br />
the force produced by the fin tobe<br />
easily modulated (Figure 2). Gross<br />
changes to the profile of the fin’s force<br />
can be made by changing the fin’s gait<br />
66 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 2<br />
Thrust (horizontal) <strong>and</strong> lift (vertical) forces for<br />
pectoral fins executing normal <strong>and</strong> modified<br />
steady swimming gaits. By making relatively<br />
small changes to fin stiffness (A) <strong>and</strong> fin motions<br />
(B), the forces can be moved throughout<br />
the thrust-lift plane. Normal, full-fin steady<br />
swimming for three levels of stiffness (A). The<br />
gait was modified slightly by altering the phase<br />
angle between fin rays by 30° (B, red) or by<br />
using just the upper or lower half of the fin<br />
(B, green). The magnitude of the force can<br />
also be altered simply by changing the frequency<br />
of the fin beat.<br />
pattern, for example, by switching<br />
from a steady swimming gait to the<br />
pattern used by the fish for a turn maneuver.<br />
Smaller changes to the force<br />
profile can be made by changing the<br />
frequency of the fin beat <strong>and</strong>/or by<br />
changing phase relationships between<br />
fin rays (Figure 2A). Considerable<br />
changes to the magnitude <strong>and</strong> direction<br />
of the force can also be made by<br />
adjusting the mechanical properties<br />
of some, or all, of the fin rays. When<br />
the mechanical properties of the fin<br />
rays are under active control, as in<br />
the fish, changes to the force profile<br />
can happen very quickly since the<br />
driven motions of the fin do not have<br />
to be changed.<br />
The designs of the robotic fins were<br />
modified <strong>and</strong> the fins implemented<br />
on a freely swimming biorobotic fish<br />
(Figure 1). Modifications included<br />
placing actuators within the fish body<br />
adjacent to each fin, using a network of<br />
microcontrollers to drive fin motions,<br />
<strong>and</strong> minimizing the number of actuated<br />
DOF for each fin ray.Themotions<br />
<strong>and</strong> orientation of the robotic<br />
fish are controlled by adjusting the<br />
propulsive forces created by five rayfinned<br />
fins. In this first implementation,<br />
the forces are modulated by<br />
switching between several fin gaits<br />
<strong>and</strong> by making predetermined changes<br />
to fin beat frequencies <strong>and</strong> to the phase<br />
relationships between fin rays.<br />
Sensory-Based Control of Fins<br />
What is clearly missing in this robotic<br />
system is the ability to automatically<br />
modulate the kinematics <strong>and</strong><br />
mechanical properties of the fins<br />
based on sensory information about<br />
the fins <strong>and</strong> their interaction with the<br />
water. The motions of the fins are adjusted<br />
based on the forces required to<br />
control the robot’s body, but sensory<br />
information is not being used to exploit<br />
the phenomena that are critical<br />
to the efficient production of force<br />
(e.g., vorticity) nor to adjust behaviors<br />
in response to changes in the flow (e.g.,<br />
speed <strong>and</strong> turbulence). This is due<br />
to the fact that very little is known<br />
about the sensory-based control of<br />
ray-finned fins (Phelan et al., 2010).<br />
The fine level of control that the sunfish<br />
has over fin motions <strong>and</strong> mechanical<br />
properties suggests strongly that<br />
there is closed-loop control of the fins.<br />
However, fundamental questions<br />
about the existence of sensory systems<br />
intrinsic to fins, about the types of<br />
stimuli that elicit responses from fins,<br />
about information in the flow that<br />
is relevant to propulsive forces, <strong>and</strong><br />
about the behavior of the fins in response<br />
to external perturbations have<br />
not yet been answered. This knowledge<br />
is vital for the development of<br />
fin-based propulsors that take advantage<br />
of the phenomena used by fish<br />
to produce forces efficiently <strong>and</strong> that<br />
automatically adjust their behavior in<br />
response to disturbances <strong>and</strong> changes<br />
in operating requirements.<br />
Experimental Methods<br />
<strong>and</strong> Equipment<br />
Experimentation<br />
Experiments were conducted to<br />
evaluate the response of the sunfish’s<br />
pectoral fins to external perturbations<br />
applied to the fin <strong>and</strong>tothefish’s body<br />
during steady swimming. Perturbations<br />
were created using a vortex generator<br />
(Figure 3), which produces a<br />
vortex ring that moves through the<br />
water<strong>and</strong>impartsashortduration<br />
impulse to the fish (Figure 4). The<br />
strength of the vortex was sufficient<br />
to deform the pectoral fin ortodisplace<br />
the fish laterally by several millimeters.<br />
The vortex is not visible, so<br />
it does not elicit a visually mediated response<br />
from the fish. The vortex does,<br />
however, produce a pressure wave that<br />
may be sensed by the fish.<br />
Two bluegill sunfish, with body<br />
lengthsof160±10mm<strong>and</strong>intact<br />
pectoral fins, were used for the experiments.<br />
For the experimental trials,<br />
FIGURE 3<br />
The vortex ring generator <strong>and</strong> vortex (left).<br />
Blue dye was added to the vortex generator’s<br />
cavity to make the vortex visible to the naked<br />
eye. The vortex generator comprises an orifice<br />
plate (1), two cavity plates (2), a latex membrane<br />
(3), <strong>and</strong> a connector plate (4), which<br />
enables the air-line to be connected to the vortex<br />
generator.<br />
July/August 2011 Volume 45 Number 4 67
FIGURE 4<br />
Sunfish in flow tank with vortex generator (A). The sunfish kindly positioned itself in the center<br />
of the test area <strong>and</strong> laser sheet (B). The laser sheet is used with PIV to characterize the vortex as<br />
it travels toward the fish.<br />
asunfish was placed in the working<br />
area (280 × 280 × 800 mm) of a<br />
600-l flow tank <strong>and</strong> was allowed to acclimate<br />
for 2 h. The flow rate was set to<br />
100 mm s −1 , which equates to a steady<br />
swimming speed of approximately<br />
0.6 body lengths s −1 .Atthisspeed,<br />
sunfish generate swimming forces<br />
using primarily their pectoral fins. The<br />
tail <strong>and</strong> the caudal, anal, dorsal, <strong>and</strong><br />
paired pelvic fins are moved very little<br />
but are important for stability. The<br />
vortex generator was positioned approximately<br />
150 mm above the tank<br />
floor <strong>and</strong> placed either perpendicular<br />
to the fish in order to perturb the fish’s<br />
body or at a 45° angle to the fish in<br />
order to perturb the pectoral fin during<br />
its outstroke. A horizontal light sheet<br />
(Figure 4) used for particle image<br />
velocimetry (PIV) was positioned so<br />
that it had the same height as the middle<br />
of the vortex generator. The fish<br />
was directed into the middle of the<br />
test area <strong>and</strong> light sheet by coaxing it<br />
with a wooden dowel. Once the fish<br />
was positioned properly, the vortex<br />
was launched to strike the fish. Vortices<br />
impacted the fish (1) on the body<br />
near the tip of the left pectoral fin<br />
while the fin rested against the body<br />
during the pause between fin beats<br />
<strong>and</strong> (2) at the tip of the left pectoral<br />
fin as the fin completed its outstroke.<br />
High-speed (500 fps), highdefinition<br />
video (1024 × 1024 pixels)<br />
was used to capture the motions of<br />
the fish <strong>and</strong> of the fish’s fins. Two<br />
cameras (Photron 1024 PCI, Photron<br />
USA, Inc., San Diego, CA) were synchronized<br />
<strong>and</strong> positioned so that the<br />
ventral <strong>and</strong> posterior views of the fish<br />
were captured.<br />
Analysis<br />
The linear <strong>and</strong> rotational velocities<br />
of the vortices were analyzed using<br />
DaVis (LaVision GmbH, Göttingen,<br />
Germany).<br />
The motions of the fish <strong>and</strong> of the<br />
fins were analyzed for two fin beats before<br />
<strong>and</strong> two beats after the impact of<br />
the vortex. The coordinates of eight<br />
points along the fish body <strong>and</strong> pectoral<br />
were digitized using Matlab (The<br />
Mathworks Inc., Natick, MA) <strong>and</strong><br />
tracked through time. Deformations<br />
<strong>and</strong> curvatures were calculated for<br />
the pectoral fin during the impact of<br />
thevortexring.Threepointsalongthe<br />
fin were selected to characterize the<br />
shape of the fin <strong>and</strong>todefine the radius<br />
of curvature.<br />
Design of the Vortex<br />
Ring Generator<br />
Vortex rings are commonly generated<br />
using a piston that moves within<br />
a cylindrical cavity <strong>and</strong> pushes a volume<br />
of fluid (the slug) out of the cavity<br />
<strong>and</strong> past an orifice with sharp<br />
edges. The movement of the piston<br />
causes the boundary layer that develops<br />
in the cavity to separate at the orifice’s<br />
edge <strong>and</strong> to roll up into a vortex<br />
ring that has a toroidal shape. The<br />
speed of the piston, the diameter of<br />
the orifice pate, <strong>and</strong> the ratio of cavity<br />
length to cavity diameter influence the<br />
formation of the vortex <strong>and</strong> the speed<br />
at which the vortex travels. Excellent<br />
discussions of vortex generation are<br />
presented in Gharib et al. (1998),<br />
Allen <strong>and</strong> Auvity (2002), Shusser<br />
et al. (2002), <strong>and</strong> Mohseni (2006).<br />
The vortex generator that was developed<br />
for our experiments is similar<br />
to a piston based vortex generator,<br />
but the design was modified so that it<br />
would be more appropriate for the<br />
testing of swimming fish. Two requirements<br />
that influenced the design were<br />
(1) the vortex generator had to be silent,<br />
so that the fish did not hear a<br />
mechanism <strong>and</strong> anticipate the arrival<br />
of the vortex, <strong>and</strong> (2) the system had<br />
to be small, so that it could be placed<br />
at the side of the flow tank without interfering<br />
with the swimming fish. The<br />
vortex generator consists of two acrylic<br />
plates(45×55×12mm)inwhich<br />
a cylindrical cavity is cut (Figure 3).<br />
The plates are covered by a 0.3-mm<br />
thick aluminum plate with either<br />
a 4.0- or 7.5-mm diameter orifice. A<br />
latex membrane is s<strong>and</strong>wiched between<br />
the cavity plates <strong>and</strong> another<br />
acrylic plate in which a cylindrical<br />
well is cut. This plate is connected via<br />
a 6-mm diameter air line (Polyurethane<br />
Tubing, NewWay Air Bearings, Aston,<br />
PA) to a 60-ml syringe (Becton<br />
Dickinson <strong>and</strong> Company, Franklin<br />
Lakes, NJ). A fast push on the syringe<br />
plunger causes the latex membrane to<br />
exp<strong>and</strong> into the cavity <strong>and</strong> to exhaust<br />
the fluid <strong>and</strong> create the vortex. The<br />
68 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
effective length of the cavity can be increased<br />
by drawing the syringe plunger<br />
back. This draws the latex membrane<br />
back into the cylindrical well. Dye<br />
was introduced into the chamber via<br />
a1.6-mmdiameterholedrilledinto<br />
the acrylic plate, radial to the cavity. A<br />
steel tube was inserted into the hole,<br />
<strong>and</strong> was connected via medical tubing<br />
(Scientific Commodities, Inc., Lake<br />
Havasu City, AZ) to a syringe filled<br />
with food-grade dye. The vortex generator<br />
was mounted to an aluminum<br />
arm (80/20 Inc., Columbia City, IN)<br />
so that it could be positioned within<br />
the flow tank.<br />
The force, impulse, <strong>and</strong> linear velocity<br />
of 12 vortices were characterized<br />
to better underst<strong>and</strong> the properties<br />
of the vortex <strong>and</strong> how best to actuate<br />
the plunger. The force generated by<br />
the impact of the vortex was measured<br />
(Figure 5b) by shooting the vortex<br />
against a plate that was connected to<br />
a2.5g force transducer (LSB200, JR<br />
S-Beam Load Cell, Irvine, CA). The<br />
plate was located 100 mm from the orifice<br />
of the vortex generator. The vortex<br />
was imaged using the high-speed<br />
camera as it travelled within the 2-mm<br />
thick light sheet. Mean values for vortices<br />
created using a 5-mm diameter<br />
cavity were: 13 mN force (0.6 mN SE),<br />
0.13 mNs impulse (0.002 mNs SE),<br />
<strong>and</strong> 0.99 m/s velocity (0.01 m/s SE).<br />
A 13-mm diameter cavity produced<br />
a more powerful but slower vortex:<br />
67 mN force (2.3 mN SE), 1.0 mNs<br />
impulse (0.02 mNs SE), <strong>and</strong> 0.85 m/s<br />
velocity (0.01 m/s SE). These values<br />
compare well with estimates we have<br />
made for the peak force <strong>and</strong> impulse<br />
created by a sunfish pectoral fins. At<br />
a swimming speed of 0.5 body length<br />
per second, average fin forces are less<br />
than approximately 10 mN <strong>and</strong> the<br />
impulse over the fin beat is less than<br />
2.5 mNs.<br />
FIGURE 5<br />
Evaluation of vortex ring’s velocity using PIV (A). Force from vortex ring during impact with rigid<br />
plate attached to force transducer (B).<br />
Response to<br />
Vortex Perturbations<br />
Perturbation experiments that<br />
involved hitting the swimming fish<br />
with a vortex ring showed that the<br />
fish did not alter the pectoral fin beat<br />
during the time course of a single fin<br />
stroke but did change the amplitude<br />
<strong>and</strong> timing of the pectoral fin beats<br />
subsequent to a vortex impact that<br />
perturbed the fish’s position.<br />
Response to Vortex Perturbations<br />
Applied to the Body<br />
Vortex perturbations that impacted<br />
the side of the fish displaced the fish<br />
FIGURE 6<br />
laterally by several millimeters (Figure<br />
6), which is significant relative to<br />
the thickness of the fish’s body (maximum<br />
of approximately 25 mm). The<br />
lateral displacement occurred whether<br />
the fish had been drifting toward or<br />
away from the vortex generator prior<br />
to the disturbance <strong>and</strong> was not accompanied<br />
by any obvious change to<br />
the roll or yaw of the fish. An active<br />
response of the fish to the vortex perturbation<br />
was evident in the fishes’<br />
motion after a short delay. The soonest<br />
the active response occurred was<br />
0.05 s, while the longest delay before<br />
a response was evident was 0.20 s. In<br />
the majority of trials, the fishes actively<br />
Distance from orifice plate of the left pectoral fin (purple), the right pectoral fin (blue), <strong>and</strong> the<br />
fish at a point between the pelvic fins (green). The distance between the fish <strong>and</strong> the orifice plate<br />
is amplified relative to the fins <strong>and</strong> is measured at the scale on the right. In this trial, the fish was<br />
moving towards the vortex generator <strong>and</strong> was hit by the vortex at about t = 1.49 (red). The active<br />
response of the fish occurred by t = 1.50 (gray). (Color versions of figures available online at:<br />
http://www.ingentaconnect.com/content/mts/mtsj/2011/00000045/00000004.)<br />
July/August 2011 Volume 45 Number 4 69
moved away from the vortex generator<br />
after being hit by the vortex (Figure 7).<br />
The movement was not particularly<br />
quick, but was always faster than the<br />
fish’s lateral velocity before the perturbation<br />
had occurred. In some cases (e.g.,<br />
Figure 6), the fish actively moved toward<br />
the vortex generator after being<br />
pushed away from the vortex generator<br />
by the impulse. This occurred only<br />
when the fish had been drifting towards<br />
the vortex generator before the perturbation.<br />
In some trials, the fish was startled<br />
by the vortex <strong>and</strong> swam out of the test<br />
area. The startled motions were not<br />
analyzed quantitatively.<br />
The motions of the pectoral fins<br />
during the fin beat subsequent to the<br />
perturbation were significantly differentfromthemotionsofthepectoral<br />
fins prior to the perturbation. However,<br />
the pectoral fins did not seem to<br />
react quickly to the stimulus. In fact,<br />
the initial movement of the fish’s body<br />
in response to the vortex generally<br />
occurred between pectoral fin beats,<br />
while the pectoral fins were against<br />
the fish body (Figures 6 <strong>and</strong> 7).<br />
Thus, the active motion of the fish<br />
was initiated by other fins, which reacted<br />
within as little as 0.05 s. Active<br />
FIGURE 7<br />
Distance from orifice plate of the left pectoral fin (purple), the right pectoral fin (blue), <strong>and</strong> the<br />
fish at a point between the pelvic fins (green). The distance between the fish <strong>and</strong> the orifice plate<br />
is amplified relative to the fins <strong>and</strong> is measured at the scale on the right. In this trial, the fish was<br />
moving away from the vortex generator <strong>and</strong> was hit by the vortex at about t = 1.25 (red). The<br />
active response of the fish occurred by t = 1.35 (gray). The fish continued to move away from<br />
the vortex generator until approximate t = 1.8 s.<br />
movement of the pectoral fins did<br />
not usually resume until 0.10-0.20 s<br />
after the vortex. The frequency of the<br />
pectoral fin beats did not change<br />
consistently after the perturbation. In<br />
three of the eight trials, the frequency<br />
of the pectoral fin beat increased from,<br />
on average, 1.37 Hz (SD = 0.15) to<br />
1.83 Hz (SD = 0.28). In the other<br />
five trials, the frequency of the fin beat<br />
decreased from, on average, 1.69 Hz<br />
(SD = 0.26) to 1.34 Hz (SD = 0.21).<br />
The amplitude of the pectoral fin motions<br />
also changed. This altered the<br />
force balance between the two pectoral<br />
fins <strong>and</strong> contributed to the movement<br />
of the fish body. In the beat after the<br />
vortex stimulus, the amplitude of the<br />
right pectoral fin (opposite the side of<br />
the impact) was consistently smaller<br />
than before the vortex. Its motion decreased<br />
in all eight trials, on average by<br />
18.9% (SD = 10.9%). The amplitude<br />
of the left pectoral fin alsochanged,<br />
but the changes were less consistent.<br />
In four trials, the amplitude decreased<br />
by, on average, 37.4% (SD = 22.8),<br />
while in the other four trials, the<br />
amplitude increased by, on average,<br />
8.9% (SD 7.0%). By the second fin<br />
beat after the perturbation, the motions<br />
of the left <strong>and</strong> right pectoral<br />
fins were much more similar to the<br />
motions before the fin beat, <strong>and</strong> were<br />
similar to each other.<br />
Response to Vortex Perturbations<br />
Applied to the Fin<br />
The pectoral fins were deformed<br />
significantly when struck by the vortex<br />
during the fin beat (Figures 8 <strong>and</strong> 9).<br />
The vortex made contact with the left<br />
pectoral fin near the end of the fin’s<br />
outstroke. The vortex bent the tips<br />
of the fin rays <strong>and</strong> progressively bent<br />
larger portions of the fin as the vortex<br />
travelled towards the fish body. The fin<br />
seemed to bend <strong>and</strong> fold as if it were<br />
made from thin paper <strong>and</strong> exhibited<br />
deformations from the tip to the<br />
base. The maximum measured curvature<br />
of the fin (alongfin ray6)increased<br />
from 0.054 mm −1 near the tip<br />
<strong>and</strong> 0.024 mm −1 near the base during<br />
unperturbed swimming to 0.113 mm −1<br />
near the tip <strong>and</strong> 0.029 mm −1 near<br />
the base when in contact with the vortex.<br />
The vortex remained in contact<br />
with the fin while it travelled towards<br />
the fish body. This resulted in the<br />
pectoral fin being pushed back to the<br />
body faster than during an unperturbed<br />
instroke. Times ranged from one third<br />
to one half of the duration of a normal<br />
instroke <strong>and</strong> were dependent on many<br />
FIGURE 8<br />
Pectoral fin perturbed by vortex during swimming.<br />
The mean fin ray curvature after impact<br />
was 14.4 mm −1 (0.05 mm SE). Reflective<br />
particles are used so that the fluid movement<br />
is visible.<br />
70 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 9<br />
Ventral view of the fish as the left pectoral fin is hit by a vortex (no dye). The left pectoral fin<br />
is hit by the vortex (1). The fin is deformed (2, 3, <strong>and</strong> 4) <strong>and</strong> is pushed to the body by the vortex.<br />
(5 <strong>and</strong> 6) The right pectoral fin continues to beat normally. The body is not deflected by the vortex.<br />
variables, including the speed of the<br />
vortex, how well contact was made<br />
with the fin, <strong>and</strong> the time of impact<br />
within the fin beat.<br />
Despite the severity with which the<br />
vortex changed the shape <strong>and</strong> trajectory<br />
of the perturbed fin, the fish did<br />
not appear to react to the perturbation<br />
or to change its behavior subsequent to<br />
the perturbation. During the perturbation,<br />
the observed motions of the<br />
unperturbed fin <strong>and</strong>ofthefish body<br />
were not visibly different from motions<br />
prior to the perturbation. Subsequent<br />
to the perturbation, the perturbed fin<br />
remained against the fish body until<br />
the unperturbed fin completed its<br />
instroke. Both fins then resumed<br />
what appeared to be a normal fin<br />
beat. Small differences in the pectoral<br />
fin beat <strong>and</strong> the use of other fins likely<br />
occurred to accommodate for differences<br />
in propulsive forces produced<br />
during the perturbation, but these<br />
changes were not visible. Nor were<br />
there changes in the motion of the<br />
fish body, which was not observed to<br />
move laterally or to rotate in yaw.<br />
Discussion<br />
Theobjectiveoftheexperiments<br />
was to determine how sunfish respond<br />
to perturbations applied to the body<br />
<strong>and</strong> fins during steady swimming.<br />
These experiments provided a contextual<br />
underst<strong>and</strong>ing of sensory based<br />
modulation of pectoral fin function.<br />
The experiments produced a mix of<br />
expected <strong>and</strong> surprising results.<br />
As expected, the fishes did alter the<br />
amplitude <strong>and</strong> timing of the pectoral<br />
fin beats subsequent to a perturbation<br />
that disturbed the lateral position of<br />
the fish body. However, the pectoral<br />
fins did not respond quickly to the<br />
disturbance, but remained against the<br />
fish body for durations that were only<br />
slightly different from the pauses<br />
between fin beats prior to the disturbance.<br />
Active movement of the fish’s<br />
body after the disturbance occurred<br />
with a latency of as little as 0.05 s,<br />
which is similar to the 0.08 s latency<br />
measured by Webb (2004) in response<br />
to roll disturbances. The movement of<br />
the fish body is believed to have been<br />
caused by fins other than the pectoral<br />
fins, since the pectoral fins remained<br />
against the body for 0.10–0.20 s after<br />
the perturbation. When the pectoral<br />
fins were moved, the amplitudes of<br />
the fins seemed to have been adjusted<br />
to help equilibrate the movement of<br />
the fish. By the second fin beatafter<br />
the disturbance, the motions of the<br />
two pectoral fins were synchronized<br />
<strong>and</strong> had amplitudes similar to those<br />
before the disturbance.<br />
The delay in the response of the<br />
pectoral fins to the vortex <strong>and</strong> lateral<br />
disturbance is different from the response<br />
of the fins during experiments<br />
where an obstacle was placed in front<br />
of the swimming fish (Gottlieb et al.,<br />
2010). In those experiments, sunfish<br />
altered the motions of the left<br />
<strong>and</strong> right pectoral fins during the outstroke<br />
of a steady swimming beat.<br />
The changes were not subtle, <strong>and</strong> the<br />
fish did not seem to wait for the next<br />
cycle as in the present studies. The pectoral<br />
fin onthesideoftheobstacle<br />
stiffened <strong>and</strong> the fin rays were moved<br />
through trajectories that were very different<br />
from steady swimming. The fin<br />
on the side opposite to the obstacle<br />
nearly stopped <strong>and</strong> served to stabilize<br />
the motion of the fish. The difference<br />
in the pectoral fins’ response to the<br />
obstacle <strong>and</strong> to the vortex <strong>and</strong> lateral<br />
displacement may be related to the<br />
fish’s perception of the stimuli. The<br />
obstacle may have been more threatening<br />
than the vortex, which the fish may<br />
have interpreted as a common fluidic<br />
event. The fish therefore disrupted<br />
the steady swimming gait in order to<br />
produce large lateral forces that turned<br />
the fish away from an unknown obstacle<br />
that may have posed a threat.<br />
In contrast, the disturbance in motions<br />
caused by a fluidic stimulus could be<br />
accommodated simply by adjusting<br />
motions of the fins within their normal<br />
gaits. This would allow the central<br />
pattern generator that drives the motions<br />
of pectoral fins (Westneat et al.,<br />
2004) to continue to produce similar<br />
output characteristic rather than having<br />
to switch between gaits.<br />
Most surprising was the lack of reaction<br />
to the vortex when the vortex<br />
deformed the pectoral fin attheend<br />
July/August 2011 Volume 45 Number 4 71
of the fin’s outstroke <strong>and</strong> throughout<br />
the instroke. Nerves <strong>and</strong> free nerve<br />
endings exist throughout the fin rays<br />
<strong>and</strong> the fin webbing (experimental<br />
findings, M. Hale, University of<br />
Chicago), <strong>and</strong> so it was expected that<br />
at least one of the phenomena that<br />
the vortex created—pressure, impact,<br />
bending—would have elicited a sensory<br />
mediated response. The vortex<br />
was in contact with the fin for over<br />
100ms,<strong>and</strong>sothedurationofthe<br />
stimulus was certainly sufficient for a<br />
sensory-mediated response to occur.<br />
It was also surprising that neither the<br />
motions of the body, nor subsequent<br />
beats of the pectoral fins, were clearly<br />
different from those before the vortex<br />
perturbation. It is highly likely that the<br />
left pectoral fin, while being deformed,<br />
produced forces that were different<br />
from normal. During a normal steady<br />
swimming gait, each pectoral fin will<br />
produce lateral forces that are similar<br />
in magnitude to thrust <strong>and</strong> lift. Since<br />
the fins typically beat synchronously,<br />
the lateral forces from the left <strong>and</strong><br />
right fins balance <strong>and</strong> cancel. This<br />
would not have been the case when<br />
the left pectoral fin was deformed,<br />
<strong>and</strong> the unbalanced forces should<br />
have accelerated the fish body laterally<br />
<strong>and</strong>/or in roll <strong>and</strong> yaw. The lack of obvious<br />
lateral motion <strong>and</strong> adjustment to<br />
the pectoral fin beat may be due simply<br />
to the fish being insensitive to lateral<br />
forces. To move the fish laterally,<br />
forces must accelerate the mass of the<br />
fish <strong>and</strong> also overcome drag forces<br />
<strong>and</strong> the load from the mass of water<br />
against which the side of the fish<br />
pushes. Thus, the loss of lateral force<br />
during a single fin beat can be easily<br />
tolerated because it is difficult for the<br />
fish to move sideways. So although<br />
studies of biorobotic models of the<br />
pectoral fins have shown that the<br />
fin’s kinematics <strong>and</strong> mechanical properties<br />
must be controlled very carefully<br />
to produce forces like the fish<br />
(Tangorra et al., 2007, 2010), the<br />
mechanics of the fish body do not<br />
necessarily require the careful control<br />
of forces at all times in all directions.<br />
Conclusions<br />
Perturbation experiments which<br />
involved hitting the swimming sunfish<br />
with a vortex ring showed that the<br />
fish did not alter the pectoral fin beat<br />
during the time course of a single fin<br />
stroke but did change the amplitude<br />
<strong>and</strong> timing of its motions in beats subsequent<br />
to an impact that disturbed<br />
the fish’s position. Vortices that struck<br />
the pectoral fin during the fin’s outstroke<br />
deformed the fin extensively,<br />
but the perturbations did not cause<br />
the stroke of the unaffected pectoral<br />
fin to change, nor did the perturbation<br />
cause changes in the motions of the<br />
fish body or in the subsequent strokes<br />
of either pectoral fin.<br />
These outcomes suggest that the<br />
kinematics of the pectoral fins is modulated<br />
by sensory information only<br />
when a perturbation results in a disturbance<br />
to the fish body, which is the system<br />
that the fins are working to control.<br />
The pectoral fins did not react quickly<br />
when the vortex displaced the fish’s<br />
body but modulated their motions to<br />
help stabilize the displaced fish after<br />
other fins had already been engaged.<br />
The pectoral fins also did not react reflexively<br />
to vortex perturbations that<br />
deformed the fins’ webbing <strong>and</strong> fin<br />
rays. The fin did not appear to move<br />
away from the vortex or to resist the deformation<br />
by stiffening. The compliant<br />
fin allowed itself to bend <strong>and</strong> perhaps to<br />
shed the load from the vortex, <strong>and</strong> then<br />
altered its motions during the course of<br />
the subsequent fin beat.<br />
Theresultsillustrateabenefit of<br />
compliant mechanisms within a highly<br />
controllable system. The fins of rayfinned<br />
fish have the ability to control<br />
forces precisely by altering the kinematics<br />
<strong>and</strong> mechanical properties of<br />
individual fin rays. Small changes in<br />
either the trajectories or stiffness of<br />
fin rays can significantly alter the<br />
force that is transferred to the fish<br />
(Tangorra et al., 2010). However, it<br />
is not always necessary to regulate the<br />
fins precisely. By maintaining its flexibility<br />
<strong>and</strong> allowing itself to be deformed,<br />
the fin was able to be hit by<br />
the vortex—which had sufficient<br />
forces to displace the fish—without<br />
transferring the full impact of the perturbation<br />
to the fish body. Therefore,<br />
the fin’s passive mechanics made it unnecessary<br />
for the fins to be modulated<br />
in order to restore the fish to equilibrium.<br />
However, when the fish does want to<br />
move the quickly—as in a maneuver<br />
away from the obstacle—the pectoral<br />
fin can be stiffened, the gait changed,<br />
<strong>and</strong> large lateral forces from the fin<br />
can be transferred to the fish body.<br />
Acknowledgments<br />
This work was supported by the Office<br />
of Naval Research grant N00014-<br />
0910352 on fin neuromechanics monitored<br />
by Dr. Thomas McKenna <strong>and</strong><br />
by the National Science Foundation<br />
EFRI 0938043. We are very grateful<br />
to the members of the Lauder <strong>and</strong> the<br />
Tangorra laboratories for many helpful<br />
discussions about fish fins <strong>and</strong> robotic<br />
fins <strong>and</strong> for assistance in executing<br />
experiments <strong>and</strong> analyzing results.<br />
Lead Author:<br />
James Tangorra<br />
3141 Chestnut St.,<br />
R<strong>and</strong>ell 115<br />
Drexel University,<br />
Philadelphia, PA 19104<br />
Email: tangorra@coe.drexel.edu<br />
72 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
References<br />
Akhtar, I., Mittal, R., Lauder, G., & Drucker,<br />
E. 2007. Hydrodynamics of a biologically<br />
inspired t<strong>and</strong>em flapping foil configuration.<br />
Theoret Comput Fluid Dynamics. 21(3):<br />
155-70. doi: 10.1007/s00162-007-0045-2.<br />
Allen, J., & Auvity, B. 2002. Interaction of a<br />
vortex ring with a piston vortex. J Fluid Mech.<br />
465(1):353-78.<br />
Dong, H., Bozkurttas, M., Mittal, R.,<br />
Madden, P., & Lauder, G. 2010. Computational<br />
modelling <strong>and</strong> analysis of the hydrodynamics<br />
of a highly deformable fish pectoral<br />
fin. J Fluid Mech. 645:345-73. doi: 10.1017/<br />
S0022112009992941.<br />
Fish, F.E., Nusbaum, M.K., Beneski, J., &<br />
Ketten, D. 2006. Passive cambering <strong>and</strong><br />
flexible propulsors: cetacean flukes. Bioinspir<br />
Biomim. 1(4):S42-48. doi: 10.1088/<br />
1748-3182/1/4/S06.<br />
Gharib, M., Rambod, E., & Sharriff, K.<br />
1998. A universal time scale for vortex ring<br />
formation. J Fluid Mech. 360(1):121-40.<br />
doi: 10.1017/S0022112097008410.<br />
Gottlieb, J., Tangorra, J., Esposito, C., &<br />
Lauder, G. 2010. A biologically derived<br />
pectoral fin for yaw turn manoeuvres.<br />
Appl Bionics Biomech. 7(1):41-55.<br />
doi: 10.1080/11762320903537782.<br />
Lauder, G.V., & Madden, P.G. 2007. Fish<br />
locomotion: kinematics <strong>and</strong> hydrodynamics<br />
of flexible foil-like fins. Exp Fluids. 43(5):<br />
641-53. doi: 10.1007/s00348-007-0357-4.<br />
Lauder, G.V., Madden, P.G., Mittal, R.,<br />
Dong, H., & Bozkurttas, M. 2006. Locomotion<br />
with flexible propulsors: I. Experimental<br />
analysis of pectoral fin swimming in sunfish.<br />
Bioinspir Biomim. 1:S25. doi: 10.1088/<br />
1748-3182/1/4/S04.<br />
Liu, P., & Bose, N. 1997. Propulsive performance<br />
from oscillating propulsors with<br />
spanwise flexibility. P Roy Soc Lond A Mat.<br />
453(1963):1763.<br />
McHenry, M. 1995. Mechanical control of<br />
swimming speed: stiffness <strong>and</strong> axial wave<br />
form in undulating fish models. J Exp Biol.<br />
198(11):2293-305.<br />
Mittal, R., Dong, H., Bozkurttas, M., Lauder,<br />
G., & Madden, P.G. 2006. Locomotion<br />
with flexible propulsors: II. Computational<br />
modeling of pectoral fin swimming in<br />
sunfish. Bioinspir Biomim. 1(4):S35-41.<br />
doi: 10.1088/1748-3182/1/4/S05.<br />
Mohseni, K. 2006. Pulsatile vortex generators<br />
for low-speed maneuvering of small underwater<br />
vehicles. Ocean Eng. 33(16):2209-23.<br />
doi: 10.1016/j.oceaneng.2005.10.022.<br />
Nicholson, J. W., & Healey, A.J. 2008. The<br />
present state of autonomous underwater<br />
vehicle (AUV) applications <strong>and</strong> technologies.<br />
Mar Technol Soc. 42(1):8.<br />
Office of the Secretary of Defense, USA.2009.<br />
Unmanned Systems Roadmap (2009-2034).<br />
Phelan, C.T., Tangorra, J.L., Lauder, G.V.,<br />
& Hale, M.E. 2010. A biorobotic model of<br />
the sunfish pectoral fin for investigations of fin<br />
sensorimotor control. Bioinspir Biomim. 5.<br />
doi: 10.1088/1748-3182/5/3/035003.<br />
Prempraneerach, P., Hover, F.S., &<br />
Triantafyllou, M.S. 2003. The effect of<br />
chordwise flexibility on the thrust <strong>and</strong><br />
efficiency of a flapping foil. Autonomous<br />
Undersea Systems Institute. In: 13th Int.<br />
Symp. Unmanned Untethered Submersible<br />
Techn. Durham, NH.<br />
Shoele, K., & Zhu, Q. 2009. Fluid–structure<br />
interactions of skeleton-reinforced fins: performance<br />
analysis of a paired fin in lift-based<br />
propulsion. J Exp Biol. 212(16):2679.<br />
doi: 10.1242/jeb.030023.<br />
Shusser, M., Gharib, M., Rosenfeld, M., &<br />
Mohseni, K. 2002. On the effect of pipe<br />
boundary layer growth on the formation of<br />
a laminar vortex ring generated by a piston/<br />
cylinder arrangement. Theoret Comput Fluid<br />
Dynamics. 15(5):303-16. doi: 10.1007/<br />
s001620100051.<br />
Tangorra, J., Lauder, G.V., Hunter, I.W.,<br />
Mittal, R., Madden, P.G., & Bozkurttas, M.<br />
2010. The effect of fin ray flexural rigidity on<br />
the propulsive forces generated by a biorbotic<br />
fish pectoral fin. J Exp Biol. 213:4043-54.<br />
doi: 10.1242/jeb.048017.<br />
Tangorra, J., Phelan, C.T., Esposito, C.J., &<br />
Lauder, G.V. 2011. Use of biorobotic models<br />
of highly deformable fins for studying the<br />
mechanics <strong>and</strong> control of fin forces in fishes.<br />
Integr Comp Biol. in press. doi: 10.1093/<br />
icb/icr036.<br />
Tangorra, J.L., Davidson, S.N., Hunter, I.W.,<br />
Madden, P.G., Lauder, G.V., Dong, H., …<br />
Mittal, R. 2007. The development of a biologically<br />
inspired propulsor for unmanned<br />
underwater vehicles. IEEE J Oceanic Eng.<br />
32(3):533-50. doi: 10.1109/JOE.2007.903362.<br />
Triantafyllou, M., Techet, A., Zhu, Q.,<br />
Beal, D., Hover, F., & Yue, D. 2002.<br />
Vorticity control in fish-like propulsion <strong>and</strong><br />
maneuvering. Integr Comp Biol. 42(5):1026.<br />
doi: 10.1093/icb/42.5.1026.<br />
Triantafyllou, M.S., Hover, F.S., Techet, A.,<br />
& Yue, D. 2005. Review of hydrodynamic<br />
scaling laws in aquatic locomotion <strong>and</strong> fishlike<br />
swimming. Appl Mech Rev. 58(4):226-37.<br />
doi: 10.1115/1.1943433.<br />
Webb, P. 2004. Response latencies to postural<br />
disturbances in three species of teleostean<br />
fishes. J Exp Biol. 207(6):955-61. doi: 10.1242/<br />
jeb.00854.<br />
Westneat, M., Thorsen, D., Walker, J., &<br />
Hale, M. 2004. Structure, function, <strong>and</strong><br />
neural control of pectoral fins in fishes. IEEE J<br />
Oceanic Eng. 29(3):674-83. doi: 10.1109/<br />
JOE.2004.833207.<br />
Zhu, Q., & Shoele, K. 2008. Propulsion<br />
performance of a skeleton-strengthened fin.<br />
J Exp Biol. 211(13):2087. doi: 10.1242/<br />
jeb.016279.<br />
Zhu, Q., Wolfgang, M.J., Yue, D.K.P., &<br />
Triantafyllou, M.S. 2002. Three-dimensional<br />
flow structures <strong>and</strong> vorticity control in fishlike<br />
swimming. J Fluid Mech. 468:1-28.<br />
doi: 10.1017/S002211200200143X.<br />
July/August 2011 Volume 45 Number 4 73
PAPER<br />
Bioinspired Design Process for an Underwater<br />
Flying <strong>and</strong> Hovering Vehicle<br />
AUTHORS<br />
Jason D. Geder<br />
John S. Palmisano<br />
Ravi Ramamurti<br />
Marius Pruessner<br />
Banahalli Ratna<br />
Naval Research Laboratory<br />
William C. S<strong>and</strong>berg<br />
Science Applications<br />
International Corporation<br />
Background<br />
Biologists <strong>and</strong> zoologists have<br />
been studying fish swimming for<br />
many decades, <strong>and</strong> several comprehensive<br />
texts <strong>and</strong> papers exist (Breder,<br />
1926; Lindsey, 1978; Alex<strong>and</strong>er,<br />
1983; Azuma, 1992; Blake, 1983;<br />
Webb, 1975, 1984; Videler, 1993;<br />
Sfakiotakis et al., 1999). The experimental<br />
studies of fish swimming biodynamics<br />
have become increasingly<br />
quantitative as measurement technology<br />
has improved. The use of highspeed<br />
photography shed light upon<br />
the details of fin deformation (Gibb<br />
et al., 1994; Walker & Westneat,<br />
1997). Laser light scattering techniques<br />
enabled observations of not<br />
only the fish body <strong>and</strong> fin dynamics<br />
but also the velocity field about<br />
the fish <strong>and</strong> in the wake (Drucker &<br />
Lauder, 1999, 2002; Gharib et al.,<br />
2002; Bartol et al., 2003). Vehicle designers<br />
are able to draw upon such rich<br />
data sets as they embark upon designs.<br />
But where does one begin<br />
A bioinspired vehicle design should<br />
begin, according to Webb (2004), by<br />
specifying the performance goals of<br />
ABSTRACT<br />
We review here the results obtained during the past several years in a series of<br />
computational <strong>and</strong> experimental investigations aimed at underst<strong>and</strong>ing the origin of<br />
high-force production in the flapping wings of insects <strong>and</strong> the flapping <strong>and</strong> deforming<br />
fins of fish <strong>and</strong> the incorporation of that information into bioinspired vehicle<br />
designs. We summarize the results obtained on pectoral fin force production, flapping<br />
<strong>and</strong> deforming fin design, <strong>and</strong> the emulation of fish pectoral fin swimming in<br />
unmanned vehicles. In particular, we discuss the main results from the computational<br />
investigations of pectoral fin force production for a particular coral reef fish,<br />
the bird wrasse (Gomphosus varius), whose impressive underwater flight <strong>and</strong> hovering<br />
performance matches our vehicle mission requirements. We describe the<br />
tradeoffs made between performance <strong>and</strong> produceability during the bio-inspired<br />
design of an actively controlled curvature pectoral fin <strong>and</strong> the incorporation of it<br />
into two underwater flight vehicles: a two-fin swimming version <strong>and</strong> four-fin swimming<br />
version. We describe the unique computational approach taken throughout<br />
the fin <strong>and</strong> vehicle design process for relating fin deformation time-histories to<br />
specified desired vehicle dynamic behaviors. We describe the development of the<br />
vehicle controller, including hardware implementation, using actuation of the multiple<br />
deforming flapping fins as the only means of propulsion <strong>and</strong> control. Finally,<br />
we review the comparisons made to date between four-fin vehicle experimental<br />
trajectory measurements <strong>and</strong> controller simulation predictions <strong>and</strong> discuss the<br />
incorporation of those comparisons into the controller design.<br />
Keywords: bio-inspired robotics, pectoral fin, unmanned systems, computational<br />
fluid dynamics<br />
the desired mission first <strong>and</strong> then examining<br />
those living creatures whose<br />
performance is relevant. These creatures<br />
have evolved to meet all their<br />
needs, <strong>and</strong> the maneuvering enabled<br />
by these needs may intersect with<br />
the performance requirements driving<br />
a vehicle design. However, since the<br />
living creatures selected for study are<br />
most likely not optimized for the mobility<br />
characteristics that are driving<br />
the design, one should not copy nature<br />
but instead be guided by it.<br />
This point of caution to designers<br />
has been made many times by biologists<br />
(Combes & Daniel, 2001;<br />
Wainwright et al., 2002; Collar et al.,<br />
2008).<br />
The mission selected for the Naval<br />
Research Laboratory (NRL) swimming<br />
vehicle, which we describe<br />
below, requires precise low-speed maneuvering<br />
<strong>and</strong> excellent hovering in a<br />
complex near-shore environment in<br />
addition to excellent position-keeping<br />
in tidal currents. Cost <strong>and</strong> mechanical<br />
simplicity constraints dem<strong>and</strong>ed a<br />
rigid hull. This eliminated undulatory<br />
swimming fish as a primary means of<br />
design inspiration, <strong>and</strong> instead we<br />
were lead to consider fin-based swimming<br />
creatures. Kato <strong>and</strong> Furushima<br />
74 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
(1996) had already proceeded down<br />
the path of paired fin swimming, drawing<br />
inspiration from the black bass to<br />
design rigid pectoral fins <strong>and</strong> incorporate<br />
them into a test-bed vehicle.<br />
He subsequently pursued low-speed<br />
maneuvering using rigid pectoral fins<br />
(Kato, 2000) <strong>and</strong> then went on to develop<br />
a passive flexible fin <strong>and</strong> an active<br />
pneumatic actuator pectoral fin (Kato<br />
et al., 2008). Barrett <strong>and</strong> Triantafyllo<br />
(1995) on the other h<strong>and</strong> had taken<br />
their inspiration from the undulating<br />
body <strong>and</strong> oscillating caudal fin ofthe<br />
tuna to design <strong>and</strong> build their flexible<br />
“Robotuna.” We looked for a fish<br />
which possessed the dynamic performance<br />
characteristics we needed <strong>and</strong><br />
for which experimental measurements<br />
of swimming dynamics, including fin<br />
kinematics, already existed. Experimental<br />
observations of pectoral fin<br />
muscle activity, kinematics, <strong>and</strong> dynamics<br />
in coral reef wrasses (Westneat<br />
& Walker, 1997; Walker & Westneat,<br />
1997) have shown that the body is<br />
essentially held rigid during straightline<br />
motion, thus satisfying our rigid<br />
hull constraint. Very rapid (∼10 body<br />
lengths per second) translational<br />
motions were observed (Walker &<br />
Westneat, 2000) to give comparable<br />
swimming performance to that seen<br />
in body-caudal fin swimmers of comparable<br />
size even though there was no<br />
contribution from body undulation<br />
or caudal fin oscillation in the wrasses.<br />
This solution offered the potential of<br />
fast forward swimming (or positionkeeping<br />
in strong currents) while<br />
avoiding the complexity of a flexible<br />
hull. In addition, Walker <strong>and</strong><br />
Westneat (1997) had observed high<br />
maneuverability by these swimmers<br />
in their complex reef habitats. These<br />
habitats are reasonable representations<br />
of our vehicle’s projected operating<br />
environment, hence we saw the possibility<br />
of obtaining both high forward<br />
speed <strong>and</strong> excellent low-speed maneuvering<br />
<strong>and</strong> hovering performance as<br />
well.<br />
The fish we selected to inspire our<br />
design was one of the coral reef pectoral<br />
fin swimmers, the bird wrasse<br />
(Gomphosus varius), shown in Figure 1.<br />
FIGURE 1<br />
Bird wrasse (Gomphosus varius).<br />
Controlled Deformation<br />
Fin <strong>Technology</strong><br />
Development<br />
Living creatures, such as insects,<br />
birds, <strong>and</strong> pectoral fin swimmers,<br />
generate lift <strong>and</strong> thrust by executing<br />
large-amplitude wing/fin flapping,<br />
often with substantial shape deformation<br />
from root to tip <strong>and</strong> leading edge<br />
to trailing edge. The flow for these<br />
motions is three-dimensional <strong>and</strong> unsteady,<br />
<strong>and</strong> conventional steady-state<br />
aerodynamics is unable to correctly<br />
compute the corresponding time history<br />
of flapping-force generation.<br />
Comprehensive reports on the research<br />
carried out to study the fluid dynamics<br />
of flapping fins <strong>and</strong> wings<br />
(Rozhdestvensky & Ryzhov, 2003)<br />
<strong>and</strong> of biomimetic fins for underwater<br />
vehicles (Triantafyllo et al., 2004)<br />
exist, in addition to an extensive number<br />
of studies, too great to list here, on<br />
unsteady lift production by flapping<br />
insect wings. Three-dimensional unsteady<br />
computations are necessary to<br />
correctly predict the lift <strong>and</strong> thrust variation<br />
throughout the flapping stroke<br />
cycle. Such computations, for creaturesorvehicleswithmoving<strong>and</strong><br />
deforming surfaces, provide the timevarying<br />
pressure distribution on all<br />
surfaces, which in turn can provide insights<br />
into how the flapping forces <strong>and</strong><br />
maneuvering moments are being generated.<br />
This information can be coupled<br />
with computational visualization<br />
of the time-varying flow about the<br />
fish to analyze the origin of body <strong>and</strong><br />
fin vorticity, its growth, <strong>and</strong> eventual<br />
shedding into the wake. This is the<br />
approach we developed for tuna caudal<br />
fin force production analysis<br />
(Ramamurti et al., 1996, 1999), subsequently<br />
validated against wrasse experimental<br />
data (Ramamurti et al.,<br />
2002), <strong>and</strong> which we also followed<br />
throughout our fin <strong>and</strong> vehicle development<br />
efforts described below.<br />
There are 13 multiply bifurcating<br />
fin rays in the bird wrasse pectoral fin<br />
shown in Figure 2, each contributing<br />
to the fin curvature time-variation<br />
throughout the stroke cycle. For ease<br />
of design, manufacture, actuation,<br />
<strong>and</strong> control, it is ideal to have the fewest<br />
possible number of rays (which we<br />
refer to as ribs) <strong>and</strong> have each of them<br />
FIGURE 2<br />
Bird wrasse fin structure (from Walker &<br />
Westneat, 1997).<br />
July/August 2011 Volume 45 Number 4 75
e as simple in shape <strong>and</strong> structure<br />
as possible. But for more effective fin<br />
propulsion, it is ideal to maximize the<br />
number of ribs since more control<br />
points result in a smoother fit to<br />
desired fin curvature time-histories.<br />
Our computational investigations<br />
(Ramamurti et al., 2004) have shown<br />
that the loss of force magnitude over<br />
the stroke cycle is very small if we<br />
considerably reduce the number of<br />
ribs, as long as we maintain the ability<br />
to properly modify the surface curvature.<br />
The computations showed that<br />
when the fin was made rigid by specifying<br />
the motion with just the leading<br />
edge of the fin tip, the thrust produced<br />
during the upstroke was less than half<br />
of the peak thrust produced by the<br />
flexible fin computations. During the<br />
downstroke, the computations for<br />
the rigid <strong>and</strong> nearly rigid fin produced<br />
no positive thrust, while the<br />
partially <strong>and</strong> fully flexible cases produced<br />
substantial thrust. In the case<br />
of the rigid fin, there was also a substantial<br />
penalty in lift during the<br />
FIGURE 3<br />
Effect of fin flexibility on the time variation of thrust forces (from Ramamurti et al., 2004).<br />
(Color versions of figures available online at: http://www.ingentaconnect.com/content/mts/<br />
mtsj/2011/00000045/00000004.)<br />
FIGURE 4<br />
Representative rib cross-sectional geometry<br />
<strong>and</strong> bending analysis showing a 20° rib<br />
deflection.<br />
upstroke. An example from these<br />
computational investigations is shown<br />
below in Figure 3.<br />
Assessment of these findings led us<br />
to reduce the number of ribs from 13<br />
to 5. Five was selected since it enabled<br />
reduced fin complexity, thus substantially<br />
reducing fin size <strong>and</strong> weight,<br />
while maintaining the critically important<br />
flexural capability. Each of the five<br />
fin rays were individually designed<br />
<strong>and</strong> constructed from compliant ABS<br />
plastic material using a 3-D printer to<br />
achieve the desired tip deflection with<br />
an achievable linear actuation force applied<br />
to each individual rib at the root<br />
(Trease et al., 2003), as illustrated in<br />
Figure 4. The pushing <strong>and</strong> pulling of<br />
the ribs at the root of the fin is similar<br />
to how fish bend their ribs using muscle<br />
actuation.<br />
The root section of the fin was selected<br />
to be of rectangular cross section<br />
with rounded leading edges <strong>and</strong> a tapered<br />
trailing edge in order to accommodate<br />
the actuators at the base of the<br />
ribs as shown in Figure 5a. A translucent<br />
silicone rubber membrane skin,<br />
optimized in thickness to approximately<br />
0.5 mm via finite-element analysis<br />
(Palmisano et al., 2007), provided<br />
the continuous surface covering for<br />
the rays as shown in Figure 5b.<br />
After several parametric 3-D unsteady<br />
computational fluid dynamics<br />
(CFD) studies had been carried out<br />
(Ramamurti & S<strong>and</strong>berg, 2006), various<br />
fin parameters were chosen that<br />
optimized thrust performance, given<br />
the mechanical constraints of the fin.<br />
An example of these computations<br />
showing the variation of lift <strong>and</strong><br />
thrust as a function of fin flexibility,<br />
stroke amplitude, <strong>and</strong> fin strokebias<br />
is shown in Figure 6.<br />
The angle of attack of the root section<br />
of the fin was chosen to be 20°, the<br />
amplitude of the oscillation to be 114°,<br />
the flapping frequency to be 1 Hz, <strong>and</strong><br />
the rib spacing to be the minimum<br />
possible value of 1.2 cm dictated by<br />
the size of the actuators. These specific<br />
values, selected for maximizing fin<br />
forceproduction,areforasolitary<br />
flapping <strong>and</strong> deforming fin. Incorporation<br />
of the fin into a vehicle design<br />
presents challenges to retain fin performance<br />
while maintaining vehicle<br />
simplicity.<br />
Two-Fin Test-Bed Vehicle<br />
A simple vehicle was designed <strong>and</strong><br />
built to serve as a test-bed for evaluating<br />
the performance of all aspects of<br />
76 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 5<br />
Mechanical fin (a) CAD image without skin <strong>and</strong> (b) actual with skin.<br />
the controlled curvature fin technology,<br />
including its capability for vehicle<br />
propulsion, low-speed maneuvering,<br />
<strong>and</strong> hovering. It was, therefore, intentionally,<br />
a minimalist design that<br />
served to house the fins, actuators,<br />
<strong>and</strong> a battery. We described above<br />
how the parameters that govern the<br />
force production by the fin were chosen.<br />
However, due to mechanical<br />
constraints <strong>and</strong> a desire to easily<br />
manufacture a vehicle prototype, additional<br />
3-D unsteady CFD studies of<br />
the flapping fins incorporated in the<br />
FIGURE 6<br />
test-bed vehicle led us to modify the<br />
fin-alone values. The angle of attack<br />
of the root section of the fin was chosentobe0°,<strong>and</strong>theribspacingwas<br />
reduced to 0.8 cm. It is during construction<br />
tradeoffs of this type that<br />
one relies upon what has been learned<br />
from the studies of nature to meet operational<br />
performance goals while balancing<br />
that with the desire to create a<br />
vehicle that is producible at a reasonable<br />
cost. Fixing the fin root at a specific<br />
angle deviates from nature, but<br />
the construction simplicity <strong>and</strong> cost<br />
Mean fin generated (a) thrust <strong>and</strong> (b) lift as functions of non-dimensionalized stroke amplitude<br />
<strong>and</strong> fin flex, or curvature. The surfaces indicate a bias in the fin stroke angle of 0° (red), 20°<br />
(green), <strong>and</strong> 40° (blue).<br />
savings are so substantial that some<br />
performance penalty was accepted.<br />
The flapping stroke amplitude <strong>and</strong><br />
frequency, as well as the phasing of<br />
the individual fin tipdeflection time<br />
histories, were retained as controllable<br />
parameters, <strong>and</strong> we have performed<br />
CFD analyses of force production<br />
with this finconfiguration (Ramamurti<br />
et al., 2010).<br />
In keeping with the review nature<br />
of this paper, we are emphasizing the<br />
process carried out for our specific<br />
bio-inspired vehicle design. The details<br />
of the controlled curvature fin design,<br />
the linear actuator design <strong>and</strong> construction,<br />
the isolated fin construction<br />
<strong>and</strong> testing, the vehicle design, the vehicle<br />
construction, the experimental<br />
testing, <strong>and</strong> the validation of computations<br />
were previously reported<br />
(Palmisano et al., 2007, 2008; S<strong>and</strong>berg<br />
& Ramamurti, 2008). The fin technology<br />
test-bed demonstration vehicle<br />
incorporated two actively controlled deformation<br />
fins. The two-fin vehicle is<br />
shown in Figure 7.<br />
Four-Fin Test-Bed Vehicle<br />
The test results for the two-fin vehicle<br />
demonstrated that the controlled<br />
deformation fin force production was<br />
capable of meeting our vehicle propulsion<br />
(position-keeping in a current)<br />
requirements (Geder et al., 2008).<br />
However, by design, this test-bed fin<br />
technology demonstration vehicle<br />
had restricted options for sensor payload<br />
<strong>and</strong> did not have the fore-aft<br />
force production capability needed<br />
for heave-pitch control. Hence, the<br />
design of a larger 41-cm long four-fin<br />
vehicle was initiated (Figure 8). The<br />
fore-aft symmetry of the four-fin design<br />
enables hover <strong>and</strong> higher precision<br />
positioning capabilities by decoupling<br />
vehicle pitch <strong>and</strong> heave control.<br />
July/August 2011 Volume 45 Number 4 77
FIGURE 7<br />
Two-fin technology test-bed demonstration vehicle: (a) exterior view <strong>and</strong> (b) interior view.<br />
The current four-fin vehicle design shown here employs a water-tight cylinder<br />
for housing the power source <strong>and</strong> electronics with a flooded space in the nose <strong>and</strong><br />
tail for buoyancy trimming <strong>and</strong> supplemental sensors. Hardware control <strong>and</strong> all<br />
computations are performed by a 16-MHz ATmega2560 microcontroller. Computations<br />
(Ramamurti et al., 2010) <strong>and</strong> experimental tests carried out to date<br />
(Geder et al., 2011) characterized how changes in fin stroke amplitude, frequency,<br />
bias angle, <strong>and</strong> curvature affect the thrust <strong>and</strong> lift forces for zero free stream flow<br />
speed. Further testing <strong>and</strong> computations are ongoing to fully characterize the fin<br />
forces <strong>and</strong> vehicle dynamics. However, current models have the necessary fidelity<br />
to accurately predict vehicle performance as outlined in the following sections.<br />
Four-Fin Vehicle Control<br />
With the vehicle state variables defined as in Figure 8, the vehicle dynamics can<br />
be written as,<br />
Mv →̇ þ Cv ð<br />
→ Þv → þ Dv ð →<br />
Þv → þ g →<br />
ðη →<br />
Þ ¼ τ → ;<br />
where M is a matrix of rigid body mass <strong>and</strong> inertial terms, C is a matrix of centripetal<br />
<strong>and</strong> Coriolis terms, D is a matrix of hydrodynamic lift <strong>and</strong> drag terms, g is a<br />
vector of hydrostatic terms, v =[uvwpqr] T , η =[xyzϕθψ] T is the position <strong>and</strong><br />
orientation vector in the earth-fixed frame where ϕ, θ, <strong>and</strong> ψ are roll, yaw, <strong>and</strong><br />
pitch angles, <strong>and</strong> τ is a vector of all forces <strong>and</strong> moments external to the rigid body.<br />
The portion of the vector, τ, that is effected by the fins is represented as,<br />
→<br />
τfins ¼<br />
2<br />
6<br />
4<br />
3<br />
f T ;LF þ f T ;LB þ f T ;RF þ f T ;RB<br />
0<br />
f L;LF f L;LB f L;RF f L;RB<br />
<br />
<br />
y L f L;LF þ f L;LB y R f L;RF þ f L;RB ; ð2Þ<br />
7<br />
x F f L;LF þ f L;RF þ xB f L;LB þ f L;RB<br />
5<br />
<br />
f T ;LF þ f T ;LB f T ;RF þ f T ;RB<br />
y L<br />
y R<br />
ð1Þ<br />
where f T is fin thrust <strong>and</strong> f L is fin lift.<br />
Subscripts ‘LF’, ‘LB’, ‘RF’, <strong>and</strong>‘RB’<br />
identify the left front, left back, right<br />
front, <strong>and</strong> right back fins, respectively.<br />
The x-position of the center of pressure<br />
on the fins is denoted by x F for the<br />
front fins <strong>and</strong> x B for the back fins.<br />
The y-position of the center of pressure<br />
on the fins is denoted by y L for the left<br />
fins <strong>and</strong> y R for the right fins.<br />
Mathematical models representing<br />
thedynamicperformanceofthefins<br />
have been developed to include the effects<br />
on force production of inflow velocities<br />
to the leading edge (or trailing<br />
edge for reverse motion) <strong>and</strong> to include<br />
the effects of fin interactions with each<br />
other, namely the effects of the trailing<br />
vortices off the front fins on the inflow<br />
to the back fins (Geder et al., 2011).<br />
Other fin dynamic representations<br />
modeled controllable parameters including<br />
fin curvature <strong>and</strong> stroke amplitude(Ramamurtietal.,2010).In<br />
the earlier 3-D unsteady CFD studies,<br />
these two key parameters were found<br />
to have a direct relationship with thrust<br />
generation—increasing stroke amplitude<br />
or fin curvature increased thrust<br />
(Ramamurti & S<strong>and</strong>berg, 2006).<br />
However, since both stroke amplitude<br />
<strong>and</strong> flapping frequency are limited mechanically<br />
in the vehicle, optimal combinations<br />
of amplitude <strong>and</strong> frequency<br />
were experimentally found for high<br />
thrust <strong>and</strong> lift fin gaits. The best mix<br />
of these parameters for our vehicle<br />
was determined to be 100° for stroke<br />
amplitude <strong>and</strong> 1.8 Hz flapping frequency,<br />
which yielded not only high<br />
force output but also relatively low<br />
power consumption (Palmisano et al.,<br />
2007). These findings allowed us to fix<br />
stroke amplitude <strong>and</strong> frequency as<br />
constants <strong>and</strong> to focus on fin curvature<br />
as the primary thrust control parameter.<br />
Further, biasing the fin stroke up<br />
or down, as in Figure 9, affects fin lift<br />
78 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 8<br />
Four-fin technology test-bed vehicle, (a) exterior view, (b) interior view.<br />
generation while maintaining constant<br />
thrust (Geder et al., 2008).<br />
In previous work we evaluated the<br />
benefits of two vehicle control methods<br />
(Geder et al., 2008). The first<br />
method, called weighted gait combination,<br />
used combinations of thrustgenerating<br />
<strong>and</strong> lift-generating fin<br />
gaits to produce vectored propulsive<br />
forces. The second method, called<br />
mean bulk angle bias (MBAB), used<br />
weighted forward-reverse gait control<br />
with stroke bias angle control. Between<br />
these two control methods,<br />
our results showed that MBAB better<br />
decoupled control over body-fixed<br />
thrust <strong>and</strong> lift forces <strong>and</strong> yielded better<br />
vehicle response characteristics in simulation.<br />
As such, the MBAB method<br />
is used to control the four-fin vehicle.<br />
The vehicle controller comm<strong>and</strong>s<br />
changes to the fins to effect changes<br />
in the forces <strong>and</strong> moments imparted<br />
on the vehicle, as shown in equation 2.<br />
These fin comm<strong>and</strong>sarebasedon<br />
errors in the vehicle dynamic states,<br />
computed as the comm<strong>and</strong>ed values<br />
minus the computed values. States<br />
are computed onboard the vehicle<br />
using a suite of sensors (three axes<br />
of accelerometers, three axes of rate<br />
gyros, magnetic compass, <strong>and</strong> pressure<br />
sensor) <strong>and</strong> sensor fusion <strong>and</strong> filtering<br />
schemes (Geder et al., 2009). Errors in<br />
surge motion (x-axis translation) dictate<br />
comm<strong>and</strong>s for fin thrust changes<br />
to all fins. Errors in heave motion<br />
(z-axis translation) dictate comm<strong>and</strong>s<br />
for fin lift changes to all fins. Errors<br />
in roll motion (x-axis rotation) dictate<br />
comm<strong>and</strong>s for differential lift changes<br />
between left <strong>and</strong> right fins. Errors in<br />
FIGURE 10<br />
Four-fin vehicle operating in a Naval Research Laboratory test facility.<br />
pitch motion (y-axis rotation) dictate<br />
comm<strong>and</strong>s for differential lift changes<br />
in forward <strong>and</strong> back fins. Errors in yaw<br />
motion (z-axis rotation) dictate comm<strong>and</strong>s<br />
for differential thrust changes<br />
in left <strong>and</strong> right fins. The vehicle has<br />
no direct control over sway motion<br />
( y-axis translation), <strong>and</strong> instead comm<strong>and</strong>s<br />
yaw motion changes to move<br />
in this direction. The direction of<br />
yaw motion depends on the sway<br />
error. The output of a proportionalintegral-derivative<br />
(PID) controller<br />
for each vehicle state is used to determine<br />
forward <strong>and</strong> reverse fin gait percentage<br />
<strong>and</strong> bulk angle comm<strong>and</strong>s.<br />
Four-Fin Vehicle<br />
Performance<br />
Initial measurements of the dynamic<br />
performance of the four-fin vehicle<br />
have been conducted in two test<br />
facilities, one a 6 × 2.5 × 2 foot water<br />
tank <strong>and</strong> the other a 50-foot diameter<br />
by 50-foot deep water tank (Figure 10).<br />
The experimental measurements have<br />
served to validate the vehicle dynamic<br />
model <strong>and</strong> to begin the assessment of<br />
vehicle performance.<br />
FIGURE 9<br />
Vehicle images showing all four fins with<br />
strokes (a) biased down to produce positive<br />
lift <strong>and</strong> (b) biased up to produce negative lift.<br />
July/August 2011 Volume 45 Number 4 79
FIGURE 11<br />
FIGURE 12<br />
Comparison of experimental <strong>and</strong> simulated open-loop heading angle responses (from Geder et al.,<br />
2011).<br />
Comparison of experimental <strong>and</strong> simulated closed-loop heading angle responses with (a) proportional<br />
control <strong>and</strong> (b) PID control (from Geder et al., 2011). The solid curve represents the actual<br />
simulated heading response of the vehicle based on the modeled dynamics. The dashed curve<br />
represents the actual experimental heading response observed in vehicle testing. The square<br />
data represent the measured heading response of the vehicle based on the output of sensor models.<br />
The circle data represent the measured experimental vehicle heading from the output of onboard<br />
sensors.<br />
An open loop test was conducted to<br />
characterize vehicle heading angle response<br />
<strong>and</strong> to compare model simulation<br />
performance with experimental<br />
performance (Geder et al., 2011).<br />
The right finsweresetatfullreversekinematics,<br />
<strong>and</strong> the left fins were set to<br />
closely match the opposite thrust of<br />
the right fins. At t =11s,thegaitweighting<br />
inputs were reversed. The simulated<br />
<strong>and</strong> experimental responses show very<br />
good agreement (Figure 11) with a maximum<br />
difference between the two responses<br />
at any given time of 5°. Both<br />
simulated <strong>and</strong> experimental results exhibit<br />
a 30°/s maximum turning rate,<br />
4 s time from zero to maximum speed,<br />
<strong>and</strong> braking angle of 35°—the amount<br />
of residual turning distance after fin<br />
kinematics are reversed.<br />
After validating the four-fin vehicle<br />
dynamic model in yaw motion <strong>and</strong> implementing<br />
state feedback control,<br />
initial closed-loop experiments were<br />
done to test heading angle control<br />
(Geder et al., 2011). In Figure 12, a<br />
comparison of experimental <strong>and</strong> simulated<br />
results is given. For a simple proportional<br />
control algorithm (Figure 12a),<br />
we see the results match well with a<br />
30-40° amplitude <strong>and</strong> 6.5-s period. Differences<br />
between measured experimental<br />
heading from onboard sensors<br />
<strong>and</strong> simulated heading can be attributed<br />
to the noise <strong>and</strong> sensitivity characteristics<br />
of the sensors (Geder et al.,<br />
2009). External magnetic disturbances<br />
in our test facility cause errors up to<br />
10° in heading measurements, also factoring<br />
into the variation between experimental<br />
<strong>and</strong> simulated heading<br />
responses, as our calibration of the onboard<br />
compass was not perfect. Adding<br />
in a derivative gain to damp the<br />
heading response <strong>and</strong> a small integral<br />
gain to eliminate any steady-state error,<br />
we see the response to a 180° step<br />
comm<strong>and</strong> in heading in Figure 12b.<br />
With PID control over heading angle,<br />
the response is nearly critically damped<br />
with a rise time of 7 s. We also see close<br />
agreement between measured <strong>and</strong> actual<br />
angles, <strong>and</strong> again differences in<br />
the responses can be attributed to sensor<br />
noise <strong>and</strong> sensitivity, as well as compass<br />
calibration errors.<br />
80 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
Conclusions<br />
We have reviewed the history<br />
of our bioinspired vehicle design process.<br />
The process began by collaborating<br />
with biologists in order to<br />
underst<strong>and</strong> the dynamics of flapping<br />
<strong>and</strong> deforming fins in pectoral fin<br />
swimmers. The detailed kinematics<br />
they measured in fish swimming experiments<br />
provided the time-varying<br />
fin surface curvature data necessary<br />
for computing the 3-D unsteady flow<br />
about the swimming fish. Examination<br />
of the computed flow variations<br />
about the flapping <strong>and</strong> deforming<br />
fins <strong>and</strong> the fish body provided insights<br />
into the relationship between the fin<br />
flows <strong>and</strong> the fin force time-histories<br />
throughout the stroke cycle. Parametric<br />
variations of key stroke parameters in<br />
3-D unsteady flow computations<br />
yielded the sensitivity of the timevarying<br />
fin forces, which in turn provided<br />
the information needed to modify<br />
our fin designfromthatofthebird<br />
wrasse. It is during such computations<br />
that insights from nature can be<br />
blended with design constraints to<br />
yield a range of possible bio-inspired<br />
designs. A controlled curvature fin utilizing<br />
individually designed fin ribs,<br />
each actuated at the rib root by a linear<br />
actuator was built <strong>and</strong> tested. The successful<br />
fin tests were followed by the<br />
design <strong>and</strong> construction of a two-fin<br />
test-bed vehicle to demonstrate the<br />
mobility potential of the concept. Incorporation<br />
of the fins into a vehicle<br />
necessitated further compromises<br />
where we balanced biomimetic performance<br />
with produceability <strong>and</strong> cost.<br />
These test-bed vehicle tests were followed<br />
by the design <strong>and</strong> construction<br />
of a more capable four-fin vehicle, incorporating<br />
the same fins. A series of<br />
controllers were developed <strong>and</strong> built<br />
to enable assessment of vehicle propulsion<br />
<strong>and</strong> maneuvering performance as<br />
a function of the varying kinematics of<br />
each fin. Deforming fin force time histories,<br />
including fin-fin unsteady interactions,<br />
have been incorporated into<br />
the vehicle dynamic model used for<br />
controller development. The preliminary<br />
four-fin vehicle dynamic performance<br />
measurements indicate very<br />
good agreement with computed performance.<br />
These initial results are<br />
very encouraging <strong>and</strong> indicate that<br />
our efforts to emulate the positionkeeping,<br />
low-speed maneuvering, <strong>and</strong><br />
hovering performance of the bird<br />
wrasse into a producible <strong>and</strong> low cost<br />
vehicle are bearing fruit.<br />
Lead Author:<br />
Jason D. Geder<br />
Laboratory for Computational<br />
Physics <strong>and</strong> Fluid Dynamics<br />
Naval Research Laboratory<br />
Overlook Avenue, SW,<br />
Washington, DC<br />
Email: jgeder@lcp.nrl.navy.mil<br />
References<br />
Alex<strong>and</strong>er, R.M. 1983. The history of fish<br />
biomechanics. In: Fish Biomechanics, eds.<br />
Webb, P.W., & Weihs, D., 1-35.<br />
New York: Praeger.<br />
Azuma, A. 1992. The Biokinetics of Flying<br />
<strong>and</strong> Swimming. Tokyo: Springer-Verlag.<br />
265 pp.<br />
Barrett, D.S., & Triantafyllo, M.S. 1995.<br />
The design of a flexible hull undersea vehicle<br />
propelled by an oscillating foil. In: Proc.<br />
9th Int. Symp. on Unmanned Untethered<br />
Submersible <strong>Technology</strong>. Durham, NH:<br />
Autonomous Undersea Systems Institute<br />
(AUSI).<br />
Bartol, I.K., Gharib, M., Weihs, D., Webb,<br />
P.W., Hove, J.R., & Gordon, M.S. 2003.<br />
Hydrodynamic stability of swimming in<br />
ostraciid fishes: role of the carapace in the<br />
smooth trunkfish Lactophrys triqueter<br />
(Teleostei: Ostraciidae). J Exp Biol. 206:<br />
725-44. doi: 10.1242/jeb.00137.<br />
Blake, R.W. 1983. Fish Locomotion.<br />
Cambridge, UK: Cambridge University<br />
Press. 228 pp.<br />
Breder, C.M. 1926. The locomotion of fishes.<br />
Zoologica. 4:159-297.<br />
Collar, D., Wainwright, P., & Alfaro, M.<br />
2008. Integrated diversification of locomotion<br />
<strong>and</strong> feeding in labrid fishes. Biol Lett. 4:84-6.<br />
doi: 10.1098/rsbl.2007.0509.<br />
Combes, S., & Daniel, T. 2001. Shape,<br />
flapping, <strong>and</strong> flexion: wing <strong>and</strong> fin design<br />
for forward flight. J Exp Biol. 204:2073-85.<br />
Drucker, E.G., & Lauder, G.V. 1999.<br />
Locomotor forces on a swimming fish:<br />
Three-dimensional vortex wake dynamics<br />
quantified using digital particle image<br />
velocimetry. J Exp Biol. 203:2393-412.<br />
Drucker, E.G., & Lauder, G.V. 2002.<br />
Experimental hydrodynamics of fish<br />
locomotion: Functional insights from wake<br />
visualization. Integr Comp Biol. 42:243-57.<br />
doi: 10.1093/icb/42.2.243.<br />
Geder, J.D., Palmisano, J., Ramamurti, R.,<br />
S<strong>and</strong>berg, W.C., & Ratna, B. 2008. Fuzzy<br />
logic PID-based control design <strong>and</strong> performance<br />
for pectoral fin-propelled unmanned<br />
underwater vehicle. In: Proc. Int. Conf on<br />
Control, Automation <strong>and</strong> Systems. Seoul,<br />
Korea: IEEE. doi: 10.1109/ICCAS.2008.<br />
4694526.<br />
Geder, J.D., Ramamurti, R., Palmisano, J.,<br />
Pruessner, M., Ratna, B., & S<strong>and</strong>berg, W.C.<br />
2009. Sensor data fusion <strong>and</strong> submerged<br />
test results of a pectoral fin propelled UUV.<br />
In: Proc. 16th Intl Symp on Unmanned<br />
Untethered Submersible <strong>Technology</strong>. Durham,<br />
NH: Autonomous Undersea Systems Institute<br />
(AUSI).<br />
Geder, J.D., Ramamurti, R., Palmisano, J.,<br />
Pruessner, M., S<strong>and</strong>berg, W.C., & Ratna, B.<br />
2011. Four-fin bio-inspired UUV: modeling<br />
July/August 2011 Volume 45 Number 4 81
<strong>and</strong> control solutions. ASME Intl Mech Eng<br />
Congress Exposition. IMECE. 2011-64005.<br />
Gharib, M., Pereira, F., Dabiri, D., Hove,<br />
J.R., & Modarress, D. 2002. Quantitative<br />
flow visualization: Toward a comprehensive<br />
flow diagnostic tool. Integr Comp Biol.<br />
42:964-70. doi: 10.1093/icb/42.5.964.<br />
Gibb, A.C., Jayne, B.C., & Lauder, G.V.<br />
1994. Kinematics of pectoral fin locomotion<br />
in the bluegill sunfish Lepomis Macrochirus.<br />
J Exp Biol. 189:133-61.<br />
Kato, N. 2000. Control performance in<br />
the horizontal plane of a fish robot with<br />
mechanical pectoral fins. IEEE J Oceanic Eng.<br />
25:125-9. doi: 10.1109/48.820744.<br />
Kato, N., Ando, Y., Tomokazu, A., Suzuki, H.,<br />
Suzumori, K., K<strong>and</strong>a, T., & Endo, S. 2008.<br />
Elastic pectoral fin actuators for biomimetic<br />
underwater vehicles. In: Bio-mechanisms<br />
of Swimming <strong>and</strong> Flying, eds. Kato, N.,<br />
Kamimura, S., 271-82. Tokyo: Springer.<br />
Kato, N., & Furushima, M. 1996. Pectoral<br />
fin model for maneuver of underwater vehicles.<br />
In: Proc. Symp. on Autonomous<br />
Underwater Vehicle <strong>Technology</strong>. Monterrey,<br />
CA: IEEE. doi: 10.1109/AUV.1996.532400.<br />
Lindsey, C.C. 1978. Form, function, <strong>and</strong><br />
locomotory habits in fish. In: Fish Physiology<br />
Vol. II, eds. Hoar, W.S., & R<strong>and</strong>all, D.J.,<br />
1-100. New York: Academic Press.<br />
Palmisano, J., Geder, J.G., Ramamurti, R.,<br />
Liu, K.J., Cohen, J.J., Mengesha, T., …<br />
Ratna, B. 2008. Design, development <strong>and</strong><br />
testing of flapping fin with actively-controlled<br />
curvature for an unmanned underwater vehicle.<br />
In: Bio-mechanisms of Swimming <strong>and</strong><br />
Flying, eds. Kato, N., & Kamimura, S.,<br />
283-94. Tokyo: Springer. doi: 10.1007/<br />
978-4-431-73380-5_23.<br />
Palmisano, J., Ramamurti, R., Liu, K.J.,<br />
Cohen, J., S<strong>and</strong>berg, W.C., & Ratna, B.<br />
2007. Design of a bio-mimetic controlledcurvature<br />
robotic pectoral fin. In: Proc.<br />
IEEE Intl. Conf. on Robotics <strong>and</strong> Automation.<br />
Rome, Italy: IEEE. doi: 10.1109/ROBOT.<br />
2007.363110.<br />
Ramamurti, R., Geder, J.D., Palmisano, J.,<br />
Ratna, B., & S<strong>and</strong>berg, W.C. 2010. Computations<br />
of flapping flow propulsion for unmanned<br />
underwater vehicle design. AIAA J.<br />
48(1):188-201. doi: 10.2514/1.43389.<br />
Ramamurti, R., Lohner, R., & S<strong>and</strong>berg,<br />
W.C. 1996. Computation of the unsteady<br />
flow past a tuna with caudal fin oscillation.<br />
In: Advances in Fluid Mechanics, eds. Rahman,<br />
M.,&Brebbia,C.,vol.9,169-78.Southampton,<br />
UK: Computational Mechanics Publishing.<br />
Ramamurti, R., Lohner, R., & S<strong>and</strong>berg,<br />
W.C. 1999. Computation of the 3-D<br />
unsteady flow past deforming geometries.<br />
Int J Comput Fluid D. 13:83-99.<br />
doi: 10.1080/10618569908940891.<br />
Ramamurti, R., & S<strong>and</strong>berg, W.C. 2006.<br />
Computational fluid dynamics study for<br />
optimization of a fin design. In: Proc. 24th<br />
AIAA Applied Aerodynamics Conference.<br />
AIAA. AIAA-2006-3658.<br />
Ramamurti, R., S<strong>and</strong>berg, W.C., & Lohner, R.<br />
2004. The influence of fin rigidity <strong>and</strong><br />
gusts on the force production in fish <strong>and</strong><br />
insects: a computational study. In: Proc. 42nd<br />
AIAA Aerospace Sciences Meeting <strong>and</strong> Exhibit.<br />
Am Inst Aeronautics <strong>and</strong> Astronautics.<br />
AIAA-2004-0404.<br />
Ramamurti, R., S<strong>and</strong>berg, W.C., Lohner, R.,<br />
Walker, J.A., & Westneat, M.W. 2002. Fluid<br />
dynamics of flapping aquatic flight in the<br />
bird wrasse: 3-D unsteady computations with<br />
fin deformation. J Exp Biol. 205:2997-3008.<br />
Rozhdestvensky, K.V., & Ryzhov, V.A.<br />
2003. Aerohydrodynamics of flapping-wing<br />
propulsors. Prog Aerosp Sci. 39:585-633.<br />
doi: 10.1016/S0376-0421(03)00077-0.<br />
S<strong>and</strong>berg, W.C., & Ramamurti, R. 2008.<br />
3-D Unsteady computations of flapping flight<br />
in insects, fish, <strong>and</strong> unmanned vehicles.<br />
In: Bio-mechanisms of Swimming <strong>and</strong><br />
Flying, eds. Kato, N., & Kamimura, S.,<br />
205-217. Tokyo: Springer. doi: 10.1007/<br />
978-4-431-73380-5_17.<br />
Sfakiotakis, M., Lane, D.M., & Davies, J.B.C.<br />
1999. Review of fish swimming modes for<br />
aquatic locomotion. J Exp Biol. 24:241-52.<br />
Trease, B.P., Liu, K.J., & Kota, S. 2003.<br />
Biomimetic compliant system for actuatordriven<br />
aquatic propulsor: preliminary results.<br />
In: ASME Intl. Mech. Eng. Congress <strong>and</strong><br />
Exposition. IMECE 2003-41446.<br />
Triantafyllo, M.S., Techet, A.H., & Hover,<br />
F.S. 2004. Review of experimental work in<br />
biomimetic foils. IEEE J Oceanic Eng.<br />
29:585-94. doi: 10.1109/JOE.2004.833216.<br />
Videler, J.J. 1993. Fish Swimming. London,<br />
UK: Chapman <strong>and</strong> Hall. 260 pp.<br />
Wainwright, P., Bellwood, D., & Westneat,<br />
M.W. 2002. Ecomorphology of locomotion<br />
in labrid fishes. Environ Biol Fish. 65:47-62.<br />
doi: 10.1023/A:1019671131001.<br />
Walker, J.A., & Westneat, M.W. 1997.<br />
Labriform propulsion in fishes: Kinematics<br />
of flapping aquatic flight in the bird wrasse<br />
Gomphosus varius (Labridae). J Exp Biol.<br />
200:1549-69.<br />
Walker, J.A., & Westneat, M.W. 2000.<br />
Mechanical performance of aquatic rowing<br />
<strong>and</strong> flying. P Roy Soc Lond B Bio. 267:<br />
1875-81. doi: 10.1098/rspb.2000.1224.<br />
Webb, P.W. 1975. Hydrodynamics <strong>and</strong><br />
energetics of fish propulsion. Bull Fish Res Bd<br />
Can. 190:1-158.<br />
Webb, P.W. 1984. Form <strong>and</strong> function in fish<br />
swimming. Sci Amer. 251:58-68.<br />
Webb, P.W. 2004. Maneuverability—<br />
General issues. IEEE J Oceanic Eng.<br />
29:547-53. doi: 10.1109/JOE.2004.833220.<br />
Westneat, M.W., & Walker, J.A. 1997.<br />
Motor patterns of underwater flight: an<br />
electromyographic study of the pectoral<br />
muscles in the bird wrasse, Gomphosus varius.<br />
J Exp Biol. 200:1881-93.<br />
82 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
PAPER<br />
A Twistable Ionic Polymer-Metal Composite<br />
Artificial Muscle for <strong>Marine</strong> Applications<br />
AUTHORS<br />
Kwang J. Kim<br />
David Pugal<br />
Active Materials <strong>and</strong> Processing<br />
Laboratory, Department of<br />
Mechanical Engineering,<br />
University of Nevada-Reno<br />
Kam K. Leang<br />
Electroactive Systems <strong>and</strong> Controls<br />
Laboratory, Department of<br />
Mechanical Engineering,<br />
University of Nevada-Reno<br />
Introduction<br />
I<br />
onic polymer-metal composite<br />
(IPMC) material is one of the most<br />
promising active (smart) materials<br />
for developing novel soft biomimetic<br />
actuators <strong>and</strong> sensors, preferably for<br />
underwater applications (Shahinpoor<br />
et al., 1998; Shahinpoor & Kim, 2001;<br />
Kim & Shahinpoor, 2003; Shahinpoor<br />
& Kim, 2005). The advantages of<br />
the IPMC include low driving voltage<br />
(
FIGURE 1<br />
(a) Scanning electron microscope image of the cross-section of a Nafion-based IPMC. (b) Illustrative<br />
movement of cations <strong>and</strong> water molecules inside of an IPMC.<br />
saturated in a polar solvent (such as<br />
water) <strong>and</strong> then an electric field is applied<br />
across the electrodes, the composite<br />
bends. The bending is caused<br />
by induced swelling on the cathode<br />
side of the composite <strong>and</strong> shrinking<br />
on the anode side [see Figure 1(b)]<br />
due to a sudden flux of cations <strong>and</strong><br />
polar solvent (such as water). An oppositely<br />
applied voltage causes bending in<br />
the opposite direction. Conversely,<br />
when an IPMC is mechanically deformed,<br />
charges develop on the electrodes<br />
<strong>and</strong> thus IPMCs can function<br />
as current or voltage sensor (Alici<br />
et al., 2008; Pugal et al., 2010a).<br />
The electromechanical behavior of<br />
IPMCs have many noteworthy applications.<br />
Due to their biocompatibility,<br />
IPMC actuators show great promise in<br />
biomedical devices such as active endoscopes<br />
(Yoon et al., 2007) <strong>and</strong> smart<br />
catheters (Fang et al., 2007). Strips of<br />
IPMCs can be used as sensors in<br />
h<strong>and</strong> prostheses (Biddiss & Chau,<br />
2006). But one of the most promising<br />
applications is innovative propulsion<br />
systems for underwater autonomous<br />
systems (Fish et al., 2008; Kamamichi<br />
et al., 2006; Kim et al., 2005,<br />
2007; Lauder, 2007). Specifically,<br />
IPMCs can replace or enhance the<br />
design of propulsors for underwater<br />
walking <strong>and</strong> swimming machines that<br />
are currently based on traditional actuators<br />
such as DC motors (Ayers &<br />
Witting, 2007; Bozkurttas et al., 2008;<br />
Buchholz et al., 2008; Kato, 1998;<br />
Krieg & Mohseni, 2008; Tangorra<br />
et al., 2007), pneumatic actuators<br />
(Cai et al., 2010), <strong>and</strong> magnetic actuators<br />
(Tortora et al., 2010). In fact,<br />
strips of IPMCs have been used to construct<br />
artificial tentacles for a jellyfishlike<br />
robot (Guo et al., 2007). The<br />
walking speed of the jellyfish robot<br />
was controlled through the frequency<br />
of the input voltage applied to the<br />
IPMC-based legs. Likewise, an artificial<br />
caudal fin to propel a robotic fish<br />
was created from an IPMC actuator<br />
(Chen et al., 2010; Guo et al., 2006;<br />
Aureli et al., 2010a). The achievable<br />
peak swimming speed of the robotic<br />
fish in (Chen et al., 2010) was reported<br />
at 22 mm/s. In terms of performance,<br />
the maximum (stall) torque to weight<br />
ratio between a prototype twistable<br />
IPMC AM with dimensions 50 mm ×<br />
25 mm × 1 mm is comparable to a<br />
small ungeared DC motor as illustrated<br />
in the comparison shown in Figure 2.<br />
The comparable performance of<br />
IPMCs <strong>and</strong> traditional actuators suggests<br />
that IPMCs can play a critical<br />
role in the development of highly<br />
maneuverable <strong>and</strong> efficient marine<br />
systems, e.g., the system described in<br />
Menozzi et al. (2008).<br />
Due to the nature of the material<br />
deformation caused by cation <strong>and</strong><br />
solvent flux, bending motion is the<br />
most commonly studied <strong>and</strong> applied<br />
for IPMCs (Kim et al., 2007). Particularly,<br />
when an IPMC strip is mounted<br />
in the cantilever configuration, with<br />
one end fixed <strong>and</strong> the other free, an<br />
applied electric field to the IPMC as<br />
illustrated causes the actuator to bend<br />
as illustrated in Figure 3(a). This single<br />
degree-of-freedom bending motion<br />
has wide applications, such as a singlelink<br />
(Aureli et al., 2010a) <strong>and</strong> multilink<br />
(Yim et al., 2007) oscillatory<br />
propulsor. However, IPMCs with<br />
multiple degrees of freedom are highly<br />
desirable to create control surfaces,<br />
which can undergo complex motion<br />
<strong>and</strong> deformation, for both station<br />
keeping (B<strong>and</strong>yopadhyay et al.,<br />
2008) as well as propulsion <strong>and</strong> maneuvering.<br />
It has been observed that<br />
the propulsion <strong>and</strong> maneuvering<br />
capabilities of the Bluegill (Lepomis<br />
marcrochirus) sunfish is primarily due<br />
to its highly deformable pectoral fin<br />
(Bozkurttas et al., 2008). Thus, multiple<br />
degrees-of-freedom IPMC actuator<br />
technology offers many possibilities for<br />
mimicking such behavior to create<br />
more efficient <strong>and</strong> maneuverable<br />
underwater systems (Chen & Tan,<br />
2010). As depicted in Figure 3(b),<br />
twisting motion can be achieved by<br />
patterning sectored electrodes combined<br />
with independent control each<br />
isolated region. By carefully creating<br />
electrodes on the surface of the<br />
84 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 2<br />
Comparison of torque-to-weight ratios for traditional motors to twistable IPMC AM fin. IPMC AM fin<br />
was fixed on one end, then 5-V DC input was applied to each electrode, with opposing polarity, <strong>and</strong><br />
the blocking torque of the free end was measured with an ATI Nano17 force/torque sensor.<br />
polymer with proper electrical isolation<br />
between adjacent units, sections<br />
of the AM fin can be independently<br />
controlled to achieve complex shapes<br />
FIGURE 3<br />
IPMC AM fin motion: (a) bending <strong>and</strong> (b) twisting.<br />
<strong>and</strong> deformations. Additionally, isolated<br />
electrodes can be patterned for<br />
sensing motion <strong>and</strong> deformation of<br />
the AM fin (Kruusamae et al., 2009).<br />
IPMC Manufacturing<br />
Basic Structure of IPMC <strong>and</strong><br />
Nafion Membrane Fabrication<br />
A basic IPMC consists of an ion exchange<br />
polymer, such perfluorinated<br />
alkenes or styrene/divinylbenzenebased<br />
polymers, s<strong>and</strong>wiched between<br />
two noble metallic electrodes as<br />
shown in Figure 1(a). The conducting<br />
media can be palladium, silver, gold,<br />
carbon, graphite, <strong>and</strong> even nanotubes;<br />
however, platinum is the most commonly<br />
used. The metal electrodes are<br />
often chemically deposited on the<br />
polymer’s surface through a reduction<br />
process (Kim & Shahinpoor, 2003).<br />
Conductive paint can also be applied<br />
to the surface of the membrane to<br />
serve as an electrode; however, this approach<br />
is not as robust <strong>and</strong> effective as<br />
electrochemical plating.<br />
The commonly used ion exchange<br />
membrane Nafion (Dupont) for manufacturing<br />
IPMCs is easily available<br />
from distributors. In the case of<br />
Nafion, the typical chemical structure<br />
is shown in Figure 4, <strong>and</strong> it consists of<br />
fluorocarbons, oxygen, sulfonate<br />
groups, <strong>and</strong> a mobile cation, which<br />
are typically either hydrogen, sodium,<br />
or lithium. Commercially available<br />
Nafion membrane such as Nafion<br />
115, Nafion 117, <strong>and</strong> Nafion 1110<br />
for fabricating IPMCs have nominal<br />
dry thicknesses of 127, 178, <strong>and</strong><br />
254 μm, respectively (Aureli et al.,<br />
2010a). Methods to enhance the performance<br />
of IPMCs include boosting<br />
the capacitance of the composite<br />
(Akle et al., 2005; Aureli et al.,<br />
2009), where this is motivated by studies<br />
that correlate actuation <strong>and</strong> sensing<br />
performancewithcapacitance(Akle<br />
et al., 2005). Enhancements in performance<strong>and</strong>blockingforcehavealso<br />
been made by incorporating nanoparticulates<br />
into the polymer matrix<br />
July/August 2011 Volume 45 Number 4 85
FIGURE 4<br />
The chemical structure of Nafion, which includes fluorocarbons, oxygen, sulfonate groups, <strong>and</strong><br />
a mobile cation X + that can be hydrogen, sodium, or lithium. The K is usually 5–11, <strong>and</strong> the L is<br />
usually 1 (Shahinpoor & Kim, 2001).<br />
(Nam et al., 2003; Nguyen et al., 2007).<br />
In these studies, the nanocompositebased<br />
IPMCs were observed to have<br />
higher water uptake <strong>and</strong> slower water<br />
loss, thus leading to larger bending<br />
displacement <strong>and</strong> blocking force. It<br />
was also found that by using a dispersing<br />
agent in the reduction process<br />
to form fine platinum polycrystals<br />
whichsubsequentlyleadtodeeper<br />
penetration of the platinum layer, the<br />
blocking force was increased significantly<br />
(Shahinpoor & Kim, 2001).<br />
More recent work to improve output<br />
force is by increasing the thickness of<br />
the Nafion membrane (Kim &<br />
Shahinpoor, 2002). Since relatively<br />
thick Nafion membrane is not readily<br />
available, researchers have explored the<br />
solution casting process (Kim &<br />
Shahinpoor, 2002; Kim et al., 2003;<br />
Pak et al., 2004; Shan & Leang,<br />
2009) <strong>and</strong> the hot pressing technique<br />
(Lee et al., 2006). The output force<br />
enhancement for thicker IPMCs is evident<br />
by considering two IPMC strip<br />
actuators, both having the same length<br />
<strong>and</strong> width, but each have a different<br />
thicknesses, such as t <strong>and</strong> 2t (twice as<br />
thick). Assuming that for both actuators<br />
the same tip displacement is required,<br />
then the required strain for<br />
the thick actuator is 2ɛ. As the stress<br />
tensor for linear beam with thickness<br />
of t, width b, <strong>and</strong> length L can be expressed<br />
as (Kim & Shahinpoor, 2002)<br />
σ t ¼ 6F tL<br />
bt 2 ; ð1Þ<br />
the ratio of the stresses is<br />
σ 2t<br />
σ t<br />
≈ 2 ¼ F 2t<br />
2 2 F t<br />
; ð2Þ<br />
FIGURE 5<br />
hence F 2t =8F t .Therefore,athicker<br />
IPMC will produce a larger blocking<br />
force, motivating the need to create<br />
thicker membranes for fabricating<br />
IPMCs. At the same time, the measurements<br />
show that actuation speed<br />
reduces with the increase of the thickness.<br />
Therefore, further study is necessary<br />
to find an optimal thickness of the<br />
membrane for marine applications.<br />
Pretreatment <strong>and</strong> Platinum<br />
Plating Process<br />
The manufacturing of IPMCs begins<br />
with the pretreatment of the ion<br />
exchange membrane (Nafion) <strong>and</strong><br />
the platinum plating process as<br />
outlined in the flowchart shown in<br />
Figure 5. First, the surface of the membrane<br />
is either mechanically roughened<br />
or chemically etched (Yoon et al.,<br />
2007) to either enhance the capacitance<br />
or to improve adhesion of the<br />
metal electrode to the surface. Then,<br />
organic <strong>and</strong> metallic impurities on<br />
thebareNafion membrane are removed<br />
through a pretreatment process<br />
by initially chemically cleaning the<br />
Nafion membrane in 3% hydrogen<br />
peroxide (H 2 O 2 ). Next, the cleaned<br />
membrane is rinsed in 0.5 M sulfuric<br />
acid (H 2 SO 4 )at80°C.Afterwards,<br />
the pretreated <strong>and</strong> cleaned Nafion<br />
membrane is immersed in an appropriate<br />
metal salt solution such as<br />
tetraamineplatinum (II) chloridemonohydrate<br />
[(Pt(NH 3 ) 4 )Cl 2 H 2 O] for<br />
2 h, followed by several washings in<br />
de-ionized water. Platinum particles<br />
are metalized on the surface of the<br />
Nafion membrane by reducing the<br />
Left: The process flow for pretreating the Nafion membrane <strong>and</strong> applying platinum electrodes.<br />
Right: fabricated IPMCs from commercially available Nafion membrane.<br />
86 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
membrane in a sodium borohydride<br />
(NaBH 4 ) or lithium borohydride<br />
(LiBH 4 ) solution for 3 h. Finally, the<br />
platinum-plating process is repeated to<br />
achieve at least three layers of platinum<br />
on the surface of the Nafion membrane<br />
to enhance surface conductivity<br />
<strong>and</strong> overall performance. The platinum<br />
particulate layer is often buried<br />
1-20 μm withintheIPMCsurface<br />
<strong>and</strong> is highly dispersed.<br />
Electrode Patterning Process<br />
The sectored electrodes for the<br />
twistable IPMC AM fin can be created<br />
by masking, surface machining, or ablating<br />
the electrode using a high-power<br />
laser. The first approach involves the<br />
use of a mask to cover areas of the<br />
membrane that should not be exposed<br />
to the platinum-plating process,<br />
subsequently creating an isolation<br />
region between adjacent electrodes.<br />
The second approach uses a precision<br />
computer-controlled milling machine<br />
to mechanically remove the platinum<br />
electrodes from the surface of an IPMC<br />
membrane to create the isolation region.<br />
Masking Technique: The masking<br />
technique to create IPMCs with<br />
sectored electrodes involves the use<br />
of UHMW (Ultra-high-molecularweight)<br />
polyethylene tape (3M). The<br />
tape is used to cover specific regions<br />
of the bare Nafion membrane to inhibit<br />
the plating of platinum on the surface<br />
of the membrane. For example,<br />
the tape is applied to the bare Nafion<br />
membrane, then the taped Nafion<br />
membrane is processed using the pretreatment<br />
<strong>and</strong> plating process described<br />
above <strong>and</strong> outlined in Figure 5. Sample<br />
IPMCs with sectored electrodes are<br />
shown in Figure 6(a), <strong>and</strong> an outline<br />
of the process for the masking technique<br />
is illustrated in Figure 7(a).<br />
Machining Technique: The<br />
surface machining method utilizes a<br />
computer-controlled milling machine,<br />
such as an automated circuit board<br />
router (e.g., ProtoMat S42, LPKF) to<br />
mechanically remove the plated platinum<br />
metal on the surface of the Nafion<br />
membrane. The surface machining<br />
process is outlined in Figure 7(b) <strong>and</strong><br />
described as follows:<br />
1. Create the machining path to create<br />
the electrode pattern using CAD<br />
software (e.g., Solidworks) or a circuit<br />
board layout program (e.g.,<br />
Eagle). The result of this step is a<br />
CAD/CAM file, which is ran by<br />
the milling machine.<br />
2. Attach an IPMC sample (with platinum<br />
electrodes) to the working<br />
surface of the milling machine<br />
using an adhesive layer, for example<br />
double-sided tape (see Figure 7(b))<br />
or a vacuum system. Air bubbles<br />
trapped underneath the IPMC<br />
sample should be removed. Additionally,<br />
the locations of the corners<br />
of the IPMC on the working surface<br />
are marked for aligning the<br />
sample.<br />
3. Load the CAD/CAM file onto the<br />
milling machine <strong>and</strong> start the milling<br />
process.<br />
4. Remove the machined IPMC sample,<br />
then flip it over <strong>and</strong> attach the<br />
sample to the working surface making<br />
sure the corners are aligned with<br />
the markings created in Step 2.<br />
Repeat Steps 2 <strong>and</strong> 3.<br />
5. Remove the machined IPMC sample<br />
<strong>and</strong> trim away excess material.<br />
Sample IPMCs with sectored<br />
electrodes created by the surface<br />
machining process are shown in Figures<br />
6(b1)-6(b4). During the machining<br />
process, care is taken to<br />
avoid removing too much material<br />
for thin IPMC membranes. In general,<br />
the machine should be set to remove<br />
only the platinum material (approximately<br />
25-50 μm deep). Comparing<br />
the masking <strong>and</strong> surface-machining results,<br />
the machining process allows<br />
better control of the shape <strong>and</strong> pattern<br />
of the electrode. One of the major<br />
challenges of the masking process is<br />
ensuring that the tape adheres to the<br />
IPMC membrane’s surface all through<br />
the plating process. During machining,<br />
if the depth of cut is not well-controlled,<br />
removal of excess material affects the<br />
mechanical properties, for example,<br />
stiffness, of the IPMC actuator.<br />
Modeling<br />
Electromechanical<br />
Transduction<br />
Two approaches are discussed to<br />
model the electromechanical bending<br />
<strong>and</strong> twisting response of the sectoredelectrode<br />
IPMCs: (1) a simplified<br />
finite-element analysis (FEA) model<br />
derived from the piezoelectric effect<br />
<strong>and</strong> (2) a more comprehensive physicsbased<br />
model. The former model has<br />
the advantage of being easy to implement<br />
due to fact that software packages<br />
are available <strong>and</strong> specifically tailored to<br />
solve the problem. Additionally, the<br />
established algorithms are computationally<br />
efficient. However, the first<br />
method lacks consideration of the<br />
physical processes that governs the behavior<br />
of the IPMC. The latter model,<br />
on the other h<strong>and</strong>, considers the underlying<br />
physics that includes electrostatic<br />
forces, osmotic pressure, charge<br />
imbalance <strong>and</strong> the effects of local strains<br />
(Kim & Shahinpoor, 2003; Nemat-<br />
Nasser & Jiang, 2000; Tadokoro<br />
et al., 2000; Chen & Tan, 2008; Leo<br />
et al., 2005). This type of model offers<br />
valuable insight on the physical behavior<br />
of the composite material <strong>and</strong><br />
the model can be used to help guide<br />
the development of the material on<br />
many levels. Despite being more<br />
realistic—<strong>and</strong> sometimes more<br />
accurate—the physics-based model<br />
July/August 2011 Volume 45 Number 4 87
FIGURE 6<br />
Fabricated IPMCs with sectored electrodes: samples created using (a) the masking technique <strong>and</strong> (b1)-(b4) the surface-machining approach.<br />
Dimensions are in millimeters (mm).<br />
is more computationally dem<strong>and</strong>ing<br />
to solve. Simulation results comparison<br />
with experimental data are<br />
presented in Performance Characterization<br />
<strong>and</strong> Discussion<br />
Simplified FEA Model for<br />
Electromechanical Response<br />
Asimplified finite-element model<br />
is created to predict <strong>and</strong> gain insight<br />
on the bending <strong>and</strong> twisting capabilities<br />
of the sectored-electrode IPMC<br />
AM fin. Such a model can also be<br />
applied to optimize the design of an<br />
AM fin for specific applications. The<br />
key feature of this simple model is<br />
the deformation is estimated using an<br />
equivalent bimorph beam model analogous<br />
to a piezoelectric actuator (Kim<br />
& Tadokoro, 2007). As a result the<br />
model only captures the basic electromechanical<br />
behavior <strong>and</strong> is thus easy to<br />
implement <strong>and</strong> computationally efficient.<br />
Figure 8 shows the boundary conditions<br />
for the finite-element model,<br />
where the ‘clamped area’ (25 mm ×<br />
5 mm) is considered fixed. The bimorph<br />
finite-element model consists of two<br />
electro-mechanical actuation layers,<br />
layers (i) <strong>and</strong> (iii), as shown in Figure 8.<br />
88 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 7<br />
Electrode patterning process: (a) the masking <strong>and</strong> (b) the surface-machining technique.<br />
The material separating patterned areas<br />
of the material are modeled by layer (ii).<br />
The IPMC material is treated as a<br />
homogenous material (uniform stiffness<br />
<strong>and</strong> density). The st<strong>and</strong>ard stiffness<br />
matrix k, piezoelectric strain matrix d,<br />
<strong>and</strong> permittivity matrix ɛ are used in<br />
the modeling (Moheimani & Fleming,<br />
2006). The material properties are as<br />
follows: density is 2930 kg/m 3 ;<br />
Poisson’s ratio is 0.49; E=1.16GPa;<br />
d 31 = d 32 =4.11×10 -6 m/V (bending);<br />
d 31 = d 32 =2.67×10 −7 m/V (twisting).<br />
The finite-element model is created<br />
using ANSYS software (Canonsburg,<br />
PA). A SOLID98 tetrahedral coupled<br />
field solid element is chosen for the<br />
electro-mechanical material, while a<br />
SOLID187 tetrahedral structural solid<br />
element is chosen for the electrode separation<br />
material. Both material properties<br />
are assigned a tetrahedral solid to compensate<br />
for the irregular mess interface<br />
caused by large aspect ratio between the<br />
IPMCs thickness <strong>and</strong> length. By selecting<br />
these elements, a uniform tetrahedral<br />
mesh throughout the entire model is<br />
achieved.<br />
Physics-Based Model for<br />
Electromechanical Transduction<br />
When a voltage is applied to the<br />
electrodes of an IPMC, the freely movable<br />
cations inside the polymer start<br />
migrating due to the imposed electric<br />
FIGURE 8<br />
Finite-element model structure. Layers (i) <strong>and</strong> (iii) are the bimorph (electro-mechanical) layers, <strong>and</strong> layer (ii) is assumed to be a nonconducting<br />
electrode separation layer (absent of electro-mechanical properties).<br />
July/August 2011 Volume 45 Number 4 89
field. However, the attached anions do not migrate nor diffuse. In case of waterbased<br />
IPMCs, migrating cations drag the water molecules along, causing<br />
osmotic pressure changes <strong>and</strong> therefore swelling near the cathode <strong>and</strong> contraction<br />
of the polymer near the anode. This in turn results in bending of the material<br />
towards the anode side. Furthermore, the migrated cations cause charge imbalance<br />
near the electrodes <strong>and</strong> this possibly results in the electrostatic force contribution<br />
to the local strain. In the following, the basic underlying equations<br />
are presented <strong>and</strong> how to apply the calculated charge imbalance near the<br />
electrodes in modeling the IPMC actuation in a 3-D domain is discussed<br />
(see Figure 9).<br />
First, the ionic migration <strong>and</strong> diffusion are described in the polymer domain by<br />
Nernst-Planck <strong>and</strong> Poisson equations,<br />
∂C<br />
∂t<br />
þ ∇· ð D∇C zμFC∇ϕÞ ¼0;<br />
∇ 2 ϕ ¼ F ð C C 0Þ<br />
; ð4Þ<br />
ɛ<br />
where C is the cation concentration, μ is the mobility of counter ions, D is the diffusion<br />
constant, F is the Faraday constant, z is the charge number, ϕ is the electric<br />
potential, <strong>and</strong> C 0 is the anion concentration with initial value of C 0 =1200mol/m 3 .<br />
For the electrode, Ohm’s law for the current density is<br />
σ∇V ¼ → j; ð5Þ<br />
where σ is electric conductivity, V is the voltage, <strong>and</strong> → j is the current density in the<br />
electrode. It must be noted that the electric potential ϕ inside the polymer <strong>and</strong> the<br />
electric potential V in the electrode are different. To relate the variables of<br />
the electrodes V <strong>and</strong> → j to the ionic flux inside the polymer, the Ramo-Shockley<br />
FIGURE 9<br />
Conceptual physics-based model. Calculations are done in three domains—the polymer domain<br />
<strong>and</strong> two electrode domains.<br />
ð3Þ<br />
theorem is used (Ramo, 1939; Shockley,<br />
1938), i.e.,<br />
I ¼ 1 φ ∑ →<br />
n q nW ð<br />
→<br />
r n Þ· →v n ;<br />
ð6Þ<br />
where → r n ; → v n ,<strong>and</strong> → q n are the position vector,<br />
instantaneous velocity, <strong>and</strong> charge<br />
of cation n, respectively. The electric<br />
field that would be produced by 1 V applied<br />
potential without any charges,<br />
neither mobile nor fixed, is denoted by<br />
→<br />
W . The current in the external circuit is<br />
represented by I, <strong>and</strong>ϕ is a constant<br />
with value of 1 V. Equation (6) can be<br />
simplified for a 2-D domain with parallel<br />
electrodes<br />
j ¼ 1 h ∫ f dl;<br />
ð7Þ<br />
where f is an ionic flux in the electrode<br />
direction <strong>and</strong> has a unit of Cm 2 s <strong>and</strong> j is<br />
local current density at the inner boundary<br />
of an electrode. The term dl is the<br />
integration element along the path<br />
where the particle moves, which in this<br />
case is assumed to be from one electrode<br />
to other.<br />
The Equations (3)-(7) are used to<br />
calculate the time dependent cation<br />
concentration C <strong>and</strong> corresponding<br />
charge density ρ = C − C 0 in the polymer<br />
of IPMC in response to an applied<br />
voltage. To relate the charge density<br />
ρ to the physical bending, the force<br />
coupling similar to the one shown in<br />
Pugal et al. (2008) is used. A set of<br />
continuum mechanics equations were<br />
implemented for the polymer domain.<br />
Normal <strong>and</strong> shear strain are by<br />
definition<br />
ɛ i ¼ ∂u i<br />
; ɛ ij ¼ 1 <br />
∂u i<br />
þ ∂u <br />
j<br />
; ð8Þ<br />
∂x i 2 ∂x j ∂x i<br />
where u is the displacement vector,<br />
x denotes a coordinate <strong>and</strong> indices i<br />
90 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
̂<br />
<strong>and</strong> j are in the range of 1-3 <strong>and</strong> denote<br />
components correspondingly to x, y,or<br />
z direction. The stress-strain relationship<br />
is<br />
σ f ¼ Dɛ;<br />
ð9Þ<br />
where D is a 6 × 6 elasticity matrix,<br />
consisting of components of Young’s<br />
modulus <strong>and</strong> Poisson’s ratio. The system<br />
is in equilibrium, if the relation<br />
∇· σ f ¼ → F ð10Þ<br />
is satisfied. This is the Navier’s equation<br />
for displacement (Heinbockel,<br />
2001). The body force <strong>and</strong> charge coupling<br />
is defined as<br />
→<br />
F ¼ Aρx; ̂<br />
ð11Þ<br />
the range of hundreds of thous<strong>and</strong>s<br />
of degrees of freedom. The challenges<br />
are described in more detail in Pugal<br />
et al. (2010b). To reduce the problem<br />
size without significant loss in the<br />
calculation precision, the following<br />
modeling approach for 3-D actuation<br />
of IPMC is developed.<br />
Firstly, the cation concentration<br />
C (from which the charge density ρ<br />
can be directly calculated) <strong>and</strong> voltage<br />
ϕ are calculated in a 2-D domain.<br />
As the 2-D domain is scaled in the longitudinal<br />
(x) direction of an IPMC, the<br />
electrode conductivity value is also<br />
linearly scaled. The electrode currents<br />
<strong>and</strong> corresponding voltage gradient in<br />
the electrodes are taken into account in<br />
the 2-D model to obtain more precise<br />
cation concentration. See, for instance,<br />
Figure 10—there the molar flux f<br />
<strong>and</strong> electric current j are depicted. As<br />
a result of the calculations, spatial<br />
<strong>and</strong> temporal C <strong>and</strong> ϕ in 2-D are<br />
found <strong>and</strong> stored. This is done for<br />
each patterned electrode that is subjected<br />
to a different voltage. Due to<br />
thesmallDebyescreeninglengthof<br />
the charges near the electrodes (Porfiri,<br />
2009), the 2-D model calculations are<br />
done on a mesh that is extremely fine<br />
near the boundaries.<br />
Secondly, the calculated C <strong>and</strong> ϕ are<br />
extruded onto slightly coarser mesh in<br />
the 3-D domain as illustrated in<br />
Figure 11. For instance, Figure 12<br />
shows the extruded voltage ϕ values<br />
on the 3-D domain boundary. It<br />
should be noted that in the current<br />
model, it is not necessary to extrude<br />
the values of ϕ, but it was done for<br />
illustration purposes. Finally, the<br />
where A is a parametrically determined<br />
constant <strong>and</strong> x is IPMC’s longitudinal<br />
direction. This approach allows calculating<br />
deformation that is in the typical<br />
IPMC actuation range. When a very<br />
large deformation is expected, for instance,<br />
in case of a very thin membrane,<br />
geometric nonlinearity <strong>and</strong><br />
more precise force coupling must be<br />
used in the model. The conceptual<br />
model with the variables is illustrated<br />
in Figure 9.<br />
The finite element method is used<br />
to solve Equations (3), (4), <strong>and</strong> (5)<br />
with Equation (7) as the boundary<br />
condition between the electrode <strong>and</strong><br />
polymer domain <strong>and</strong> Equation (10).<br />
The equations are implemented in<br />
Comsol Multiphysics software package<br />
(Multiphysics, 2011). The high aspect<br />
ratio of the domain representing<br />
an IPMC <strong>and</strong> the nonlinear nature<br />
of the problem make it difficult to<br />
directly solve the equations in a<br />
full-scale 3-D domain. Namely, the<br />
problem size would be very large, in<br />
FIGURE 10<br />
Electric current streamlines in the electrodes <strong>and</strong> total ionic flux streamlines in the polymer<br />
A<br />
domain. The color depicts the total current density m2<br />
in the electrodes. (Color versions of<br />
figures available online at: http://www.ingentaconnect.com/content/mts/mtsj/2011/00000045/<br />
00000004.)<br />
July/August 2011 Volume 45 Number 4 91
FIGURE 11<br />
Extruding the cation concentration C that has been calculated for each<br />
time step in 2-D into a 3-D domain.<br />
FIGURE 12<br />
Extruded data from 2-D domain. The surface depicts the extruded voltage<br />
values on the electrodes. Notice the voltage gradient in the longitudinal<br />
x direction.<br />
coupling Equations (11) <strong>and</strong> (10) are<br />
carried out to calculate the timedependent<br />
bending of an IPMC in 3-D.<br />
Performance<br />
Characterization<br />
<strong>and</strong> Discussion<br />
The bending <strong>and</strong> twisting performance<br />
of the fabricated IPMC AM<br />
fins are characterized <strong>and</strong> then compared<br />
to the electromechanical models.<br />
The experimental setup consists of<br />
control electronics connected to the<br />
IPMC AM fin <strong>and</strong> a measurement system<br />
(laser displacement sensors <strong>and</strong><br />
data acquisition system) for collecting<br />
the output response. Each set of electrodes<br />
are independently controlled<br />
by a separate custom-designed voltage<br />
amplifier as depicted in Figure 13. A<br />
complete description of the experimental<br />
setup is described in Riddle<br />
et al. (2010). All measurements are<br />
taken while the subject IPMC is actuated<br />
in de-ionized water.<br />
Measured Bending <strong>and</strong><br />
Twisting Performance<br />
Themeasuredresponseforaselected<br />
masked <strong>and</strong> machined electrode<br />
IPMC, Figure 6(b2), are shown in Figure<br />
14. The results in Figure 14 show<br />
that both types of actuators provided<br />
approximately the same degree of<br />
FIGURE 13<br />
twist. Furthermore, the actuators also<br />
exhibited non-smooth twisting<br />
motion, which may have been caused<br />
by the effects of the fabrication process.<br />
A peak twist angle of 7.3° is measured<br />
for the machined IPMC [see<br />
Figures 14(c) <strong>and</strong> 14(d)]. The frequency<br />
of the actuation of 1 Hz is<br />
Experimental setup for measuring bending <strong>and</strong> twisting response using two non-contact laser<br />
sensors (Micro-Epsilon, optoNCDT 1402). Voltage amplifier gain is A.<br />
92 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
sufficient for underwater propulsion<br />
(Chen et al., 2010). The bending<br />
angle could be enhanced in various<br />
ways, depending on the application<br />
of interest. For instance, rigid extensions<br />
with different shapes can be<br />
used (Anton, 2008).<br />
FIGURE 14<br />
Measured IPMC twisting response for 5 V sinusoidal input at 1 Hz: (a) input voltage applied to<br />
electrodes; (b) twisting response for masked electrode IPMC; (c) twisting response for machined<br />
electrode IPMC; (d) sequence images of machined electrode IPMC showing twisting behavior.<br />
Simulated Responses<br />
The bending <strong>and</strong> twisting response<br />
of the IPMC, Figure 6(b2), produced<br />
by the simple FEA model for an<br />
input voltage of 2 V are shown in<br />
Figures 15(a) <strong>and</strong> 15(b). For bending,<br />
the tip displacement is predicted at<br />
approximately 2.5 mm. For twisting,<br />
the angle of twist is estimated using<br />
θ ¼ tan 1 a b<br />
, where a <strong>and</strong> b are<br />
shown in Figure 15(c). The predicted<br />
maximum angle of twist is approximately<br />
0.7° for a 2-V input. In Figures<br />
15(d) <strong>and</strong> 15(e), the FEA results<br />
<strong>and</strong> the response for the bending <strong>and</strong><br />
twisting motion measured along the<br />
length of the IPMC actuator are<br />
compared. For bending, the predicted<br />
<strong>and</strong> measured results agreed well,<br />
where the maximum root-meansquared<br />
(RMS) error is approximately<br />
0.2 mm. For bending, however, the<br />
RMS error increased significantly for<br />
3 V <strong>and</strong> beyond. It is noted that the<br />
finite-element approach assumes the<br />
material is isotropic <strong>and</strong> operating<br />
within the linear-elastic range. Due to<br />
the complex nature of the IPMC<br />
material, the simplified FEA may be<br />
limited in its ability to accurately<br />
predict the performance for large<br />
electric fields. For example, as shown<br />
in Figure 14, the measured twist<br />
angle is approximately 7° for an input<br />
voltage with magnitude of 5 V. This<br />
result indicates that the IPMC’s response<br />
can be highly nonlinear for<br />
large input voltages. Therefore, one<br />
of the challenges is developing detailed<br />
enough models to aid in designing the electrode patterns to meet a specific<br />
application.<br />
To adequately model the twisting deformation with the physics-based electromechanical<br />
model presented in Physics-Based Model for Electromechanical<br />
Transduction, the sinusoidal voltages<br />
u 1 ðÞ¼5 t ½V<br />
Šsinð2πtÞ; u 2 ðÞ¼ t<br />
5½V<br />
Šsinð2πtÞ; ð12Þ<br />
are set as time-dependent boundary conditions to the electrodes 1, 2 <strong>and</strong> 3, 4,<br />
respectively (see Figure 13). The calculated displacement fields at two different<br />
times are shown in Figure 16. The model is validated against measured time response<br />
[Figure 14(c)]. The model estimation versus measurements are shown in<br />
Figure 17. It can be seen that the model predicts the twisting deformation well.<br />
Slight discrepancies in the peak values can be attributed to the fact that the model<br />
July/August 2011 Volume 45 Number 4 93
FIGURE 15<br />
Finite-element modeling results for (a) bending <strong>and</strong> (b) twisting for 2-V input. (c) Definition of twist angle θ. Comparison of finite-element model<br />
results <strong>and</strong> measured response: (d) bending <strong>and</strong> (e) twisting response for different applied voltages.<br />
does not take into account electrochemical<br />
currents <strong>and</strong> therefore does<br />
not provide correct voltage gradient on<br />
the electrodes for higher applied voltages.<br />
It must be noted that the hydrodynamic<br />
effects are not considered in<br />
the calculations due to the low frequency<br />
<strong>and</strong> rather small twisting amplitude.<br />
Conclusions<br />
IPMC AMs are suited for creating<br />
artificial fish-like propulsors that can<br />
mimic the undulation, flapping, <strong>and</strong><br />
complex motions of fish fins. A newly<br />
developed IPMC AM fin withpatterned<br />
electrodes was introduced for<br />
realizing multiple degrees-of-freedom<br />
motion, such as bending <strong>and</strong> twisting.<br />
These twistable AM fins are suited for<br />
creating artificial fish-like propulsors<br />
that can mimic the undulatory, flapping,<br />
<strong>and</strong> complex motions of fish<br />
fins as well as novel control surfaces<br />
for applications in a wide spectrum of<br />
micro-autonomous robots <strong>and</strong> marine<br />
systems. The masking <strong>and</strong> surface machining<br />
fabrication process are viable<br />
approaches to create a twistable AM<br />
fin. Experimental characterization<br />
showed that peak twisting for the<br />
sectored-electrode IPMC exceeded 7°.<br />
A finite-element bimorph beam<br />
model was used to predict the bending<br />
<strong>and</strong> twisting behavior of a selected<br />
IPMC actuator, where good agreement<br />
between the measured response <strong>and</strong><br />
model output was found for low<br />
electric fields (2 V). A full-scale 3-D<br />
94 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 16<br />
Calculated twisting of IPMC at t = 0.25 s (left) <strong>and</strong> at t = 0.75 s (right). The color shows y-directional displacement.<br />
physics-based model to simulate electromechanical<br />
twisting transduction<br />
of IPMC was developed. Comparison<br />
between the experimental results <strong>and</strong><br />
model output for the electromechanical<br />
transduction indicated that the<br />
FIGURE 17<br />
Comparison of experimental (solid line) <strong>and</strong> simulated twisting angle (dash line).<br />
model predicts the twisting deformation<br />
well. The future work will explore<br />
complex electrode patterns, integrated<br />
sensing electrodes, an alternative<br />
approach to create a twistable fin structure<br />
using a soft boot, <strong>and</strong> the development<br />
of a prototype autonomous<br />
marine system powered by the twistable<br />
IPMC AM fin.<br />
Acknowledgments<br />
The authors gratefully thank<br />
the financial support from the U.S.<br />
Office of Naval Research (grant<br />
N0001409102183) <strong>and</strong> Dr. Tom<br />
McKenna. The authors also thank<br />
S.M. Kim, Y.S. Jung, S. Song,<br />
R. Riddle, <strong>and</strong> Y. Shan for their help<br />
with the experiments. KJK expresses<br />
his special thanks to Dr. Promode<br />
B<strong>and</strong>yopadhyay of the Naval Undersea<br />
Warfare Center (NUWC) in<br />
Newport, RI, for his thoughtful<br />
encouragement.<br />
Lead Authors:<br />
Kwang J. Kim<br />
Active Materials <strong>and</strong><br />
Processing Laboratory<br />
Department of Mechanical<br />
Engineering, University<br />
of Nevada-Reno<br />
Email: kwangkim@unr.edu<br />
July/August 2011 Volume 45 Number 4 95
Kam K. Leang<br />
Electroactive Systems <strong>and</strong><br />
Controls Laboratory<br />
Department of Mechanical<br />
Engineering, University<br />
of Nevada-Reno<br />
Email: kam@unr.edu<br />
References<br />
Anton, M. 2008. Mechanical modeling<br />
of IPMC actuators at large deformations.<br />
Ph.D. Thesis, University of Tartu.<br />
Akle, B.J., Leo, D.J., Hickner, M. A., &<br />
McGrath, J.E. 2005. Correlation of capacitance<br />
<strong>and</strong> actuation in ionomeric polymer<br />
transducers. J Mater Sci. 40:3715-24.<br />
doi: 10.1007/s10853-005-3312-x.<br />
Alici, G., Spinks, G.M., Madden, J.D.,<br />
Wu, Y., & Wallace, G.G. 2008. Response<br />
characterization of electroactive polymers as<br />
mechanical sensors. IEEE/ASME T Mech.<br />
13(2):187-96.<br />
Aureli, M., Kopman, V., & Porfiri, M.<br />
2010a. Free-locomotion of underwater<br />
vehicles actuated by ionic polymer metal<br />
composites. IEEE/ASME T Mech.<br />
15(4):603-14.<br />
Aureli, M., Lin, W., & Porfiri, M. 2009.<br />
On the capacitance-boost of ionic polymer<br />
metal composites due to electroless plating:<br />
Theory <strong>and</strong> experiments. J Appl Phys.<br />
105:104911. doi: 10.1063/1.3129503.<br />
Aureli, M., Prince, C., Porfiri, M., &<br />
Peterson, S.D. 2010b. Energy harvesting from<br />
base excitation of ionic polymer metal<br />
composites in fluid environments. Smart<br />
Mater Struct. 19:015003. doi: 10.1088/<br />
0964-1726/19/1/015003.<br />
Ayers, J., & Witting, J. 2007. Biomimetic<br />
approaches to the control of underwater<br />
walking machines. Philos T R Soc A.<br />
365:273-95. doi: 10.1098/rsta.2006.1910.<br />
B<strong>and</strong>yopadhyay, P.R., Beal, D.N., &<br />
Menozzi, A. 2008. Biorobotic insights into<br />
how animals swim. J Exp Biol. 211:206-14.<br />
doi: 10.1242/jeb.012161.<br />
Bhat, N.D., & Kim, W.-J. 2004. Precision<br />
force <strong>and</strong> position control of ionic polymer-metal<br />
composite. J Syst Contr Eng. 218(6):421-32.<br />
Biddiss, E., & Chau, T. 2006. Electroactive<br />
polymeric sensors in h<strong>and</strong> prostheses: Bending<br />
response of an ionic polymer metal composite.<br />
Med Eng Phys. 28(6):568-78. doi: 10.1016/<br />
j.medengphy.2005.09.009.<br />
Bozkurttas, M., Tangorra, J., Lauder, G., &<br />
Mittal, R. 2008. Underst<strong>and</strong>ing the hydrodynamics<br />
of swimming: From fish fins to<br />
flexible propulsors for autonomous underwater<br />
vehicles. Adv Sci Tech. 58:193-202.<br />
doi: 10.4028/www.scientific.net/AST.58.193.<br />
Brufau-Penella, J., Puig-Vidal, M., Giannone,<br />
P., Graziani, S., & Strazzeri, S. 2008. Characterization<br />
of the harvesting capabilities of<br />
an ionic polymer metal composite device.<br />
Smart Mater Struct. 17(1):015009.<br />
doi: 10.1088/0964-1726/17/01/015009.<br />
Buchholz, J.H.J., Clark, R.P., & Smits, A.J.<br />
2008. Thrust performance of unsteady propulsors<br />
using a novel measurement system,<br />
<strong>and</strong> corresponding wake patterns. Exp<br />
Fluids. 45:461-72. doi: 10.1007/s00348-<br />
008-0489-1.<br />
Cai, Y., Bi, S., & Zheng, L. 2010. Design<br />
<strong>and</strong> experiments of a robotic fish imitating<br />
wow-nosed ray. J Bionic Eng. 7:120-126.<br />
doi: 10.1016/S1672-6529(09)60204-3.<br />
Chen, Z., Shatara, S., & Tan, X. 2010.<br />
Modeling of biomimetic robotic fish propelled<br />
by an ionic polymer-metal composite caudal<br />
fin. IEEE/ASME T Mech. 15:448-59.<br />
Chen, Z., & Tan, X. 2008. A controloriented<br />
<strong>and</strong> physics-based model for ionic<br />
polymer-metal composite actuators. IEEE/<br />
ASME T Mech. 13(5):519-29.<br />
Chen, Z., & Tan, X. 2010. Monolithic<br />
fabrication of ionic polymer-metal composite<br />
actuators capable of complex deformation.<br />
Sensor Actuat A-Phys. 157:246-57.<br />
doi: 10.1016/j.sna.2009.11.024.<br />
Chen, Z., Um, T., & Bart-Smith, H. 2011.<br />
A novel fabrication of ionic polymer-metal<br />
composite actuator capable of 3-dimensional<br />
kinematic motions. Sensor Actuat A-Phys.<br />
168(1):131-9. doi: 10.1016/j.<br />
sna.2011.02.034.<br />
Fang, B.-K., Ju, M.-S., & Lin, C.-C. K. 2007.<br />
A new approach to develop ionic polymermetal<br />
composites (IPMC) actuator: Fabrication<br />
<strong>and</strong> control for active catheter systems. Sensor<br />
Actuat A-Phys. 137(2):321-9. doi: 10.1016/<br />
j.sna.2007.03.024.<br />
Fish, F.E., Howle, L.E., & Murray, M.M.<br />
2008. Hydrodynamic flow control in marine<br />
mammals. Integr Comp Biol. 48:1-13.<br />
doi: 10.1093/icb/icn029.<br />
Guo, S., Ge, Y., Li, L., & Liu, S. 2006.<br />
Underwater swimming micro robot using<br />
IPMC actuator. In: Proceedings of the<br />
2006 IEEE International Conference on<br />
Mechatronics <strong>and</strong> Automation, June 25-28.<br />
pp. 249-54. Piscataway, NJ: IEEE.<br />
doi: 10.1109/ICMA.2006.257505.<br />
Guo, S., Shi, L., Ye, X., & Li, L. 2007. A<br />
new jellyfish type of underwater microrobot.<br />
In: Proc. IEEE International Conference on<br />
Mechatronics <strong>and</strong> Automation. pp. 509-14.<br />
Piscataway, NJ: IEEE.<br />
Heinbockel, J. 2001. Introduction to<br />
Tensor Calculus <strong>and</strong> Continuum Mechanics.<br />
Bloomington, IN: Trafford Publishing.<br />
Jeon, J.H., & Oh, I.K. 2009. Selective growth<br />
of platinum electrodes for MDOF IPMC<br />
actuators. Thin Solid Films. 517:5288-92.<br />
doi: 10.1016/j.tsf.2009.03.111.<br />
Jeon, J.H., Oh, I.K., Han, J.H., & Lee,<br />
S.W. 2007. Development of bio-mimetic<br />
patterned IPMC actuators with multiple<br />
electrodes. Key Eng Mat. 334:1005-8.<br />
doi: 10.4028/www.scientific.net/KEM.<br />
334-335.1005.<br />
Kamamichi, N., Yamakita, M., Asaka, K.,<br />
& Luo, Z.-W. 2006. A snake-like swimming<br />
robot using IPMC actuator/sensor.<br />
In: IEEE International Conference on<br />
Robotics <strong>and</strong> Automation. pp. 1812-7.<br />
Piscataway, NJ: IEEE.<br />
Kang, S., Shin, J., Kim, S.J., Kim, H.J., &<br />
Kim, Y.H. 2007. Robust control of ionic<br />
polymer-metal composites. Smart Mater<br />
96 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
Struct. 16(6):2457-63. doi: 10.1088/<br />
0964-1726/16/6/049.<br />
Kato, N. 1998. Locomotion by mechanical<br />
pectoral fins. J Mar Sci Technol. 3:113-21.<br />
doi: 10.1007/BF02492918.<br />
Kim, B., Kim, B.M., Ryu, J., Oh, I.-H.,<br />
Lee, S.-K., Cha, S.-E., & Pak, J. 2003.<br />
Analysis of mechanical characteristics of the<br />
ionic polymer metal composite (IPMC)<br />
actuator using cast ion-exchange film. In:<br />
Proceeding of the SPIE Smart Structures<br />
<strong>and</strong> Materials: Electroactive Polymer Actuators<br />
<strong>and</strong> Devices (EAPAD). pp. 486-95.<br />
San Diego, CA: SPIE.<br />
Kim, B., Kim, D.-H., Jung, J.J., & Park,<br />
J.-O. 2005. A biomimetic undulatory tadpole<br />
robot using ionic polymer-metal composite<br />
actuators. Smart Mater Struct. 14:1579-85.<br />
Kim, K.J., & Shahinpoor, M. 2002. A novel<br />
method of manufacturing three-dimensional<br />
ionic polymer-metal composites (ipmcs)<br />
biomimetic sensors, actuators <strong>and</strong> artificial<br />
muscles. Polymer. 43(3):797-802.<br />
doi: 10.1016/S0032-3861(01)00648-6.<br />
Kim, K.J., & Shahinpoor, M. 2003. Ionic<br />
polymer-metal composites: II. Manufacturing<br />
techniques. Smart Mater Struct. 12(1):65-79.<br />
doi: 10.1088/0964-1726/12/1/308.<br />
Kim, K.J., & Tadokoro, S. 2007. Electroactive<br />
Polymers robotics Applications: Artificial<br />
Muscles <strong>and</strong> Sensors. London: Springer-<br />
Verlag.<br />
Kim, K.J., Yim, W., Paquette, J.W., & Kim, D.<br />
2007. Ionic polymer-metal composites<br />
for underwater operation. J Intel Mat Syst<br />
Str. 18(2):123-31. doi: 10.1177/<br />
1045389X06063468.<br />
Krieg, M., & Mohseni, K. 2008. Thrust<br />
characterization of a bioinspired vortex ring<br />
thruster for locomotion of underwater<br />
robots. IEEE J Oceanic Eng. 33(2):123-32.<br />
doi: 10.1109/JOE.2008.920171.<br />
Krishen, K. 2009. Space applications for ionic<br />
polymer-metal composite sensors, actuators,<br />
<strong>and</strong> artificial muscles. Acta Astronaut.<br />
64(11-12):1160-6.<br />
Kruusamae, K., Brunetto, P., Graziani, S.,<br />
Punning, A., Di Pasquale, G., & Aabloo, A.<br />
2009. Self-sensing ionic polymer-metal<br />
composite actuating device with patterned<br />
surface electrodes. Polym Int. 59(3):300-4.<br />
doi: 10.1002/pi.2752.<br />
Lauder, G.V. 2007. How fish swim: Flexible<br />
fin thrusters as an EAP platform. In: Proceedings<br />
of the SPIE Smart Structures <strong>and</strong><br />
Materials: Electroactive Polymer Actuators<br />
<strong>and</strong> Devices (EAPAD). pp. 652402. San<br />
Diego, CA: SPIE.<br />
Lavu, B.C., Schoen, M.P., & Mahajan, A.<br />
2005. Adaptive intelligent control of ionic<br />
polymer metal composites. Smart Mater<br />
Struct. 14(4):466-74. doi: 10.1088/<br />
0964-1726/14/4/002.<br />
Lee, S.J., Han, M.J., Kim, S.J., Jho, J.Y., Lee,<br />
H.Y., & Kim, Y.H. 2006. A new fabrication<br />
method for ipmc actuators <strong>and</strong> application to<br />
artificial fingers. Smart Mater Struct. 15:<br />
1217-24. doi: 10.1088/0964-1726/15/5/008.<br />
Leo, D.J., Farinholt, K., & Wallmersperger, T.<br />
2005. Computational models of ionic<br />
transport <strong>and</strong> electromechanical transduction<br />
in ionomeric polymer transducers. In: Proceedings<br />
of the SPIE Smart Structures <strong>and</strong><br />
Materials: Electroactive Polymer Actuators<br />
<strong>and</strong> Devices (EAPAD). pp. 170-81.<br />
Menozzi, A., Leinhos, H.A., Beal, D.N., &<br />
B<strong>and</strong>yopadhyay, P.R. 2008. Open-loop control<br />
of a multifin biorobotic rigid underwater<br />
vehicle. IEEE J Oceanic Eng. 33(2):59-68.<br />
doi: 10.1109/JOE.2008.918687.<br />
Moheimani, S.O.R., & Fleming, A.J. 2006.<br />
Piezoelectric Transducers for Vibration Control<br />
<strong>and</strong> Damping (Advances in Industrial<br />
Control). London: Springer.<br />
Multiphysics, C. 2011. User’s Guide,<br />
version 4.1 a. COMSOL AB. Los Angeles, CA.<br />
Nam, J.D., Choi, H.R., Tak, Y.S., & Kim,<br />
K.J. 2003. Novel electroactive, silicate<br />
nanocomposites prepared to be used as<br />
actuators <strong>and</strong> artificial muscles. Sensor<br />
Actuat A-Phys. 105:83-90.<br />
Nemat-Nasser, S., & Jiang, Y.L. 2000.<br />
Electromechanical response of ionic<br />
polymer-metal composites. J Appl Phys.<br />
87(7):3321-31. doi: 10.1063/1.372343.<br />
Nguyen, V.K., Lee, J.W., & Yoo, Y. 2007.<br />
Characteristics <strong>and</strong> performance of ionic<br />
polymer-metal composite actuators based<br />
on nafion/layered silicate <strong>and</strong> nafion/silica<br />
nanocomposites. Sensor Actuat B. 120:<br />
529-37. doi: 10.1016/j.snb.2006.03.015.<br />
Pak, J.J., Kim, J., Oh, S.W., Son, J.H.,<br />
Cho, S.H., Lee, S.-K., … Kim, B. 2004.<br />
Fabrication of ionic-polymer-metal-composite<br />
(IPMC) micropump using a commercial<br />
nafion. In: Proceedings of the SPIE Smart<br />
Structures <strong>and</strong> Materials: Electroactive Polymer<br />
Actuators <strong>and</strong> Devices (EAPAD). pp. 272-80.<br />
San Diego, CA: SPIE.<br />
Porfiri, M. 2009. Influence of electrode<br />
surface roughness <strong>and</strong> steric effects on the<br />
nonlinear electromechanical behavior of<br />
ionic polymer metal composites. Phys Rev E.<br />
79(4):041503. doi: 10.1103/<br />
PhysRevE.79.041503.<br />
Pugal, D., Jung, K., Aabloo, A., & Kim,<br />
K.J. 2010a. Ionic polymer metal composite<br />
mechanoelectrical transduction: Review <strong>and</strong><br />
perspectives. Polym Int. 59(3):279-89.<br />
doi: 10.1002/pi.2759.<br />
Pugal, D., Kim, K.J., & Aabloo, A. 2010b.<br />
Modeling the transduction of IPMC in<br />
3D configurations. In: Proceedings of the<br />
SPIE Smart Structures <strong>and</strong> Materials <strong>and</strong><br />
Nondestructive Evaluation <strong>and</strong> Health<br />
Monitoring. pp. 76441T. San Diego, CA:<br />
SPIE.<br />
Pugal, D., Kim, K.J., Punning, A., Kasemagi,<br />
H., Kruusmaa, M., & Aabloo, A. 2008.<br />
A self-oscillating ionic polymer-metal composite<br />
bending actuator. J Appl Phys.<br />
103(8):084908-6. doi: 10.1063/1.2903478.<br />
Ramo, S. 1939. Currents induced by<br />
electron motion. Proc IRE. 27(9):584-5.<br />
doi: 10.1109/JRPROC.1939.228757.<br />
Richardson, R.C., Levesley, M.C., Brown,<br />
M.D., Hawkes, J.A., Watterson, K., &<br />
Walker, P.G. 2003. Control of ionic polymer<br />
metal composites. IEEE/ASME T Mech.<br />
8(2):245-3.<br />
July/August 2011 Volume 45 Number 4 97
Riddle, R.O., Jung, Y., Kim, S.-M., Song, S.,<br />
Stolpman, B., Kim, K.J., & Leang, K.K.<br />
2010. Sectored-electrode IPMC actuator for<br />
bending <strong>and</strong> twisting motion. In: Proceedings<br />
of the SPIE Smart Structures <strong>and</strong> Materials<br />
<strong>and</strong> Nondestructive Evaluation <strong>and</strong> Health<br />
Monitoring. San Diego, CA: SPIE.<br />
Shahinpoor, M., Bar-Cohen, Y., Simpson,<br />
J.O., & Smith, J. 1998. Ionic polymer-metal<br />
composites (IPMCs) as biomimetic sensors,<br />
actuators <strong>and</strong> artificial muscles—a review.<br />
Smart Mater Struct. 7(6):R15-30.<br />
doi: 10.1088/0964-1726/7/6/001.<br />
Shahinpoor, M., & Kim, K.J. 2001. Ionic<br />
polymer-metal composites: I. Fundamentals.<br />
Smart Mater Struct. 10:819-33.<br />
doi: 10.1088/0964-1726/10/4/327.<br />
Shahinpoor, M., & Kim, K.J. 2005. Ionic<br />
polymer-metal composites: IV. Industrial <strong>and</strong><br />
medical applications. Smart Mater Struct.<br />
14(1):197-214. doi: 10.1088/0964-<br />
1726/14/1/020.<br />
Shan, Y., & Leang, K.K. 2009. Frequencyweighted<br />
feedforward control for dynamic<br />
compensation in ionic polymer-metal<br />
composite actuators. Smart Mater Struct.<br />
18(12):125016. doi: 10.1088/0964-1726/18/<br />
12/125016.<br />
unmanned underwater vehicles. IEEE<br />
J Oceanic Eng. 32(3):533-50.<br />
doi: 10.1109/JOE.2007.903362.<br />
Tiwari, R., Kim, K.J., & Kim, S. 2008. Ionic<br />
polymer-metal composite as energy harvesters.<br />
Smart Struct Syst. 4(5):549-63.<br />
Tortora, G., Caccavaro, S., Valdastri, P.,<br />
Menciassi, A., & Dario, P. 2010. Design of an<br />
autonomous swimming miniature robot based<br />
on a novel concept of magnetic actuation.<br />
In: IEEE International Conference on<br />
Robotics <strong>and</strong> Automation. pp. 1592-7.<br />
Piscataway, NJ: IEEE. doi: 10.1109/<br />
ROBOT.2010.5509599.<br />
Yim, W., Lee, J., & Kim, K.J. 2007. An<br />
artificial muscle actuator for biomimetic<br />
underwater propulsors. Bioinspir Biomim.<br />
2:S31-41. doi: 10.1088/1748-3182/2/2/S04.<br />
Yoon, W.J., Reinhall, P.G., & Seibel, E.J.<br />
2007. Analysis of electro-active polymer<br />
bending: A component in a low cost<br />
ultrathin scanning endoscope. Sensor Actuat<br />
A-Phys. 133(2):506-17. doi: 10.1016/<br />
j.sna.2006.04.037.<br />
Shockley, W. 1938. Currents to conductors<br />
induced by a moving point charge. J Appl<br />
Phys. 9(10):635-6. doi: 10.1063/1.1710367.<br />
Smela, E., Kallenbach, M., & Holdenried, J.<br />
1999. Electrochemically driven polypyrrole<br />
bilayers for moving <strong>and</strong> positioning bulk<br />
micromachined silicon plates. IEEE/ASME<br />
T Microelectromech Syst. 8(4):373-83. doi:<br />
10.1109/84.809051.<br />
Tadokoro, S., Yamagami, S., & Takamori, T.<br />
2000. An actuator model of ICPF for robotic<br />
applications on the basis of physicochemical<br />
hypotheses. In: IEEE International Conference<br />
on Robotics <strong>and</strong> Automation. pp. 1340-6.<br />
Piscataway, NJ: IEEE.<br />
Tangorra, J.L., Davidson, S.N., Hunter,<br />
I.W., Madden, P.G.A., Lauder, G.V., Dong,<br />
H., … Mittal, R. 2007. The development<br />
of a biologically inspired propulsor for<br />
98 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
PAPER<br />
Batoid Fishes: Inspiration for the Next<br />
Generation of Underwater Robots<br />
AUTHORS<br />
Keith W. Moored<br />
Mechanical <strong>and</strong> Aerospace<br />
Engineering Department,<br />
Princeton University<br />
Frank E. Fish<br />
Department of Biology,<br />
University of West Chester<br />
Trevor H. Kemp<br />
Hilary Bart-Smith<br />
Mechanical <strong>and</strong> Aerospace<br />
Engineering Department,<br />
University of Virginia<br />
Introduction<br />
ABSTRACT<br />
For millions of years, aquatic species have utilized the principles of unsteady<br />
hydrodynamics for propulsion <strong>and</strong> maneuvering. They have evolved high-endurance<br />
swimming that can outperform current underwater vehicle technology in the areas<br />
of stealth, maneuverability <strong>and</strong> control authority. Batoid fishes, including the manta<br />
ray, Manta birostris, the cownose ray, Rhinoptera bonasus, <strong>and</strong> the Atlantic stingray,<br />
Dasyatis sabina, have been identified as a high-performing species due to their<br />
ability to migrate long distances, maneuver in spaces the size of their tip-to-tip<br />
wing span, produce enough thrust to leap out of the water, populate many underwater<br />
regions, <strong>and</strong> attain sustained swimming speeds of 2.8 m/s with low flapping/<br />
undulating frequencies. These characteristics make batoid fishes an ideal platform<br />
to emulate in the design of a bio-inspired autonomous underwater vehicle. The<br />
enlarged pectoral fins of each ray undergoes complex motions that couple spanwise<br />
curvature with a chordwise traveling wave to produce thrust <strong>and</strong> to maneuver. Researchers<br />
are investigating these amazing species to underst<strong>and</strong> the biological principles<br />
for locomotion. The continuum of swimming motions—from undulatory to<br />
oscillatory—demonstrates the range of capabilities, environments, <strong>and</strong> behaviors<br />
exhibited by these fishes. Direct comparisons between observed swimming motions<br />
<strong>and</strong> the underlying cartilage structure of the pectoral fin have been made. A<br />
simple yet powerful analytical model to describe the swimming motions of batoid<br />
fishes has been developed <strong>and</strong> is being used to quantify their hydrodynamic performance.<br />
This model is also being used as the design target for artificial pectoral<br />
fin design. Various strategies have been employed to replicate pectoral fin motion.<br />
Active tensegrity structures, electro-active polymers, <strong>and</strong> fluid muscles are three<br />
structure/actuator approaches that have successfully demonstrated pectoral-finlike<br />
motions. This paper explores these recent studies to underst<strong>and</strong> the relationship<br />
between form <strong>and</strong> swimming function of batoid fishes <strong>and</strong> describes attempts<br />
to emulate their abilities in the next generation of bio-inspired underwater vehicles.<br />
Keywords: biomimicry, bioinspired, autonomous underwater vehicle, manta ray,<br />
tensegrity structures<br />
There has been an explosion of activity<br />
in the area of biomimicry <strong>and</strong><br />
bioinspired engineering research. Biomimicry<br />
directly emulates the form<br />
<strong>and</strong> function of species to illuminate<br />
the physical principles behind nature’s<br />
designs. Bioinspired engineering takes<br />
advantage of the knowledge gained<br />
through biomimetic studies to judiciously<br />
apply novel physical principles<br />
to develop solutions with added functionality<br />
over conventional engineering<br />
approaches. Biology, biomimetics<br />
<strong>and</strong> bioinspired engineering are intimately<br />
linked. Thus it comes as no surprise<br />
that biologists <strong>and</strong> engineers are<br />
collaborating by developing biorobotic<br />
devices to: (1) elucidate key insights<br />
into biological form <strong>and</strong> function <strong>and</strong><br />
(2) develop bioinspired autonomous<br />
underwater vehicles (BAUVs) to improve<br />
functionality of autonomous<br />
underwater vehicles (AUVs).<br />
Aquatic species outperform conventional<br />
AUVs in the areas of maneuverability<br />
<strong>and</strong> control authority<br />
(B<strong>and</strong>yopadhyay, 2005), while having<br />
a low-noise signature that blends into<br />
the background <strong>and</strong> high swimming<br />
efficiencies. Batoid rays excel in all<br />
of these areas, giving them an abundance<br />
of recent attention. The focus<br />
of this paper is to present the growing<br />
body of work being done to underst<strong>and</strong><br />
<strong>and</strong> quantify the swimming performance<br />
of batoid fishes (i.e., skates,<br />
sting rays, manta rays) <strong>and</strong> the stateof-the-art<br />
in robotic mimicry. Of particular<br />
interest are the mechanisms<br />
associated with the swimming of<br />
these fishes, which employ flattened<br />
pectoral fins to propel <strong>and</strong> maneuver<br />
in the ocean <strong>and</strong> in rivers.<br />
Underst<strong>and</strong>ing Biological Foundation<br />
will discuss our current biological<br />
July/August 2011 Volume 45 Number 4 99
underst<strong>and</strong>ing of batoid rays. Rationale<br />
for Mimicking Batoid Rays will<br />
present compelling reasons for scientists<br />
<strong>and</strong> engineers to study batoids<br />
rays, highlighting their swimming<br />
characteristics that would be desired<br />
in an underwater vehicle. Bioinspired<br />
Robotics delves into the exp<strong>and</strong>ing<br />
worldofbio-inspiredunderwatervehicles,<br />
with particular emphasis on<br />
ray-like platforms. Concluding Comments<br />
<strong>and</strong> Future Directions concludes<br />
with a discussion on critical areas that<br />
need to be addressed in order for the<br />
next generation of underwater vehicles<br />
to be truly bioinspired.<br />
Underst<strong>and</strong>ing Biological<br />
Foundation<br />
Fish swim by imparting momentum<br />
to water from the movements<br />
of a variety of propulsors, which can<br />
include the body, median fins, <strong>and</strong><br />
paired fins (Sfakiotakis et al., 1999).<br />
Although primitive batoid fishes use<br />
the body <strong>and</strong> caudal fin toswim,<br />
more advanced batoids have become<br />
specialized to swim with enlarged pectoral<br />
fins. It is emphasized that pectoral<br />
fin locomotion can have significant<br />
advantages in maneuvering <strong>and</strong> stationkeeping<br />
(Sfakiotakis et al., 1999). In<br />
recent years, more attention is being<br />
paid to pectoral fin hydrodynamics as<br />
their importance is being realized in not<br />
only steady-state locomotion but also in<br />
transient maneuvers (B<strong>and</strong>yopadhyay,<br />
FIGURE 1<br />
2005; Lauder et al., 2002; Combes &<br />
Daniel, 2001; Palmisano et al., 2007).<br />
Batoid rays take pectoral fin locomotion<br />
to an evolutionary extreme<br />
(Figure 1). Rays have a dorso-ventrally<br />
flattened body with enlarged pectoral<br />
fins that are seamlessly merged with<br />
their body to form a biological blended<br />
wing-body configuration. Propulsive<br />
waves are passed through the fins by<br />
serial contraction of the appendicular<br />
musculature. The waves have their<br />
greatest amplitude toward the periphery<br />
of the fin.<br />
Even though among rays there is<br />
similar morphology, their locomotor<br />
strategies can be very different. Undulatory<br />
motion, defined as having<br />
greater than one or more waves present<br />
on a fin (Rosenberger, 2001), is one<br />
extreme of kinematic motion <strong>and</strong> was<br />
termed ‘rajiform’ by Breder (1926).<br />
These fishes swim just over the ocean<br />
floor. The other extreme is oscillatory<br />
motion, defined as having less than<br />
half of a wave present (flapping) on a<br />
fin, <strong>and</strong> was coined ‘mobuliform’ by<br />
Webb (1994). The mobuliform swimming<br />
mode appears as a wing-like flapping<br />
motion <strong>and</strong> is associated with rays<br />
that have a more pelagic existence. The<br />
various species of batoids exhibit a<br />
continuum of kinematic motions between<br />
the two extremes of undulation<br />
<strong>and</strong> oscillation (Rosenberger, 2001).<br />
Myliobatoids, i.e., the mid-water<br />
rays, including manta, eagle, bat, <strong>and</strong><br />
(a) Image of a manta ray. (b) Image of a cownose ray. Both rays are part of the batoid family <strong>and</strong><br />
swim via an oscillatory motion.<br />
cownose rays, nearly exclusively utilize<br />
oscillatory motion (Klausewitz, 1964;<br />
Sasko et al., 2006; Heine, 1992).<br />
Some research has been done to characterize<br />
the biology <strong>and</strong> behavior of<br />
myliobatoids (Schaefer & Summers,<br />
2005; Summers, 2000). Heine (1992)<br />
studied the kinematics of the cownose<br />
ray by videotaping live rays swimming<br />
in a flow tank. Rosenberger (2001)<br />
compared the kinematics of many<br />
batoid rays spanning the undulationoscillation<br />
continuum <strong>and</strong> suggested<br />
that oscillatory rays have evolved to<br />
have efficient locomotion. Klausewitz<br />
(1964) describes the kinematic motions<br />
of the manta ray while Moored<br />
et al. (Moored, 2010; Moored et al.,<br />
2011b) developed a simple yet powerful<br />
analytical model to quantify the kinematics<br />
of different species of batoid<br />
rays (such as the manta ray, Atlantic<br />
stingray, <strong>and</strong> the cownose ray). This<br />
model is used as a target deformation<br />
field for a bio-inspired fin (Moored<br />
et al., 2011b) <strong>and</strong> to calculate the<br />
swimming performance of different<br />
batoid ray species (Moored et al.,<br />
2011b; Pederzani et al., 2011).<br />
Morphometrics<br />
The greatly enlarged pectoral fins<br />
form wide lateral extensions of the<br />
body that range in morphology from<br />
a circular disc to triangular, wing-like<br />
planforms. Species of batoids show<br />
overa90-foldrangeinsizewiththe<br />
largest being the manta (Manta birostris).<br />
Rays that swim by undulations<br />
of the fin intherajiformmodehave<br />
fin shapes with relatively low aspect<br />
ratios (the ratio of span to chord). Oscillatory<br />
swimmers, using the mobuliform<br />
mode, possess higher aspect ratio<br />
fins with longer spans.<br />
The cross-sectional geometry of<br />
batoid rays has a streamlined appearance.<br />
Rajiform swimmers have a<br />
100 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
ody <strong>and</strong> pectoral fins with a flattened<br />
ventral side <strong>and</strong> low vaulted dorsum,<br />
giving a design similar to a cambered<br />
wing. Although the central portion of<br />
the body shows a slight asymmetry<br />
with a flattened ventral surface <strong>and</strong><br />
convex dorsal surface, the pectoral<br />
fins of mobuliform swimmers display<br />
symmetrical cross-sectional profiles<br />
reminiscent of engineered foils<br />
(Abbott & von Doenhoff, 1949).<br />
The internal skeleton of the pectoral<br />
fins of batoids are composed of numerous<br />
short, cylindrical cartilaginous<br />
elements (Heine, 1992; Schaefer &<br />
Summers, 2005). These cartilaginous<br />
elements are the supportive radials of<br />
the fin. The radials are stacked end to<br />
end. The radial cartilages are mineralized<br />
to varying degree depending on<br />
the species of batoid, where the mineralization<br />
is found on the exterior<br />
of the cartilaginous element with the<br />
core being unmineralized (Schaefer &<br />
Summers, 2005). Rajiform swimming<br />
rays display joint staggering with little<br />
calcification of the joints, whereas<br />
the skeleton of oscillatory swimmers<br />
shows cross-bracing <strong>and</strong> calcification.<br />
The skeleton is moved by long thin<br />
muscles that run from the exp<strong>and</strong>ed<br />
pectoral girdle along each fin rayto<br />
every radial. The range of motion of<br />
the articulated radials is small (∼15°),<br />
but the large number of components<br />
in the pectoral fins permits sufficient<br />
spanwise <strong>and</strong> chordwise flexibility<br />
for propulsion <strong>and</strong> maneuvering<br />
(Rosenberger, 2001; Klausewitz,<br />
1964; Heine, 1992; Schaefer &<br />
Summers, 2005).<br />
Rationale for Mimicking<br />
Batoid Rays<br />
With respect to pursuing bioinspired<br />
engineering, a key question<br />
to be answered is to explain/justify<br />
why a particular species is a good c<strong>and</strong>idate<br />
to emulate. These reasons can<br />
be very diverse <strong>and</strong> are motivated by<br />
the particular application envisioned.<br />
In the case of AUVs, compelling reasons<br />
to consider biology as a starting<br />
point for the development of the next<br />
generation vehicle are (1) a stealthy<br />
signature, (2) high efficiency <strong>and</strong> economy,<br />
(3) exp<strong>and</strong>ed working environment,<br />
<strong>and</strong> (4) scalability/payload<br />
capacity. Additionally, a key justification<br />
for this approach is that there are<br />
tangible improvements that can be<br />
made over current AUV technologies.<br />
The pool of species to emulate is vast.<br />
However, recent studies of batoid rays<br />
have demonstrated significant swimming<br />
abilities that would be desirable<br />
in an underwater vehicle.<br />
Stealth means either quiet operation<br />
or the ability to blend into the<br />
background noise. A biomimetic approach<br />
naturally fulfills these requirements<br />
by creating a vehicle whose<br />
minimal noise signature blends in<br />
with the environment. The noise signature<br />
of a fish is very different to<br />
that of a propeller (B<strong>and</strong>yopadhyay,<br />
2005). Even biomimetic sensor arrays<br />
such as an artificial lateral line (Yang<br />
et al., 2006) or artificial seal whiskers<br />
(Stocking et al., 2010) that are sensitive<br />
to hydrodynamic wake signatures<br />
would presumably delineate a flapping<br />
fin wake as an animal <strong>and</strong> a propeller<br />
wake as a man-made device. Given recent<br />
advances in underwater imaging<br />
technology including LIDAR systems<br />
(Jaffe et al., 2001), synthetic aperture<br />
sonar (Kocak & Caimi, 2005) <strong>and</strong><br />
biomimetic sonar systems (Dobbins,<br />
2007), the shape <strong>and</strong> movements of<br />
an AUV are becoming increasingly important.<br />
By mimicking the body form<br />
of an aquatic species, the identification<br />
of such a stealthy vehicle as a manmade<br />
device becomes difficult.<br />
Batoid rays offer an intriguing design<br />
solution for a high-endurance vehicle.<br />
Pelagic rays, such as the manta<br />
ray or cownose ray, migrate thous<strong>and</strong>s<br />
of miles a year. This suggests that these<br />
species have evolved to become highendurance<br />
swimmers. As discussed<br />
previously, myliobatoid rays have a<br />
dorso-ventrally flattened body with<br />
enlarged pectoral fins, forming a natural<br />
gliding morphology. In terms of<br />
avehicle,abatoid-inspiredUVisan<br />
advance on current underwater glider<br />
technology. This bioinspired platform<br />
wouldenableavehicletohavehighendurance<br />
capabilities like current underwater<br />
gliders, such as the Slocum<br />
AUV (Webb & Simonetti, 1999). Additionally,<br />
this system has the potential<br />
to transition to a faster, more maneuverable<br />
vehicle that can operate in dynamic<br />
environments such as the<br />
littoral zone or areas with large currents<br />
<strong>and</strong> high wave action. Observations<br />
of various rays show them to be<br />
highly maneuverable <strong>and</strong> adaptable<br />
to local conditions—for example, to<br />
station keep <strong>and</strong> even swim backwards.<br />
Their ability to control their<br />
stability via the pectoral fins, especially<br />
when compensating for challenging<br />
environments, must also be considered<br />
as a desirable characteristic to emulate<br />
in an underwater vehicle.<br />
Scalability is an attractive feature<br />
in any artificial system. With respect<br />
to bioinspired underwater vehicles, batoids<br />
display an extraordinary range of<br />
dimensions, growing in excess of<br />
9 m tip-to-tip in the case of manta<br />
rays.Thus,thesize<strong>and</strong>speedthat<br />
batoids perform at are equivalent to<br />
the operation range of marine vehicles.<br />
The size of the vehicle will very much<br />
depend on the mission requirements,<br />
but by using the batoid as the foundation,<br />
it is feasible to produce a variety<br />
of sized vehicles that can explore <strong>and</strong><br />
July/August 2011 Volume 45 Number 4 101
traverse a wide range of ocean space<br />
while performing a wide range of tasks.<br />
Moreover, a batoid-inspired vehicle<br />
would have a large planform surface<br />
area, making this platform an excellent<br />
c<strong>and</strong>idate for flexible solar cells to extend<br />
its range (Dennler et al., 2008),<br />
similar to the solar powered SAUV II<br />
vehicle (Jalbert et al., 2003). In addition,<br />
the rigid body of batoids permits<br />
space for control systems, sensory<br />
devices <strong>and</strong> increased payload.<br />
Bioinspired Robotics<br />
There has been growing use of<br />
AUVs in recent years with over<br />
240 different AUV platforms developed<br />
<strong>and</strong> used in the field (B<strong>and</strong>yopadhyay,<br />
2005). These AUVs typically are built<br />
for reconnaissance/surveying <strong>and</strong> were<br />
originally designed for endurance<br />
(Blidberg, 2001). This gave rise to<br />
the design of underwater gliders<br />
using conventional design principles<br />
(i.e., steady-state hydrodynamics)<br />
that have high endurance but little maneuverability<br />
(Webb & Simonetti,<br />
1999). From another perspective,<br />
biology has created thous<strong>and</strong>s of swimming<br />
platforms that can outmaneuver<br />
the best AUVs while still<br />
having highly efficient, high-speed<br />
<strong>and</strong> high-endurance performance.<br />
Moreover, many of these biological<br />
systems can also hover in place with no<br />
forward locomotion, generate large<br />
enough forces to hold station under<br />
adverse environmental conditions,<br />
burst with incredible acceleration <strong>and</strong><br />
have a significantly reduced noise signature<br />
compared to man-made AUVs<br />
(Fish & Lauder, 2006). In an attempt<br />
to bridge the performance gap between<br />
conventional AUVs <strong>and</strong> biological<br />
systems, engineers have been<br />
shifting focus to BAUVs, which is a<br />
highly multi-disciplinary research area<br />
(B<strong>and</strong>yopadhyay, 2005; Colgate &<br />
Lynch, 2004). To reach some of these<br />
goals, there is a spectrum of first generation<br />
BAUVs that have been developed.<br />
Some form an exotic collection<br />
mimicking lamprey (Ayers et al.,<br />
2000; Crespi et al., 2004), tuna<br />
(Barrett et al., 1996; Yu et al., 2004;<br />
Anderson & Chhabra, 2002), <strong>and</strong> dolphins<br />
(Yu et al., 2007), while others<br />
aremoreconventionalstyleAUV<br />
designs outfitted with bioinspired flapping<br />
propulsors (Fish et al., 2003; Low<br />
&Willy,2006;Listaketal.,2005;<br />
Borgen et al., 2003; Mojarrad, 2000;<br />
Licht et al., 2004). These different<br />
BAUV designs were made possible<br />
partly from advances in our underst<strong>and</strong>ing<br />
of unsteady hydrodynamics<br />
<strong>and</strong> the biology of nektonic (swimming)<br />
organisms. However, this BAUV<br />
technology still has a long way to go<br />
before the performance gap is bridged.<br />
Recently, researchers have turned<br />
to pectoral fin locomotion for inspiration.<br />
Pectoral fin motions utilized by<br />
sunfish, perch, bass <strong>and</strong> bird wrasse<br />
for low-speed swimming <strong>and</strong> maneuvering<br />
have been studied (Gibb et al.,<br />
1994; Drucker & Jensen, 1997;<br />
Lauder & Jayne, 1996; Walker &<br />
Westneat, 1997). To underst<strong>and</strong> the<br />
forces <strong>and</strong> moments produced by<br />
these pectoral fins, biorobotic solutions<br />
began with paddle-like fins that<br />
mimic the bulk kinematics of labriform<br />
swimming (Kato, 1998; Kato &<br />
FIGURE 2<br />
A biomimetic sunfish pectoral fin (Tangorra et al., 2008a, 2008b).<br />
Furushima, 2002). In recent years, devices<br />
have been constructed to more<br />
accurately replicate the kinematics<br />
utilized by the fish with the advent of<br />
actively flexible fins that can produce<br />
chordwise undulations as well as<br />
spanwise curvature (Yan et al., 2010;<br />
Palmisano et al., 2008; Kato et al.,<br />
2008; Tangorra et al., 2007). Actively<br />
flexible fins deform due to the presence<br />
of actuators instead of undergoing<br />
rigid body motions, like the heave<br />
<strong>and</strong> pitch of oscillating airfoils. The<br />
motion of actively flexible fins is fully<br />
prescribed. In contrast, passively flexible<br />
fins deform under fluid loading<br />
such that their motion is not fully prescribed,<br />
but a function of the forces applied<br />
to the fin. Tangorra et al. (2008a,<br />
2008b) advanced their artificial pectoral<br />
fin (Figure 2) by not only matching<br />
the kinematics of the sunfish but also<br />
replicating the internal fin structure<br />
<strong>and</strong> material properties, which allowed<br />
the fin to have a greater degree of passive<br />
flexibility. This fin wasusedto<br />
fully characterize how sunfish produce<br />
<strong>and</strong> manipulate fluid forces to propel<br />
themselves <strong>and</strong> maneuver. With<br />
equivalent passive flexibility as the<br />
fins of the sunfish, the artificial fin<br />
was able to produce thrust on both<br />
the outstroke <strong>and</strong> instroke of its fin<br />
beat, as observed of the animal.<br />
An excellent example demonstrating<br />
the link between biology, biomimetics,<br />
<strong>and</strong> bioinspired engineering<br />
102 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
is in the development of an artificial<br />
ghost knifefish (Curet et al., 2010).<br />
Through observation of biology, a<br />
biorobotic device was developed <strong>and</strong><br />
used to underst<strong>and</strong> the locomotion<br />
strategies of this fish. Particle image<br />
velocimetry, in conjunction with computational<br />
fluid dynamics, were employed<br />
to explore the propulsive <strong>and</strong><br />
station-keeping characteristics of this<br />
fascinating fish.<br />
Batoid-Inspired Devices<br />
There have also been attempts by<br />
researchers to develop batoid-inspired<br />
fins <strong>and</strong> AUVs. These devices mimic<br />
both the undulatory swimming seen<br />
inbenthicrays(similartothelocomotion<br />
of the ghost knifefish) as well<br />
as the oscillatory swimming seen in pelagic<br />
rays, such as the manta. Many of<br />
these batoid-inspired devices are used<br />
as a platform for exploring actuation<br />
technologies.<br />
Motors <strong>and</strong> servomotors are used<br />
in ray-like devices due to their simple<br />
controllability, high-speed operation<br />
<strong>and</strong> repeatability. Some motor-driven<br />
devices mimic undulatory rays (Low<br />
& Willy, 2006; V. y Alvarado et al.,<br />
2010), while others mimic oscillatory<br />
rays (Yang et al., 2009; Zhou &<br />
Low, 2010; Gao et al., 2007). Researchers<br />
have developed oscillatory<br />
ray-like vehicles based on pneumatic<br />
pectoral fins (Brower, 2006; Cai<br />
et al., 2010; Suzumori et al., 2007).<br />
Sfakiotakis et al. (2005) also used<br />
pneumatically driven “fin rays” to produce<br />
an undulatory ray-like device.<br />
Festo (2008) has built a BAUV called<br />
AquaRay, utilizing fluidic muscles.<br />
This robot uses an oscillatory flapping<br />
motion to swim, but no quantitative<br />
data on the performance is given.<br />
Takagi et al. (2007) utilized ionic<br />
polymer-metal composites (IPMCs)<br />
actuators to develop a stingray-like<br />
device that could achieve a swimming<br />
speed of 0.24 BL/s. Chen et al. (2011)<br />
developed a novel fabrication method<br />
to produce IPMCs that can deform<br />
with complex three-dimensional kinematics.<br />
This fabrication technology<br />
was used to produce a manta ray-like<br />
device. Shape memory alloys have<br />
also been employed in the design of artificial<br />
pectoral fins (Yong-hua et al.,<br />
2007; Wang et al., 2008). Wang<br />
et al. (2008) presented a robotic squid<br />
utilizing a rajiform mode of swimming<br />
to achieve 0.24 BL/s swimming<br />
speed. The best swimming speed performance<br />
of these actuator platforms<br />
was 1.4 BL/s achieved by the servomotor<br />
driven devices; however, the<br />
associated power cost is not given.<br />
These studies showcase the plethora<br />
of actuator technologies that can<br />
be utilized to produce deformations<br />
similar to that of rays. One concern<br />
with this approach is that the actuator<br />
choice is directly coupled with the fin<br />
technology. An alternative approach is<br />
to start with a fin design that is actuator<br />
independent <strong>and</strong> so the choice of<br />
actuator is dependent on the application<br />
of the vehicle. For instance, if actuator<br />
efficiency is not a concern but<br />
noiseless operation is of prime importance,<br />
an SMA actuator could be chosen.<br />
Also, this approach opens up the<br />
possibility of replacing current actuators<br />
with new technologies that may<br />
be superior. Solutions like this are discussed<br />
next.<br />
In a study to increase our underst<strong>and</strong>ing<br />
of the hydrodynamics of<br />
batoid locomotion, Clark <strong>and</strong> Smits<br />
(2006) designed <strong>and</strong> built an active<br />
artificial oscillating fin that was independent<br />
of the actuator. They quantified<br />
the performance of the fin by<br />
measuring the efficiency <strong>and</strong> thrust<br />
production, as well as determining an<br />
optimal traveling wave wavelength.<br />
Furthermore, by using dye flow visualization,<br />
they characterized the wake<br />
structure as a series of interacting trailing<br />
edge vortices forming a threedimensional<br />
reverse von Kármán<br />
vortex street. In free swimming tests<br />
(Moored et al., 2011a), a swimming<br />
speed of 2 BL/s <strong>and</strong> a swimming economy,<br />
ζ (ζ = U = P f ,whereU is the<br />
swimming speed <strong>and</strong> P f is the power<br />
consumption), of 0:132 BL/J was<br />
reached for an actively flexible single<br />
fin. When some passive flexibility<br />
was introduced, the swimming speed<br />
dropped to 1:7 BL/s while the economy<br />
rose to 0:18 BL/J at the same flapping<br />
frequency of 2 Hz. This work<br />
has also highlighted the prime importance<br />
of the traveling wave in ray-like<br />
propulsion.<br />
Another actuator-independent approach<br />
has been developed using active<br />
tensegrity structures. Tensegrity structures<br />
are truss-like structures where<br />
some of the rigid elements have been<br />
replaced by cable elements (Figure 3).<br />
The cable elements must be in a<br />
state of tension for the structure to<br />
have integrity, giving rise to the contraction<br />
of “tensional-integrity” to<br />
tensegrity. Tensegrity structures act<br />
as a “skeleton-tendon” foundation<br />
that can use any actuator type to support<br />
the generation of large loads,<br />
match the kinematics of batoid rays,<br />
<strong>and</strong> perform with minimal actuation<br />
energy (Moored et al., 2011b; Moored<br />
& Bart-Smith, 2009).<br />
Various tensegrity actuation strategies<br />
are explored that are capable of<br />
matching the key kinematic features<br />
of batoid-propulsion: a chordwise<br />
traveling wave coupled with a large<br />
amplitude curved spanwise deformation.<br />
The strategies involve either<br />
embedding the actuators into the tensegrity<br />
structure (embedded actuation)<br />
or migrating the actuators outside of<br />
July/August 2011 Volume 45 Number 4 103
FIGURE 3<br />
(a) Three-dimensional tensegrity structures (three, four, <strong>and</strong> six strut prismatic structures). (b) Tensegrity-based fin concept. The fin can deform<br />
with coupled curved spanwise motion <strong>and</strong> chordwise undulation to mimic the kinematics of the manta ray <strong>and</strong> the batoid family in general. The<br />
tensegrity deforms when active elements contract or exp<strong>and</strong>.<br />
the structure (remote actuation). With<br />
respect to embedded actuation, optimal<br />
solutions have been calculated<br />
that give the location <strong>and</strong> actuation<br />
strain necessary to match a target displacement<br />
field (Moored & Bart-<br />
Smith, 2007). However, embedded<br />
actuation is problematic, as it requires<br />
many actuators to match the complex<br />
ray kinematics, adds mass to the<br />
active structure (thereby requiring<br />
more power to flap) <strong>and</strong> limits the<br />
scalability of solutions to the size of<br />
the actuator. Remote actuation overcomes<br />
these limitations by placing<br />
the actuators outside of the active region<br />
<strong>and</strong> connecting to the structure<br />
via a routed cable. A general numerical<br />
model––applicable to any topology<br />
<strong>and</strong> any actuation strategy––has<br />
been derived (Moored & Bart-Smith,<br />
2009).<br />
Moored et al. (2011) derive analytical<br />
solutions for active planar tensegrity<br />
beam structures. These solutions coupled<br />
with the numerical solution are<br />
utilized to identify optimal stiffnessto-mass<br />
<strong>and</strong> strength-to-mass strategies.<br />
Structural performance metrics<br />
were calculated showing that the fin<br />
structure can closely match the kinematics<br />
of the manta ray, under external<br />
loading, using open-loop actuation of<br />
four actuators remotely located outside<br />
of the active structure (Moored et al.,<br />
2011b). In an attempt to simplify the<br />
experimental design of an artificial<br />
fin, actuated via remote actuation, a<br />
single tensegrity beam was built<br />
<strong>and</strong>placedwithinanelastomerfin.<br />
Figure 4a shows images of a single<br />
tensegrity beam as it is actuated. The<br />
beam enables leading-edge actuation<br />
of the artificial fin (Figure4b).This<br />
fin was then tested in a flow tank to<br />
observe the influence of frequency on<br />
the wake topology. Figures 4c <strong>and</strong> 4d<br />
show the actuating fin inwaterfrom<br />
the side <strong>and</strong> below. The black lines<br />
superimposed on these images represent<br />
the kinematical model for the kinematics<br />
of a cownose ray—note the<br />
excellent agreement between experiment<br />
<strong>and</strong> theory, especially with the<br />
passive response of the elastomer fin.<br />
This approach costs minimal power<br />
consumption <strong>and</strong> shows the simple design<br />
of a high-performance tensegritybased<br />
artificial pectoral fin.<br />
Concluding Comments<br />
<strong>and</strong> Future Directions<br />
The idea to look to nature for inspiration<br />
is not new, <strong>and</strong> this rich arena<br />
104 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 4<br />
Example of a tensegrity-based actuating fin. (a) Photos of a tensegrity beams employing remote<br />
actuation. (b) Dorsal view of elastomer fin with leading edge actuation via tensegrity beam.<br />
(c) Posterior view of actuating fin shown in (b). (d) Lateral view of actuating fin. Note the lines<br />
in (c) <strong>and</strong> (d) represent the mathematical model derived to describe kinematic motions of a manta<br />
ray pectoral fin.<br />
continues to be a source for engineers<br />
to aid in solving challenging problems.<br />
From Leonardo Da Vinci’s flying machine,<br />
Helical Air Screw, to leading<br />
edge whale-like tubercles on wind turbine<br />
blades to improve efficiencies<br />
(Fish et al., 2011). The opportunities<br />
to learn from nature <strong>and</strong> emulate its<br />
unique approach to overcoming challenges<br />
seem endless. In this paper, we<br />
have touched upon the challenge of<br />
creating efficient, economic, <strong>and</strong> maneuverable<br />
underwater swimming<br />
platforms. We have focused on batoid<br />
fishes for inspiration in the design of<br />
the next generation of bioinspired underwater<br />
vehicles, as emulation of its<br />
swimming characteristics—efficiency,<br />
maneuverability, stealth, working<br />
environments, scalability—has the<br />
potential to significantly improve<br />
upon current state-of-the-art in AUV<br />
technologies.<br />
The development of a batoidinspired<br />
underwater vehicle can be<br />
classified in terms of the approach<br />
taken to achieve ray-like swimming.<br />
The first is developing a batoidinspired<br />
AUV that performs as a platform<br />
to test the capabilities of a variety<br />
of traditional <strong>and</strong> novel actuating technologies.<br />
The motivation here is to<br />
demonstrate the capabilities of such<br />
devices—usually in terms of force<br />
<strong>and</strong> stroke—<strong>and</strong> is not necessarily a<br />
desire to truly replicate the biological<br />
system. These actuators include<br />
electroactive polymers, fluidic muscles,<br />
shape memory alloys, motors<br />
<strong>and</strong> servomotors. Of particular interest<br />
in testing these actuators has been the<br />
challenge of quantifying the swimming<br />
performance of the particular<br />
vehicle, many of which do not necessarily<br />
mimic the kinematics of batoid<br />
fishes. As biology has limitations due<br />
to the materials available to construct<br />
abody<strong>and</strong>theevolutionaryprocess<br />
that produces an organism, possible<br />
improvements to the basic body plan<br />
can perhaps be engineered to enhance<br />
performance beyond the capabilities of<br />
nature.<br />
The second approach considers fundamental<br />
questions associated with<br />
biology’s solutions to propulsion,<br />
maneuverability, stability, <strong>and</strong> stealth.<br />
<strong>Technology</strong> is used in this case to replicate<br />
the biology to help answer these<br />
questions. Using the underlying biology<br />
as the basis for inquiry, the<br />
mechanisms that dictate batoid swimming<br />
performance are explored. This is<br />
being done through the design <strong>and</strong> development<br />
of artificial systems—either<br />
real or virtual—that can achieve near<br />
identical kinematic motions of the rays<br />
being studied. Specifically, researchers<br />
are working to elucidate the dominant<br />
mechanisms in batoid swimming that<br />
dictate efficiency <strong>and</strong> maneuverability.<br />
A key outcome of this work is to fully<br />
explore nature’s design space <strong>and</strong><br />
beyond. By mimicking biology, we<br />
attempt to elucidate the key features<br />
that control <strong>and</strong> optimize function.<br />
Nature evolves solutions that satisfy<br />
multiple constraints; engineers <strong>and</strong><br />
scientists can design for a single desired<br />
outcome. By identifying <strong>and</strong> quantifying<br />
the key features/characteristics that<br />
dictate optimality, we can judiciously<br />
choose to build these into an artificial<br />
system, depending on the required<br />
functionality of the device. For example,<br />
one may desire a vehicle that<br />
can swim for as long as possible or<br />
as fast as possible—two different solutions<br />
may be necessary for these two<br />
requirements.<br />
As mentioned in Underst<strong>and</strong>ing<br />
Biological Foundation, there has<br />
been an extensive study of the underlying<br />
cartilage structure of various<br />
July/August 2011 Volume 45 Number 4 105
FIGURE 5<br />
batoid rays (Schaefer & Summers,<br />
2005). Biomechanical studies of the<br />
cartilage arrangement have been carried<br />
out to examine the relationship<br />
between the form of the underlying<br />
structure <strong>and</strong> its impact on the function<br />
(Russo et al., 2011). In this<br />
work, Russo et al. have taken the<br />
cartilage architecture <strong>and</strong> developed<br />
a numerical model to study the kinematic<br />
function of this form. This<br />
initial study is beginning to explore<br />
the relationship between form <strong>and</strong><br />
function.<br />
One of the most exciting developments<br />
in the creation of these bioinspired<br />
devices <strong>and</strong> vehicles is in the<br />
development of rapid prototyping fabrication<br />
(Figure 5). This has opened<br />
the possibility to build a cartilage<br />
structure that uses the same design<br />
principles observed in nature, as described<br />
by Schaefer & Summers<br />
(2005). In this physical model, cartilage<br />
elements are connected in the<br />
spanwise direction with cross-bracing<br />
in the chordwise direction to mimic<br />
the architecture of the Atlantic stingray<br />
(www.bartsmithlabs.com). This<br />
technology enables the design to be<br />
quickly <strong>and</strong> easily varied so as to answer<br />
questions regarding the influence<br />
of the architecture on kinematic<br />
performance. The images in Figure 5<br />
are compelling, as they demonstrate<br />
kinematic capabilities <strong>and</strong> possibilities<br />
of such a structure. By mimicking<br />
the underlying structure of biology,<br />
we can explore the capabilities of<br />
these species <strong>and</strong> potentially exp<strong>and</strong><br />
upon them.<br />
Significant progress has been made,<br />
but there is still much to be done. Information<br />
on the kinematics of swimming<br />
is being generated, but there is<br />
still much to be learned, especially<br />
with respect to some of the more fine<br />
motor skills observed. Also, not much<br />
Example of artificial cartilage structure design using rapid-prototyping technology. The individual<br />
elements represent the cartilage elements found in the pectoral fins of batoid rays. The arrangement<br />
<strong>and</strong> connectivity are similar to portions found in the Atlantic stingray.<br />
is known about the hydroacoustic<br />
properties of the biology. Material<br />
properties of the constituent parts of<br />
the pectoral fin are needed to improve<br />
the fidelity biomechanical models that<br />
describe form <strong>and</strong> function. Lastly,<br />
more investigation of the sensing <strong>and</strong><br />
control strategies of batoids is needed.<br />
This improved underst<strong>and</strong>ing will<br />
providevaluableinsightwhenmore<br />
sophisticated vehicles are developed.<br />
With regard to engineering a batoidinspired<br />
AUV, there are huge opportunities<br />
in actuator design <strong>and</strong> development.<br />
Structural <strong>and</strong> material design <strong>and</strong> selection<br />
also are areas that need to be<br />
addressed. For example, how do we design<br />
a skin that can accommodate both<br />
the out-of-place hydrodynamic forces<br />
<strong>and</strong> the potential in-plane stretching<br />
experienced during actuation. How<br />
do the properties of the artificial system<br />
scale with the biological properties<br />
Actuation technology is an area<br />
that has the potential to revolutionize<br />
the field of biomimicry <strong>and</strong> bioinspired<br />
engineering.<br />
In this paper, we have focused on<br />
a small subset of bio-inspired underwater<br />
vehicles. We have presented a<br />
review of work related to the development<br />
of a bio-inspired underwater<br />
robot—actuation technology integrated<br />
into a batoid-like vehicle <strong>and</strong><br />
underst<strong>and</strong>ing the biological foundation<br />
to explore the full design space.<br />
It is clear though that these two categories<br />
are very closely related. Without<br />
actuator development, it may not be<br />
possible to achieve anything close to<br />
what biology achieves. But a clear picture<br />
of biology function is needed<br />
so that actuator requirements can be<br />
quantified. Synergy between biology,<br />
biomimicry, <strong>and</strong> bio-inspired engineering<br />
is essential if we want to<br />
develop the next generation of underwater<br />
vehicles.<br />
106 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
Acknowledgments<br />
The authors would like to acknowledge<br />
funding from the Office of Naval<br />
Research through the MURI program<br />
on Biologically Inspired Autonomous<br />
Sea Vehicles (Program Manager: Dr.<br />
R. Brizzolara, Contract No. N00014-<br />
08-1-0642) <strong>and</strong> the David <strong>and</strong> Lucille<br />
Packard Foundation.<br />
References<br />
Abbott, I.H., & von Doenhoff, A.E. 1949.<br />
Theory of Wing Sections: Including a<br />
Summary of Airfoil Data. New York:<br />
McGraw-Hill.<br />
Ayers, J., Wilbur, C., & Olcott, C. 2000.<br />
Lamprey robots. In: Proceedings of 1st International<br />
Symposium on Aqua Bio-mechanisms,<br />
eds. Wu, T., & Kato, N. Honolulu, Hawaii:<br />
Tokai University Pacific Center.<br />
Anderson, J.M., & Chhabra, N.K. 2002.<br />
Maneuvering <strong>and</strong> stability performance of<br />
a robotic tuna. Integr Comp Biol. 42(1):<br />
118-26. doi: 10.1093/42.1.118.<br />
B<strong>and</strong>yopadhyay, P.R. 2005. Trends in biorobotic<br />
autonomous undersea vehicles.<br />
IEEE J Oceanic Eng. 30(1):109-139.<br />
doi: 10.1109/JOE.2005.843748.<br />
Barrett, D., Grosenbaugh, M., & Triantafyllou,<br />
M. 1996. The optimal control of a flexible<br />
hull robotic undersea vehicle propelled by<br />
an oscillating foil. In: Proceedings of the Symposium<br />
on Autonomous Underwater Vehicle<br />
<strong>Technology</strong>, AUV’96. 1-9. Monterey, CA:<br />
IEEE. doi: 10.1109/AUV.1996.532833.<br />
Blidberg, D.R. 2001. The development of<br />
autonomous underwater vehicles (AUVs):<br />
A brief summary. In: IEEE ICRA Conference,<br />
Seoul Korea, May 21-25, 2001.<br />
Borgen, M.G., Washington, G.N., & Kinzel,<br />
G.L. 2003. Design <strong>and</strong> evolution of a piezoelectrically<br />
actuated miniature swimming<br />
vehicle. IEEE-ASME T Mech. 8(1):66-76.<br />
doi: 10.1109/TMECH.2003.809131.<br />
Breder, C.M., Jr. 1926. The locomotion of<br />
fishes. Zoologica. 4:159-297.<br />
Brower, T.P.L. 2006. Design of a manta<br />
ray inspired underwater propulsive mechanism<br />
for long range, low power operation. Master’s<br />
thesis, Tufts University.<br />
Cai, Y., Bi, S., & Zheng, L. 2010. Design<br />
<strong>and</strong> experiments of a robotic fish imitating<br />
cow-nosed ray. J Bionic Eng. 7(2):120-6.<br />
doi: 10.1016/S1672-6529(09)60204-3.<br />
Chen, Z., Um, T.I., & Bart-Smith, H. 2011.<br />
A novel fabrication of ionic polymer-metal<br />
composite membrane actuator capable of<br />
3-dimensional kinematic motions. Sensor<br />
Actuat A-Phys. 168(1):131-9. doi: 10.1016/<br />
j.sna.2011.02.034.<br />
Clark, R.P., & Smits, A.J. 2006. Thrust<br />
production <strong>and</strong> wake structure of a batoidinspired<br />
oscillating fin. J Fluid Mech. 562:<br />
415-29. doi: 10.1017/S0022112006001297.<br />
Colgate, J.E., & Lynch, K.M. 2004.<br />
Mechanics <strong>and</strong> control of swimming: A<br />
review. IEEE J Oceanic Eng. 29(3):660-73.<br />
Combes, S.A., & Daniel, T.L. 2001. Shape,<br />
apping <strong>and</strong> exion: wing <strong>and</strong> fin design for<br />
forward flight. J Exp Biol. 204(12):2073-85.<br />
Crespi, A., Badertscher, A., Guignard, A., &<br />
Ijspeert, A.J. 2004. An amphibious robot<br />
capable of snake <strong>and</strong> lamprey-like locomotion.<br />
In: Proceedings of the 35th International<br />
Symposium on Robotics ISR 2004. Paris,<br />
France: ISR publisher.<br />
Curet, O.M., Patankar, N.A., Lauder, G.V.,<br />
& MacIver, M.A. 2010. Aquatic manoeuvering<br />
with counter-propagating waves: a novel<br />
locomotive strategy. J R Soc Interface.<br />
8(60):1041-50.<br />
Dennler, G., Scharber, M., Ameri, T.,<br />
Denk, P., Forberich, K., Waldauf, C., &<br />
Brabec, C. 2008. Design rules for donors<br />
in bulk-heterojunction t<strong>and</strong>em solar cells<br />
Towards 15% Energy-Conversion Eficiency.<br />
Adv Mater. 20(3):579-83. doi: 10.1002/<br />
adma.200702337.<br />
Dobbins, P. 2007. Dolphin sonar: Modelling<br />
a new receiver concept. Bioinspir Biomim. 2:19.<br />
doi: 10.1088/1748-3182/2/1/003.<br />
Drucker, E.G., & Jensen, J.S. 1997. Kinematic<br />
<strong>and</strong> electromyographic analysis of steady<br />
pectoral fin swimming in the surfperches.<br />
J Exp Biol. 200:1709-23.<br />
Festo, M. 2008. Fluidic Muscle Brochure.<br />
http://www/festo.com/cms/en-us_us/5030.htm.<br />
Fish, F.E., & Lauder, G.V. 2006. Passive <strong>and</strong><br />
active ow control by swimming fishes <strong>and</strong><br />
mammals. Annu Rev Fluid Mech. 38:193-224.<br />
doi: 10.1146/annurev.fluid.38.050304.092201.<br />
Fish, F.E., Lauder, G.V., Mittal, R., Techet,<br />
A.H., Triantafyllou, M.S., Walker, J.A., &<br />
Webb, P.W. 2003. Conceptual design for the<br />
construction of a biorobotic AUV based on<br />
biological hydrodynamics. In: Proceedings of<br />
the 13th International Symposium on Unmanned<br />
Untethered Submersible <strong>Technology</strong>.<br />
Durham, NH: AUSI (Autonomous Undersea<br />
Systems Institute).<br />
Fish, F.E., Weber, P.W., Murray, M.M., &<br />
Howle, L.E. 2011. The humpback whale’s<br />
ipper: Application of bio-inspired tubercle<br />
technology. Integr Comp Biol. 211:1857-67.<br />
Gao, J., Bi, S., Xu, Y., & Liu, C. 2007.<br />
Development <strong>and</strong> design of a robotic manta<br />
ray featuring flexible pectoral fins. In: IEEE<br />
International Conference on Robotics <strong>and</strong><br />
<strong>Biomimetics</strong>, December 15-18, 2007. ROBIO<br />
2007, 519-23. Yalong Bay of Sanya, China:<br />
IEEE.<br />
Gibb, A., Jayne, B., & Lauder, G. 1994.<br />
Kinematics of pectoral fin locomotion in the<br />
bluegill sunfish Lepomis machrochirus.<br />
J Exp Biol. 189:133-61.<br />
Heine, C. 1992. Mechanics of apping fin<br />
locomotion in the cownose ray, Rhinoptera<br />
bonasus (Elasmobranchii, Myliobatidae).<br />
Ph.D. thesis, Duke University.<br />
Jaffe, J., Moore, K., McLean, J., & Str<strong>and</strong>,<br />
M. 2001. Underwater optical imaging: status<br />
<strong>and</strong> prospects. Oceanography. 14(3):66-76.<br />
Jalbert, J., Baker, J., Duchesney, J., Pietryka,<br />
P., Dalton, W., Blidberg, D., … Holappa, K.<br />
2003. Solarpowered autonomous underwater<br />
vehicle development. In: Proceedings of the<br />
Thirteenth International Symposium on Unmanned<br />
Untethered Submersible <strong>Technology</strong>.<br />
Durham, NH: Autonomous Undersea Systems<br />
Institute.<br />
July/August 2011 Volume 45 Number 4 107
Kato, N. 1998. Locomotion by mechanical<br />
pectoral fins. J Mar Sci Technol. 3(3):113-21.<br />
doi: 10.1007/BF02492918.<br />
Kato, N., Ando, Y., Tomokazu, A., Suzuki,<br />
H., Suzumori, K., K<strong>and</strong>a, T., & Endo, S.<br />
2008. Elastic pectoral fin actuators for biomimetic<br />
underwater vehicles. In: Bio-mechanisms<br />
of Swimming <strong>and</strong> Flying, eds. Kato, N., &<br />
Kamimura, S., Part 3, 271-82. Springer.<br />
doi: 10.1007/978-4-431-73380-5_22.<br />
Kato, N., & Furushima, M. 2002. Pectoral<br />
fin model for maneuver of underwater<br />
vehicles. In: Autonomous Underwater<br />
Vehicle <strong>Technology</strong>, 1996. AUV’96,<br />
Proceedings of the 1996 Symposium on<br />
IEEE. pp. 49-56. Durham, NH: Autonomous<br />
Undersea Systems Institute.<br />
Klausewitz, W. 1964. Der lokomotionsmodus<br />
der Flugelrochen (Myliobatoidei).<br />
Zool Anz. 173:111-20.<br />
Kocak, D., & Caimi, F. 2005. The current<br />
art of underwater imaging with a glimpse<br />
of the past <strong>and</strong> vision of the future. Mar<br />
Technol Soc J. 39(3):5-26. doi: 10.4031/<br />
002533205787442576.<br />
Lauder, G.V., & Jayne, B.C. 1996. Pectoral<br />
fin locomotion in fishes: Testing drag-based<br />
models using three-dimensional kinematics.<br />
Am Zool. 36(6):567-81.<br />
Lauder, G.V., Nauen, J.C., & Drucker, E.G.<br />
2002. Experimental hydrodynamics <strong>and</strong> evolution:<br />
function of median fins in ray-finned<br />
fishes 1. Integr Comp Biol. 42(5):1009-17.<br />
doi: 10.1093/icb/42.5.1009.<br />
Licht, S., Polidoro, V., Flores, M., Hover, F.S.,<br />
& Triantafyllou, M.S. 2004. Design <strong>and</strong><br />
projected performance of a apping foil AUV.<br />
IEEE J Oceanic Eng. 29(3):786-94.<br />
doi: 10.1109/JOE.2004.833126.<br />
Listak, M., Martin, G., Pugal, D., Aabloo, A.,<br />
& Kruusmaa, M. 2005. Design of a semiautonomous<br />
biomimetic underwater vehicle for<br />
environmental monitoring. In: Computational<br />
Intelligence in Robotics <strong>and</strong> Automation.<br />
Espoo, Finl<strong>and</strong>: IEEE.<br />
Low, K.H., & Willy, A. 2006. Biomimetic<br />
motion planning of an undulating robotic<br />
fish fin. J Vib Control. 12:1337-59.<br />
doi: 10.1177/1077546306070597.<br />
Mojarrad, M. 2000. AUV biomimetic<br />
propulsion. In: OCEANS 2000 MTS/IEEE<br />
Conference <strong>and</strong> Exhibition. pp. 2141-6.<br />
Providence, RI: IEEE.<br />
Moored, K.W. August 2010. The design of a<br />
novel tensegrity-based synthetic pectoral fin<br />
for bio-inspired propulsion. Ph.D. thesis,<br />
University of Virginia.<br />
Moored, K.W., & Bart-Smith, H. 2007. The<br />
analysis of tensegrity structures for the design<br />
of a morphing wing. J Appl Mech. 74:668-76.<br />
doi: 10.1115/1.2424718.<br />
Moored, K.W., & Bart-Smith, H. 2009. Investigation<br />
of clustered actuation in tensegrity<br />
structures. Int J Solids Struct. 46:3272-81.<br />
doi: 10.1016/j.ijsolstr.2009.04.026.<br />
Moored, K.W., Dewey, P.A., Leftwich, M.C.,<br />
Bart-Smith, H., & Smits, A.J. 2011a.<br />
Bio-inspired propulsion mechanisms based<br />
on manta ray locomotion. Mar Technol Soc J.<br />
45(4):110-8.<br />
Moored, K.W., Kemp, T.H., Houle, N.E., &<br />
Bart-Smith, H. 2011b. Analytical predictions,<br />
optimization <strong>and</strong> design of a tensegrity-based<br />
artificial pectoral fin. Int J Solids Struct.<br />
(Accepted for Publication).<br />
Palmisano, J., Geder, J., Ramamurti, R.,<br />
Liu, K.J., Cohen, J., Mengesha, T., … Ratna,<br />
B. 2008. Design, development, <strong>and</strong> testing<br />
of flapping fins with actively controlled curvature<br />
for an unmanned underwater vehicle.<br />
In: Bio-mechanisms of Swimming <strong>and</strong> Flying:<br />
Fluid Dynamics, Biomimetic Robots, <strong>and</strong><br />
Sports Science, eds. Kato, N., & Kamimura,<br />
S., 283-94. Berlin: Springer.<br />
Palmisano, J., Ramamurti, R., Lu, K.J.,<br />
Cohen, J., S<strong>and</strong>berg, W., & Ratna, B. 2007.<br />
Design of a biomimetic controlled-curvature<br />
robotic pectoral fin. In: IEEE International<br />
Conference on Robotics <strong>and</strong> Automation<br />
(ICRA). 966-73. IEEE.<br />
Pederzani, J.-N., Moored, K.W., Fish, F.E.,<br />
& Haj-Hariri, H. 2011. A numerical investigation<br />
of the hydrodynamic signature of<br />
batoid swimming. In: <strong>Society</strong> for Integrative<br />
<strong>and</strong> Comparative Biology, Annual meeting.<br />
Salt Lake City, UT: Oxford Univ Press Inc.<br />
Rosenberger, L.J. 2001. Pectoral fin locomotion<br />
in batoid fishes: Undulation versus<br />
oscillation. J Exp Biol. 204(2):379-94.<br />
Russo, R.S., Blemker, S., Fish, F.E., Moored,<br />
K.W., & Bart-Smith, H. 2011. Form function<br />
relationship between ray skeletal architecture<br />
<strong>and</strong> ray locomotion. In: Annual meeting of<br />
the <strong>Society</strong> for Integrative <strong>and</strong> Comparative<br />
Biology. Salt Lake City, UT: Oxford Univ<br />
Press Inc.<br />
Sasko, D.E., Dean, M.N., Motta, P.J., &<br />
Hueter, R.E. 2006. Prey capture behavior<br />
<strong>and</strong> kinematics of the Atlantic cownose ray,<br />
Rhinoptera bonasus. Zoology. 109(3):171-81.<br />
doi: 10.1016/j.zool.2005.12.005.<br />
Schaefer, J.T., & Summers, A.P. 2005.<br />
Batoid wing skeletal structure: Novel<br />
morphologies, mechanical implications, <strong>and</strong><br />
phylogenetic patterns. J Morphol. 264(3):<br />
298-313. doi: 10.1002/jmor.10331.<br />
Sfakiotakis, M., Laue, D., & Davies, B.<br />
2005. An experimental undulating-fin device<br />
using the parallel bellows actuator. In:<br />
Robotics <strong>and</strong> Automation, Proceedings 2001<br />
ICRA. IEEE International Conference on.<br />
pp. 2356-62. Seoul, Korea: IEEE.<br />
Sfakiotakis, M., Lane, D.M., & Davies,<br />
J.B.C. 1999. Review of fish swimming modes<br />
for aquatic locomotion. IEEE J Oceanic Eng.<br />
24(2):237-52. doi: 10.1109/48.757275.<br />
Stocking, J.B., Eberhardt, W.C., Shakhsheer,<br />
J.R., Paulus, M.A., & Calhoun, B.H. 2010.<br />
A capacitance-based whisker-like artificial<br />
sensor for fluid motion sensing. In: IEEE<br />
Sensors. Waikoloa, Big Isl<strong>and</strong>, Hawaii.<br />
Summers, A. 2000. Stifiening the stingray<br />
skeleton: An investigation of durophagy in<br />
myliobatid stingrays (Chondrichthyes, Batoidea,<br />
Myliobatidae). J Morphol. 243(2):113-26.<br />
doi: 10.1002/(SICI)1097-4687(200002)<br />
243:23.0.CO;2-A.<br />
Suzumori, K., Endo, S., K<strong>and</strong>a, T., Kato, N.,<br />
& Suzuki, H. 2007. A bending pneumatic<br />
rubber actuator realizing soft-bodied manta<br />
swimming robot. In: 2007 IEEE International<br />
108 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
Conference on Robotics <strong>and</strong> Automation.<br />
pp. 4975-980. IEEE.<br />
Takagi, K., Yamamura, M., Luo, Z., Onishi,<br />
M., Hirano, S., Asaka, K., & Hayakawa, Y.<br />
2007. Development of a rajiform swimming<br />
robot using ionic polymer artificial muscles.<br />
In: Intelligent Robots <strong>and</strong> Systems, 2006<br />
IEEE/RSJ International Conference on.<br />
pp. 1861-6. Beijing, China: IEEE.<br />
Tangorra, J., Anquetil, P., Fofonoff, T., Chen,<br />
A., Zio, M., & Hunter, I. 2007. The application<br />
of conducting polymers to a biorobotic<br />
fin propulsor. Bioinspir Biomim. 2:S6.<br />
doi: 10.1088/1748-3182/2/2/S02.<br />
Tangorra, J., Davidson, S., Hunter, I.,<br />
Madden, P., Lauder, G., Haibo, D., …<br />
Mittal, R. 2008a. The development of a<br />
biologically inspired propulsor for unmanned<br />
underwater vehicles. IEEE J Oceanic Eng.<br />
32(3):533-50.<br />
Tangorra, J., Lauder, G., Madden, P., Mittal, R.,<br />
Bozkurttas, M., & Hunter, I. 2008b. A<br />
biorobotic apping fin for propulsion <strong>and</strong><br />
maneuvering. In: Robotics <strong>and</strong> Automation,<br />
2008. ICRA 2008. IEEE International<br />
Conference on. pp. 700-5. Pasadena, CA:<br />
IEEE.<br />
V. y Alvarado, P., Chin, S., Larson, W.,<br />
Mazumdar, A., & Youcef-Toumi, K. 2010. A<br />
soft body under-actuated approach to multi<br />
degree of freedom biomimetic robots: A<br />
stingray example. In: Biomedical Robotics <strong>and</strong><br />
Biomechatronics (BioRob), 2010 3rd IEEE<br />
RAS <strong>and</strong> EMBS International Conference<br />
on. pp. 473-8. Tokyo, Japan: IEEE.<br />
Walker, J.A., & Westneat, M.W. 1997.<br />
Labriform propulsion in fishes: kinematics<br />
of flapping aquatic flight in the bird wrasse<br />
Gomphosus varius (Labridae). J Exp Biol.<br />
200:1549-69.<br />
Wang, Z., Hang, G., Wang, Y., Li, J., &<br />
Du, W. 2008. Embedded SMA wire actuated<br />
biomimetic fin: a module for biomimetic<br />
underwater propulsion. Smart Mater<br />
Struct. 17:025039. doi: 10.1088/0964-<br />
1726/17/2/025039.<br />
underwater glider. In: International Symposium<br />
on Unmanned Untethered Submersible<br />
<strong>Technology</strong>. 75-85. Durham, NH: Autonomous<br />
Undersea Systems Institute.<br />
Webb, P.W. 1994. Mechanics <strong>and</strong> Physiology<br />
of Animal Swimming: The Biology of Fish<br />
Swimming. Cambridge, UK: Cambridge<br />
University Press. 45-62 pp.<br />
Yan, Q., Zhang, S., & Yang, J. 2010. Initial<br />
implementation of basic actuated unit of a<br />
flexible pectoral fin driven by SMA. In:<br />
Mechatronics <strong>and</strong> Automation (ICMA),<br />
2010 International Conference on.<br />
pp. 899-904. Xi’an, China: IEEE.<br />
Yang, S., Qiu, J., & Han, X. 2009. Kinematics<br />
modeling <strong>and</strong> experiments of pectoral<br />
oscillation propulsion robotic fish. J Bionic<br />
Eng. 6(2):174-9. doi: 10.1016/S1672-6529<br />
(08)60114-6.<br />
Yang, Y., Chen, J., Engel, J., P<strong>and</strong>ya, S.,<br />
Chen, N., Tucker, C., … Liu, C. 2006.<br />
Distant touch hydrodynamic imaging with<br />
an artificial lateral line. P Natl Acad Sci.<br />
103(50):18891.<br />
Yong-hua, Z., Shi-wu, Z., Jie, Y., & Low, K.<br />
2007. Morphologic optimal design of bionic<br />
undulating fin based on computational fluid<br />
dynamics. In: Mechatronics <strong>and</strong> Automation,<br />
2007. ICMA 2007. International Conference<br />
on. pp. 491-6. Harbin, Heilongjian, China:<br />
IEEE.<br />
Yu, J., Hu, Y., Fan, R., Wang, L., & Huo, J.<br />
2007. Mechanical design <strong>and</strong> motion<br />
control of a biomimetic robotic dolphin.<br />
Adv Robotics. 21(3):499-513.<br />
doi: 10.1163/156855307780131974.<br />
Yu, J., Tan, M., Wang, S., & Chen, E. 2004.<br />
Development of a biomimetic robotic fish<br />
<strong>and</strong> its control algorithm. IEEE Sys Man<br />
Cybern B. 34(4), 1798-810.<br />
Zhou, C., & Low, K. 2010. Better endurance<br />
<strong>and</strong> load capacity: An improved design of<br />
manta ray robot. J Bionic Eng. 7:S137-44.<br />
doi: 10.1016/S1672-6529(09)60227-4.<br />
Webb, D., & Simonetti, P. 1999. The<br />
Slocum AUV: An environmentally propelled<br />
July/August 2011 Volume 45 Number 4 109
PAPER<br />
Bioinspired Propulsion Mechanisms<br />
Based on Manta Ray Locomotion<br />
AUTHORS<br />
Keith W. Moored<br />
Peter A. Dewey<br />
Mechanical <strong>and</strong> Aerospace<br />
Engineering, Princeton University<br />
Megan C. Leftwich<br />
Physics Division, Los Alamos<br />
National Laboratory<br />
Hilary Bart-Smith<br />
Mechanical <strong>and</strong> Aerospace<br />
Engineering, University of Virginia<br />
Alex<strong>and</strong>er J. Smits<br />
Mechanical <strong>and</strong> Aerospace<br />
Engineering, Princeton University<br />
Introduction<br />
I<br />
n recent years, there has been<br />
considerable interest in developing<br />
novel underwater vehicles that use<br />
propulsion systems inspired by biology<br />
(Colgate & Lynch, 2004;<br />
B<strong>and</strong>yopadhyay, 2005). Such vehicles<br />
have the potential to open up new mission<br />
capabilities <strong>and</strong> improve maneuverability,<br />
efficiency, <strong>and</strong> speed (Fish<br />
et al., 2003, 2011). Here we explore<br />
how various aspects of biological locomotion<br />
relate to performance in the<br />
particular case of ray-like swimming,<br />
with the aim of informing the design<br />
of new vehicles.<br />
The kinematic motion of batoid fish<br />
(rays) is based on the chordwise traveling<br />
wave that is a hallmark of their motion<br />
(Rosenberger, 2001). Species are classified<br />
as being oscillatory if the traveling<br />
wave wavelength is longer than the<br />
chord of their fin <strong>and</strong> undulatory if the<br />
wavelength is less than the chord of<br />
ABSTRACT<br />
Mobuliform swimmers are inspiring novel approaches to the design of underwater<br />
vehicles. These swimmers, exemplified by manta rays, present a model for new classes<br />
of efficient, highly maneuverable, autonomous undersea vehicles. To improve our<br />
underst<strong>and</strong>ing of the unsteady propulsion mechanisms used by these swimmers,<br />
we report detailed studies of the performance of robotic swimmers that mimic aspects<br />
of the animal propulsive mechanisms. We highlight the importance of the undulatory<br />
aspect of producing efficient manta ray propulsion <strong>and</strong> show that there is a strong interaction<br />
between the propulsive performance <strong>and</strong> the flexibility of the actuating surfaces.<br />
Keywords: mobuliform, manta ray, unsteady, swimming, flexible actuators<br />
their fin. The manta ray is an example<br />
of an oscillatory swimmer. Previously,<br />
Clark <strong>and</strong> Smits (2006) explored the<br />
thrust production <strong>and</strong> efficiencies of an<br />
artificial pectoral fin thatcapturedthe<br />
traveling wave motion <strong>and</strong> observed<br />
efficiencies upwards of 50% for an oscillatory<br />
motion at a fixed flow velocity.<br />
To study the swimming of mantas,<br />
we use artificial or robotic devices that<br />
generate a simple baseline motion that<br />
approximates biological kinematics.<br />
The complexity of the motion is then<br />
progressively increased by adding more<br />
kinematic features until the motion<br />
resembles the biology very closely. At<br />
each level of complexity, various performance<br />
metrics are measured. We explore<br />
theroleofspanwisecurvature,theeffects<br />
ofaspanwisetravelingwave<strong>and</strong>tip<br />
speed modulation, which have not been<br />
previously investigated, <strong>and</strong>, the role<br />
of a chordwise traveling wave motion<br />
is investigated in the performance of<br />
an actively <strong>and</strong> passively flexible fin.<br />
Experiments<br />
Two biorobotic devices were developed<br />
<strong>and</strong> tested. First, an artificial pectoral<br />
fin able to produce root-fixed<br />
pure heaving motions was developed<br />
<strong>and</strong> will be referred to as the heaving<br />
fin. Thefin wascastfromaflexible<br />
plastic <strong>and</strong> actuated with variable degrees<br />
of spanwise curvature. The measurements<br />
on the heaving fin were<br />
conducted in a tow tank (Figure 1).<br />
This facility tows a fin throughstill<br />
water at a fixed velocity, U, inatank<br />
measuring 5-m long, 1-m deep, <strong>and</strong><br />
1.5-m wide, <strong>and</strong> we directly measure<br />
thenetforceproduced,T, <strong>and</strong>the<br />
power imparted into the fluid, P f ,<br />
by a flapping fin structure. The<br />
propulsive efficiency of the motion,<br />
η p ¼ TU= ‐ P ‐ f , can then be calculated<br />
from the thrust <strong>and</strong> power measurements<br />
averaged over a cycle,<br />
‐<br />
T <strong>and</strong> P ‐ f , respectively. If the fin were<br />
unconstrained <strong>and</strong> free-swimming,<br />
then the net force would cause the<br />
fin to accelerate or decelerate to a<br />
new velocity where there is no average<br />
net force. Constraining the fin allows<br />
for the measurement of force production<br />
<strong>and</strong> is a commonly used experimental<br />
approach (Anderson et al.,<br />
1998).<br />
110 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 1<br />
Tow tank facility consisting of an artificial pectoral fin, tow tank, motor, carriage, <strong>and</strong> a control<br />
<strong>and</strong> measurement system.<br />
Second, an artificial pectoral fin capable<br />
of generating a chordwise traveling<br />
wave motion was developed <strong>and</strong><br />
will be referred to as the traveling<br />
wave fin. This fin was used to measure<br />
the free-swimming performance in a<br />
tank measuring 6.7-m long, 1-m<br />
deep, <strong>and</strong> 1-m wide. This fin was<br />
also cast from a flexible plastic, but it<br />
was activated using a number of rigid<br />
spars in the spanwise direction. By reducing<br />
the number of actuating spars,<br />
the degree of passive flexibility of the<br />
fin could be varied. The traveling<br />
wave wavelength was controlled by<br />
changing the phase differences between<br />
adjacent spars. The steady<br />
swimming speed U <strong>and</strong> mean power<br />
input over a cycle P ‐ f were measured,<br />
<strong>and</strong> hence, the energy economy ζ<br />
could be found, where ζ ¼ U = P ‐ f .<br />
Energy economy is the inverse of<br />
cost of transport, both of which are<br />
used extensively in the biological literature<br />
(Schmidt-Nielsen, 1972; Fish<br />
et al., 1991; Liao et al., 2003; Liao,<br />
2004). Energy economy, however, is<br />
a more appropriate engineering metric<br />
as the dimensions are distance/per unit<br />
energy (the units could be miles per<br />
gallon for instance). Efficiency is also<br />
an appropriate performance measure,<br />
but from an experimental point-ofview<br />
it can only be measured when<br />
there is net thrust production or the<br />
fin is not in a free-swimming mode.<br />
Thus, for the experiments in the tow<br />
tank, efficiency was a measurable<br />
performance metric, but in the freeswimming<br />
cases, economy is used instead<br />
because the efficiency could not<br />
be directly measured.<br />
The performance is measured as a<br />
function of the Strouhal number,<br />
St = fA/U, where f is the frequency of<br />
motion, A is the peak-to-peak trailingedge<br />
amplitude of motion at the midspan,<br />
<strong>and</strong> U is the free-stream velocity.<br />
The Strouhal number is a measure<br />
of the lateral to streamwise spacing of<br />
the shed vortices in the wake <strong>and</strong> to a<br />
large extent governs the structure of<br />
the wake. It has been shown to be a<br />
critical parameter in describing the efficient<br />
propulsion of oscillating foils<br />
<strong>and</strong> plates (Anderson et al., 1998;<br />
Buchholz & Smits, 2008) <strong>and</strong> for<br />
swimming (Clark & Smits, 2006;<br />
Borazjani & Sotiropoulos, 2008,<br />
2009) <strong>and</strong> flying animals (Taylor<br />
et al., 2003).<br />
Fixed Velocity Experiments:<br />
Heaving Motion<br />
The skeletal structure of the heaving<br />
fin is composed of three connected<br />
hinged plates. The angular position of<br />
each hinge is individually controlled by<br />
a linear actuator. The fin allows for<br />
out-of-plane motion with no pitching<br />
or undulation in the chordwise direction.<br />
The skeletal structure is embedded<br />
into a compliant PVC polymer.<br />
The PVC is molded around the structure<br />
into a fin with a trapezoidal planform<br />
shape. This shape was chosen to<br />
be a simple representative shape of the<br />
manta ray as well as to maximize spacing<br />
for the internal structure. The fin<br />
has a span length of b =28cm<strong>and</strong><br />
an average chord length of ― c =19<br />
cm, with an aspect ratio, AR = b 2 /S,<br />
of 1.47, wherein S is the planform<br />
area. The cross-sectional shape is a<br />
NACA 0020 airfoil. The trailing edge<br />
is stiffened by a thin metal sheet attached<br />
to the main structure.<br />
Three flapping mode shapes were<br />
explored: flat root-fixed heave (Figure<br />
2(a)), curved root-fixed heave (Figure<br />
2(b)), <strong>and</strong> curved root-fixed heave<br />
with a span-wise traveling wave (Figure<br />
2(c)). For each of these mode<br />
shapes the tip speed of the fin can be<br />
modulated. Previous kinematics studies<br />
of ray locomotion (Heine, 1992;<br />
Rosenberger, 2001) found that in<br />
order to swim faster oscillatory rays<br />
do not vary their beat frequency, but<br />
instead they vary the tip speed of<br />
their fin while holding frequency <strong>and</strong><br />
amplitude constant. This can be<br />
viewed as modifying the actuation<br />
waveform to suit a particular mode of<br />
swimming. To implement this mode<br />
in our experiments, the time-varying<br />
waveform was varied from a pure sinusoid<br />
towards an almost square wave<br />
form (Figure 3). This allows the frequency<br />
<strong>and</strong> amplitude to be held<br />
fixed while the maximum fin tip<br />
speed is increased. The thrust <strong>and</strong> propulsive<br />
efficiency were measured for<br />
each prescribed motion.<br />
July/August 2011 Volume 45 Number 4 111
FIGURE 2<br />
Swimming modes ranging from an artificial/simple motion that approximates manta ray locomotion<br />
to a more complex biologically inspired motion: (a) flat root-fixed heaving motion, (b) curved<br />
root-fixed heaving motion, <strong>and</strong> (c) curved root-fixed heaving motion with a span-wise traveling<br />
wave (tip lag effect).<br />
Free-Swimming Experiments:<br />
Traveling Wave Motion<br />
The traveling wave fin was tested<br />
under free-swimming conditions in a<br />
stationary tank to explore the role<br />
of flexibility in ray-like propulsion.<br />
Low-friction carts were attached to<br />
the fin actuation mechanism to form<br />
a carriage that was mounted on tracks<br />
above the tank (see Figure 4). The<br />
chord of the fin was aligned parallel<br />
to the tracks, which allowed for a single<br />
degree of freedom <strong>and</strong> made the<br />
swimming direction of the carriage<br />
unidimensional, <strong>and</strong> no transverse<br />
motion of the carriage was permitted.<br />
The root of the fin abutted against an<br />
acrylic sheet that was in contact with<br />
thefreesurfaceofthewatertominimize<br />
surface waves. Upon actuating<br />
the fin, the cart propelled itself down<br />
the length of the tank.<br />
FIGURE 3<br />
An elliptical planform fin withan<br />
aspect ratio of 1.6 <strong>and</strong> a NACA 0020<br />
cross-section was cast using a flexible<br />
PVC plastic. Four aluminum spars<br />
were embedded into the fin to provide<br />
actuation. A push-rod connected each<br />
spar to a gear in a gear-train, driven by<br />
a DC motor, that produced a sinusoidal<br />
rotation of the spar about a pivot<br />
point located at the root chord of the<br />
fin. This results in a linearly increasing<br />
amplitude of motion along the span of<br />
the fin (Figure 5). The wavelength of<br />
the traveling wave λ could be varied<br />
by changing the phase difference between<br />
the actuating spars.<br />
In addition, the number of actuating<br />
spars could be varied to allow for a<br />
certain degree of passive flexibility in<br />
the fin. When four actuating spars<br />
are used, the locomotion of the fin<br />
was prescribed for the entire chord of<br />
Holding frequency <strong>and</strong> amplitude constant while modulating tip speed can be achieved by varying<br />
the actuation waveform from a sine wave to a square wave.<br />
the fin, with a traveling wave wavelength<br />
λ a , <strong>and</strong> this fin will be referred<br />
to as the active fin. When the two trailing<br />
edge spars are removed, the trailing<br />
half of the chord of the fin willpassively<br />
respond to the leading edge actuation<br />
<strong>and</strong> external fluid forces. This fin<br />
will be referred to as the passive fin<br />
(Figure 5). For the passive fin, the<br />
traveling wave along the chord is generated<br />
by the first two gears in the gear<br />
train <strong>and</strong> the passive response of the<br />
trailing edge. Due to these compounding<br />
factors, the precise wavelength<br />
along the chord for the passive fin is<br />
unknown, <strong>and</strong> we define an analogous<br />
wavelength, λ p such that λ a =<br />
λ p when the offset between the first<br />
two gears is identical for both the active<br />
<strong>and</strong> passive fin. We define the dimensionless<br />
wavelengths λ a ¼ λ a=C<br />
<strong>and</strong> λ p ¼ λ p=C, whereC is the root<br />
chord of the fin. This study focuses on<br />
cases with λ a ; λ p > 1, representative<br />
of oscillatory swimmers (Rosenberger,<br />
2001).<br />
Results <strong>and</strong> Discussion<br />
Flat <strong>and</strong> Curved Modes<br />
The first two modes of swimming,<br />
the flat mode (Figure 2(a)) <strong>and</strong><br />
thecurvedmode(Figure2(b)),were<br />
studied using the heaving fin.<br />
The thrust coefficient is defined by<br />
C T ¼ T = 1 2 ρU 2 S,whereinT is the<br />
thrust, ρ is the water density, <strong>and</strong> S<br />
is the planform area of the fin. Similarly,<br />
the power coefficient is defined<br />
by C p ¼ P= 1 2 ρU 3 S, wherein P is the<br />
power input to the water (as defined<br />
by Clark & Smits, 2006). Figure 6(a)<br />
shows that the thrust production <strong>and</strong><br />
power input increases as the Strouhal<br />
number increases, as found in previous<br />
studies by Anderson et al. (1998) <strong>and</strong><br />
Dong et al. (2006). The flat mode of<br />
swimming produces more thrust <strong>and</strong><br />
112 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 4<br />
Tank facility for the chordwise traveling wave experiments: (a) perspective view <strong>and</strong> (b) front<br />
view. Drawings not to scale.<br />
FIGURE 5<br />
Traveling wave actuation system for (left) active fin <strong>and</strong> (right) passive fin.<br />
FIGURE 6<br />
Flat mode compared to curved mode: (a) thrust performance as a function of St <strong>and</strong> (b) power<br />
coefficient as a function of St.<br />
uses more power than the curved mode<br />
throughout the Strouhal range. This<br />
results is not unexpected as the flat<br />
mode of swimming sweeps out a larger<br />
volume of fluid than the curved<br />
mode of swimming, causing the overall<br />
increase in the thrust <strong>and</strong> power<br />
coefficients.<br />
Because the two modes of swimming<br />
are dissimilar, comparing the<br />
thrust or efficiency as a function of<br />
Strouhal number may be misleading.<br />
For a fixed Strouhal number, different<br />
amounts of thrust <strong>and</strong> power are produced<br />
by each swimming mode. A<br />
better comparison is the thrust or efficiency<br />
as a function of the power<br />
coefficient because each swimming<br />
mode can be compared at a fixed<br />
power input.<br />
Figure 7a shows this comparison.<br />
For both modes of swimming the<br />
power input increases as the thrust increases,<br />
although at higher power<br />
input the thrust increases at a slower<br />
rate. For all power inputs, more thrust<br />
is produced for the flat mode of swimming<br />
than for the curved mode of<br />
swimming, while the thrust tends to<br />
the drag of the motionless fin asthe<br />
power input tends to zero. Interestingly,<br />
observations of swimming<br />
manta rays do not indicate that they<br />
use this higher-performance flat<br />
mode <strong>and</strong> instead use the curved<br />
mode for swimming. This suggests<br />
that the curved mode, coupled with<br />
another mechanism (perhaps a chordwise<br />
traveling wave), more fully characterizes<br />
the swimming mechanics of<br />
manta rays.<br />
Figure 7(b) shows the propulsive<br />
efficiency as a function of Strouhal<br />
number. The efficiency of the curved<br />
mode first rises quickly rise with<br />
increasing Strouhal number, <strong>and</strong> then<br />
apeakinefficiency is attained, followed<br />
by a slow decline in efficiency<br />
July/August 2011 Volume 45 Number 4 113
FIGURE 7<br />
Flat mode compared to curved mode: (a) thrust coefficient as a function of the power coefficient<br />
<strong>and</strong> (b) efficiency as a function of Strouhal number.<br />
at the higher Strouhal numbers. This<br />
trend is characteristic of efficiency<br />
curves for oscillating foils <strong>and</strong> plates<br />
(Clark & Smits, 2006; Anderson<br />
et al., 1998; Heathcote et al., 2006a,<br />
2006b; Heathcote & Gursul, 2007;<br />
Buchholz & Smits, 2008). In general,<br />
the efficiency crosses from negative<br />
(net drag) to positive (net thrust) <strong>and</strong><br />
continues to increase while at high<br />
values of St there is a decline in efficiency<br />
that follows potential flow theory<br />
(Jones et al., 1998). Thus, a peak<br />
in efficiency is expected at an intermediate<br />
Strouhal number. The peak efficiency<br />
for the curved mode is about<br />
20% occurring at a St = 0.2, which is<br />
in the range of 0.2 < St
FIGURE 9<br />
Tip lag mode: (a) efficiency as a function of St <strong>and</strong> (b) peak efficiency dependence on tip lag.<br />
function was used to modulate the tip speed while fixing the frequency <strong>and</strong><br />
amplitude, as given by<br />
xt ðÞ¼at ð ϕÞ 4 þbt ð ϕÞ 2 þA; 0 ≤ t ≤ T =2<br />
a ¼<br />
U sqr<br />
max<br />
2ϕ 3 þ A ϕ 4 ; b ¼ 2A<br />
ϕ 2<br />
ϕ ¼ 1=4f ; U sqr<br />
max ¼ 2παU ∞ St<br />
þ U sqr<br />
max<br />
2ϕ<br />
The amplitude, A, the frequency, f, <strong>and</strong> the maximum tip speed, U sqr<br />
max, all determine<br />
the shape of the quartic function. The period is T. By holding the frequency<br />
<strong>and</strong> amplitude constant the tip speed can be modulated by varying α<br />
between 1 <strong>and</strong> 2.667. When α is 1, the quartic function matches a sine wave,<br />
but when α is 2.667, the maximum tip speed is 2.667 times faster than the maximum<br />
tip speed of a sine wave without varying the frequency or amplitude. It<br />
should be noted, however, that many animals increase their swimming speed<br />
by modulating their frequency of motion while fixing the amplitude. Frequency<br />
modulation will increase the Strouhal number by increasing frequency.<br />
Tests were conducted at two Strouhal numbers, 0.2 <strong>and</strong> 0.25, <strong>and</strong> the amplitude<br />
was fixed at A/b = 0.44 for α =1-2.4. Figure 10a shows the thrust coefficient<br />
dependence on the tip speed, α, for a Strouhal number of 0.2. As the tip speed is<br />
increased the thrust coefficient increases linearly, indicating that tip speed modulation<br />
can be used to increase thrust production, as exhibited by rays (Heine,<br />
1992; Rosenberger, 2001).<br />
Figure 10(b) shows the thrust coefficient plotted against the power coefficient<br />
for both frequency modulation <strong>and</strong> tip speed modulation. Tip speed modulation<br />
produces less thrust than frequency modulation for the same input power, suggesting<br />
that tip speed modulation is not an efficiency strategy. However, incorporating<br />
a chordwise traveling wave may change this outcome. Additionally, as the<br />
value of α is increased, the fin has an increasing period of effectively no motion. In<br />
a free-swimming test, the increasing resting time for the fin would result in an<br />
unpowered gliding period over part of the flapping cycle. This would result in a<br />
burst-<strong>and</strong>-coast behavior that could improve the economy since forward motion<br />
would occur without any input power for part of the flapping cycle. Alternatively,<br />
(1)<br />
tip speed modulation could an effective<br />
flight/sprinting mode, where efficiency<br />
is not important.<br />
Chordwise Traveling Wave<br />
We now investigate the effects of a<br />
chordwise traveling wave, using the<br />
traveling wave fin (the experimental<br />
arrangement was described in Free-<br />
Swimming Experiments: Traveling<br />
Wave Motion).<br />
Clark <strong>and</strong> Smits (2006) investigated<br />
the thrust <strong>and</strong> efficiency of an artificial<br />
pectoral fin using a chordwise<br />
traveling wave motion <strong>and</strong> found efficiencies<br />
peaking near 50% for optimal<br />
conditions (St ≈ 0.25 <strong>and</strong> λ a ≈ 4‐6).<br />
The efficiencies were measured at predetermined<br />
Strouhal numbers. For the<br />
current work, this constraint is not imposed,<br />
<strong>and</strong> the fin was instead actuated<br />
at a given frequency <strong>and</strong> wavelength<br />
<strong>and</strong> allowed to freely swim down the<br />
length of a tow tank. In doing so, the<br />
fin attains its self-propelled swimming<br />
speed <strong>and</strong> Strouhal number for that<br />
frequency <strong>and</strong> wavelength.<br />
Figure 11 shows the steady velocity<br />
achieved as a function of input flapping<br />
frequency for different wavelengths<br />
of actuation. The velocity is<br />
giveninroot-chordlengths(CL)per<br />
second, where C root = 0.254 m. For<br />
the active fin, an almost linear increase<br />
in velocity is observed with increasing<br />
frequency, a result in agreement with<br />
previous studies that found thrust<br />
coefficients increasing with frequency<br />
(Anderson et al., 1998). Peak velocities<br />
were found to occur for λ a ¼ 6 at the<br />
highest flapping frequencies. In these<br />
instances the velocity was upwards of<br />
2 CL/s, corresponding to a dimensional<br />
velocity of 0.51 m/s, highlighting<br />
that this form of propulsion may<br />
prove fruitful for future underwater vehicle<br />
designs that dem<strong>and</strong> relatively<br />
high speeds.<br />
July/August 2011 Volume 45 Number 4 115
FIGURE 10<br />
(a) Thrust coefficient increasing with tip speed increase <strong>and</strong> (b) tip speed modulation compared<br />
to frequency modulation. The parameter α is the ratio of maximum tip speed of the square wave<br />
actuation compared to the maximum tip speed of a sine wave of the same frequency <strong>and</strong> amplitude,<br />
α ¼ Umax=U sqr<br />
max sin .<br />
Figure 11 also reveals the role of<br />
passive flexibility. Here, the fin was actuated<br />
using only the two anterior<br />
spars, leaving the posterior half of the<br />
fin to respond passively to the forcing<br />
by the actuators <strong>and</strong> the fluid forces.<br />
The velocity of this fin still increases<br />
with increasing frequency, but the<br />
trend is no longer linear. Tests in still<br />
water indicate that the passive fin has a<br />
resonant frequency of about 2.4 Hz,<br />
defined as the frequency at which the<br />
trailing edge amplitude is maximized.<br />
FIGURE 11<br />
Fin velocity in chord lengths per second as a<br />
function of flapping frequency. The wavelength<br />
λ a refers to the active fin whileλ p refers to the<br />
passive fin. (Color versions of figures available<br />
online at: http://www.ingentaconnect.com/<br />
content/mts/mtsj/2011/00000045/00000004.)<br />
FIGURE 12<br />
Strouhal number as a function of steady-state<br />
swimming velocity for the active fin compared<br />
with biological data of the manta ray<br />
(Fish, 2010).<br />
As resonance is approached, the amplitude<br />
of the trailing edge motion increases,<br />
resulting in the fin velocity<br />
increasing at an enhanced rate (in comparison<br />
to a linear trend). It should be<br />
noted that the swimming velocities for<br />
the passive fin were approximately<br />
80% of those of the active fin.<br />
The free-swimming Strouhal number<br />
for the active fin, along with manta<br />
ray field data (Fish, 2010), are displayed<br />
in Figure 12 (the Strouhal<br />
number for the passive fin isnot<br />
shown since the trailing edge excursion<br />
of the passive fin is unknown). The experimental<br />
data collapse onto a single<br />
curve, indicating that the Strouhal<br />
number for the freely swimming active<br />
fin does not depend on the wavelength.<br />
As the swimming velocity increases,<br />
the Strouhal number begins<br />
to enter the regime presumed to be efficient<br />
(St = 0.2-0.4; Taylor et al.,<br />
2003). Hence, the optimal swimming<br />
speed for the traveling wave fin, from<br />
the perspective of efficiency, is likely<br />
to be ≥2 CL/s. The biological <strong>and</strong><br />
experimental data display the same<br />
trend, whereby an increase in swimming<br />
velocity yields a decrease in<br />
Strouhal number. Borazjani <strong>and</strong><br />
Sotiropoulos (2008, 2009) were able<br />
to show, for carangiform <strong>and</strong> anguiliform<br />
swimming, that the Strouhal<br />
number for self-propulsion approaches<br />
the efficient regime observed in nature<br />
only with increasing swimming velocity.<br />
The current study supports this<br />
conclusion for mobuliform swimming<br />
as well.<br />
The energy economy, ζ =V c /P f , for<br />
the active <strong>and</strong> passive fins is shown in<br />
Figure 13. In the case of the shortest<br />
wavelength(λ* = 3), the active fin has<br />
a higher energy economy than the passive<br />
fin, but for the longer wavelengths<br />
the energy economy for the passive fin<br />
FIGURE 13<br />
Energy economy as a function of flapping frequency.<br />
The wavelength λ a refers to the active<br />
fin while λ p refers to the passive fin.<br />
116 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
exceeds that of the active fin despite a<br />
decrease in the overall swimming speed<br />
(Figure 11). Clearly, by removing the<br />
two trailing edge spars in creating the<br />
passive fin, the power consumption<br />
decreases significantly compared to<br />
the active fin. The highest energy<br />
economy recorded was 0.18 CL/J<br />
for the passive fin withλ p ¼ 12 <strong>and</strong><br />
f = 2 Hz, but its economy is still trending<br />
upward with increasing frequency,<br />
which may reflect the fact that the<br />
maximum test frequency was still<br />
below the resonant frequency of the<br />
fin (2.4Hz).Leftwich<strong>and</strong>Smits<br />
(2010) found thrust production of a<br />
passively flexible artificial lamprey tail<br />
to increase as resonance is approached.<br />
The current work suggests that the energy<br />
economy also benefits by a system<br />
exploiting the resonant modes of a fin.<br />
The increase in energy economy<br />
with frequency may also be a Strouhal<br />
number effect. Figure 12 indicates that<br />
the fin begins to enter the “efficient”<br />
regime (St = 0.2-0.4) with increasing<br />
swimming velocity (which occurs at<br />
higher flapping frequencies; see Figure<br />
11). While the energy economy is<br />
not a direct measure of efficiency, the<br />
two parameters are inherently linked,<br />
so it is not altogether surprising that<br />
the energy economy increases as the<br />
Strouhal number approaches the supposedly<br />
optimal range. The active fin<br />
with λ a ¼ 12 displays a maximum at<br />
f = 1.6 Hz. It is believed that this<br />
peak is related to changes in the wake<br />
structure that result from the threedimensionality<br />
of the wake associated<br />
with this fin. Dewey et al. (2011)<br />
found that an increasing wavelength<br />
causes the wake to bifurcate into a double<br />
wake structure that is less efficient,<br />
<strong>and</strong> it is believed to be responsible for<br />
the maximum observed in Figure 13.<br />
It may be that the other cases will<br />
eventually reach a maximum due to a<br />
similar mechanism, but further testing<br />
is required to support this suggestion.<br />
Conclusions<br />
This study has explored various kinematic<br />
modes associated with biological<br />
propulsion based on the manta ray.<br />
The results highlight the interdependence<br />
of the kinematic motions <strong>and</strong><br />
fluid–structure interactions on the performance<br />
characteristics of the animal.<br />
A purely heaving fin was used to examine<br />
three kinematic modes of swimming<br />
(flat, curved, <strong>and</strong> tip lag) as well<br />
as variation in the actuation waveform<br />
(sine waveform to square waveform). It<br />
was found that the flat mode of swimming<br />
produces higher efficiency <strong>and</strong><br />
thrust compared to the curved mode<br />
of swimming. Furthermore, there was<br />
no performance benefit foundbyincorporating<br />
tip lag into the motion or<br />
by modulating tip speed instead of frequency.<br />
These results are counter to<br />
the hypothesis that as the motion becomes<br />
more biologically similar the<br />
thrust or efficiency performance will<br />
increase. The maximum thrust coefficient<br />
was found to be about 0.7 at<br />
St =0.45,<strong>and</strong>themaximumpropulsive<br />
efficiency was about 22% at<br />
St = 0.15. These performance metrics<br />
were low, as expected, due to the absenceofachordwisetravelingwave<br />
motion.Whatwasnotexpectedwas<br />
that the performance would be insensitive<br />
to curvature, tip lag, <strong>and</strong> tip<br />
speed variations. Without flexibility,<br />
the motion of the fin may be too constrained<br />
<strong>and</strong> not “natural” enough to<br />
achieve the performance benefits of<br />
the kinematic variations explored.<br />
The presence of a chordwise traveling<br />
wave to generate an undulatory<br />
motion appears to be of prime importance<br />
for efficient propulsion. For example,<br />
Clark <strong>and</strong> Smits (2006) found<br />
efficiencies upwards of 50%. In freeswimming<br />
experiments of artificial<br />
fins similar to that studied by Clark<br />
<strong>and</strong> Smits, we found that they were<br />
able to generate speeds similar to<br />
those observed in nature, of the order<br />
2 CL/s. When the fin was actively actuated,<br />
that is to say that the traveling<br />
wave motion was defined for all points<br />
on the chord of the fin, the steady-state<br />
Strouhal number obtained by the oscillating<br />
fin was found to be independent<br />
of the wavelength <strong>and</strong> exhibited<br />
thesametrendasthemantarayin<br />
nature. That is, at low swimming<br />
velocities, both the manta ray <strong>and</strong> the<br />
artificial fin display high steady-state<br />
Strouhal numbers, but the Strouhal<br />
numbers decrease with increasing<br />
swimming speed <strong>and</strong> they approach<br />
the regime where efficient propulsion<br />
is hypothesized to exist (St =0.2-<br />
0.4). Introducing passive flexibility<br />
into the fin, by restricting the actuationtotheleadingedge<strong>and</strong>letting<br />
the rest of the fin respond passively to<br />
the actuation <strong>and</strong> the external fluid<br />
forces, improved the energy economy.<br />
It was found that the passively actuated<br />
fin achieved steady state swimming<br />
speeds that were approximately 80%<br />
of that of the actively actuated fin,<br />
but because there was a significant decrease<br />
in the power required to propel<br />
the passively actuated fin the energy<br />
economy increased.<br />
Acknowledgments<br />
The authors would like to thank<br />
Daphne Rein-Weston, Dan Quinn,<br />
<strong>and</strong> Dr. Melissa Green for their aid<br />
in developing the low-friction carriage<br />
experiment. We would also like to<br />
thank Professor Frank Fish for correspondence<br />
regarding manta rays in<br />
nature. The authors would like to acknowledge<br />
funding from the Office<br />
July/August 2011 Volume 45 Number 4 117
of Naval Research through the<br />
MURI program on Biologically-<br />
Inspired Autonomous Sea Vehicles<br />
(grant N0001408-1-0642), the<br />
David <strong>and</strong> Lucille Packard Foundation,<br />
the National Science Foundation<br />
(grant CMS-0384884), <strong>and</strong> the Virginia<br />
Space Grant Consortium.<br />
Corresponding Author:<br />
Alex<strong>and</strong>er J. Smits<br />
Mechanical <strong>and</strong> Aerospace<br />
Engineering, Princeton University<br />
Princeton, NJ 08544<br />
Email: asmits@princeton.edu<br />
References<br />
Anderson, J.M., Streitlien, K., Barrett, D.S.,<br />
& Triantafyllou, M.S. 1998. Oscillating<br />
foils of high propulsive efficiency.<br />
J Fluid Mech. 360:41-72. doi: 10.1017/<br />
S0022112097008392.<br />
B<strong>and</strong>yopadhyay, P.R. 2005. Trends in<br />
biorobotic autonomous undersea vehicles.<br />
IEEE J Oceanic Eng. 30(1):109-39.<br />
doi: 10.1109/JOE.2005.843748.<br />
Borazjani, I., & Sotiropoulos, F. 2008.<br />
Numerical investigation of the hydrodynamics<br />
of carangiform swimming in the transitional<br />
<strong>and</strong> inertial flow regimes. J Exp Biol.<br />
211:1541-58. doi: 10.1242/jeb.015644.<br />
Borazjani, I., & Sotiropoulos, F. 2009.<br />
Numerical investigation of the hydrodynamics<br />
of anguilliform swimming in the transitional<br />
<strong>and</strong> inertial flow regimes. J Exp Biol.<br />
212:576-92. doi: 10.1242/jeb.025007.<br />
Buchholz, J.H.J., & Smits, A.J. 2008.<br />
The wake structure <strong>and</strong> thrust performance<br />
of a rigid low-aspect-ratio pitching panel.<br />
J Fluid Mech. 603:331-65. doi: 10.1017/<br />
S0022112008000906.<br />
Clark, R.P., & Smits, A.J. 2006. Thrust<br />
production <strong>and</strong> wake structure of a<br />
batoid-inspired oscillating fin. J Fluid<br />
Mech. 562:415-29. doi: 10.1017/<br />
S0022112006001297.<br />
Colgate, J.E., & Lynch, K.M. 2004. Mechanics<br />
<strong>and</strong> control of swimming: A review.<br />
IEEE J Oceanic Eng. 29(3):660-73.<br />
doi: 10.1109/JOE.2004.833208.<br />
Dewey, P.A., Carriou, A., & Smits, A.J.<br />
2011. On the relationship between efficiency<br />
<strong>and</strong> wake structures of a batoid-inspired<br />
oscillating fin. J Fluid Mech. Under review.<br />
Dong, H., Mittal, R., & Najjar, F.M. 2006.<br />
Wake topology <strong>and</strong> hydrodynamic performance<br />
of low-aspect-raio flapping foils.<br />
J Fluid Mech. 566:309-43. doi: 10.1017/<br />
S002211200600190X.<br />
Fish, F.E. 2010. Private Communication.<br />
Fish, F.E., Fegely, J.F., & Xanthopoulos,<br />
C.J. 1991. Burst-<strong>and</strong>-coast swimming in<br />
schooling fish (Notemigonus crysoleucas) with<br />
implications for energy economy. Comp<br />
Biochem Phys A Phys. 100:633-7.<br />
doi: 10.1016/0300-9629(91)90382-M.<br />
Fish, F.E., Haj-Hariri, H., Smits, A.J.,<br />
Bart-Smith, H., & Iwasaki, T. 2011.<br />
Biomimetic swimmer inspired by the<br />
manta ray. In: <strong>Biomimetics</strong>: Nature-Based<br />
Innovation, ed. Bar-Cohen, Y., 495-523.<br />
Boca Raton, FL: CRC Press.<br />
Fish, F.E., Lauder, G.V., Mittal, R., Techet,<br />
A.H., Triantafyllou, M.S., Walker, J.A., &<br />
Webb, P.W. 2003. Conceptual design for the<br />
construction of a biorobotic AUV based on<br />
biological hydrodynamics. In: Proceedings<br />
of the 13th International Symposium on<br />
Unmanned Untethered Submersible<br />
<strong>Technology</strong>. Durham, NH: Autonomous<br />
Systems Undersea Institute.<br />
Heathcote, S., & Gursul, I. 2007. Jet<br />
switching phenomenon for a periodically<br />
plunging airfoil. Phys Fluids. 19(2):027104.<br />
doi: 10.1063/1.2565347.<br />
Heathcote, S., Wang, Z., & Gursul, I. 2006a.<br />
Effect of spanwise flexibility on flapping wing<br />
propulsion. In: 36th AIAA Fluid Dynamics<br />
Conference <strong>and</strong> Exhibit. Reston, VA: American<br />
Institute of Aeronautics <strong>and</strong> Astronautics Inc.<br />
Heathcote, S., Wang, Z., & Gursul, I.<br />
2006b. Flexible flapping airfoil propulsion<br />
at low Reynolds numbers. AIAA J. 45(5):<br />
1066-79. doi: 10.2514/1.25431.<br />
Heine, C. 1992. Mechanics of flapping fin<br />
locomotion in the cownose ray, Rhinoptera<br />
bonasus (Elasmobranchii, Myliobatidae).<br />
Ph.D. Thesis Duke University.<br />
Jones, K.D., Dohring, C.M., & Platzer,<br />
M.F. 1998. Experimental <strong>and</strong> computational<br />
investigation of the KnollerBetz effect.<br />
Zool Anz. 173:1240-6.<br />
Klausewitz, W. 1964. Der lokomotionsmodus<br />
der Flugelrochen (Myliobatoidei). AIAA J.<br />
37:1240-6.<br />
Leftwich, M.C., & Smits, A.J. 2010. Thrust<br />
production by a mechanical swimming<br />
lamprey. Exp Fluids. 50(5):1349-55.<br />
doi 10.1007/s00348-010-0994-x.<br />
Liao, J.C. 2004. Neuromuscular control<br />
of trout swimming in a vortex street: Implications<br />
for energy economy during the<br />
Karman gait. J Exp Biol. 207:3495-506.<br />
doi: 10.1242/jeb.01125.<br />
Liao, J.C., Beal, D.N., Lauder, G.V., &<br />
Triantafyllou, M.S. 2003. Fish exploiting<br />
vortices decrease muscle activity. Science.<br />
302:1566-9. doi: 10.1126/science.1088295.<br />
Rosenberger, L.J. 2001. Pectoral fin locomotion<br />
in batoid fishes: Undulation versus<br />
oscillation. J Exp Biol. 204(2):379-94.<br />
Schmidt-Nielsen, K. 1972. Locomotion:<br />
Energy cost of swimming, flying, <strong>and</strong> running.<br />
Science. 177:222-8. doi: 10.1126/science.<br />
177.4045.222.<br />
Taylor, G.K., Nudds, R.L., & Thomas, A.L.R.<br />
2003. Flying <strong>and</strong> swimming animals cruise<br />
at a Strouhal number tuned for high power<br />
efficiency. Nature. 425(6959):707-11.<br />
doi: 10.1038/nature02000.<br />
118 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
PAPER<br />
Inspired by Sharks: A Biomimetic Skeleton<br />
for the Flapping, Propulsive Tail of an<br />
Aquatic Robot<br />
AUTHORS<br />
John H. Long, Jr.<br />
Department of Biology,<br />
Vassar College<br />
Tom Koob<br />
MiMedx Group, Inc.<br />
Justin Schaefer<br />
David Geffen School of Medicine,<br />
University of California<br />
Adam Summers<br />
Friday Harbor Labs,<br />
University of Washington<br />
Kurt Bantilan<br />
Department of Biology,<br />
Vassar College<br />
Sindre Grotmol<br />
Department of Biology,<br />
University of Bergen<br />
Marianne Porter<br />
Department of Biology,<br />
Vassar College<br />
Propulsive Functions of<br />
Vertebral Columns<br />
I<br />
n sharks <strong>and</strong> other fish, the<br />
body’s primary skeleton is the vertebral<br />
column, which runs from the<br />
head to the caudal fin (Summers &<br />
Long, 2006). The vertebral column is<br />
a jointed framework to which muscles<br />
attach <strong>and</strong> on which the muscles pull<br />
to create the traveling waves of flexure<br />
that transfer momentum from the<br />
body to the surrounding fluid. The<br />
vertebral column is composed of rigid<br />
ABSTRACT<br />
The vertebral column is the primary stiffening element of the body of fish. This<br />
serially jointed axial support system offers mechanical control of body bending<br />
through kinematic constraint <strong>and</strong> viscoelastic behavior. Because of the functional<br />
importance of the vertebral column in the body undulations that power swimming,<br />
we targeted the vertebral column of cartilaginous fishes—sharks, skates, <strong>and</strong> rays—<br />
for biomimetic replication. We examined the anatomy <strong>and</strong> mechanical properties of<br />
shark vertebral columns. Based on the vertebral anatomy, we built two classes of<br />
biomimetic vertebral column (BVC): (1) one in which the shape of the vertebrae<br />
varied <strong>and</strong> all else was held constant <strong>and</strong> (2) one in which the axial length of the<br />
invertebral joint varied <strong>and</strong> all else was held constant. Viscoelastic properties of<br />
the BVCs were compared to those of sharks at physiological bending frequencies.<br />
The BVCs with variable joint lengths were then used to build a propulsive tail, consisting<br />
of the BVC, a vertical septum, <strong>and</strong> a rigid caudal fin. The tail, in turn, was<br />
used as the propeller in a surface-swimming robot that was itself modeled after a<br />
biological system. As the BVC becomes stiffer, swimming speed of the robot increases,<br />
all else being equal. In addition, stiffer BVCs give the robot a longer stride<br />
length, the distance traveled in one cycle of the flapping tail.<br />
Keywords: biomimetics, robot, vertebral column, propulsion<br />
elements, called vertebrae, connected<br />
by flexible intervertebral joints<br />
(Grotmol et al., 2003; Koob &<br />
Long, 2000). The joints <strong>and</strong> their<br />
adjoining vertebrae limit the body’s kinematic<br />
degrees of freedom, constraining<br />
bending primarily to the lateral<br />
direction in response to loads imposed<br />
by muscle, inertia, <strong>and</strong> external fluid<br />
forces (Grotmol et al., 2006; Porter<br />
et al., 2009; Symmons, 1979; Schmitz,<br />
1995). In this way, the vertebral<br />
column functions to control dynamic<br />
reconfigurations of the self-propelling<br />
body.<br />
In roll-stable sharks <strong>and</strong> fish, lateral<br />
body bending is characterized by the<br />
overlay of harmonic <strong>and</strong> transient motions<br />
that range from small traveling<br />
flexures to large-amplitude st<strong>and</strong>ing<br />
waves (Long et al., 2010). These bending<br />
motions produce the propulsive<br />
forces that create forward swimming,<br />
turning maneuvers, <strong>and</strong> rapid accelerations.<br />
Across many different kinds of<br />
fish, the stiffness of the bending joints,<br />
measured as the apparent Young’smodulus,<br />
E, rangesfrom0.1to8MPa<br />
(Long et al., 2002). The E is a mechanical<br />
property, sometimes called<br />
the “material stiffness,” that measures<br />
the contribution of the material, independent<br />
of its geometric arrangement<br />
in the structure, to the structure’s resistance<br />
to changing shape when an external<br />
load is applied to it.<br />
July/August 2011 Volume 45 Number 4 119
Joints with any appreciable stiffness<br />
at all may at first seem to be a paradox:<br />
why not have low-stiffness joints that<br />
cost little, in terms of mechanical<br />
work, to bend The answer seems to<br />
be two-fold: (1) Stiff joints increase<br />
their resistance, in terms of the absolute<br />
bending moment, M (in units of<br />
Nm), in proportion to the magnitude<br />
of bending curvature, κ (m −1 ). This resistance,<br />
which can grow nonlinearly<br />
with κ, serves as a brake, limiting<br />
lateral bending (Long et al., 2002).<br />
(2) Stiff joints store <strong>and</strong> release more<br />
mechanical work, so-called “elastic<br />
energy,” than flexible joints (Long,<br />
1992). The amount of work released<br />
in elastic recoil is also in proportion<br />
to E <strong>and</strong> to the square of κ. Hence,<br />
the vertebral column functions best<br />
as a spring when muscles have reached<br />
their functional limits, at the end of a<br />
FIGURE 1<br />
large-κ bend when connective tissues<br />
are stiffest.<br />
To explore the mechanical design<br />
space of vertebral columns, we created<br />
biomimetic vertebral columns (BVCs).<br />
The BVCs are modeled after the vertebral<br />
column of sharks. We chose<br />
sharks’ vertebral columns since they<br />
are structurally simple, compared to<br />
those of bony fish, consisting of cylindrical<br />
centra, small neural <strong>and</strong> hemal<br />
arches, <strong>and</strong> thin intervertebral joints<br />
(Figure 1). Together, a centrum <strong>and</strong><br />
its arches are called a vertebra, <strong>and</strong> in<br />
the cartilaginous sharks, skates, <strong>and</strong><br />
rays, the vertebrae (plural form of<br />
‘vertebra’) are composed of mineralized<br />
cartilage. The compressive stiffness<br />
of these vertebrae, measured by<br />
E, ranges from 25 to 500 MPa, overlapping<br />
the lower range of E for bone<br />
(Porter et al., 2006). Compared to the<br />
Vertebral columns of two species of sharks. For scale, each centrum shown here is between 4- <strong>and</strong><br />
6-mm long in the head-to-tail direction. Head is to the left; tail is to the right.<br />
bone of mammals, for a given value of E<br />
the vertebrae of sharks are stronger, where<br />
strength is measured in terms of breaking<br />
stress (Porter & Long, 2010). Thus,<br />
in some ways, vertebral columns made of<br />
mineralized cartilage perform better than<br />
vertebral columns made of bone.<br />
In summary, vertebral columns serve<br />
at least three important propulsive functions<br />
during swimming: (1) they control<br />
dynamic reconfigurations of the body<br />
by limiting the kinematic degrees of<br />
freedom, (2) they brake high-amplitude<br />
bends by virtue of their stiffness, <strong>and</strong><br />
(3) they integrate muscle work over<br />
time by recoiling elastically.<br />
To build a BVC that can function as<br />
part of an aquatic propulsion system,<br />
we (1) characterized the morphology<br />
(size <strong>and</strong> shape) of the vertebral columns<br />
of sharks, (2) measured the mechanical<br />
properties of those vertebral<br />
columns as they underwent sinusoidal<br />
bending, (3) used that information<br />
about morphology <strong>and</strong> mechanical<br />
properties to design BVCs, (4) tested<br />
BVCs as they underwent the dynamic<br />
bending characteristic of swimming<br />
<strong>and</strong> propulsion, <strong>and</strong> (5) tested the<br />
BVCs as the primary skeleton in the<br />
flapping tail of an aquatic robot.<br />
Morphology of Sharks’<br />
Vertebral Columns<br />
In three individuals of the blacktip<br />
shark, Carcharinus limbatus, <strong>and</strong><br />
the bonnethead shark, Sphryna<br />
tiburo, we measured, from radiographs,<br />
the following features from<br />
the head to the beginning of the caudal<br />
fin (Figure 2): (1) the length of centrum,<br />
c, (2), the diameter of the centrum,<br />
d, (3) the length of the intervertebral<br />
joint, j, <strong>and</strong> (4) the cone angle, Ξ,<br />
of the capsule of the joint. Blacktip<br />
sharks, members of the family<br />
Carcharhinidae, were chosen because<br />
120 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 2<br />
Measuring vertebral morphology of blacktip <strong>and</strong> bonnethead sharks. Representative X-rays<br />
show the heavily mineralized vertebral centra, which possess an “X” shapeinthistwo-dimensional<br />
view that is from cone-shaped joint capsules. The dark space between vertebrae is the intervertebral<br />
joint. The morphology of each vertebra <strong>and</strong> intervertebral joint was measured from digitized l<strong>and</strong>marks<br />
(blue dots). (Color versions of figures available online at: http://www.ingentaconnect.com/<br />
content/mts/mtsj/2011/00000045/00000004.)<br />
they are known to be fast swimming<br />
predators of fish. In contrast,<br />
bonnethead sharks, members of the<br />
hammerhead family Sphyrnidae, are<br />
known less for their speed <strong>and</strong> more<br />
for their maneuverability <strong>and</strong> ability<br />
to find <strong>and</strong> eat crustaceans. Of similar<br />
adult body size, the two species<br />
represent contrasting swimming styles<br />
<strong>and</strong> ecologies. Three individuals of<br />
each species were used for this study.<br />
In the bonnethead shark, all the<br />
morphological features, except d,<br />
increased in size from the head to the<br />
end of the abdomen <strong>and</strong> then decreased<br />
towards the caudal fin (Figure 3). In<br />
blacktip shark, only Ξ <strong>and</strong> c varied<br />
from head to caudal fin. The significance<br />
(α = 0.05) of the morphological<br />
variation was determined with a<br />
multivariate analysis of covariance<br />
(MANCOVA), with species <strong>and</strong> position<br />
as main effects <strong>and</strong> individual as<br />
the covariate (JMP 8.0.2., SAS Institute,<br />
Cary, NC). Following an identity<br />
model MANCOVA, univariate<br />
ANCOVAs were also run.<br />
The variation in the morphology of<br />
these vertebral columns was used to<br />
guide the construction of the BVCs.<br />
For each of the sharks’ morphological<br />
features, we indicated which dimensions<br />
were used (see black arrows on<br />
the ordinates, Figure 3).<br />
Mechanical Properties of<br />
Sharks’ Vertebral Columns<br />
Using the same freshly dissected<br />
vertebral columns from which morphology<br />
was measured, we conducted<br />
3-point dynamic bending tests using<br />
an MTS model Mini Bionix 858<br />
(Eden Prairie, MN) with a 500 N<br />
load cell. For blacktip sharks, each<br />
vertebral column was cut into five<br />
segments of 19 vertebrae each. For<br />
bonnethead sharks, each vertebral<br />
column was cut into five segments of<br />
14 vertebrae each. The number of segments<br />
in each species was varied to<br />
keep the absolute length of each test<br />
segment approximately equal. Each<br />
segment was subjected to sinusoidal<br />
bending at a frequency, f (Hz), <strong>and</strong><br />
maximum curvature, κ (m −1 ), varied<br />
to hold constant the time rate of<br />
change of κ, which is equivalent to<br />
the strain rate (actuator displacement<br />
amplitude of 2 mm s −1 ).<br />
To characterize the viscoelastic<br />
properties of the vertebral column during<br />
bending, the apparent storage <strong>and</strong><br />
loss moduli, E′ <strong>and</strong> E″ (MPa), respectively,<br />
were measured at each combination<br />
of two species, five segment<br />
positions, <strong>and</strong> three κ. TheE′ measures<br />
the purely elastic component of<br />
the stiffness; it is the force proportional<br />
to the magnitude of the bending of the<br />
vertebral column. The E″ measures the<br />
purely viscous component of the stiffness;<br />
it is the force proportional to the<br />
velocity of the bending of the vertebral<br />
column. These properties were calculated<br />
from the following formulae:<br />
E ′ ¼ E*cos δ <strong>and</strong> E ″ ¼ E*sin δ,<br />
wherein E* ¼ F maxL 3<br />
48Iy max<br />
<strong>and</strong> δ is the<br />
phase lag (radians) between the displacement<br />
<strong>and</strong> load signals. Moreover,<br />
F max is the force (N) measured at the<br />
load cell, L is the gauge length of the<br />
specimen (m), I is the specimen’s<br />
July/August 2011 Volume 45 Number 4 121
FIGURE 3<br />
Vertebral morphology of sharks. Four dimensions were used to characterize the size <strong>and</strong> shape of<br />
the vertebral centra <strong>and</strong> the intervertebral joints. The means of three individuals for each species<br />
are shown; individuals ranged from 0.59 to 0.91 m in overall body length. The error bars indicate<br />
the st<strong>and</strong>ard error of the mean. Black arrows show the specific dimensions represented in our<br />
BVCs. MANCOVA, using the identity method, calculated a significant Wilkes λ (p < 0.0001),<br />
with a significant interaction of species <strong>and</strong> position <strong>and</strong> significant main effect of species; the<br />
covariate, individual, was also significant. Partial correlations among the response variables<br />
ranged from a low of 0.38 between j <strong>and</strong> d toahighof0.79betweenΞ <strong>and</strong> c. Significance<br />
level is indicated (*p < 0.05, **p < 0.01, ***p < 0.001).<br />
second moment of area (m 4 ), <strong>and</strong> y max<br />
(m) is the distance from the presumed<br />
neutral plane of bending (transverse<br />
center of specimen) <strong>and</strong> the lateralmost<br />
fibers of the specimen.<br />
In blacktip sharks, the E′ <strong>and</strong> E″<br />
values were of greater magnitude<br />
( p < 0.05) than those of bonnethead<br />
sharks (Figure 4). In both species,<br />
E′ increased towards the tail, an effect<br />
that is amplified at higher values of κ,<br />
as indicated by a significant (p < 0.05)<br />
interaction term. The significance of<br />
the variation in E′ <strong>and</strong> E″ was determined<br />
using ANOVA, with species,<br />
position, <strong>and</strong> κ as main effects (JMP<br />
8.0.2., SAS Institute, Cary, NC).<br />
Since the data blacktip <strong>and</strong> bonnethead<br />
sharks were taken at a single amplitude<br />
of strain rate (2.0 mm s −1 ), we<br />
sought additional information about<br />
how E′ <strong>and</strong> E″ vary with changes in<br />
strain rate. We also wanted to test the<br />
hypothesis that the intervertebral capsule,<br />
which contains liquid under<br />
above-ambient pressure, uses its internal<br />
fluid pressure to alter the apparent<br />
E′ <strong>and</strong> E″ of the vertebral column. Because<br />
blacktip <strong>and</strong> bonnethead sharks<br />
were not available for these tests,<br />
spiny dogfish, Squalus acanthias, were<br />
used. Fresh 10-vertebrae segments<br />
were removed from the region of the<br />
first dorsal fin inthreedogfish. Each<br />
segment was pressure-clamped at the<br />
terminal vertebrae <strong>and</strong> end-loaded<br />
with bending moments, M (for experimental<br />
configuration, see Long et al.,<br />
2011). The bending motion was delivered<br />
via moment arms attached to a<br />
single-axis linear actuator using an<br />
MTS model Tytron 250 (Eden Prairie,<br />
MN) <strong>and</strong> a 50-N load cell. To test the<br />
effects of both f <strong>and</strong> κ on E′ <strong>and</strong> E″,<br />
each segment was bent sinusoidally at<br />
each combination of five f values <strong>and</strong><br />
three κ values. In addition, to test the<br />
effects of the integrity of the fluid-filled<br />
intervertebral joint capsule on E′ <strong>and</strong><br />
E″, we repeated this suite of tests<br />
after (a) puncturing a single joint capsule<br />
located in the middle of the segment<br />
<strong>and</strong> (b) puncturing three joint<br />
capsules, including the first one punctured<br />
<strong>and</strong> two adjoining capsules.<br />
Increases in f increased only E′<br />
( p < 0.05) while increases in κ<br />
increased both E′ <strong>and</strong> E″ (Figure 5).<br />
The only significant effect of puncturing<br />
the intervertebral capsule was<br />
when three capsules were punctured,<br />
<strong>and</strong> even then only E″ increased. The<br />
122 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 4<br />
Mechanical properties of the vertebral columns of sharks in sinusoidal bending. Points are means<br />
from three individuals. Error bars are the st<strong>and</strong>ard error of the mean. Size of the symbol indicates<br />
the relative magnitude of the curvature, κ. Significance level is indicated (n.s. = not significant,<br />
*p < 0.05, **p < 0.01, ***p < 0.001).<br />
significance of the variation in E′ <strong>and</strong><br />
E″ was determined using ANCOVA,<br />
with puncture, f, <strong>and</strong> κ as main effects<br />
<strong>and</strong> individual as the covariate (JMP<br />
8.0.2., SAS Institute, Cary, NC).<br />
In summary, the vertebral columns<br />
of sharks have mechanical properties<br />
that are highly variable. As species<br />
<strong>and</strong> anatomical position change, so,<br />
too, do E′ <strong>and</strong> E″. Within a given vertebral<br />
segment, the apparent storage<br />
modulus, E′, <strong>and</strong> the apparent loss<br />
modulus, E″, can be altered by the<br />
bending that they undergo. Increasing<br />
the segment’s curvature,κ, increases<br />
both E′ <strong>and</strong> E″; increasing the segment’s<br />
frequency of bending, f, increases the<br />
E′. Knowing the mechanical behavior<br />
of shark vertebral columns under realistic<br />
bending conditions creates specifications<br />
for BVCs.<br />
Designing BVCs<br />
To begin to underst<strong>and</strong> how to<br />
control the mechanical behavior of<br />
BVCs, we built two classes of sharkinspired<br />
BVC: (1) BVC with variable<br />
cone angle, Ξ (BVC Ξ ): vertebrae<br />
were created with variable Ξ <strong>and</strong> the<br />
BVC had constant joint length, j,<br />
<strong>and</strong> (2) BVC with variable joint length,<br />
j (BVC j ): vertebrae were created with<br />
aconstantΞ <strong>and</strong> the BVC had variable<br />
j. In addition to exploring the effects<br />
of the structures Ξ <strong>and</strong> j, we<br />
also varied the amount of cross-linking<br />
of the hydrogel material forming the<br />
joint. Thus, we explored the BVC<br />
“morphospace,” the variety of designs<br />
described by three dimensions: Ξ, j,<br />
<strong>and</strong> cross-linking. Part of this exploration<br />
involved the challenge of making<br />
composite structures that concatenate<br />
flexible <strong>and</strong> rigid elements. After fabrication<br />
<strong>and</strong> mechanical testing of both<br />
classes of BVC, we selected a single<br />
class, the BVC j , for performance testing<br />
in a tail-flapping aquatic robot.<br />
In the BVC Ξ , vertebrae were designed<br />
in software (SolidWorks,<br />
Dassault Systèmes SolidWorks Corp.,<br />
Concord, MA) to have the following<br />
values of Ξ: 15°, 30°, <strong>and</strong> 45° (Figure<br />
6). These values correspond to<br />
low, medium, <strong>and</strong> high values of Ξ<br />
measured in sharks (see Figure 3).<br />
The diameter, d, <strong>and</strong>axiallength,c,<br />
of the vertebrae were fixed at 1 cm<br />
for both. The j of the column was<br />
fixed at 0.25 cm.<br />
Vertebrae were fabricated with a<br />
rapid prototyper (Z-Corp, model<br />
310), which produced a porous, plaster<br />
part that was subsequently infiltrated<br />
with cyanoacrylate (EZ bond<br />
5cps, K&R International, Diamond<br />
Bar, CA). This process yielded vertebrae<br />
with mean compressive moduli,<br />
E (MPa) of 43, 50, <strong>and</strong> 61 for vertebrae<br />
with values of Ξ at 15°, 30°,<br />
<strong>and</strong> 45°, respectively. These values of<br />
E are within the range measured for<br />
shark vertebrae (Porter et al., 2006).<br />
Vertebrae of a given Ξ were assembled<br />
into a BVC Ξ in two stages. First,<br />
seven vertebrae were linked together,<br />
spaced at the fixed j, with eight horse<br />
hairs (E in tension of 900 MPa)<br />
arrayedaxially<strong>and</strong>affixed to the<br />
outer circumference of the vertebrae.<br />
These horse hairs served as first<br />
July/August 2011 Volume 45 Number 4 123
FIGURE 5<br />
Mechanical properties of the vertebral column vary as a function of cycle frequency, f, <strong>and</strong> the integrity<br />
of the intervertebral joint in the spiny dogfish, Squalus acanthias. Points are means from<br />
three individuals. Error bars are the st<strong>and</strong>ard error of the mean. Size of the symbol indicates<br />
the relative magnitude of the curvature, κ. Significance level is indicated (n.s. = not significant,<br />
*p < 0.05, **p < 0.01, ***p < 0.001).<br />
10% porcine gelatin fixed in 2.5%<br />
glutaraldehyde (Long et al., 2006).<br />
Vertebrae were slid onto the hydrogel,<br />
spaced evenly at the desired j, <strong>and</strong> affixed<br />
to the hydrogel with cyanoacrylate<br />
adhesive. A total of 12 different<br />
types of BVC j were produced, one<br />
type for each of the 12 different values<br />
of j. Three replicates of each type were<br />
produced <strong>and</strong> tested. Please note that<br />
in the BVC j horse hairs were omitted<br />
because at all but the smallest values<br />
of j, the hairs cut into the hydrogel during<br />
bending.<br />
approximations of the intervertebral<br />
ligaments found in the vertebral columns<br />
of sharks. Second, a 10% porcine<br />
gelatin solution was injected in<br />
between the vertebrae; the gelatin was<br />
solidified at 4° C. Once solidified, each<br />
BVC Ξ was then subjected to one<br />
of three fixation treatments: 0, 1%,<br />
or 5 % glutaraldehyde, a chemical<br />
agent that cross-links the collagen in<br />
the hydrogel. A total of nine different<br />
types of BVC Ξ were produced, with<br />
each possible pairwise combination of<br />
Ξ <strong>and</strong> glutaraldehyde concentration.<br />
In the BVC j , vertebrae were designed<br />
to have a Ξ of 90°, which created<br />
ring-shaped vertebrae (Figure 7).<br />
The d <strong>and</strong> c of the vertebrae were<br />
fixed at 0.5 <strong>and</strong> 1.0 cm, respectively.<br />
The overall length of the BVC j was<br />
fixed at 8.4 cm. As the number of<br />
nonterminal vertebrae were varied<br />
from 0 to 11, j varied from 720 to<br />
0.5 mm. These values of j created a<br />
range that extended below <strong>and</strong> above<br />
the range of j measured in sharks (see<br />
Figure 3).<br />
Vertebrae were milled from<br />
Delrin, a polyoxymethylene thermoplastic.<br />
Delrin has a compressive E<br />
of 3.1 GPa (Delrin Design Guide,<br />
Module III, from DuPont), which<br />
liesinthemiddleoftherangeof<br />
E values reported for shark vertebrae<br />
(Porter et al., 2006).<br />
The ring vertebrae had an inner diameter<br />
of 0.8 cm, which matched the<br />
outer diameter of hydrogels made from<br />
Mechanical Properties<br />
of BVCs<br />
The E′ <strong>and</strong> E″ of the BVCs were<br />
measured in two different kinds of<br />
sinusoidal bending test, which corresponded<br />
to the tests performed on<br />
sharks’ vertebral columns. In the<br />
BVC Ξ , 3-point bending tests were<br />
conducted in a manner identical with<br />
those on the blacktip <strong>and</strong> bonnethead<br />
sharks. In the BVC j , end-loaded bending<br />
tests were conducted in a manner<br />
identical with those on the spiny dogfish<br />
sharks. The E′ <strong>and</strong> E″ data for the<br />
BVC j have been analyzed previously<br />
(Long et al., 2011). In the analysis<br />
here, the data have been reanalyzed<br />
to calculate the mechanical work required<br />
to bend the BVC j <strong>and</strong> the mechanical<br />
work recovered as recoil.<br />
In the BVC Ξ , both E′ <strong>and</strong> E″ increased<br />
as the glutaraldehyde concentration<br />
increased, E′ <strong>and</strong> E″ decreased<br />
as the Ξ increased, <strong>and</strong> E′ increased<br />
<strong>and</strong> E″ decreased as κ increased (Figure8).Thesignificance<br />
of the variation<br />
in E′ <strong>and</strong> E ″ was determined<br />
using ANOVA, with glutaraldehyde<br />
concentration, Ξ, <strong>and</strong>κ. asmain<br />
effects (JMP 8.0.2., SAS Institute,<br />
Cary, NC).<br />
124 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 6<br />
BVCs (BVC Ξ ) with variable cone angles, Ξ, <strong>and</strong> constant joint length, j.<br />
FIGURE 7<br />
BVCs (BVC j ) with variable joint lengths, j, <strong>and</strong> constant cone angle, Ξ.<br />
Compared to the mechanical properties<br />
of shark vertebral columns, the<br />
BVC Ξ have values of E′ that have a<br />
wider range, overlapping the lower<br />
values <strong>and</strong> exceeding the sharks’ higher<br />
values by an order of magnitude (compare<br />
Figures 8 <strong>and</strong> 4). In contrast, the<br />
E″ values of the BVC Ξ overlap only<br />
with those of the bonnethead shark;<br />
the BVC Ξ has much lower values of<br />
E″ than either the blacktip or spiny<br />
dogfish shark. Moreover, the E′ for<br />
BVC Ξ decreases as κ increases; we<br />
measured the opposite trend in sharks<br />
(see Figures 4 <strong>and</strong> 5). Hence, the<br />
BVC Ξ is not a good biological model<br />
in this sense. Our hypothesis as to the<br />
source of this strain softening is that<br />
the horse hairs force the column to<br />
bend primarily by compression, rather<br />
than by a combination of tension <strong>and</strong><br />
compression.<br />
In the BVC j ,bothE′ <strong>and</strong> E″ decreased<br />
nonlinearly as j increased<br />
(Figure 9). Compared to the mechanical<br />
properties of dogfish vertebral<br />
columns, the BVC j span a nearly identical<br />
range of E′ <strong>and</strong> E″ values. The<br />
greatest sensitivity to changes in j occurred<br />
at the smallest values of j (Figure<br />
9) in the region that corresponds<br />
to the j measured in the vertebral columns<br />
of sharks (Figure 3). In data<br />
shown elsewhere (Long et al., 2011),<br />
E ′ <strong>and</strong> E ″ of the BVC j increased<br />
with increasing κ, just as in sharks<br />
(see Figure 5 herein). Moreover the E′<br />
increased with increasing f, aslikewise<br />
seen in sharks (Figure 5).<br />
The mechanical work to bend the<br />
BVC j increased with increasing κ <strong>and</strong><br />
increasing E′ (Figure 9). The mechanical<br />
work recovered as elastic recoil,<br />
W recoil , is a function of the resilience,<br />
R, which, over all the testing<br />
conditions <strong>and</strong> sizes of joints, averaged<br />
76%.<br />
BVCs in Aquatic Vehicles<br />
The flexible skeletons of sharks <strong>and</strong><br />
fish are inspiring the design of novel<br />
propulsive systems <strong>and</strong> aquatic vehicles<br />
(for review, see Fish, 2006; Long,<br />
2007, 2011). Fins with life-like flexibility<br />
have been built to propel a 0.7 m<br />
long robotic turtle (Long et al., 2006)<br />
<strong>and</strong>a0.4mlongroboticknifefish<br />
(Curet et al., 2011). Bodies with<br />
life-like flexibility have been built to<br />
July/August 2011 Volume 45 Number 4 125
FIGURE 8<br />
The mechanical properties of the BVCs (BVC Ξ ) with variable cone angles, Ξ, <strong>and</strong> constant joint<br />
length, j. Horizontal bars indicate the median, the lower <strong>and</strong> upper limits of the box indicate the<br />
25th <strong>and</strong> 75th percentiles, respectively, <strong>and</strong> the whiskers indicate the range.<br />
FIGURE 9<br />
The mechanical properties of the BVCs (BVC j ) with variable joint length, j, <strong>and</strong> constant cone<br />
angle, Ξ. Top row: points represent the means of E′ <strong>and</strong> E″ pooled across f <strong>and</strong> κ; error bars<br />
are one st<strong>and</strong>ard error of the mean. Bottom row: points are not pooled.<br />
propel a 0.7-m long robotic electric<br />
ray (Krishnamurthy et al., 2010),<br />
a 0.5-m long robotic trout (Kruusmaa<br />
et al., 2011), a 0.12-m long mechanical<br />
sunfish (McHenry et al., 1995),<br />
<strong>and</strong> a 0.5-m long mechanical pickerel<br />
(Conte et al., 2010). Of these<br />
self-propelled aquatic vehicles, only<br />
the mechanical pickerel has anything<br />
resembling a vertebral column:<br />
a piece of spring steel designed to<br />
release mechanical work to power<br />
accelerations.<br />
The BVC j presented here was invented<br />
to propel a surface swimming,<br />
0.3-m long tadpole robot (Long<br />
et al., 2006; Doorly et al., 2009),<br />
known as Tadro4 (Figure 10). BVC j<br />
were attached to a servo motor that<br />
created a sinusoidally varying pitching<br />
motion of a tail. That pitch bent the<br />
BVC j , creating a bending moment<br />
that propagated down the length of<br />
the BVC j in a traveling wave that, in<br />
turn, oscillated the terminal caudal<br />
fin. In this configuration, without<br />
distributed muscles, the BVC j acts<br />
as both a transmission system, transferring<br />
momentum from the servo motor<br />
to the caudal fin, <strong>and</strong> as a propeller,<br />
directly transferring momentum to<br />
the surrounding fluid.<br />
Since Tadro4 was built to behave<br />
reactively, with sensorimotor feedback<br />
systems creating foraging <strong>and</strong> predator<br />
avoidance, we needed a version that<br />
could be programmed to swim straight<br />
using a constant flapping frequency of<br />
the tail, f, <strong>and</strong> lateral amplitude of the<br />
caudal fin. That modified version of<br />
Tadro4 was called MARMT (Mobile<br />
Autonomous Robot for Mechanical<br />
Testing), <strong>and</strong> it had a hull length of<br />
17 cm <strong>and</strong> a tail length of 10 cm<br />
(Long et al., 2011).<br />
Outfitted with a given BVC j ,<br />
MARMT’s steady swimming performance<br />
was measured as swimming<br />
126 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 10<br />
The aquatic robot, Tadro4, is propelled by a BVC (BVC j ). Tadro4 is a fully-autonomous surfaceswimmer<br />
with a flattened circular body <strong>and</strong> propulsive undulatory tail. It is modeled after fish like<br />
the extinct Drepanaspis <strong>and</strong> the living electric ray, Narcine. Using sensory input from photoresistors<br />
<strong>and</strong> IR proximity detectors, Tadro4 searches for <strong>and</strong> swims up light gradients while avoiding<br />
collisions. Tadro4 is propelled by its submerged BVC j , which is wrapped in a thin membrane, attached<br />
to a caudal fin, <strong>and</strong> actuated by an oscillating servo motor. Tadro4 was developed by Doorly<br />
et al. (2009). Photo of Drepanaspis specimen 8462, American Museum of Natural History. Photo<br />
of adult Narcine is courtesy of Dr. Steve Kajiura.<br />
of E′ (Figure 11, top panel). The stride<br />
length of MARMT increased initially,<br />
doubling as E′ doubled, before tapering<br />
off.<br />
When the BVC j operates in the<br />
flapping tail, MARMT’s swimming<br />
performance is clearly linked to the<br />
mechanical properties of the BVC j , E′<br />
in the case shown here. Those mechanical<br />
properties are, in turn, under<br />
the control of the structure of the<br />
BVC j . Thus these experiments, taken<br />
together, demonstrate the functional<br />
relationship between the structure of<br />
the BVC j <strong>and</strong> the performance of a<br />
self-propelled aquatic robot.<br />
speed, U, <strong>and</strong> stride length, the slope<br />
ofthelineofU regressed onto f,<br />
which measures the distance the<br />
robot travels over one period of the<br />
flapping tail (Figure 11). As f increased<br />
for any BVC j , so, too, did the U. The<br />
BVC j with greater values of E′ produced<br />
a more rapid increase in U,<br />
over the same range of change in f,<br />
compared to BVC j , with smaller values<br />
Summary<br />
Using the morphology <strong>and</strong> mechanical<br />
properties of the vertebral<br />
columns of sharks as our biological<br />
target, we built <strong>and</strong> tested a series of<br />
BVC. The mechanical behavior of<br />
the BVCs, measured by the storage<br />
<strong>and</strong> loss moduli over a range of bending<br />
frequencies <strong>and</strong> curvatures, can be<br />
altered by changing (1) the material<br />
properties of the hydrogel that makes<br />
up the intervertebral joint, (2) the<br />
length of the intervertebral joint, or<br />
(3) the shape of the vertebrae. BVCs<br />
are sufficient to function as propulsive<br />
elements in swimming aquatic robots:<br />
in Tadro4 <strong>and</strong> MARMT the<br />
BVC converts a simple pitch oscillation<br />
from a servo motor into a wave<br />
of bending that drives the caudal fin<br />
laterally.<br />
Having identified variables that influence<br />
the mechanical behavior of<br />
BVCs, we offer a few observations for<br />
those wishing to build jointed, flexible<br />
biomimetic skeletons for use in flexible,<br />
flapping propulsive systems:<br />
(1) Design of biomimetic systems:<br />
Engineered systems that are much<br />
July/August 2011 Volume 45 Number 4 127
FIGURE 11<br />
Swimming performance of a surface-swimming robot, MARMT, propelled by a tail with the BVC j as<br />
the primary skeleton. MARMT is a version of Tadro4 (Figure 10) modified for mechanical testing<br />
over a range of flapping frequencies of the tail, f (Hz). For all types of BVC j tested, swimming speed<br />
of MARMT, U, increased linearly with increases in f (all R 2 values > 0.92). The rate of change of<br />
U with respect to f is the stride length (distance traveled per period of the flapping cycle); it was<br />
greatest with BVC j having larger storage moduli, E′. Three replicates of each kind of BVC j were<br />
tested (N = 36). Means (N = 12) are shown here. Data reanalyzed from Long et al., 2011.<br />
functional morphology. Next, test<br />
the morphology’s mechanical behavior<br />
under physiologically relevant testing<br />
conditions. Finally, build <strong>and</strong> test<br />
simple biomimetic models of the system<br />
that change just a single structural<br />
variable over a wide range. Repeat this<br />
process with different variables, ceteris<br />
paribus, until the designer knows<br />
which variables permit the natural system<br />
<strong>and</strong> its operational range to be<br />
mimicked or extended in biomimetic<br />
form.<br />
Acknowledgments<br />
We thank Carl Bertsche, Nicole<br />
Doorly, Carina Frias, Andres Gutierrez,<br />
Jonathan Hirokawa, Kira Irving, Doug<br />
Pringle, Foster Ranney, Hannah<br />
Rosenblum, Hassan Sahktah, Sonia<br />
Roberts, Elise Stickles, Josh Sturm,<br />
<strong>and</strong> Janese Trimaldi for their help in<br />
designing, building, <strong>and</strong> testing vertebral<br />
columns, BVCs, <strong>and</strong> the aquatic<br />
robots. This work was supported by<br />
the National Science Foundation<br />
of the USA (DBI-0442269 <strong>and</strong><br />
IOS-0922605).<br />
simpler than the targeted biological<br />
system can match <strong>and</strong> extend<br />
the targeted range of mechanical<br />
behaviors.<br />
(2) Control of mechanical properties:<br />
The spacing of rigid elements<br />
in a flexible matrix is more important<br />
than the shape of the rigid elements<br />
or the material properties of<br />
the flexible material.<br />
(3) Control of reconfiguration: Because<br />
of the strain- <strong>and</strong> strain-ratedependence<br />
of viscoelastic materials,<br />
no passive, flexible propulsive system,<br />
if its E′ <strong>and</strong> E″ matches that<br />
of the vertebral column of sharks,<br />
will produce constant motions over<br />
a wide range of motor inputs.<br />
This work is a straight-forward example<br />
of one method of biomimetic<br />
design (Fish, 2006; Long, 2007):<br />
describe, test, build, <strong>and</strong> test. Start<br />
by identifying a specific operational<br />
context—aquatic undulatory propulsion<br />
in this case. Then describe,<br />
quantitatively, the biological system’s<br />
Lead Author:<br />
John H. Long, Jr.<br />
Department of Biology,<br />
Vassar College<br />
124 Raymond Avenue,<br />
Poughkeepsie, NY 12604-0513<br />
Email: jolong@vassar.edu<br />
References<br />
Conte, J., Modarres-Sadeghi, Y., Watts, M.N.,<br />
Hover, F.S., & Triantafyllou, M.S. 2010. A<br />
fast-starting mechanical fish that accelerates<br />
at 40 m s-2. Bioinspir Biomim. 5(3):1-9.<br />
doi: 10.1088/1748-3182/5/3/035004.<br />
Curet, O.M., Patankar, N.A., Lauder, G.V.,<br />
& MacIver, M.A. 2011. Aquatic manoeuvering<br />
with counter-propagating waves: A novel<br />
128 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
locomotive strategy. J R Soc Interface.<br />
8(60):1041-50. (10.1098/rsif.2010.0493).<br />
Doorly, N., Irving, K., McArthur, G.,<br />
Combie, K., Engel, V., Sakhtah, H., …<br />
Long, J.H., Jr. 2009. Biomimetic evolutionary<br />
analysis: Robotically-simulated vertebrates in<br />
a predator-prey ecology. In: Proc. 2009<br />
IEEE Symp. Artificial Life. 147-54. Nashville,<br />
TN: IEEE (Institute of Electrical <strong>and</strong><br />
Electronic Engineers). doi: 10.1109/<br />
ALIFE.2009.4937706.<br />
Fish, F.E. 2006. Limits of nature <strong>and</strong><br />
advances of technology: What does biomimetics<br />
have to offer to aquatic robots Appl.<br />
Bionics Biomech. 3(1):49-60. doi: 10.1533/<br />
abbi.2004.0028.<br />
Grotmol, S., Kryvi, H., Keynes, R., Krossoy,<br />
C., Nordvik, K., & Totl<strong>and</strong>, G.K. 2006.<br />
Stepwise reinforcement of the notochord<br />
<strong>and</strong> its intersection with the myoseptum:<br />
An evolutionary path leading to the development<br />
of the vertebra J Anat. 209:339-57.<br />
doi: 10.1111/j.1469-7580.2006.00618.x.<br />
Grotmol, S., Kryvi, H., Nordvik, K., &<br />
Totl<strong>and</strong>, G.K. 2003. Notochord segmentation<br />
may lay down the pathway for the<br />
development of the vertebral bodies in the<br />
Atlantic salmon. Anat Embryol. 207:263-72.<br />
doi: 10.1007/s00429-003-0349-y.<br />
Koob, T.J., & Long, J.H., Jr. 2000. The<br />
vertebrate body axis: Evolution <strong>and</strong> mechanical<br />
function. Am Zool. 40(1):1-18.<br />
doi: 10.1668/0003-1569(2000)040[0001:<br />
TVBAEA]2.0.CO;2.<br />
Krishnamurthy, P., Khorrami, F., de Leeuw,<br />
J., Porter, M., Livingston, K., & Long,<br />
J.H., Jr. 2010. An electric ray inspired<br />
biomimetic autonomous underwater vehicle.<br />
Am Control Conf. 2010:5224-9.<br />
Kruusmaa, M., Salumäe, T., Toming, G.,<br />
Ernits, A., & Ježov, J. 2011. Swimming speed<br />
control <strong>and</strong> on-board flow sensing of an<br />
artificial trout. In: Proceedings of IEEE<br />
International Conference of Robotics <strong>and</strong><br />
Automation (IEEE ICRA 2011). Shanghai,<br />
China: IEEE (Institute of Electrical <strong>and</strong><br />
Electronic Engineers).<br />
Long, J.H., Jr. 1992. Stiffness <strong>and</strong> damping<br />
forces in the intervertebral joints of blue marlin,<br />
(Makaira nigricans). J Exp Biol. 162:131-55.<br />
Long, J.H., Jr. 2007. Biomimetic robotics:<br />
Building autonomous, physical models to test<br />
biological hypotheses. Proc. Inst. Mech. Eng,<br />
C, J. Mech. Eng. Sci. 221:1193-200.<br />
Long, J.H., Jr. 2011. <strong>Biomimetics</strong>: Robotics<br />
based on fish swimming. In: Encyclopedia of<br />
Fish Physiology: From Genome to Environment,<br />
ed. Farrell, A.P., 1, 603-12. San Diego:<br />
Academic Press.<br />
Long, J.H., Jr., Koob-Emunds, M., Sinwell,<br />
B., & Koob, T.J. 2002. The notochord of<br />
hagfish, Myxine glutinosa: Viscoelastic properties<br />
<strong>and</strong> mechanical functions during steady<br />
swimming. J Exp Biol. 205:3819-31.<br />
Long, J.H., Jr., Koob, T.J., Irving, K.,<br />
Combie, K., Engel, V., Livingston, N., …<br />
Schumacher, J. 2006. Biomimetic evolutionary<br />
analysis: Testing the adaptive value of vertebrate<br />
tail stiffness in autonomous swimming<br />
robots. J Exp Biol. 209(23):4732-46.<br />
doi: 10.1242/jeb.02559.<br />
Long, J.H., Jr., Krenitsky, N., Roberts, S.,<br />
Hirokawa, J., de Leeuw, J., & Porter, M.E.<br />
2011. Testing biomimetic structures in<br />
bioinspired robots: How vertebrae control<br />
the stiffness of the body <strong>and</strong> the behavior<br />
of fish-like swimmers. Integr Comp Biol.<br />
51(1):158-75. doi: 10.1093/icb/icr020.<br />
Long, J.H., Jr., Porter, M.E., Liew, C.W., &<br />
Root, R.G. 2010. Go reconfigure: How fish<br />
change shape as they swim <strong>and</strong> evolve. Integr<br />
Comp Biol. 50(6):1120-39. doi: 10.1093/icb/<br />
icq066.<br />
McHenry, M.J., Pell, C.A., & Long, J.H., Jr.<br />
1995. Mechanical control of swimming speed:<br />
Stiffness <strong>and</strong> axial wave form in an undulatory<br />
fish model. J Exp Biol. 198:2293-305.<br />
Porter, M.E., Beltran, J.L., Koob, T.J., &<br />
Summers, A.P. 2006. Material properties<br />
<strong>and</strong> biochemical composition of mineralized<br />
vertebral cartilage in seven elasmobranch<br />
species (Chondrichthyes). J Exp Biol.<br />
209:2920-8. doi: 10.1242/jeb.02325.<br />
Porter, M.E., & Long, J.H., Jr. 2010.<br />
Vertebrae in compression: Mechanical<br />
behavior of arches <strong>and</strong> centra in the gray<br />
smooth-hound shark (Mustelus californicus).<br />
J Morph. 271:366-75.<br />
Porter, M.E., Roque, C.M., & Long, J.H., Jr.<br />
2009. Turning maneuvers in sharks: Predicting<br />
body curvature from body <strong>and</strong> vertebral<br />
morphology. J Morph. 270:954-65.<br />
doi: 10.1002/jmor.10732.<br />
Schmitz, R.J. 1995. Ultrastructure <strong>and</strong><br />
function of cellular components of the<br />
intercentral joint in the percoid vertebral<br />
column. J Morph. 226(1):1-24.<br />
doi: 10.1002/jmor.1052260102.<br />
Summers, A.P, & Long, J.H., Jr. 2006. Skin<br />
<strong>and</strong> bones, sinew <strong>and</strong> gristle: The mechanical<br />
behavior of fish skeletal tissues. In: Fish<br />
Biomechanics, eds. Shadwick, R.E., & Lauder,<br />
G.V., 141-77. San Diego: Academic Press.<br />
Symmons, S. 1979. Notochordal <strong>and</strong> elastic<br />
components of the axial skeleton of fishes<br />
<strong>and</strong> their functions in locomotion. J Zool.<br />
189(2):157-206. doi: 10.1111/j.1469-<br />
7998.1979.tb03958.x.<br />
July/August 2011 Volume 45 Number 4 129
PAPER<br />
Lateral-Line-Inspired Sensor Arrays for<br />
Navigation <strong>and</strong> Object Identification<br />
AUTHORS<br />
Vicente I. Fern<strong>and</strong>ez<br />
Audrey Maertens<br />
Department of Mechanical<br />
Engineering, Massachusetts<br />
Institute of <strong>Technology</strong> (MIT)<br />
Frank M. Yaul<br />
Department of Electrical<br />
Engineering <strong>and</strong> Computer<br />
Science, Massachusetts<br />
Institute of <strong>Technology</strong><br />
Jason Dahl<br />
Department of Mechanical<br />
Engineering, Massachusetts<br />
Institute of <strong>Technology</strong><br />
Jeffrey H. Lang<br />
Department of Electrical<br />
Engineering <strong>and</strong> Computer<br />
Science, Massachusetts<br />
Institute of <strong>Technology</strong><br />
Michael S. Triantafyllou<br />
Department of Mechanical<br />
Engineering, Massachusetts<br />
Institute of <strong>Technology</strong><br />
Introduction<br />
The lateral line organ is a unique<br />
sensory mechanism in fish, enabling<br />
complex behaviors based on the interpretation<br />
of local fluid mechanics. For<br />
example, the blind Mexican cavefish<br />
(Astyanax fasciatus) is able to navigate<br />
new environments at high speed<br />
without collision <strong>and</strong> to identify <strong>and</strong><br />
remember features of the environment<br />
(Montgomery et al., 2001; von<br />
Campenhausen et al., 1981). This surprising<br />
feat is accomplished relying<br />
primarily on its lateral line organ for<br />
ABSTRACT<br />
The lateral line is a critical component of fish sensory systems, found to affect<br />
numerous aspects of behavior, including maneuvering in complex fluid environments<br />
with poor visibility. This sensory organ has no analog in modern ocean vehicles,<br />
despite its utility <strong>and</strong> ubiquity in nature, <strong>and</strong> could fill the gap left by sonar<br />
<strong>and</strong> vision systems in turbid, cluttered environments.<br />
To emulate the lateral line <strong>and</strong> characterize its object-tracking <strong>and</strong> shape recognition<br />
capabilities, a linear array of pressure sensors is used along with analytic<br />
models of the fluid in order to determine position, shape, <strong>and</strong> size of various objects<br />
in both passive <strong>and</strong> active sensing schemes. We find that based on pressure information,<br />
tracking a moving cylinder can be effectively achieved via a particle filter.<br />
Using principal component analysis, we are also able to reliably distinguish between<br />
cylinders of different cross section <strong>and</strong> identify the critical flow signature information<br />
that leads to the shape identification. In a second application, we employ pressure<br />
measurements on an artificial fish <strong>and</strong> an unscented Kalman filter to successfully<br />
identify the shape of an arbitrary static cylinder.<br />
Based on the experiments, we conclude that a linear pressure sensor array for<br />
identifying small objects should have a sensor-to-sensor spacing of less than 0.03<br />
(relative to the length of the sensing body) <strong>and</strong> resolve pressure differences of at<br />
least 10 Pa. These criteria are used in the development of an artificial lateral line<br />
adaptable to the curved hull of an underwater vehicle, employing conductive polymer<br />
technologies to form a flexible array of small pressure sensors.<br />
Keywords: underwater sensing, artificial lateral line, pressure sensor arrays<br />
sensory feedback. All fish have this<br />
organ, although not all use it to the<br />
extent of the blind cavefish (see<br />
Montgomery et al., 2001). Many fundamental<br />
behaviors in fish have been<br />
identified by biologists to be lateralline-mediated,<br />
including tracking<br />
prey by their wake (Pohlmann et al.,<br />
2004) <strong>and</strong> recognizing nearby physical<br />
objects (von Campenhausen et al.,<br />
1981). Although the lateral line is a biological<br />
organ, its functionality would<br />
translate well to needs in underwater<br />
vehicle design, such as with object detection,<br />
navigation, <strong>and</strong> flow sensing.<br />
The lateral line organ consists of<br />
two subsystems responding separately<br />
to velocity <strong>and</strong> to pressure gradients<br />
on the surface of the fish, both using<br />
the same underlying sensory element,<br />
the neuromast, which responds directly<br />
to flow velocity. In the case of the<br />
pressure gradient measurements, these<br />
velocity sensors are embedded underneath<br />
the skin in canals periodically<br />
opening via pores to the external flow<br />
(van Netten, 2006). Biological studies<br />
have demonstrated the ability of the<br />
fish to use their lateral line to interrogate<br />
their environment through both<br />
active <strong>and</strong> passive sensing. In active<br />
sensing, a fish uses the flow generated<br />
by repeatedly gliding near new objects<br />
at a short distance (von Campenhausen<br />
130 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
et al., 1981) to interrogate their shape.<br />
In particular, blind cave fish were found<br />
to detect <strong>and</strong> discriminate between<br />
stationary objects or openings of different<br />
geometries in still water (von<br />
Campenhausen et al., 1981; Weissert<br />
& von Campenhausen, 1981; Burt de<br />
Perera, 2004). During passive sensing, a<br />
moving object generates a flow field that<br />
is detected by the stationary lateral line of<br />
astillfish. Vogel <strong>and</strong> Bleckmann (2000)<br />
demonstrated that goldfish use their<br />
lateral line to passively detect <strong>and</strong> discriminate<br />
the size, velocity <strong>and</strong> shape of<br />
passing rods in still water. One key element<br />
emerging from these studies is that<br />
in both active <strong>and</strong> passive settings, the<br />
object identification behavior is tied to<br />
the pressure gradient measurements of<br />
the lateral line <strong>and</strong> appears independent<br />
of the velocity measurements.<br />
While the lateral line organ’s role in<br />
many fish behaviors is becoming progressively<br />
better understood, there<br />
still remain many questions about the<br />
level of information detail available via<br />
the lateral line. The recent advances in<br />
underst<strong>and</strong>ing the central processing<br />
of the lateral line have all been with respect<br />
to the oscillating dipole stimulus<br />
(Curcic-Blake & van Netten, 2006,<br />
Goulet et al., 2007), which is inapplicable<br />
to both passive <strong>and</strong> active object<br />
sensing. In the passive case, a dipole<br />
model neglects the wake that forms<br />
about a moving object. In the active<br />
sensing situation, the interaction betweenthetwobodiesinstillwater<br />
leads to very different pressure distributions.<br />
Due to the difficulty in studying<br />
the neurological aspects of how a<br />
fish utilizes the stimulus of the lateral<br />
line, an artificial representation of the<br />
lateral line can give insight on the use<br />
of pressure sensing for biologically inspired,<br />
engineered applications. This<br />
paper investigates the ability of pressure<br />
sensor arrays, emulating the lateral<br />
line organ, to distinguish the shapes of<br />
physical objects through both active<br />
<strong>and</strong> passive sensing, while also identifying<br />
physical requirements for a constructed<br />
artificial lateral line to be used<br />
for underwater vehicle navigation applications.<br />
The results demonstrate<br />
the promise of a lateral-line-like sensor<br />
for autonomous underwater vehicles<br />
(AUVs) in severe environments.<br />
Previous approaches with artificial<br />
lateral lines have taken a biomimetic<br />
approach of reproducing the lateral<br />
line at the structural level, by recreating<br />
a sophisticated neuromast-like<br />
cantilevered sensors <strong>and</strong> using a canal<br />
system similar to that of the fish in<br />
order to obtain pressure gradient<br />
measurements (Chen et al., 2006;<br />
Yang et al., 2008). Yet, at its core,<br />
the portion of the biological lateral<br />
line associated with the behaviors of<br />
interest can be viewed as a more elementary<br />
processing unit, taking distributed<br />
pressure measurements as<br />
inputs <strong>and</strong>, via some processing that<br />
remains to be understood, extracting<br />
information about the environment.<br />
Thus, the current work instead takes<br />
a bioinspired approach in which the<br />
lateral line is abstracted by a linear<br />
pressure sensor array <strong>and</strong> the resulting<br />
information is processed relying on<br />
modern inference algorithms. In addition<br />
to providing potential clues as to<br />
the fundamental processes taking<br />
place in the biological sensory system,<br />
the present abstracted approach, by directly<br />
focusing on the desired functionality<br />
of the artificial lateral line,<br />
bears significant engineering benefits.<br />
As a sparse artificial lateral line can be<br />
readily implemented, making use of<br />
diaphragm-based pressure sensors, the<br />
initial focus is able to shift from sensor<br />
fabrication to the inference <strong>and</strong> processing.<br />
In turn, insights from the inference<br />
results are then used to develop<br />
specifications for optimal sensor design.<br />
As another benefit of using pressure<br />
sensors to approximate the lateral<br />
line, note that the available pressure<br />
information includes, but is not limited<br />
to, discrete pressure gradients (as in the<br />
lateral line canal system).<br />
First, we present two experiments<br />
demonstrating passive object detection<br />
through the use of an artificial lateral<br />
line. In this study, an artificial lateral<br />
line is used to track the size <strong>and</strong> location<br />
of a moving cylinder <strong>and</strong> separately to<br />
identify the shape of a moving cylinder<br />
between known possibilities. Second,<br />
we present an experiment demonstrating<br />
active object detection, where a<br />
moving artificial lateral line is used to<br />
identify the completely unknown<br />
shape of an object. In this study, the<br />
sensor arrangement is studied with<br />
relation to the applied problem in<br />
order to identify constraints of sensor<br />
spacing. Finally, we discuss the construction<br />
<strong>and</strong> testing of a prototype artificial<br />
lateral line using conductive<br />
polymer materials for use in object detection<br />
<strong>and</strong> navigation applications.<br />
Passive Cylinder<br />
Identification<br />
In passive object identification, an<br />
interaction between the object <strong>and</strong> the<br />
environment generates a pressure field<br />
that stimulates the sensor array. In the<br />
examples considered here, this interaction<br />
is between a moving cylinder<br />
<strong>and</strong> still water. The sensor array is also<br />
stationary in the water. The problem of<br />
identifying a moving cylinder via the<br />
pressure along a lateral-line-like sensor<br />
is addressed in two stages. First, we discuss<br />
an approach for determining the<br />
position <strong>and</strong> size of a cylinder, focusing<br />
on a single shape, <strong>and</strong> subsequently, we<br />
consider the question of distinguishing<br />
between different shapes.<br />
July/August 2011 Volume 45 Number 4 131
Experimental Setup<br />
Two experiments were used to analyze<br />
passive object detection techniques.<br />
In the first experimental setup<br />
(see Figure 1A), the position tracking<br />
ofacircularcylinderwasexperimentally<br />
tested using off-the-shelf pressure<br />
sensors (Honeywell 19C015PG4K)<br />
arranged in a linear array. Seven sensors<br />
were embedded along a 46-cm<br />
square flat plate spaced 1.9 cm<br />
(0.75 inch) apart, as shown in Figure<br />
1A. Using a mechanical stage<br />
with millimeter accuracy in the<br />
x-y plane, a 32-mm diameter circular<br />
cylinder was passed in front of the sensor<br />
array at constant velocity <strong>and</strong><br />
known position. These experiments<br />
were performed in a 3.6 m × 1.2 m ×<br />
1.2 m water tank at the Singapore-<br />
MIT Alliance for Research <strong>and</strong> <strong>Technology</strong><br />
(SMART) Centre. Pressure<br />
signals were amplified by a factor of<br />
1000 immediately adjacent to the sensors<br />
using AD620 instrumentation<br />
amplifiers <strong>and</strong> were next sampled<br />
using an NI USB-6210 analog-todigital<br />
converter. The sensors were<br />
calibrated using static pressure.<br />
In the second passive detection experimental<br />
setup, as with the first, a<br />
verticallymountedcylinderpassesin<br />
front of a stationary linear pressure<br />
sensor array at constant velocity (Figure<br />
1B). In this experiment, four<br />
Honeywell 242PC15M pressure sensors<br />
were enclosed in a streamlined<br />
cylindrical enclosure. The sensors<br />
were mounted rigidly with a small diameter<br />
outlet in order to minimize<br />
noise. Due to onboard amplification,<br />
the sensors were only amplified by a<br />
factor of 10 before data acquisition<br />
(NI USB 6210). These experiments<br />
were performed in the MIT towing<br />
tank facility (36.6 m × 2.4 m ×<br />
1.2 m). As the goal of these experiments<br />
<strong>and</strong> subsequent analysis is to<br />
FIGURE 1<br />
Schematics of two experimental apparatuses for passive object identification <strong>and</strong> sample pressure<br />
data from one apparatus. (A) A moving circular cylinder that is towed past an array of seven stationary<br />
pressure sensors on a flat wall. (B) An array of four pressure sensors mounted on a stationary streamlined<br />
body with a square cylinder towed past the body. (C) A representative data set corresponding to<br />
the layout in part B, filtered with a cut-off frequency of 100 Hz. (Color versions of figures available<br />
online at: http://www.ingentaconnect.com/content/mts/mtsj/2011/00000045/00000004.)<br />
distinguish between two cylinder<br />
cross-section shapes known apriori,<br />
both square <strong>and</strong> round cross-section<br />
cylinders were towed past the sensor<br />
array. In order to provide a reliable<br />
common denominator for shape classification,<br />
the data was gathered from<br />
a variety of speeds, cylinder sizes, <strong>and</strong><br />
distances from the sensor array as<br />
shown in Table 1. Calibration of the<br />
pressure sensors was completed in situ<br />
by comparing the amplitude of pressure<br />
oscillations due to regularly generated<br />
waves, using a wavemaker, to the<br />
theoretical amplitude based on linearized<br />
surface wave theory.<br />
Passive Cylinder Tracking<br />
Tracking the position <strong>and</strong> size of a<br />
circular cylinder can be interpreted as<br />
the first step or lowest level of shape<br />
identification. This task is complicated<br />
by the wake that forms when a cylinder<br />
moves. In order to track the position<br />
132 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
<strong>and</strong> size, an accurate <strong>and</strong> preferably<br />
simple model for relating the cylinder<br />
state to the pressure measurements is<br />
needed as well as a technique for estimating<br />
the state in real time. Potential<br />
flow solutions are an immediate c<strong>and</strong>idate<br />
for modeling an object in a fluid<br />
since they provide an analytic solution<br />
with relatively few parameters to define<br />
the complete flow field.<br />
The base model chosen for tracking<br />
a circular cylinder in the present experimentsisaRankinehalf-bodyin<br />
potential flow. This well-known structure<br />
is described by the superposition<br />
of a source <strong>and</strong> a free stream. A mirror<br />
image of the half-body provides the<br />
necessary boundary conditions to<br />
define the wall in the model. This<br />
model was found to approximate the<br />
pressure field of a cylinder with a<br />
wake when compared with pressure<br />
measurements obtained through a viscous<br />
numerical simulation. The radius<br />
of the cylinder was taken as the distance<br />
from the location of the source<br />
to the nearby forward stagnation<br />
point in the flow. It is important to<br />
note that the dipole model frequently<br />
used in a vibratory setting in studies<br />
with fish (e.g., Coombs & Janssen,<br />
1990) does not model the flow well<br />
about a steadily translating body due<br />
to the absence of the wake in the<br />
model. It is not surprising that a potential<br />
flow model like the Rankine halfbody<br />
approximates the experimental<br />
situation, since the flow outside the<br />
separated region in the wake can be<br />
reasonably considered as potential<br />
flow. Although the real situation is<br />
not steady <strong>and</strong> the pressure far from<br />
the cylinder was found to converge<br />
more slowly than in numerical simulations,<br />
a Rankine model captures the<br />
amplitude <strong>and</strong> shape of the pressure response<br />
well in the immediate vicinity<br />
of the cylinder. Using the velocity potential<br />
from this model, the pressure at<br />
the sensor locations is calculated using<br />
Bernoulli’s equation.<br />
Although the pressure sensors used<br />
in these experiments measure pressure<br />
with respect to a fix reference value,<br />
one key element in the analysis for<br />
tracking a cylinder is to consider the<br />
difference in pressure between adjacent<br />
pressure sensors instead of the absolute<br />
pressure. Long-wavelength,<br />
small-amplitude disturbances at the<br />
free surface of the tank, caused by the<br />
motion of test cylinder, were found to<br />
cause pressure fluctuations on the<br />
order of 20 Pa, significant compared<br />
to the 100 Pa pressure signals from<br />
the cylinder. Due to the long wavelength,<br />
taking the difference in pressure<br />
signals effectively removes this contamination.<br />
As mentioned in the Introduction,<br />
the portion of the lateral line that<br />
has been associated with object identification<br />
responds to pressure differences<br />
(Coombs, 2001), so it is interesting<br />
that the use of pressure differences is<br />
necessary even when absolute pressure<br />
measurements are available.<br />
For the purpose of tracking the<br />
position <strong>and</strong> size of a cylinder, the forward<br />
model relating the hidden parameters<br />
of interest to the available<br />
measurements has been defined by<br />
the Rankine half-body representation<br />
of the cylinder <strong>and</strong> wake, in conjunction<br />
with the Bernoulli equation to<br />
obtain the pressure at the sensors. To<br />
tackle the inverse problem <strong>and</strong> estimate<br />
the cylinder state based on the<br />
pressure measurements, several additional<br />
issues require consideration.<br />
First, the variables of interest are the<br />
position (x <strong>and</strong> y as labeled in Figure<br />
1A) <strong>and</strong> radius (R) of the cylinder.<br />
In order to cast the problem in the<br />
form of a hidden Markov model (see,<br />
for example, Cappe et al., 2005), in<br />
which the state at each timestep depends<br />
solely on the previous iteration,<br />
the velocity (u) must to be included in<br />
the state vector. Using this hidden<br />
Markov model formulation allows for<br />
the use of well-known estimation algorithms<br />
for solving the inverse problem,<br />
such as the Kalman filter extensions for<br />
nonlinear models (Gelb, 1974; Julier<br />
& Uhlmann, 2004). Second, the forward<br />
model requires a strict constraint<br />
on two of the state variables, R <strong>and</strong> y, in<br />
order to correspond to a physical realization:<br />
0 < R < y. Unfortunately, there<br />
is no natural boundary in the measurement<br />
model that corresponds to these<br />
physical constraints, as the potential<br />
flow model works uniformly well <strong>and</strong><br />
there is no sudden shift in the predicted<br />
pressure. As a consequence of this, an estimation<br />
technique such as the extended<br />
Kalman filter fails when it is applied to<br />
naturally noisy experimental data.<br />
The successful tracking of the position<br />
<strong>and</strong> size of a cylinder was accomplished<br />
instead using a particle filter<br />
(see Branko et al., 2004). This general<br />
inverse problem technique operates by<br />
using a number of samples, which represent<br />
possible states. To each sample,<br />
there is a corresponding weight that<br />
roughly tracks the quality of that sample.<br />
At each time step, each sample is<br />
updated to a new value r<strong>and</strong>omly chosen<br />
from a distribution based on the<br />
previous sample, <strong>and</strong> the weight of<br />
that sample is updated based on how<br />
likely it is to have generated the measurements.<br />
The key advantage of this<br />
approach is that, in specifying the<br />
noise distributions that govern the update<br />
of the samples, it is possible to<br />
limit the range of state vectors to<br />
those that correspond to a physical system.<br />
In the present case, the x position<br />
<strong>and</strong> the velocity u are assumed to have<br />
Gaussian noise distributions, which is<br />
the typical assumption for an unconstrained<br />
variable where the noise is<br />
July/August 2011 Volume 45 Number 4 133
generated by the accumulation of<br />
many small r<strong>and</strong>om influences. The<br />
noise distribution associated with the<br />
y position is assumed to be a lognormal<br />
distribution, <strong>and</strong> for the radius<br />
it is assumed to be a Gaussian distribution<br />
that is truncated at 0 <strong>and</strong> y <strong>and</strong> renormalized.<br />
These distributions satisfy<br />
the constraints between the variables<br />
<strong>and</strong> have intuitive limiting behaviors.<br />
When the mean is far from zero with<br />
respect to the st<strong>and</strong>ard deviation, the<br />
lognormal is approximately symmetric<br />
about the mean. In the case of the<br />
truncated Gaussian distribution, if<br />
the previous value of the radius is<br />
greater than the new y value, then the<br />
distribution is weighted heavily to<br />
larger values <strong>and</strong> has a maximum<br />
likelihood of y. Themaincostto<br />
using a particle filter is that it frequently<br />
requires a large number of<br />
samples to generate repeatable accurate<br />
results. For this application, it was<br />
found that approximately 150 particles<br />
are sufficient.<br />
A typical result of the particle filter<br />
implementation on experimental data<br />
is shown in Figure 2. In the figure,<br />
the black line corresponds to the estimate<br />
of the variable being output<br />
from the filter, <strong>and</strong> the red line corresponds<br />
to the true value. Note that<br />
the filter is initialized with a uniform<br />
distribution <strong>and</strong> does not begin to<br />
converge until there are significant<br />
pressure differences. As seen from the<br />
results, the particle filter estimate generally<br />
tracks the x position <strong>and</strong> radius<br />
well but underestimates the y position.<br />
The velocity is also underestimated in<br />
this example, although both overestimates<br />
<strong>and</strong> underestimates of the velocity<br />
were recorded in general. These<br />
results demonstrate that using a steady<br />
potential flow model, the radius <strong>and</strong><br />
aspects of the position can be accurately<br />
tracked. With further refinements of<br />
the model, the underestimation of<br />
the y position could likely be overcome.<br />
The use of a simple analytical<br />
model here makes it more likely that<br />
equivalent information is also available<br />
to a fish via its lateral line, <strong>and</strong> similarly<br />
likely that it could be adapted for<br />
quickly tracking objects near a hull.<br />
TABLE 1<br />
Matrix of experiments for passive cylinder shape classification.<br />
Passive Cylinder<br />
Shape Classification<br />
The second set of experiments is<br />
concerned with the more complicated<br />
problemofshapeclassification. Fundamentally,<br />
we consider whether it is<br />
possible to distinguish between two<br />
similar shapes after a single pass of an<br />
object past the sensor array, using a<br />
robust test for the decision. The test<br />
in question must identify whether<br />
a passing cylinder has a square or<br />
round cross section. Since the decision<br />
is based on a single pass, the stochastic<br />
nature of the flow in the wake of the<br />
cylinder must be overcome.<br />
The large data set of pressure obtained<br />
from the experimental setup in<br />
Figure 1B, corresponding to cylinders<br />
of different sizes, velocities, distances,<br />
<strong>and</strong> shapes provides a broad range of<br />
parameters for verifying the ability of<br />
any decision rule. In addition, the<br />
large number of runs for each point<br />
in the test matrix allows for an accurate<br />
comparison of the mean pressure responses.<br />
As shown in Figure 3, there<br />
are distinct differences between the<br />
average pressure measured due to the<br />
circular <strong>and</strong> square cross sections; analogous<br />
differences exist for comparisons<br />
in all the tested sizes <strong>and</strong> velocities. It is<br />
notable that the most easily distinguished<br />
differences in the pressure signals<br />
occur after the zero crossing point<br />
(circled).<br />
The analysis presented here develops<br />
a test to determine, after the fact,<br />
whether a square or circular cylinder<br />
has passed the sensors. Based on the<br />
observations of the mean pressure<br />
traces, it is tempting to choose features<br />
such as the location of the minimum<br />
pressure to classify the shape of the<br />
cylinder. This approach is limited in<br />
that the features would be localized<br />
in the pressure trace <strong>and</strong>, therefore,<br />
highly susceptible to the type of noise<br />
observed in Figure 1C. Instead, we use<br />
a principal component analysis (PCA)<br />
to identify a limited number of features<br />
in the form of weighted sums over the<br />
full pressure data from a single run.<br />
This type of feature depends on all the<br />
data <strong>and</strong> is less susceptible to localized<br />
perturbations such as those in the<br />
wake. By carefully applying the PCA<br />
to a training data subset, the results<br />
can be directed towards a decision rule<br />
in classifying the shape of the cylinders.<br />
In implementation, PCA works as a<br />
singular value decomposition of the<br />
7.62 cm Diameter 5.08 cm Diameter<br />
0.5 m/s 0.75 m/s 0.5 m/s 0.75 m/s<br />
Square 100 runs 100 runs 100 runs 100 runs<br />
Round 100 runs 100 runs 100 runs 100 runs<br />
Each of the experiments listed corresponds to the cylinder passing at a distance of 5.1 mm from the<br />
sensors. Additional data, not listed in the table, were collected with the 5.08-cm-diameter cylinders passing<br />
1.27 cm from the sensors at 0.75 m/s.<br />
134 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 2<br />
Results of a cylinder tracking experiment using a particle filter. In the first four parts, each vertical<br />
slice in the surface corresponds to the probability density function for that variable at the time of<br />
the slice, given the previous pressure measurements. In these sections, the red line represents the<br />
true value of the parameter at each time instant, <strong>and</strong> the black line corresponds to the expected<br />
value of the distribution. The final part shows the corresponding pressure difference measurements,<br />
filtered at 60 Hz for anti-aliasing.<br />
sample covariance matrix of the data<br />
(Jolliffe, 2002; Jackson, 2003). One important<br />
detail in the implementation is<br />
that the mean of each initial feature (a<br />
sample point in the pressure trace)<br />
must be removed. This mean is taken<br />
across all training data, not over each<br />
class. With the mean removed, the<br />
sample covariance is straightforward<br />
to calculate. Each resulting principal<br />
component is a vector that accounts<br />
for the maximum possible variance,<br />
subject to having unit area <strong>and</strong> being<br />
uncorrelated with all the previous principal<br />
components.<br />
Since only four sensors were available<br />
for this experiment (as opposed<br />
to seven in the previous cylinder tracking<br />
experiments) <strong>and</strong> the cylinder<br />
stimuli are translating at constant velocity,<br />
an equivalence between the<br />
pressure time history <strong>and</strong> the spatial<br />
pressure field was used in the analysis.<br />
This approximation is very accurate in<br />
regions in front of the cylinder, but less<br />
so in the unsteady wake areas. In order<br />
to compare data sets with different cylinder<br />
velocities on the same spatial<br />
scale, the 0.5 m/s velocity data is<br />
down-sampled appropriately. In addition,<br />
the data is aligned by the point<br />
at which the pressure crosses zero<br />
(marked in Figure 3), which forms a<br />
reliable internal placement marker.<br />
Finally, the PCA approach does not<br />
naturally produce features which capture<br />
the cylinder shape. The decomposition<br />
of the covariance matrix<br />
generates principal components that<br />
capture the majority of the data variance<br />
in the first few components.<br />
Therefore, the resulting principal components<br />
will be most useful if the primary<br />
cause of differences in the data is<br />
due to the shape of the cylinders.<br />
While the variation in the data due to<br />
the flow separation <strong>and</strong> the r<strong>and</strong>om<br />
phase of the wake is impossible to<br />
July/August 2011 Volume 45 Number 4 135
FIGURE 3<br />
A comparison of the mean pressure measured at one experimental configuration reveals clear<br />
differences between the two cylinder shapes. The st<strong>and</strong>ard deviation is also plotted, offset by<br />
−2.0 kPa.<br />
remove, the velocity of the cylinder<br />
strongly correlates with the pressure<br />
signal amplitude <strong>and</strong> masks the response<br />
to shape. Normalizing each<br />
pressure trace by the maximum pressure<br />
effectively removes this masking<br />
effect since the pressure amplitude is<br />
consistently proportional to the velocity<br />
squared.<br />
Using the first three principal components<br />
based on a training set of data<br />
from a single sensor <strong>and</strong> covering all of<br />
the experimental parameters, a highly<br />
successful classification test is obtained.<br />
These three principal components<br />
form a space in which the data<br />
points generated from experimental<br />
runs are grouped into two roughly<br />
ellipsoidal clouds, which can be optimally<br />
separated by a decision plane,<br />
minimizing the sum of squared error<br />
on the training data. Since the projections<br />
of the data points on each axis are<br />
based on a linear combination of the<br />
pressure data with the corresponding<br />
FIGURE 4<br />
principal component, a single vector<br />
of weights can be found that corresponds<br />
to an axis perpendicular to<br />
the decision plane (Figure 4B). This<br />
results in a single linear combination<br />
using these weights with the<br />
normalized pressure measurements<br />
to determine the score of an experimental<br />
run, with a positive value<br />
implying a square cross section. When<br />
applied to the test data (excluding the<br />
training data), this test produced<br />
a very small misclassification rate<br />
of 1.2%.<br />
While the decision test encapsulated<br />
in the first two parts of Figure 4<br />
results in a surprisingly high accuracy,<br />
the difficulty with using principal<br />
components is that there is generally<br />
little corresponding intuition for why<br />
it works so well. The key element of<br />
the decision test is given by the decision<br />
weights, illustrated in Figure 4B.<br />
There are three distinct main sections<br />
The elements of the PCA-based classification test (A <strong>and</strong> B), <strong>and</strong> the accuracy of the test using<br />
different subsections of the full decision coefficients (C).<br />
136 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
to the decision weights based on the<br />
PCA results, labeled II, III, <strong>and</strong> IV in<br />
Figure 4B. Region II is the smallest in<br />
magnitude <strong>and</strong> corresponds to the region<br />
between the maximum pressure<br />
<strong>and</strong> the zero crossing. Region III corresponds<br />
to the area of largest difference<br />
between the two shapes shown in Figure<br />
3. Region IV weighs an area of the<br />
pressure response that is directly influenced<br />
by the wake of the cylinder. This<br />
region has considerable variation in the<br />
data, but it is unclear whether it is related<br />
to the cylinder shape. Regions I<br />
<strong>and</strong> IV on either side are very lightly<br />
weighted. By zeroing out regions, it is<br />
possible to examine the importance of<br />
the remaining portion of the decision<br />
weights to the accuracy in classifying<br />
the shape. In this zeroing process (Figure<br />
4C), we find that data on the rear<br />
side of the zero-crossing point is the<br />
most important in classifying cylinder<br />
shape. In fact, the data before the zerocrossing<br />
point appears to add almost<br />
no new information for classifying<br />
the shape.<br />
The emphasis on region IV implies<br />
that the wake of the cylinder contains a<br />
substantial portion of the information<br />
being used to classify the shape. Initially,<br />
this may appear to be through<br />
vortex spacing, governed by the Strouhal<br />
number. Although this could help<br />
in classifying the shape if particular<br />
sizes were being compared, in these experiments<br />
the large round cylinder <strong>and</strong><br />
the small square cylinder had nearly<br />
equal vortex spacing, meaning that<br />
half of the data would be difficult to<br />
separate by this trait. This observation is<br />
inconsistent with the 1% error rate.<br />
The results based on principal components<br />
demonstrate that it is possible<br />
to distinguish between two relatively<br />
similar cross sections of moving cylinders<br />
based on an artificial lateral line<br />
sensor. This has been hinted at based<br />
on experiments with fish, but the<br />
clear identification of shape without<br />
regard to changes in velocity or size<br />
has not been demonstrated. Given the<br />
importance of the wake in identifying<br />
the shape in the experiment, it is possible<br />
that the ability of the lateral line,<br />
<strong>and</strong> any artificial analogs, to identify<br />
shapes does not extend to arbitrary<br />
shapes. In particular, the bluntness<br />
of the leading face of the cylinder<br />
has a strong impact on the location<br />
<strong>and</strong> flow direction at the point of<br />
flow separation. If two shapes with<br />
similar sharp flow separation are employed,<br />
it may be much harder to distinguish<br />
them.<br />
Active Cylinder<br />
Identification<br />
In active object identification, we<br />
use pressure measurements from an artificial<br />
lateral line sensor array to locate<br />
<strong>and</strong> identify stationary objects in still<br />
water without prior information on<br />
their shape. The problem is considered<br />
from a two-dimensional perspective<br />
in which a moving fish-like<br />
body glides at constant speed past a<br />
column-like stationary object of<br />
unknown location, size, <strong>and</strong> crosssection<br />
geometry. The moving body<br />
is equipped with a finite number of<br />
pressure sensors distributed over its<br />
surface. The viscous effects are<br />
ignored in the model used to calculate<br />
the pressure <strong>and</strong> the limits of this assumption<br />
are discussed.<br />
Hassan (1985, 1992) showed that<br />
in the case of a fish gliding towards a<br />
cylinder or a wall, the pressure increases<br />
noticeably in the front region<br />
when the fish gets close to the obstacle.<br />
He also showed (Hassan, 1985) that<br />
for a fish gliding past a cylinder, there<br />
is a univocal relationship between<br />
spacio-temporal hydrodynamic signature<br />
<strong>and</strong> size <strong>and</strong> distance to the cylinder,<br />
an indication that an artificial<br />
lateral line may be able to distinguish<br />
these features. More recent studies<br />
(Sichert et al., 2009; Bouffanais et al.,<br />
2011) have considered parameterizations<br />
of the shape of an object relevant<br />
to potential flow models, giving an<br />
analytic representation of the flow<br />
field that can be easily related to pressure<br />
<strong>and</strong> velocity measurements.<br />
Experimental Setup<br />
Experiments were conducted in the<br />
SMART Centre water tank. The complete<br />
setup is shown in Figure 5A. The<br />
FIGURE 5<br />
Active sensing experimental layout. (A) A picture<br />
of the experimental set-up used for active object<br />
identification (the foil is moving towards<br />
the photographer). (B) A schematic of the<br />
cross-section of the same experiment showing<br />
the location of the pressure ports.<br />
July/August 2011 Volume 45 Number 4 137
moving body used to both generate the flow <strong>and</strong> sense pressure was a NACA<br />
0018 foil (chord c = 15 cm <strong>and</strong> span s = 60 cm) cast with internal 0.318 cm PVC<br />
tubing to transmit pressure from taps at the foil’s midspan to the top. Honeywell<br />
19C015PG4K pressure sensors were mounted on top of the foil, <strong>and</strong> measurements<br />
were collected at a sampling rate of 500 Hz via a NI USB-6289 DAQ.<br />
The location of the sensor ports is shown in Figure 5B. The foil was dragged at<br />
velocity v = 0.5 m/s past a static cylinder of elliptical cross section oriented at various<br />
angles. At its closest point, the foil was 5-10 mm away from the cylinder.<br />
The Forward <strong>and</strong> Inverse Problem<br />
It is believed that blind cave fish can encode separately the distance, size <strong>and</strong><br />
shape of objects (Burt de Perera, 2004). Therefore, a convenient <strong>and</strong> potentially<br />
biologically relevant way to parameterize the problem is to characterize an object<br />
by two parameters accounting for its position, one size parameter <strong>and</strong> several shape<br />
parameters. Another desirable feature of this parameterization is that the number<br />
of shape parameters needed to account for the pressure decreases with the<br />
distance to the object. Bouffanais et al. (2011) proposed such a characterization:<br />
<br />
<br />
SðθÞ ¼ a þ R e iθ ∞<br />
þ ∑ μk e ikθ<br />
k¼1<br />
¼ x þ iy þ R e iθ þ ∑<br />
∞ jμk je ik ð θ<br />
k¼1<br />
α kÞ<br />
<br />
; θ ∈ ½0; 2πŠ<br />
wherein a (a = x + iy) refers to the location of the object, R to its size <strong>and</strong> each μ k<br />
(μ k =|μ k |e ikα k<br />
)termisassociatedwitha(k +1)-gonaltypeofperturbationofthe<br />
shape from that of a circle. As k increases, the impact of the μ k term on the pressure<br />
field decays very quickly with the distance from the cylinder. For an ellipse,<br />
as in these experiments, only the first shape parameter μ 1 is non-zero <strong>and</strong> therefore<br />
the subscripts will be dropped.<br />
Given the moving object, its trajectory, <strong>and</strong> the location, size <strong>and</strong> shape (a,<br />
R, μ) of the stationary object, the velocity potential anywhere in the flow field<br />
can be expressed in terms of a singularity distribution over the surface of the objects.<br />
The problem is solved numerically using a source panel method: the surface<br />
of each object is broken up into line segments of constant source strength (364 line<br />
segments were used for the moving body <strong>and</strong> 150 for the stationary object). The<br />
pressure at the sensor locations on the surface of the moving body can then be<br />
expressed in terms of the potential at these points using the unsteady Bernoulli<br />
equation. The combination of the parameterization of the object shape <strong>and</strong><br />
the numerical potential flow model defines the forward methodology, relating<br />
the pressure measurements to the variables of interest (location, size, <strong>and</strong><br />
shape). In experiments, the pressure measured by the sensors is corrupted<br />
by noise. The two main sources of noise are laboratory noise (electrical <strong>and</strong><br />
mechanical) <strong>and</strong> the background noise of the fluid flow (which includes viscous<br />
effects).<br />
For proper inversion, the technique must be (1) robust to noise, (2) capable of<br />
h<strong>and</strong>ling non-linearity since the pressure does not depend linearly on the characterizing<br />
parameters of the stationary object, <strong>and</strong> (3) dynamic, in order to be used<br />
for navigation of underwater vehicles.<br />
A particularly suitable algorithm for<br />
such inversion is the unscented Kalman<br />
filter (UKF) ( Julier & Uhlmann,<br />
2004), which is a robust dynamic<br />
probabilistic signal filtering technique<br />
for highly nonlinear systems. The<br />
UKF is more accurate than the more<br />
traditional extended Kalman filter for<br />
highly nonlinear problems <strong>and</strong> does<br />
not require the computation of derivatives<br />
for which no analytic expressions<br />
are available. It also propagates the statistics<br />
with fewer samples than the<br />
more powerful particle filter, which<br />
makes it better suited for real time<br />
applications.<br />
Subtle modifications to the UKF<br />
are necessary before applying it to our<br />
problem. Non-physical configurations,<br />
such as bodies intersecting each<br />
other cannot be identified by an UKF<br />
assuming Gaussian distributions. This<br />
can skew statistics, producing unusable<br />
results. To avoid this problem, if after<br />
updating the location of the moving<br />
object, the estimated mean configuration<br />
is not valid, the estimated position<br />
is shifted ‘out of the way.’ If the mean<br />
configuration is valid, the parameter α<br />
that determines the spread of the samples<br />
around the mean in the unscented<br />
transform (Wan & van der Merwe,<br />
2001) is chosen small enough (for<br />
each time step) that all the samples<br />
are valid.<br />
TheUKF<strong>and</strong>theforwardmodel<br />
are combined to solve the inverse problem:<br />
locating <strong>and</strong> identifying a cylinder<br />
using pressure measurements. In the<br />
results <strong>and</strong> analysis discussed here,<br />
the steady pressure due to the constant<br />
velocity of the foil has been subtracted<br />
from the pressure signal. The measurement<br />
covariance matrix used in the<br />
UKF was calculated for each run<br />
based on the pressure measured 0.5-<br />
0.3 s before the characteristic drop<br />
138 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
of pressure at the second sensor (see<br />
Figure 7C).<br />
FIGURE 6<br />
Results of a cylinder detection <strong>and</strong> identification in two passes (A <strong>and</strong> B) using an UKF. The colored<br />
dashed lines show the true value of the parameters. (a–f) The corresponding shape estimate (red<br />
circle or ellipse), the actual cylinder (green dotted ellipse), <strong>and</strong> the position of the foil at various<br />
times. (Color versions of figures available online at: http://www.ingentaconnect.com/content/mts/<br />
mtsj/2011/00000045/00000004.)<br />
Active Sensing Results<br />
<strong>and</strong> Analysis<br />
Due to the amount of noise in the<br />
experiments <strong>and</strong> the fact that it was<br />
highly correlated, all attempts to simultaneously<br />
fully identify the location<br />
<strong>and</strong> geometry of the cylinder<br />
were unsuccessful. However, fish<br />
have been observed to pass several<br />
times in front of new objects (von<br />
Campenhausen et al., 1981), <strong>and</strong> it<br />
seems reasonable to assume that they<br />
first locate the objects <strong>and</strong> estimate<br />
their size before refining their estimate<br />
of the shape. A similar approach is used<br />
here: the first pass is used to get a first<br />
estimate of the position (x <strong>and</strong> y) <strong>and</strong><br />
size (R) of the cylinder. A second pass<br />
using the same data refines the first estimates<br />
<strong>and</strong> estimates the shape parameters<br />
(|μ| <strong>and</strong>α). This approach is<br />
consistent with the dynamic hierarchical<br />
access associated with the shape<br />
characterization parameters used by<br />
Bouffanais et al. (2011). An example<br />
of such a process of cylinder geometry<br />
reconstruction is shown in Figure 6 for<br />
a cylinder with shape parameters R =<br />
3.81 cm <strong>and</strong> μ = 0.2i <strong>and</strong> the foil passing<br />
6.5 mm away from it.<br />
Both behavioral studies (Weissert<br />
& von Campenhausen, 1981) <strong>and</strong><br />
mathematical modeling (Hassan,<br />
1985) have shown that the presence<br />
of the static object in still water cannot<br />
be detected until the fish (or here the<br />
foil) is very close to it (on the order<br />
of the width of the moving body). As<br />
canbeseeninFigure6A,assoonas<br />
the foil is close enough to the cylinder,<br />
the position (x <strong>and</strong> y) <strong>and</strong> size (R) estimates<br />
converge steadily towards the<br />
true value of the parameters. The second<br />
run allows for very good orientation<br />
(α) estimation (Figure 6B). The<br />
aspect ratio of the ellipse (|μ|) is the<br />
more difficult feature to reconstruct.<br />
The estimated parameter |μ| first w<strong>and</strong>ers<br />
around the actual value of the<br />
parameter before decreasing towards<br />
a lower value. As can be seen in Figure<br />
6F even though the shape of the<br />
ellipse has not been exactly reconstructed,<br />
the estimate of the half of<br />
the ellipse that is closest to the foil is<br />
reasonably accurate. This observation<br />
confirms that object identification<br />
based on pressure sensing becomes<br />
less reliable for features further from<br />
the sensors.<br />
The results demonstrate that object<br />
localization <strong>and</strong> object recognition are<br />
possible with experimental pressure<br />
measurements; however, there are limitations<br />
to the proposed method. The<br />
main limitation is the use of potential<br />
flow models to establish the governing<br />
equations for the filter <strong>and</strong> the simulations<br />
(the importance of viscosity on<br />
the stimulus to the lateral line system<br />
of fish is also discussed in Windsor &<br />
McHenry, 2009). A fourth pressure<br />
sensor placed near the tail of the foil<br />
measured an oscillating pressure signal<br />
characteristic of the wake, unlike the<br />
first three sensors. Particle image<br />
velocimetry was used to visualize the<br />
flow <strong>and</strong> compare it to the flow predicted<br />
by the inviscid model. As can<br />
be seen in Figures 7A <strong>and</strong> 7B, the<br />
two flows are nearly identical almost<br />
everywhere, but as the foil passes the<br />
cylinder, separation occurs over a<br />
small region on the foil (in the orange<br />
ellipse). Comparing the pressure measurements<br />
<strong>and</strong> simulations (Figure<br />
7C), we can see that the potential<br />
flow model gives very good approximations<br />
of the pressure measurements<br />
until the second sensor passes the cylinder<br />
(black dotted line). After that<br />
point at which separation occurs, the<br />
inviscid assumption is violated <strong>and</strong><br />
the model is no longer valid. Therefore,<br />
only the measurements before<br />
the black dotted line on the figure<br />
July/August 2011 Volume 45 Number 4 139
FIGURE 7<br />
Comparison between simulated (A) <strong>and</strong> experimental (B) flow field. The green <strong>and</strong> orange ellipses<br />
show where the flows differ. (C) The pressure data (plain line) is filtered with a cut-off frequency of<br />
100 Hz. (Color versions of figures available online at: http://www.ingentaconnect.com/content/<br />
mts/mtsj/2011/00000045/00000004.)<br />
have been used to locate <strong>and</strong> identify<br />
the cylinder.<br />
some of the experiments that still provided<br />
successful tracking results. The<br />
pressure data for the active sensing experiments<br />
was of a similar scale. Any<br />
sensors used for an artificial lateral<br />
line application must provide the sensitivity<br />
<strong>and</strong> correspondingly low noise<br />
floor to distinguish small changes in<br />
pressure of at least 10 Pa.<br />
The experiments in active sensing<br />
ofobjectsprovideaguidetotheoptimal<br />
density of sensors needed for<br />
similar applications, since the curved<br />
surface of the foil <strong>and</strong> the self-generated<br />
flow increase in the importance<br />
of the instantaneous measurements<br />
of the spatial pressure distribution.<br />
Simulations were performed to examine<br />
the affect of sensor density on the<br />
ability to identify a stationary object.<br />
The simulations consisted of the<br />
same foil described earlier gliding at<br />
0.5 m/s <strong>and</strong> passing 7 mm away from<br />
a cylinder with the geometry parameters<br />
R 0 = 1.5 cm <strong>and</strong> μ =0.2i.<br />
White Gaussian noise of st<strong>and</strong>ard<br />
deviation 2 Pa was added to the<br />
simulated pressure measurements. Between<br />
10 <strong>and</strong> 70 pressure measurement<br />
points were evenly distributed<br />
along the front three quarters of the<br />
foil (see Figure 8C), <strong>and</strong> the sampling<br />
Sensor Array Constraints<br />
The use of off-the-shelf sensors severely<br />
limits the density which can be<br />
achieved in the sensor array while<br />
maintaining the sensitivity to deviations<br />
from the static pressure needed<br />
for engineering applications. From<br />
the perspective of sensor sensitivity,<br />
the weakest signals are those associated<br />
with tracking the position <strong>and</strong> size of a<br />
moving cylinder. The maximum absolute<br />
pressure was as small as 30 Pa in<br />
FIGURE 8<br />
Convergence time (A) <strong>and</strong> final error (B) of the object identification as a function of sensor<br />
spacing. (C) The location of the sensors for a spacing of 6 mm.<br />
140 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
ate was chosen such that f = (number of sensors) / 20,000. The error was calculated<br />
as E ¼ 1 ∇R<br />
5 R 0<br />
þ ∇x<br />
R 0<br />
þ ∇y<br />
R 0<br />
þ ∇μ x<br />
þ ∇μ y<br />
. For each simulation, the convergence<br />
time (time elapsed before E < 0.2) <strong>and</strong> the final error were calculated.<br />
Each case was simulated 12 times, <strong>and</strong> the mean <strong>and</strong> st<strong>and</strong>ard deviations of the<br />
time of convergence <strong>and</strong> final error were computed (Figures 8A <strong>and</strong> 8B). The<br />
goal of this procedure was to identify a sensor density below which the lack of<br />
spatial density cannot be compensated by a higher sampling frequency.<br />
Both the convergence time <strong>and</strong> final error plots suggest that, at least for the<br />
configuration considered, the performance of the object identification decreases<br />
as the sensor spacing exceeds 5 mm, which corresponds to a spacing of 0.03 relative<br />
to the length of the foil. The optimal spacing that these observations suggest<br />
scales favorably with the actual spacing of lateral line canal neuromasts in the<br />
trunk canal of the blind Mexican cave fish, which is roughly 0.02 body lengths<br />
(measured from image in Windsor & McHenry, 2009). This optimal spacing<br />
derived from simulations is specific to identifying the shape of cylindrical objects<br />
in which the cross section size is on the same order or slightly smaller than the<br />
body length. In general, the optimal spacing for measuring the pressure distribution<br />
may scale with the size of the object being detected as well as the body<br />
length. In addition, based on the observed separation that occurs on the surface<br />
of the sensing body, the pressure sensors should be distributed only over roughly<br />
the forward half of the body in order to maximize the effectiveness of the sensor<br />
array without waste.<br />
A New Approach for a Flexible Pressure Sensor Array<br />
In order to achieve the optimal spatial distributions <strong>and</strong> sensitivity of pressure<br />
sensors in an effective artificial lateral line, it is necessary to develop sensor<br />
arrays on smaller scales than those possible using off-the-shelf technologies. In<br />
addition to the spatial <strong>and</strong> sensitivity constraints, it would be necessary that the<br />
sensor be applied directly on a surface in order for a lateral line sensor to be<br />
practical for ocean vehicles, instead of fabricating the sensing body around<br />
the sensor as was necessary in the experiments reported here. To address<br />
these problems, we have begun to develop a thin, flexible, one-dimensional<br />
array of pressure sensors to meet the pressure <strong>and</strong> spatial resolution requirements.<br />
The array is designed to conform to an AUV’s hull without protruding<br />
significantly.<br />
While MEMS pressure sensors are typically made with a silicon substrate<br />
(Senturia, 2001), the flexible sensor array described here is made entirely of a<br />
silicone elastomer material, allowing it to be rugged, waterproof, <strong>and</strong> flexible.<br />
A conductive polymer is used as the pressure sensitive element in order to function<br />
while maintaining mechanical compatibility with the rest of the flexible<br />
structure.<br />
The conductive polymer is composed of the silicone Polydimethylsiloxane<br />
(PDMS) doped with conductive carbon black particles (Ding et al., 2007). This<br />
material has been used for chemical sensors (Andreadis et al., 2007) but not for a<br />
high-resolution pressure sensor. Prior work involving large pressure sensing arrays<br />
has concentrated on tactile pressure sensing, which requires reduced sensitivity<br />
<strong>and</strong> greater dynamic range (Someya et al., 2004; Harsanyi, 2000). In contrast,<br />
the carbon black pressure sensors are designed specifically for the lateral line<br />
application, where small pressure variations<br />
in the tens of Pascals are of<br />
interest.<br />
Pressure Sensor Design<br />
<strong>and</strong> Fabrication<br />
An individual pressure sensing cell<br />
is depicted <strong>and</strong> diagrammed in Figure<br />
9. Its active components consist<br />
of a resistive strain gauge patterned<br />
on the surface of a 10-mm-wide,<br />
1-mm-thick elastomer membrane,<br />
as shown in Figures 1A <strong>and</strong> 1B.<br />
PDMS was chosen as the membrane<br />
material due to its low tensile modulus<br />
(Schneider et al., 2008), which<br />
improves sensitivity. A differential<br />
pressure across the surfaces of the<br />
membrane causes the membrane to<br />
deflect. For small pressures, the deflection<br />
is linear (Senturia, 2001). The deflection<br />
then induces strain in the<br />
resistive strain gauge. The resulting resistance<br />
change is measured using the<br />
four terminals of the strain gauge,<br />
as depicted in Figures 9B <strong>and</strong> 9C. A<br />
constant current is applied through<br />
the outer terminals, <strong>and</strong> the voltage is<br />
measured across the inner voltage-tap<br />
terminals. This four-point probe measurement<br />
desensitizes the device to variations<br />
in contact resistance between<br />
the wires <strong>and</strong> the strain gauge. The<br />
voltage taps are positioned to capture<br />
the greatest resistance change at the<br />
central edge of the membrane where<br />
the greatest strain occurs (Senturia,<br />
2001).<br />
For a lateral line application on an<br />
underwater vehicle, the differential<br />
pressure between sensors in the array<br />
is of interest, but the depth of the<br />
aquatic vehicle, which corresponds to<br />
the absolute pressure, is not. The<br />
slow-varying absolute pressure may<br />
be cancelled out by low frequency<br />
pressure equilibration using the air<br />
July/August 2011 Volume 45 Number 4 141
FIGURE 9<br />
Diagrams <strong>and</strong> photo of a single pressure sensing cell with a 10-mm square membrane. The device<br />
is 3-mm thick.<br />
channel diagrammed in Figure 9A.<br />
This equilibration prevents damage<br />
to the sensors caused by the water<br />
depth at which the vehicle operates.<br />
It also relaxes the design requirements<br />
on the sensor, allowing it to have a<br />
thin membrane that is sensitive while<br />
not requiring it to withst<strong>and</strong> large<br />
pressures.<br />
In order to verify the performance<br />
of the carbon black sensing element<br />
in the context of a flexible pressure sensor,<br />
a single pressure sensing cell was<br />
fabricated (Figure 9C). PDMS was<br />
cast in a mold to form the membrane<br />
structure overhanging a cavity as<br />
showninFigure9A.Toformthe<br />
strain gauge, conductive carbon black<br />
particles approximately 1 μm in diameter<br />
were uniformly mixed into<br />
PDMS. Patterning of the strain gauge<br />
was accomplished with a stencil mask,<br />
which yielded a 100-μm-thick layer.<br />
The carbon black strain gauge terminals<br />
were bonded to aluminum wire<br />
with conductive epoxy. For testing<br />
purposes, a glass slide was used to simulate<br />
the rigid hull.<br />
Sensor Performance<br />
The dynamic response of the single<br />
test pressure sensor was characterized<br />
using a manual pressure source independently<br />
measured using a Honeywell<br />
pressure sensor. The resistance of<br />
the test sensor’s carbon black strain<br />
gauge was recorded as the pressure<br />
varied. Figures 10A <strong>and</strong> 10B show<br />
the resistance of the strain gauge as a<br />
function of time, compared to the<br />
pressure measured by the off-the-shelf<br />
sensor.<br />
From the dynamic response data in<br />
Figure 10, the sensitivity can be estimated<br />
as 0.55% change in resistance<br />
over 100 Pa. An amplification circuit<br />
was used to cancel the DC offset voltage<br />
<strong>and</strong> amplify the small resistance<br />
changes by 100, resulting in a roughly<br />
55% change in voltage over 100 Pa.<br />
This is more than sufficient for detecting<br />
pressure variations on the order of<br />
10 Pa, so it is adequate for the lateral<br />
line application.<br />
Figure 10C depicts the applied<br />
pressure plotted against the measured<br />
resistance for the dynamic tests in Figures<br />
10A <strong>and</strong> 10B, as well as a static<br />
test. The static test was performed by<br />
applying a series of pressure steps<br />
with a 1-min hold time <strong>and</strong> recording<br />
the resistance at the end. The pressure<br />
was stepped up, down, <strong>and</strong> up again to<br />
characterize sensor hysteresis. The<br />
slope of the curves represents the sensitivity<br />
of the sensor, <strong>and</strong> the asymmetry<br />
between the up <strong>and</strong> down steps<br />
is indicative of hysteresis. This is expected<br />
because the PDMS membrane<br />
is a viscoelastic material that experiences<br />
both creep <strong>and</strong> stress relaxation<br />
(Schneider et al., 2008). Both behaviors<br />
are byproducts of its low tensile modulus<br />
but are well understood <strong>and</strong> can be<br />
modeled <strong>and</strong> accounted for using signal<br />
processing techniques. The creep<br />
causes increased strain for a given pressure,<br />
resulting in the static response<br />
having a greater slope than the dynamic<br />
response. The dynamic responses are<br />
more linear <strong>and</strong> consistent from cycle<br />
to cycle than the static response, because<br />
the creep is not significant on<br />
the timescale of these tests.<br />
In addition to the polymer creep,<br />
there is the time-dependent relaxation<br />
behavior inherent in the conductive<br />
polymer. From the data in Figures<br />
10A <strong>and</strong> 10B, there is a very<br />
short time constant in the carbon<br />
black strain gauge resistance data<br />
when the pressure is being stepped<br />
up, but a much longer time constant<br />
when the pressure is being stepped<br />
down. This is thought to be a product<br />
of the way the carbon black conductive<br />
polymer responds to strain (Ding et al.,<br />
2007). However, as shown in dynamic<br />
response transfer curves of Figure 10C,<br />
142 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 10<br />
Dynamic <strong>and</strong> static responses of the single cell pressure sensor, normalized by the strain gauge<br />
resistance at atmospheric pressure R initial ∼ 20 kΩ. Parts A <strong>and</strong> B compare the output of an offthe-shelf<br />
Honeywell sensor with the output of the carbon black pressure sensor when the same<br />
pressure is applied to both. Part C presents dynamic transfer curves using the dynamic data<br />
from parts A <strong>and</strong> B as well as static data obtained by applying a constant pressure to the sensor<br />
<strong>and</strong> holding it for 1 min at each data point.<br />
simulations. Ultimately, the membrane<br />
thickness <strong>and</strong> the width of the<br />
strain gauge patterns will be the limiting<br />
factors. The carbon black–based<br />
flexible sensor array is a promising<br />
technology for the lateral line application<br />
because it meets the pressure <strong>and</strong><br />
spatial resolution criteria, it has repeatable<br />
<strong>and</strong> predictable performance, <strong>and</strong><br />
it can conform to the hull of an aquatic<br />
vehicle without protruding.<br />
this behavior is repeatable <strong>and</strong> consistent<br />
during cyclic loading, so future<br />
work will consist of using signal processing<br />
to compensate for this effect.<br />
Creating a Sensor Array<br />
Extending the single carbon black<br />
sensor cell into an array involves both<br />
miniaturization of each sensor cell<br />
<strong>and</strong> routing of the carbon black strain<br />
gauges. A prototype array of four pressure<br />
sensing cells is shown in Figure<br />
11. The four sensors share two<br />
common current terminals, <strong>and</strong> each<br />
individual sensor has two voltage<br />
taps, reducing the necessary wiring.<br />
Each cell has a 5-mm-wide, 0.5-mmthick<br />
square membrane. The dimensions<br />
of the cells have been scaled<br />
down by a factor of 2 from the single<br />
test pressure sensor. The spatial resolution<br />
of the array is determined by the<br />
sensor cell spacing, which is 7-mm<br />
center-to-center for this array.<br />
Further miniaturization is possible<br />
by reducing the dimensions of the<br />
membrane <strong>and</strong> strain gauge pattern<br />
in order to achieve the sub-5 mm spacing<br />
found optimal in the active sensing<br />
Conclusion<br />
This paper has considered several<br />
aspects of object identification based<br />
entirely on a lateral-line-like pressure<br />
sensor array, with the aim of examining<br />
its potential for application<br />
in underwater vehicles. Object identification<br />
was divided into passive object<br />
identification, where an external flow<br />
interacts with the object to stimulate<br />
the sensor array, <strong>and</strong> active object<br />
identification, where self-generated<br />
flow is used to interrogate an object<br />
in still water. In each case, experiments<br />
were used to examine the problem of<br />
estimating the position, size, <strong>and</strong><br />
shape of a cylinder based on a linear<br />
array of pressure sensors in a realistic<br />
noise environment. Constraints with<br />
off-the-shelf sensors in the experimental<br />
implementation of an artificial lateral<br />
line, have led to the development<br />
of a flexible artificial lateral line that<br />
makes use of a new polymer sensing<br />
technology.<br />
The passive sensing of a moving<br />
cylinder separated the identification<br />
of the position <strong>and</strong> size from the<br />
shape into different experiments. The<br />
wake generated by the motion of<br />
the cylinder was a critical component<br />
in the success of each part. For the cylinder<br />
tracking, a simple steady potential<br />
model of the wake was sufficient<br />
to allow a particle filter to track the<br />
July/August 2011 Volume 45 Number 4 143
FIGURE 11<br />
Diagram <strong>and</strong> photo of an array of four pressure sensors, each with a 5-mm transparent square<br />
membrane. The sensors share common current terminals. Each individual sensor has two voltage<br />
tap terminals to measure the resistance of the part of the carbon black strain gauge located on the<br />
corresponding membrane.<br />
position <strong>and</strong> size in real time. In order<br />
to distinguish between a square <strong>and</strong><br />
circular cylinder, a decision rule derived<br />
from PCA was highly successful.<br />
This rule demonstrated robustness in<br />
dealing with cylinders of different<br />
sizes, distances, <strong>and</strong> speeds. In analyzing<br />
the decision rule, it was found that<br />
pressure data in the highly variable<br />
region near <strong>and</strong> aft of the flow separation<br />
is vital in determining the correct<br />
shape. This implies limitations<br />
in the types of shapes that can be distinguished<br />
passively, as dissimilar<br />
shapes with similar wakes may cause<br />
difficulty.<br />
By recreating an active sensing encounter<br />
with a foil st<strong>and</strong>ing in for the<br />
vehicle or fish, we demonstrated the<br />
possibility of identifying the location<br />
<strong>and</strong> shape of a cylinder without prior<br />
knowledge of its shape. A potential<br />
flow model was chosen for simplicity<br />
<strong>and</strong> to avoid heavy calculations that<br />
would make it impossible to run the algorithm<br />
in real time. Using this model<br />
with an UKF allowed us to obtain a<br />
very good object location estimate<br />
<strong>and</strong> reasonable ellipse identification<br />
using three pressure sensors. The results<br />
stress the importance of the<br />
head lateral line of the fish (that had<br />
been emphasized in the case of the<br />
fish moving towards an obstacle by<br />
Hassan, 1986) when it passes an object,<br />
due to the timing <strong>and</strong> location<br />
of flow separation. To make more use<br />
of a biomimetic trunk lateral line, the<br />
next step is to develop a model that<br />
takes the separation into account to<br />
be able to extract information from<br />
the second half of the data.<br />
For both active <strong>and</strong> passive sensing,<br />
we have found that information about<br />
thelocation<strong>and</strong>sizeofacylinderis<br />
available via an artificial lateral line.<br />
Beyond this point, however, there are<br />
strong differences in the two scenarios.<br />
In the case of passive sensing, the flow<br />
separation <strong>and</strong> wake are integral components<br />
to the pressure distribution<br />
<strong>and</strong> must be accounted for from the<br />
beginning. In fact, it appears the<br />
shape information may largely be available<br />
through these components. In<br />
contrast, for active sensing flow separation<br />
does not occur immediately,<br />
allowing it to be ignored in the initial<br />
shape estimation as done in this<br />
paper. Using pre-separation data limits<br />
the number of measurements<br />
available but allows the use of a general<br />
model for identifying arbitrary<br />
shapes.<br />
Simulations based on the active<br />
sensing experiments have shown that<br />
increasing the sensor spacing of<br />
approximately 0.03 body lengths<br />
achieves the fastest <strong>and</strong> most accurate<br />
object identification for cylindrical objects<br />
with a diameter on the same scale<br />
as the sensing body. Coupled with the<br />
sensitivity (less than 10 Pa) required to<br />
suitably measure the pressure signals in<br />
the experiments, this defines the specifications<br />
for an artificial lateral line<br />
that could be used to identify or locate<br />
similar objects for an underwater<br />
vehicle. Of significant additional concern<br />
is the need for a sensor that can<br />
be easily mounted to a hull without<br />
significantly disturbing the flow.<br />
Based on the design <strong>and</strong> test results<br />
shown, the carbon black-based flexible<br />
sensor array is a promising technology<br />
for the lateral line application. Used as<br />
a sensing element, it can meet the pressure<br />
<strong>and</strong> spatial resolution criteria, it<br />
has repeatable <strong>and</strong> predictable performance,<br />
<strong>and</strong> it is able to conform to<br />
the smooth curves of a hull while protruding<br />
only 3 mm.<br />
144 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
Acknowledgments<br />
The authors gratefully acknowledge<br />
the support of the Singapore-MIT<br />
Alliance for Research <strong>and</strong> <strong>Technology</strong><br />
(SMART) program’s Center for Environmental<br />
Sensing <strong>and</strong> Modeling<br />
<strong>and</strong> that of the National Oceanic <strong>and</strong><br />
Atmospheric Administration’s Sea<br />
Grant program under project number<br />
R/RT-2/RCM-17.<br />
Lead Author:<br />
Vicente I. Fern<strong>and</strong>ez<br />
Department of<br />
Mechanical Engineering,<br />
Massachusetts Institute of <strong>Technology</strong><br />
5-424, 77 Massachusetts Avenue<br />
Cambridge, MA 02139<br />
Email: vicentef@mit.edu<br />
References<br />
Andreadis, N., Chatz<strong>and</strong>roulis, S., Goustouridis,<br />
D.,Kosma,V.,Beltsios,K.,&Raptis,I.2007.<br />
Fabrication of conductometric chemical sensors<br />
by photolithography of conductive polymer<br />
composites. J Microelec Eng. 84(5-8):<br />
1211-4. doi: 10.1016/j.mee.2007.01.050.<br />
Bouffanais, R., Weymouth, G.D., & Yue,<br />
D.K.P. 2011. Hydrodynamic object recognition<br />
using pressure sensing. P R Soc A. 467(2125):<br />
19-38. doi: 10.1098/rspa.2010.0095.<br />
Branko, R., Arulampalam, S., & Gordon, N.<br />
2004. Beyond the Kalman Filter: Particle<br />
Filters for Tracking Applications. Boston:<br />
Artech House. 318 pp.<br />
Burt de Perera, T. 2004. Spatial parameters<br />
encoded in the spatial map of the blind<br />
Mexican cave fish, Astyanax fasciatus.<br />
Anim Behav. 68(2):291-5. doi: 10.1016/<br />
j.anbehav.2003.11.009.<br />
Cappe, O.,Moulines,E.,&Ryden,T.2005.<br />
InferenceinHiddenMarkovModels.NewYork:<br />
Springer Science+Business Media, Inc. 652 pp.<br />
Chen, J., Engel, J., Chen, N., P<strong>and</strong>ya, S.,<br />
Coombs, S., & Liu, C. 2006. Artificial lateral<br />
line <strong>and</strong> hydrodynamic object tracking. In:<br />
Micro Electro Mechanical Systems 2006,<br />
MEMS 2006 Istanbul. 19th IEEE International<br />
Conference on Micro Electro Mechanical<br />
Systems. pp. 694-7. Istanbul: IEEE.<br />
doi: 10.1109/MEMSYS.2006.1627894.<br />
Coombs, S. 2001. Smart skins: Information<br />
processing by lateral line flow sensors. Auton<br />
Rob. 11(3):255-61.<br />
Coombs, S., & Janssen, J. 1990. Behavioral<br />
<strong>and</strong> neurophysiological assessment of the lateral<br />
line sensitivity in the mottled sculpin,<br />
Cottus bairdi. J Comp Physiol A. 167(4):<br />
557-67. doi: 10.1007/BF00190827.<br />
Curcic-Blake, B., & van Netten, S. 2006.<br />
Source location encoding in the fish lateral<br />
line canal. J Exp Biol. 209(8):1548-59.<br />
doi: 10.1242/jeb.02140.<br />
Ding, T., Wang, L., & Wang, P. 2007.<br />
Changes in electrical resistance of carbonblack-filled<br />
silicone rubber composite during<br />
compression. J Polym Sci Part B Polym<br />
Phys. 45(19):2700-6. doi: 10.1002/<br />
polb.21272.<br />
Gelb, A. 1974. Applied Optimal Estimation.<br />
Massachusetts: MIT Press. 382 pp.<br />
Goulet, J., Engelmann, J., Chagnaud, B.,<br />
Franosch, J., Suttner, M., & Hemmen, J.<br />
2007. Object localization through the lateral<br />
line system of fish: Theory <strong>and</strong> experiment.<br />
J Comp Phys A 194(1):1-17. doi: 10.1007/<br />
s00359-007-0275-1.<br />
Harsanyi, G. 2000. Polymer films in sensor<br />
applications: A review of present uses <strong>and</strong><br />
future possibilities. Sens Rev. 20(2):98-105.<br />
doi: 10.1108/02602280010319169.<br />
Hassan, E.S. 1985. Mathematical analysis<br />
of the stimulus for the lateral line organ.<br />
Biol Cybern. 52(1):23-36. doi: 10.1007/<br />
BF00336932.<br />
Hassan, E.S. 1986. On the discrimination<br />
of spatial intervals by the blind cave fish<br />
(Anoptichthys jordani). J Comp Physiol A.<br />
159(5):701-10. doi: 10.1007/BF00612042.<br />
Hassan, E.S. 1992. Mathematical description<br />
of the stimuli to the lateral line system derived<br />
from a three dimensional flow field analysis:<br />
I. The case of moving in open water <strong>and</strong> of<br />
gliding toward a plane surface. Biol Cybern.<br />
66:443-52. doi: 10.1007/BF00197725.<br />
Jackson, J.E. 2003. A User’s Guide to<br />
Principal Components. New Jersey: Wiley.<br />
592 pp.<br />
Jolliffe, I.T. 2002. Principal Component<br />
Analysis. New York: Springer. 502 pp.<br />
Julier, S.J., & Uhlmann, J.K. 2004. Unscented<br />
filtering <strong>and</strong> nonlinear estimation.<br />
Proc IEEE. 92(3):401-22. doi: 10.1109/<br />
JPROC.2003.823141.<br />
Montgomery, J.C., Coombs, S., & Baker,<br />
C.F. 2001. The mechanosensory lateral line<br />
system of the hypogean form of Astyanax<br />
fasciatus. Eviron Biol Fishes. 62(1-3):87-96.<br />
doi: 10.1023/A:1011873111454.<br />
Pohlmann, K., Atema, J., & Breithaupt, T.<br />
2004. The importance of the lateral line<br />
in nocturnal predation of piscivorous<br />
catfish. J Exp Biol. 207(17):9271-8.<br />
doi: 10.1242/jeb.01129.<br />
Schneider, F., Fellner, T., Wilde, J., &<br />
Wallrabe, U. 2008. Mechanical properties<br />
of silicones for MEMS. J Micromech<br />
Microeng. 18(6):065008. doi: 10.1088/<br />
0960-1317/18/6/065008.<br />
Senturia, S.D. 2001. Microsystem Design.<br />
Boston: Kluwer Academic Publishers.<br />
720 pp.<br />
Sichert, A., Bamler, R., & van Hemmen, J.<br />
2009. Hydrodynamic object recognition:<br />
When multipoles count. Phys Rev Lett.<br />
102(5):058104. doi: 10.1103/<br />
PhysRevLett.102.058104.<br />
Someya, T., Sekitani, T., Iba, S., Kato, Y.,<br />
Kawaguchi, H., & Sakurai, T.A. 2004.<br />
Large-area, flexible pressure sensor matrix<br />
with organic field-effect transistors for artificial<br />
skin applications. P Natl Acad Sci.<br />
101(27):9966-70. doi: 10.1073/pnas.<br />
0401918101.<br />
van Netten, S. 2006. Hydrodynamic detection<br />
by cupulae in a lateral line canal:<br />
functional relations between physics <strong>and</strong><br />
July/August 2011 Volume 45 Number 4 145
physiology. Biol Cybern. 94(1):67-85.<br />
doi: 10.1007/s00422-005-0032-x.<br />
Vogel, D., & Bleckmann, H. 2000. Behavioral<br />
discrimination of water motions caused<br />
by moving objects. J Comp Physiol A. 186(12):<br />
1107-17. doi: 10.1007/s003590000158.<br />
von Campenhausen, C., Riess, I., &<br />
Weissert, R. 1981. Detection of stationary<br />
objects by the blind cave fish Anoptichthys<br />
jordani (Characidae). J Comp Physiol A Sens<br />
Neural Behav Physiol. 143(3):369-74.<br />
doi: 10.1007/BF00611175.<br />
Wan, E., & van der Merwe, R. 2001.<br />
The unscented Kalman filter. In: Kalman<br />
filtering <strong>and</strong> neural networks, ed. Haykin, S.,<br />
221-80. New York: Wiley. doi: 10.1002/<br />
0471221546.ch7.<br />
Weissert, R., & von Campenhausen, C.<br />
1981. Discrimination between stationary<br />
objects by the blind cave fish Anoptichthys<br />
jordani (characidae). J Comp Physiol A<br />
Sens Neural Behav Physiol. 143(3):375-81.<br />
doi: 10.1007/BF00611176.<br />
Windsor, S.P., & McHenry, M.J. 2009. The<br />
influence of viscous hydrodynamics on the<br />
fish lateral-line system. Integr Comp Biol.<br />
49(6):691-701. doi: 10.1093/icb/icp084.<br />
Yang, Y., Chen, N., Tucker, C., P<strong>and</strong>ya, S.,<br />
Jones, D., & Liu, C. 2008. Biomimetic flow<br />
sensing using artificial lateral lines. ASME<br />
Conf Proc. 2007(43025):1331-8.<br />
146 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
PAPER<br />
A Conserved Neural Circuit-Based<br />
Architecture for Ambulatory <strong>and</strong><br />
Undulatory Biomimetic Robots<br />
AUTHORS<br />
Joseph Ayers<br />
Anthony Westphal<br />
Daniel Blustein<br />
Department of Biology <strong>and</strong><br />
<strong>Marine</strong> Science Center,<br />
Northeastern University<br />
Introduction<br />
The innate behavior of underwater<br />
animals provides an effective model for<br />
the adaptive behavior of unmanned<br />
underwater vehicles (Ayers, 2004).<br />
Underwater animals must respond to<br />
a broad variety of environmental challenges<br />
including turbidity, hydrodynamic<br />
flow, heterogeneous <strong>and</strong><br />
highly structured bottom types <strong>and</strong><br />
impediment. Their relatively neutral<br />
buoyancy renders them especially susceptible<br />
to hydrodynamic perturbation.<br />
As a result, they have evolved a<br />
behavioral set that includes a broad<br />
variety of compensatory responses to<br />
perturbation. This behavioral set results<br />
from layered exteroceptive reflexes<br />
responding to exteroceptive<br />
sensor input resulting from changes<br />
in orientation relative to gravity, impediment,<br />
chemical cues, <strong>and</strong> hydrodynamic<br />
<strong>and</strong> optical flow (Ayers,<br />
2004; Blustein & Ayers, 2010).<br />
These layered exteroceptive reflexes can<br />
form taxic responses to point sources<br />
of sound or chemicals (Westphal<br />
et al., 2011). As the point sources<br />
form motivational cues for goal achieving<br />
behavioral sequences, they can<br />
ABSTRACT<br />
The adaptive capabilities of underwater organisms result from layered exteroceptive<br />
reflexes responding to gravity, impediment, <strong>and</strong> hydrodynamic <strong>and</strong> optical<br />
flow. In combination with taxic responses to point sources of sound or chemicals,<br />
these reflexes allow reactive autonomy in the most challenging of environments.<br />
We are developing a new generation of lobster <strong>and</strong> lamprey-based robots that operate<br />
under control by synaptic networks rather than algorithms. The networks,<br />
based on the comm<strong>and</strong> neuron, coordinating neuron, <strong>and</strong> central pattern generator<br />
architecture, code sensor input as labeled lines <strong>and</strong> activate shape memory alloybased<br />
artificial muscles through a simple interface that couples excitation to contraction.<br />
We have completed the lamprey-based robot <strong>and</strong> are adapting this sensor,<br />
board, <strong>and</strong> actuator architecture to a new generation of the lobster-based robot. The<br />
networks are constructed from discrete time map-based neurons <strong>and</strong> synapses <strong>and</strong><br />
are instantiated on the digital signal processing chip. A sensor board integrates inputs<br />
from a short baseline sonar array (for beacon tracking <strong>and</strong> supervisory control),<br />
accelerometer, a compass, antennae, <strong>and</strong> optionally chemosensors. Actuator<br />
control is mediated by pulse-width duty cycle coding generated by the electronic<br />
motor neurons <strong>and</strong> a comparator <strong>and</strong> power field-effect transistor (FET) system<br />
housed on low- <strong>and</strong> high-current driver boards. These circular boards are stacked<br />
in a tubular hull with the processor <strong>and</strong> batteries. This system can readily mimic the<br />
biomechanics of the model organisms by the addition of hydrodynamic control surfaces.<br />
The behavioral set results from chaining sequences of exteroceptive reflexes<br />
released by sensory feedback from the environment.<br />
Keywords: biomimetic, robot, UUV, lobster, lamprey<br />
guide reactive autonomy in the most<br />
challenging of environments. The<br />
task is to capture these performance<br />
advantages in engineered devices.<br />
The Biological Model<br />
We are developing a new generation<br />
of lobster <strong>and</strong> lamprey-based<br />
robots that operate under control by<br />
synaptic networks rather than algorithms.<br />
Previous generations of these<br />
vehicles were controlled by finite<br />
state machines that were organized<br />
around the elements of the corresponding<br />
neurobiological models (Ayers<br />
et al., 2000; Ayers & Witting, 2007).<br />
The neuronal circuits that control<br />
our current generation of vehicles are<br />
based on the comm<strong>and</strong> neuron, coordinating<br />
neuron, central pattern generator<br />
(CCCPG) architecture (Figures 1a<br />
<strong>and</strong> 1b) of innate animal behavior<br />
(Kennedy & Davis, 1977; Stein, 1978;<br />
Pearson, 1993). The networks are organized<br />
into segmental central pattern<br />
generators (CPGs) that control appendages<br />
or axial body musculature in the<br />
July/August 2011 Volume 45 Number 4 147
FIGURE 1<br />
CCCPG architecture with exteroceptive reflex. Labeled circles represent neurons; synapses are<br />
shown as connecting lines with triangular (excitatory) or circular (inhibitory) endpoints. (a) Neuronal<br />
circuit-based controller for a walking lobster robot. The effector organs of each body segment<br />
are controlled by CPGs that contain a neuronal oscillator, a pattern generator <strong>and</strong> sets of<br />
motor neuron pools. The CPGs are coordinated among themselves by a set of coordinating neurons<br />
(CoN) that provide information about the activity status of a governing oscillator to a governed<br />
oscillator. The CPGs are brought into operation by a set of comm<strong>and</strong> neurons (CN) that<br />
initiates their operation <strong>and</strong> controls their average frequency <strong>and</strong> amplitude. (b) Neuronal<br />
circuit-based controller for a swimming lamprey robot. Slight modification of the CCCPG architecture<br />
<strong>and</strong> effectors transforms the system’s motor output from walking to swimming. (c) The<br />
CNs are organized into exteroceptive reflexes that are released by neuronally coded sensor information<br />
(rounded rectangles: heading from a compass, target orientation from SBA) through sensory<br />
interneurons, which mediate in place rotation <strong>and</strong> yaw during locomotion.<br />
animal models <strong>and</strong> the robots (Ayers<br />
et al., 2010). The CPGs are coordinated<br />
among themselves by a category<br />
of neurons called coordinating neurons<br />
that pass status information<br />
from a governing CPG to a governed<br />
CPG that alters its period to remain<br />
coordinated at a particular phase,<br />
depending on the ratio of intrinsic frequencies<br />
of the governing <strong>and</strong> governed<br />
CPGs (Selverston & Ayers,<br />
2006). This temporal resetting occurs<br />
on a cycle-by-cycle basis to entrain the<br />
CPGs in a particular gait in the case of<br />
walking or to ensure propagation of a<br />
wave of flexion down the body during<br />
undulation (Figure 1).<br />
The CPGs are brought into operation<br />
<strong>and</strong> modulated by a category of<br />
neurons called comm<strong>and</strong> neurons<br />
(Kupfermann & Weiss, 1978). Comm<strong>and</strong><br />
neurons generally constitute the<br />
locus at which the decision to evoke a<br />
behavioral act is made <strong>and</strong> project<br />
from the brain through the central nervous<br />
system to bring the segmental<br />
CPGs into operation. They typically<br />
perform this process through the<br />
mechanism of neuromodulation<br />
through second messengers that alter<br />
both the cellular properties <strong>and</strong> synaptic<br />
connectivity within the CPG<br />
(Dickinson, 2006). By this mechanism<br />
or through direct synaptic<br />
modulation, the same CPG can often<br />
produce variations on a behavioral act<br />
in response to different comm<strong>and</strong>s<br />
(Selverston & Ayers, 2006).<br />
The sensors we employ are configured<br />
to encode sensory input as a labeled<br />
line code (Bullock, 1968). In this<br />
form of coding, each sensory neuron is<br />
a unique source of information. The<br />
information consists of (1) the nature<br />
of the sensory stimulus (light, optical<br />
flow, chemicals, bumps, etc.), (2) the<br />
receptive field or position of the stimulus<br />
on the body <strong>and</strong> (3) the magnitude<br />
of the stimulus coded as an<br />
action potential train where the frequency<br />
of the action potentials is<br />
proportional to the logarithm of the<br />
stimulus intensity. We configure<br />
these sensory elements in networks<br />
that filter out features of the environment<br />
through lateral inhibition,<br />
range fractionation <strong>and</strong> motion detection.<br />
These filtered outputs provide<br />
input to the comm<strong>and</strong> neurons to<br />
release behavior.<br />
We have completed the lampreybased<br />
robot <strong>and</strong> are adapting this sensor,<br />
board, <strong>and</strong> actuator architecture to<br />
a new generation of the lobster-based<br />
robot (Figure 2). The lamprey robot<br />
features an electronic nervous system<br />
that we are adapting to the new<br />
lobster-based robot. The key feature<br />
of this architecture is that it is generalizable<br />
between all animal models.<br />
Electronic Nervous Systems<br />
Thediscretetimemap-based<br />
(DTM) neuron <strong>and</strong> synapse equations<br />
(Rulkov, 2002) phenomenologically<br />
modelneuronalactivity.Themodel<br />
has two state variables x <strong>and</strong> y, two<br />
FIGURE 2<br />
Biomimetic robots. (a) Fourth generation lobsterbased<br />
robot. (b) Second generation lampreybased<br />
robot.<br />
148 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
control parameters α <strong>and</strong> σ, <strong>and</strong>a<br />
parameter β for integrated synaptic<br />
input. Variations in α <strong>and</strong> σ can configure<br />
neurons into a silent type, a<br />
spiking type, a bursting type <strong>and</strong> a<br />
chaotic type (Ayers & Rulkov, 2007).<br />
Similar control parameters for the synapse<br />
instruments determine the synaptic<br />
strength, relaxation rate, release<br />
threshold, <strong>and</strong> reversal potential that<br />
determine whether the synapse is excitatory<br />
or inhibitory. The electronic nervous<br />
systems are first prototyped in<br />
the National Instruments LabVIEW<br />
software. Neuron <strong>and</strong> synapse instruments<br />
are configured with different<br />
properties, <strong>and</strong> the modeled<br />
neurons <strong>and</strong> synapses are wired together<br />
in LabVIEW.<br />
Figure 3 demonstrates a simple<br />
CPG circuit configured to illustrate<br />
operation of the four types of neurons<br />
in our CPGs. Here, a comm<strong>and</strong> neuron<br />
(1) initiates an oscillation between<br />
a bursting neuron (2) <strong>and</strong> a spiking<br />
follower (3) using a slow modulatory<br />
synapse. A fourth coordinating neuron<br />
(4) can be activated in bursts to entrain<br />
the bursting pattern evoked by the<br />
comm<strong>and</strong>. In contrast to coordinating<br />
neurons that reset the timing of the oscillation<br />
on a cycle-by-cycle basis by<br />
perturbation, comm<strong>and</strong> neurons initiate<br />
operation of the circuits <strong>and</strong> modulate<br />
their average frequency <strong>and</strong> amplitude<br />
as parameters (Figure 3b–c).<br />
FIGURE 3<br />
DTM network integration. (a) The modeled neuronal circuit: (1) comm<strong>and</strong> neuron, (2) bursting<br />
neuron, (3) spiking neuron, (4) coordinating neuron. (b) Parametric modulation of 2 <strong>and</strong> 3 generates<br />
an antagonistic bursting pattern. Voltage vs. simulation iteration traces are shown for each<br />
component of the network. The trace below neuron 1 shows the current injected to initiate activity.<br />
(c) Perturbation of the bursting pattern in 2 by a coordinating neuron (3) to entrain the evoked<br />
rhythm. The trace below 4 shows the injected current into that neuron. Adapted from Ayers <strong>and</strong><br />
Rulkov (2007).<br />
FIGURE 4<br />
Configuration of robots. (a) Current implementation of board set of the lamprey-based robot. The<br />
sonar stack processes the analog hydrophone signals from the short baseline array (SBA). The<br />
electronic nervous system is instantiated on a Texas Instruments TMS320C6727 chip on the CNS<br />
DSP board. Sensors <strong>and</strong> the SBA processor are housed on the sensor array board. Low- <strong>and</strong> highcurrent<br />
drivers provide the current pulse trains that activate the nitinol actuators. The robot operates<br />
on a 12-V, 4.5-Ah NiMH battery pack. (b) Configuration of the lamprey robot. A hinge in the<br />
pitch module allows the undulator to alter its pitch relative to the hull to dive or climb. (c) Configuration<br />
of the lobster robot. A tubular hull similar to the lamprey robot houses the electronics<br />
<strong>and</strong> batteries. Anterior claw-like <strong>and</strong> posterior abdomen-like hydrodynamic control surfaces provide<br />
a thrust vector into the substrate to increase traction. Externally mounted sensors include<br />
optical flow detectors, hydrodynamic flow sensors on the antennae, <strong>and</strong> sonar transducers for<br />
beacon tracking <strong>and</strong> supervisory input.<br />
Board Architecture<br />
The networks that control the robots<br />
are constructed from DTM neurons<br />
<strong>and</strong> synapses in procedural C <strong>and</strong><br />
are instantiated on a Texas Instruments<br />
digital signal processing (DSP)<br />
chip. A common board architecture<br />
is used to control both the swimming<br />
<strong>and</strong> walking robots (Figure 4a).<br />
The board set consists of four types:<br />
July/August 2011 Volume 45 Number 4 149
(1) The DSP board houses the DSP<br />
chip <strong>and</strong> interconnects to sensor <strong>and</strong><br />
actuator boards. (2) A sensor array<br />
houses a compass, inclinometers, accelerometers<br />
<strong>and</strong> a processor board to<br />
derive azimuth <strong>and</strong> inclination deviation<br />
signals from the short baseline<br />
array stack. (3) A low-current driver<br />
board receives logic signals from the<br />
motor neuron output from the DSP<br />
chip <strong>and</strong> in turn controls (4) a highcurrent<br />
driver that applies current to<br />
heat the individual nitinol actuators.<br />
Artificial muscles constructed from<br />
the shape memory alloy nitinol move<br />
both the walking legs <strong>and</strong> the undulatory<br />
body axis. The nitinol is operated<br />
on a thermal cycle. When cooled by<br />
seawater, it can be deformed into martensite<br />
state that is associated with<br />
about a 5% length increase. When<br />
heated by electrical current, it transforms<br />
into the austenite state <strong>and</strong> contracts<br />
rapidly. Increases in the length<br />
of one muscle are produced by the<br />
contraction of its antagonist. Both<br />
the amplitude <strong>and</strong> velocity of the contractions<br />
can be graded by pulse-width<br />
duty cycle modulation of the drive<br />
pulses. Excitation-contraction coupling<br />
with the motor neurons is mediated<br />
by a comparator circuit on the low<br />
current boards that thresholds the<br />
action potentials to generate a square<br />
wave pulse that controls a power on<br />
the high-current boards to activate<br />
the actuators. Changes in the firing frequency<br />
of the motor neurons provide<br />
the duty cycle modulation.<br />
A common DSP board (Figure 4a)<br />
interfaces the sensor array board to<br />
the current driver boards. A separate<br />
regulator board provides the load to<br />
these boards via a 12 V NiMh battery<br />
pack. Feed-through connectors in the<br />
end caps lead the current conductors<br />
to the actuators. The boards are<br />
stacked in a tubular hull whose length<br />
can be varied to accommodate a variety<br />
of mission packages.<br />
Behavioral Set<br />
This system can readily mimic the<br />
biomechanics of the model organisms<br />
by the addition of hydrodynamic control<br />
surfaces. Turns in the undulatory<br />
robot are mediated by modulation<br />
of the amplitude of the flexions to<br />
thetwosidesasintheanimalmodel<br />
(Ayers, 1989). The direction of propagation<br />
of the flexion waves along the<br />
body axis can be reversed to mediate<br />
backward swimming. Dives <strong>and</strong><br />
climbs can be mediated by alteration<br />
of the pitch of the hull relative to the<br />
undulator (Figure 4b). Dorsal flexion<br />
of the hull generates a low-pressure<br />
area above the hull to mediate a climb<br />
while ventral flexion generates a lowpressure<br />
area below the hull to mediate<br />
diving.<br />
The primary response to hydrodynamic<br />
flow in the lobster is to orient<br />
into the flow, lower the anterior control<br />
surfaces <strong>and</strong> elevate the posterior<br />
control surfaces. As the lobster is only<br />
slightly negatively buoyant, this creates<br />
a thrust vector into the substrate <strong>and</strong><br />
increases the traction of the legs on<br />
the bottom. The three degree of freedom<br />
walking legs of the lobster robot<br />
allow the vehicle to walk in all directions<br />
(Ayers & Witting, 2007). Alterations<br />
in the degree of depression can<br />
regulate the height above the substrate,<br />
while variations along the long body<br />
axis regulates pitch. Biasing the depression<br />
on the two sides can correspondingly<br />
regulate roll to maintain primary<br />
orientation on tilted substrates.<br />
The behavioral set of both robots<br />
is organized around exteroceptive reflexes<br />
(Kennedy & Davis, 1977). An<br />
innate releasing mechanism composed<br />
of sensory neurons <strong>and</strong> interneurons<br />
filters incoming information to extract<br />
relevant features of the environment<br />
such as bumps, tilt, hydrodynamic<br />
<strong>and</strong> optical flow (Figures 1c <strong>and</strong> 5a).<br />
These sensory releasers are coded in interneurons<br />
that, in turn, activate comm<strong>and</strong><br />
systems. The interneurons use<br />
lateral inhibition from low-threshold<br />
to high-threshold elements to produce<br />
range fractionation so that different<br />
ranges of a scalar input are coded by<br />
different sensory neurons, providing<br />
the capability for detailed circuit logic.<br />
An example of such exteroceptive<br />
reflexes are those involved in the mediation<br />
of the yaw plane responses to<br />
hydrodynamic flow that occur during<br />
rheotaxis. Lobsters typically walk<br />
with their antenna projected to the<br />
front (Figure 5a, I). If wave surge<br />
occurs from the side, it bends the<br />
upstream antenna medially <strong>and</strong> the<br />
downstream antenna laterally (Figure<br />
5a, II). Our hypothesis is that<br />
this perturbation activates a rheotaxic<br />
interneuron that activates the backward<br />
walking comm<strong>and</strong> on the upstream<br />
side <strong>and</strong> the forward walking<br />
comm<strong>and</strong> on the downstream side.<br />
This would cause the animal/vehicle<br />
to rotate in place into the flow. While<br />
orienting into flow, the animals project<br />
their antenna laterally, which would<br />
switch control to another bilateral<br />
pair of surge interneurons (Figure 5a,<br />
III). As the most upstream antennae<br />
would be bent more than the more<br />
downstream antenna, <strong>and</strong> these interneurons<br />
project to the contralateral<br />
forward walking comm<strong>and</strong>s, the<br />
animal/vehicle would continue to<br />
yaw into the flow until current to the<br />
two antenna is balanced ensuring<br />
proper orientation into the flow for<br />
maximal hydrodynamic stability. The<br />
hydrodynamic control surfaces can<br />
then ensure proper traction to overcome<br />
the perturbation.<br />
150 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 5<br />
Neural simulation of rheotaxis. (a) Epochs of a lobster’s rheotaxic behavioral response to water<br />
surge shown with the predominant active neural reflex circuit. In the network diagrams, top circles<br />
represent sensory neurons corresponding to high (H), medium (M), or low (L) antennal bending in<br />
the lateral or medial direction. Black ovals represent interneurons that project to bilateral comm<strong>and</strong>s<br />
for forward (F) or backward (B) walking. In I, as a lobster walks forward, bilaterally balanced<br />
low lateral bending of the antennae is observed, which serves to sustain forward locomotion. In II,<br />
left-to-right water surge (blue arrows) causes a high medial bend of the ipsilateral antenna <strong>and</strong> a<br />
high lateral bend of the contralateral antenna eliciting rheotaxis. In III, bilaterally asymmetrical<br />
lateral antennal bending mediates yawing upstream during forward walking. (b) Voltage vs.<br />
time traces for the neurons of the rheotaxis circuit. Dashed lines distinguish the behavioral epochs<br />
shown in a.<br />
alter locomotory outputs of a network,<br />
as we have shown to produce both<br />
swimming <strong>and</strong> walking. Even hybrid<br />
systems can be achieved, such as the<br />
gait of an alligator resulting from a<br />
combination of the lamprey <strong>and</strong> lobster<br />
CPG networks. Sensor packages<br />
can be adapted with little restructuring.<br />
Layered exteroceptive reflex<br />
networks provide capabilities for navigation,<br />
investigation <strong>and</strong> obstacle negotiation<br />
in unpredictable near-shore<br />
marine environments with a minimum<br />
of supervisory control.<br />
Lead Author:<br />
Joseph Ayers<br />
Department of Biology <strong>and</strong><br />
<strong>Marine</strong> Science Center<br />
Northeastern University,<br />
East Point, Nahant MA 01908<br />
Email: lobster@neu.edu<br />
This overall control scheme applies<br />
to a variety of environmental circumstances<br />
<strong>and</strong> perturbations in the yaw,<br />
pitch, <strong>and</strong> roll planes. Many exteroceptive<br />
reflexes form taxic systems.<br />
For example, the three hydrophone<br />
short baseline sonar array (SBA) on<br />
the lamprey robot reports the deviation<br />
of the sonar beacon relative to<br />
the hull orientation in terms of inclination<br />
<strong>and</strong> azimuth (Westphal et al.,<br />
2011). The azimuthal signal modulates<br />
swim comm<strong>and</strong> systems to<br />
cause the vehicle to yaw toward the<br />
beacon (Figure 1c) while the inclination<br />
signal modulates the pitch system<br />
to cause the vehicle to climb/dive<br />
toward the beacon. Taken together<br />
these layered reflexes will cause the<br />
vehicle to home on a sonar beacon. A<br />
similar 2-D SBA is planned for the<br />
lobster robot to control yaw taxis.<br />
The sonar transducers also provide<br />
a capability for supervisory control.<br />
The vehicles will be sent supervisory<br />
comm<strong>and</strong>s that specify a heading <strong>and</strong><br />
odometry information for distance.<br />
The comm<strong>and</strong> will include a propensity<br />
to negotiate or investigate obstacles<br />
depending on the mission. At the<br />
end of the search vector, the vehicle<br />
would annunciate its location to a<br />
long baseline sonar array <strong>and</strong> be sent<br />
a new search vector. By this mechanism<br />
an operator could supervise several<br />
robots simultaneously.<br />
Conclusion<br />
The neuronal mechanisms of<br />
innate behavior can be applied to a<br />
broad variety of biomimetic underwater<br />
robots. Minimal modification<br />
of neuronal components can easily<br />
References<br />
Ayers, J. 1989. Recovery of oscillator function<br />
following spinal regeneration in the Sea<br />
Lamprey. In: Neuronal <strong>and</strong> Cellular Oscillators,<br />
ed. Jacklett, J., 371-406. New York: Marcel<br />
Dekker Inc.<br />
Ayers, J. 2004. Underwater walking. Arthropod<br />
Struct Dev. 33(3):347-60. doi: 10.1016/<br />
j.asd.2004.06.001.<br />
Ayers, J., & Rulkov, N. 2007. Controlling<br />
biomimetic underwater robots with electronic<br />
nervous systems. In: Bio-mechanisms of<br />
Animals in Swimming <strong>and</strong> Flying, eds.<br />
Kato, N., & Kamimura, S., 295-306.<br />
Tokyo: Springer-Verlag.<br />
Ayers, J., Rulkov, N., Knudsen, D., Kim,<br />
Y-B., Volkovskii, A., & Selverston, A. 2010.<br />
Controlling underwater robots with electronic<br />
nervous systems. Appl Bionics Biomech.<br />
7:57-67. doi: 10.1080/11762320903244843.<br />
Ayers, J., Wilbur, C., & Olcott, C. 2000.<br />
Lamprey robots. In: Proceedings of the<br />
International Symposium on Aqua<br />
July/August 2011 Volume 45 Number 4 151
Biomechanisms, eds. Wu, T., & Kato, N.,<br />
Tokai University.<br />
Ayers, J., & Witting, J. 2007. Biomimetic<br />
approaches to the control of underwater<br />
walking machines. Philos T R Soc Lond A.<br />
365:273-95. doi: 10.1098/rsta.2006.1910.<br />
Westphal, A., Rulkov, N., Ayers, J., Brady, D.,<br />
& Hunt, M. 2011. Controlling a lampreybased<br />
robot with an electronic nervous system.<br />
Smart Struct Syst. 7(6):471-84.<br />
Blustein, D., & Ayers, J. 2010. A conserved<br />
network for control of arthropod exteroceptive<br />
optical flow reflexes during locomotion. Lect<br />
Notes Artif Int. 6226:72-81.<br />
Bullock, T.H. 1968. Representation of information<br />
in neurons <strong>and</strong> sites for molecular<br />
participation. P Natl Acad Sci. 60:1058-68.<br />
doi: 10.1073/pnas.60.4.1058.<br />
Dickinson, P.S. 2006. Neuromodulation of<br />
central pattern generators in invertebrates <strong>and</strong><br />
vertebrates. Curr Opin Neurobiol. 16(6):<br />
604-14. doi: 10.1016/j.conb.2006.10.007.<br />
Kennedy, D., & Davis, W.J. 1977. Organization<br />
of invertebrate motor systems. In:<br />
H<strong>and</strong>book of Physiology, eds. Geiger, S.R.,<br />
K<strong>and</strong>el, E.R., Brookhart, J.M., & Mountcastle,<br />
V.B., Section 1, Vol. 1, Pt. 2:1023-87.<br />
Bethesda, MD: American Physiological<br />
<strong>Society</strong>. Section 1, Part 2.<br />
Kupfermann, I., & Weiss, K. 1978. The<br />
comm<strong>and</strong> neuron concept. Behav Brain<br />
Sciences. 1:3-39.<br />
Pearson, K.G. 1993. Common principles of<br />
motor control in vertebrates <strong>and</strong> invertebrates.<br />
Annu Rev Neurosci. 16:265-97. doi: 10.1146/<br />
annurev.ne.16.030193.001405.<br />
Rulkov, N.F. 2002. Modeling of spikingbursting<br />
neural behavior using two-dimensional<br />
map. Phys Rev E. 65:041922. doi: 10.1103/<br />
PhysRevE.65.041922.<br />
Selverston, A., & Ayers, J. 2006. Oscillations<br />
<strong>and</strong> oscillatory behavior in small neural circuits.<br />
Biol Cybern. 95:537-54. doi: 10.1007/<br />
s00422-006-0125-1.<br />
Stein, P.S.G. 1978. Motor systems, with<br />
specific reference to the control of locomotion.<br />
Annu Rev Neurosci. 1:61-81. doi: 10.1146/<br />
annurev.ne.01.030178.000425.<br />
152 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
PAPER<br />
A Hybrid Class Underwater Vehicle:<br />
Bioinspired Propulsion, Embedded<br />
System, <strong>and</strong> Acoustic Communication<br />
<strong>and</strong> Localization System<br />
AUTHORS<br />
Michael Krieg<br />
Peter Klein<br />
Robert Hodgkinson<br />
Department of Mechanical<br />
<strong>and</strong> Aerospace Engineering,<br />
University of Florida<br />
Kamran Mohseni 1<br />
Departments of Mechanical<br />
<strong>and</strong> Aerospace Engineering<br />
<strong>and</strong> Computer Engineering,<br />
Institute for Cyber Autonomous<br />
Systems, University of Florida<br />
Introduction<br />
Traditionally unmanned underwater<br />
vehicles fall into one of two categories.<br />
One class of vehicles (torpedo<br />
like) are built to travel long distances<br />
with minimal energy <strong>and</strong> are usually<br />
characterized by a long slender body,<br />
a rear propeller for propulsion <strong>and</strong> a<br />
set of fins to provide maneuvering<br />
forces. This type of vehicle is poorly<br />
suited for missions requiring a high degree<br />
of positioning accuracy because<br />
the control surfaces provide little to<br />
no maneuvering force at low forward<br />
velocity. The other class of vehicle<br />
similar to remotely operated vehicles<br />
(ROVs) is designed to operate in<br />
these situations, which do require<br />
1 This work was started while the group was at<br />
the University of Colorado.<br />
ABSTRACT<br />
Inspired by the natural locomotion of jellyfish <strong>and</strong> squid, a series of compact<br />
thrusters series is developed for propulsion <strong>and</strong> maneuvering of underwater vehicles.<br />
These thrusters successively ingest <strong>and</strong> expel jets of water in a controlled manner<br />
at high frequencies to generate propulsive forces. The parameters controlling<br />
the performance of the thrusters are reviewed <strong>and</strong> investigated to achieve higher<br />
thrust levels. The thrusters are compact <strong>and</strong> can be placed completely inside a vehicle<br />
hull providing the desired maneuvering capability without sacrificing a sleek<br />
hydrodynamic shape for efficient cruising. The system design of a prototype hybrid<br />
vehicle, called CephaloBot, utilizing these thrusters, is also presented. A compact<br />
<strong>and</strong> custom-developed embedded system is also designed for the CephaloBot. Key<br />
features of the system include a base set of navigational sensors, an acoustic<br />
system for localization <strong>and</strong> underwater communication, Xbee RF transceiver for<br />
communication above water, <strong>and</strong> a LabVIEW programmed processing board.<br />
Keywords: AUV, thruster, bioinspired, communication<br />
high positioning accuracy, <strong>and</strong> incorporate<br />
several thrusters at various locations<br />
to provide maneuvering forces<br />
in all directions. However, this class<br />
of vehicle typically has a very high<br />
drag coefficient due to the abundance<br />
of external thrusters <strong>and</strong> cannot travel<br />
to remote locations without additional<br />
support.<br />
The abundance of remote marine<br />
research sites requiring high positioning<br />
accuracy for inspection, as well as<br />
the desire to create fully autonomous<br />
vehicle sensor networks, has inspired<br />
significant research in a hybrid class<br />
of vehicles with the efficient cruising<br />
characteristics of the torpedo class<br />
<strong>and</strong> the maneuvering abilities of the<br />
ROV class. Some take a mechanical<br />
approach moving the maneuvering<br />
propellers into tunnels which run<br />
through the hull of the vehicle<br />
(Mclean, 1991; Torsiello, 1994) or<br />
into the fins themselves (Dunbabin<br />
et al., 2005). Others observe that nature’s<br />
swimmers have a healthy balance<br />
of long-distance endurance <strong>and</strong> highaccuracy<br />
low-speed maneuvering. Vehicles<br />
have been designed to use fins<br />
for both high-speed maneuvering as<br />
well as mimic the low-speed flapping<br />
of turtles <strong>and</strong> marine mammals (Licht<br />
et al., 2004; Licht, 2008; Kato,<br />
2011); <strong>and</strong> some use tail fins as a primary<br />
means of propulsion (Barrett<br />
et al., 1999). Our inspiration comes<br />
from the cyclical jet propulsion seen<br />
in jellyfish, scallops, octopus, squid<br />
<strong>and</strong> other cephalopod. Squid jet propulsion<br />
produces the fastest swimming<br />
July/August 2011 Volume 45 Number 4 153
velocities seen in aquatic invertebrates<br />
(O’Dor & Webber, 1991; Anderson<br />
& Grosenbaugh, 2005).<br />
Jetting locomotion begins when<br />
the squid inhales seawater through a<br />
pair of ostia behind the head, filling<br />
the mantle cavity (see Figure 1). The<br />
mantle then contracts forcing seawater<br />
out through the funnel that rolls into a<br />
high-momentum vortex ring <strong>and</strong> imparts<br />
the necessary propulsive force<br />
(Anderson & Grosenbaugh, 2005).<br />
The versatility of the system permits<br />
two distinct gaits, cruising <strong>and</strong> escape<br />
jetting (Bartol et al., 2008). During<br />
cruising, squid swim at nominal<br />
speed with a greater efficiency than escape<br />
jetting, which involves a hyperinflation<br />
of the mantle followed by a<br />
fast powerful contraction to impart<br />
significant acceleration at the cost of<br />
both muscular <strong>and</strong> fluid dynamic<br />
losses. Bartol et al. (2009) report cruising<br />
mode efficiency at 69% (±14%)<br />
averaged over several species <strong>and</strong><br />
swimming speeds <strong>and</strong> 59% (±14%)<br />
for escape jetting. Additionally, propulsive<br />
efficiency was seen to rise as<br />
high as 78% in adult L. brevis swimming<br />
at high velocities <strong>and</strong> averaged<br />
87% (±6.5%) for paralarvae (Bartol<br />
et al., 2008), challenging the notion<br />
that a low-volume high-velocity jet<br />
inherently negates a high propulsive<br />
efficiency.<br />
The locomotion of jellyfish tends<br />
to be very similar to that of squid<br />
with some key differences, primarily<br />
that the refilling phase of jellyfish<br />
FIGURE 1<br />
Diagram of squid layout <strong>and</strong> locomotion.<br />
swimming uses the same bell opening<br />
as the jetting phase. Despite the fact<br />
that squid do not use the funnel during<br />
refilling, the inlet vents are still<br />
on the anterior side of the mantle cavity,<br />
meaning that locomotion for both<br />
organisms is quite different from traditional<br />
pumping mechanisms. Jellyfish<br />
use the cyclic jetting process for<br />
feeding as well as locomotion as is<br />
evidenced by Lagrangian coherent<br />
structures <strong>and</strong> particle tracer analysis<br />
(Lipinski & Mohseni, 2009; Wilson<br />
et al., 2009); in addition, both squid<br />
<strong>and</strong> jellyfish utilize jetting for respiration,<br />
taking advantage of the large<br />
fluid flow rates. Both of these factors<br />
can make it difficult to determine<br />
which swimming behaviors are optimized<br />
for propulsion versus secondary<br />
functions. Similar to the different gaits<br />
seen in squid locomotion, different<br />
species of jellyfish generally fall into<br />
two categories of swimmers based on<br />
the ‘quality’ of vortex ring they produce.<br />
Jellyfish like moon jellyfish<br />
have a very large bell opening, <strong>and</strong><br />
the jetting motion is similar to a paddling<br />
type motion. Box jellyfish <strong>and</strong><br />
otherfasterswimmingjellyfish have<br />
smaller bell openings with nozzle-like<br />
flaps<strong>and</strong>haveamuchmoredistinct<br />
jet. Jellyfish morphology during swimming<br />
has been digitally captured from<br />
experiment, <strong>and</strong> the body motions<br />
were imported into numerical simulations<br />
to predict body forces on the<br />
swimming jellyfish, determining drastically<br />
different swimming efficiencies.<br />
Froude propulsive efficiency of jellyfish<br />
was directly calculated by Sahin<br />
<strong>and</strong> Mohseni (2008, 2009; Sahin et al.,<br />
2009) to be 37% for Aeqorea victoria<br />
<strong>and</strong> 17% for Sarsia tubulosa. It should<br />
be noted that both species of jellyfish<br />
most likely do not use vortex<br />
generation for the sole purpose of<br />
locomotion. Aeqorea victoria uses vortex<br />
generation for feeding <strong>and</strong> Sarsia<br />
tubulosa as an escape mechanism. Empirical<br />
data gathered through digital<br />
particle image velocimetry (DPIV)<br />
measurements of several species shows<br />
similar efficiency characteristics for<br />
the different swimming patterns<br />
(Dabiri et al., 2010).<br />
The general concept of propelling<br />
water craft by ejecting a high-velocity<br />
water jet is centuries old, was hypothesized<br />
by both Bernoulli <strong>and</strong> Benjamin<br />
Franklin, <strong>and</strong> was utilized in a rudimentary<br />
sense in one of the first steam<br />
boat designs by James Rumsey (Allen,<br />
2010). Continuously pumped jets are<br />
usedforpropulsioninmodernwater<br />
craft-like jet skis <strong>and</strong> bow thrusters<br />
of motorboats; however, this type of<br />
jet propulsion is inherently different<br />
from the propulsion of squid <strong>and</strong> jellyfish,<br />
which create distinct vortex rings.<br />
The thrusters of this paper also produce<br />
finite jets, which form arrays of<br />
vortex rings, <strong>and</strong> should be considered<br />
fundamentally different than continuous<br />
jet thrusters.<br />
This paper showcases a complete<br />
hybrid class vehicle that demonstrates<br />
added maneuvering capabilities utilizing<br />
a set of bio-inspired jet thrusters.<br />
The manuscript will focus on three<br />
primary systems of the vehicle: a bioinspired<br />
thruster system <strong>and</strong> fundamentals<br />
of thruster mechanics, an<br />
acoustic system, which serves the<br />
dual purpose of communication <strong>and</strong><br />
localization, <strong>and</strong> a compact embedded<br />
control system. The manuscript is<br />
154 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
organized as follows. The mechanics<br />
ofthethrusteraswellasthethrust<br />
dynamics are described in ‘Vortex<br />
Ring Thrusters’ section. A brief history<br />
of thruster <strong>and</strong> vehicle prototypes is<br />
given in ‘Thruster <strong>and</strong> CephaloBot<br />
Evolution’. The‘Hybrid Vehicle Description’<br />
section gives basic requirements<br />
for a hybrid class vehicle, <strong>and</strong> the<br />
subsystem components of CephaloBot<br />
are described in more details in ‘Communication<br />
Localization System’,<br />
‘Embedded System’, <strong>and</strong>‘Sensors’<br />
sections.<br />
Vortex Ring Thrusters<br />
Ourthrusterinspiredbyjellyfish<br />
<strong>and</strong> squid propulsion consists of an internal<br />
fluid cavity, with a semi-flexible<br />
plunger used to drive fluid motion,<br />
<strong>and</strong> a small circular orifice exposed to<br />
the external fluid. See Figure 2 for a<br />
diagram of the thruster layout. The<br />
cavity of the thruster provides the<br />
same functionality as the squid mantle<br />
or the jellyfish bell, exp<strong>and</strong>ing <strong>and</strong><br />
contracting to cycle water in <strong>and</strong> out<br />
of the circular orifice (functionally<br />
similar to the squid funnel/siphon).<br />
Since the thruster generates propulsion<br />
by creating energetic vortex rings, it<br />
FIGURE 2<br />
Conceptual diagram of the thruster key components.<br />
Jet shown as hypothetical slug of<br />
fluid.<br />
is termed the ‘vortex ring thruster’<br />
(VRT).<br />
Since VRTs are contained internal<br />
to the vehicle (with only a small opening<br />
on the vehicle surface), they do not<br />
significantly affect the vehicle’s forward<br />
drag profile, which means that a<br />
vehicle equipped with a set of VRTs<br />
for low-speed maneuvering <strong>and</strong> a rear<br />
propeller for primary propulsion will<br />
have a sleek aerodynamic shape allowing<br />
fast efficient cruising to a site of<br />
interest but still maintain full maneuverability<br />
(even at zero forward speed)<br />
upon reaching that site of interest. See<br />
Krieg <strong>and</strong> Mohseni (2010) for ‘parallel<br />
parking’ capability of an earlier version<br />
of our vehicle. Additionally, since<br />
the VRT only needs a single opening<br />
(unlike tunnel thrusters or traditional<br />
pumps, which extend from one end<br />
of the hull to the other), it allows for<br />
a greater degree of freedom for internal<br />
system arrangement.<br />
The impulse generated by this type<br />
of device can be modeled as if the jet<br />
acts like a solid slug of fluid with a uniform<br />
velocity across the nozzle opening<br />
(Mohseni, 2004, 2006; Krieg &<br />
Mohseni, 2008). Properties of fluid<br />
slug have been investigated by several<br />
groups (Glezer, 1988; Gharib et al.,<br />
1998; Shariff & Leonard, 1992;<br />
Mohseni & Gharib, 1998; Krieg &<br />
Mohseni, 2008). The total impulse<br />
generated for a single pulsation under<br />
slug assumptions is<br />
I slug ðÞ¼ρπ=4∫ t<br />
t 0 u2 ðτÞD 2 dτ;<br />
ð1Þ<br />
Krueger <strong>and</strong> Gharib (2003) showed<br />
that the impulse created by a cylinder<br />
piston type vortex generator was consistently<br />
higher than the impulse predicted<br />
by the slug model. This added<br />
impulse was attributed to a pressure<br />
gradient at the nozzle exit plane,<br />
referred to as ‘nozzle overpressure.’<br />
Adding I p (the impulse due to overpressure),<br />
we get an equation for the<br />
impulse, in terms of the nozzle pressure,<br />
p (which is a function of time<br />
<strong>and</strong> radial position), <strong>and</strong> the stagnation<br />
pressure, p ∞ .<br />
IðÞ¼I t slug ðÞþI t p ðÞ t<br />
I p t<br />
ðÞ¼∫ t 0∫ A pr; τ<br />
½ ð Þ p ∞ ŠdAdτ ð2Þ<br />
We performed initial testing on a<br />
thruster which periodically ingested<br />
<strong>and</strong> expelled jets with a sinusoidal velocity<br />
program; which was chosen for<br />
simplicity of fabrication. Assuming<br />
that there is no net momentum transfer<br />
during the ingestion phase (fluid<br />
being taken into the cavity starts at<br />
rest outside of the thruster <strong>and</strong> ends<br />
at rest inside the cavity) <strong>and</strong> ignoring<br />
the pressure impulse (which will be<br />
accounted for with a coefficient term<br />
post analysis), then the average thrust<br />
produced over a full cycle can be calculated<br />
in terms of the thruster frequency<br />
to be (Krieg & Mohseni, 2008)<br />
<br />
‐<br />
T ¼ ρπ3 L 2<br />
16 D4 f 2 ð3Þ<br />
D<br />
where u is the piston velocity (mass<br />
flux across thruster opening divided<br />
by nozzle area), D is the nozzle Diameter,<br />
ρ is the fluid density, t is the time<br />
at which the impulse is evaluated, <strong>and</strong><br />
τ is a dummy variable for time initialized<br />
at the beginning of pulsation.<br />
Here f is the frequency of actuation,<br />
<strong>and</strong> the term L=D is the jet stroke<br />
ratio. If the jet maintained its shape as<br />
a solid cylinder the stroke ratio would<br />
be the ratio of length to diameter of<br />
that cylinder (see Figure 2). The stroke<br />
ratio has also been called the formation<br />
July/August 2011 Volume 45 Number 4 155
FIGURE 3<br />
Thruster static testing environment.<br />
time since it is equivalent to the time<br />
since initiation of the jet flow scaled<br />
by jet velocity <strong>and</strong> nozzle diameter,<br />
t* ¼ L=D ¼ ∫ t 0 u ð τ Þdt=D. Theformation<br />
time is closely related to the<br />
jet formation dynamics. When a jet is<br />
expelled into a stationary fluid, the viscous<br />
forces cause the initial portion of<br />
the jet to roll into a tightly wound vortex<br />
ring. As more fluid is expelled, it<br />
feeds the growing vortex ring until<br />
a critical saturation point is reached<br />
<strong>and</strong> the vortex ring can no longer support<br />
the added circulation. At this<br />
point, the vortex ring separates from<br />
the remaining shear flow. The formation<br />
time when the jet has achieved the<br />
same circulation as the final vortex ring<br />
is known as the formation number.<br />
Gharib et al. (1998) demonstrated<br />
that impulsively started vortex rings<br />
have a universal formation number<br />
(≈3.6-4.2) independent of jet velocity<br />
<strong>and</strong> diameter. However, numerical<br />
studies have shown the formation<br />
number to be drastically lower for<br />
jets created with a parabolic velocity<br />
profile (Rosenfeld et al., 1998) or a<br />
2-D jet velocity like those produced<br />
in conical nozzles (Rosenfeld et al.,<br />
2009).<br />
The slug model predicts that the<br />
average thrust is proportional to both<br />
the square of the actuation frequency<br />
<strong>and</strong> the square of the stroke ratio/<br />
formation time. To test this assertion,<br />
the thruster was placed in a static fluid<br />
reservoir <strong>and</strong> suspended from a load<br />
cell, <strong>and</strong> the thrust output was measured<br />
directly. This testing setup is<br />
shown in Figure 3 (Krieg & Mohseni,<br />
2008). It should be noted that if the<br />
vehicle is moving during thruster<br />
pulsation there will inherently be a<br />
non-zero momentum transfer during<br />
ingestion; however, as was previously<br />
mentioned these thrusters are primarily<br />
intended for low-speed maneuvering<br />
(vehicle velocities well below<br />
jetting velocities), so that the momentum<br />
transfer during ingestion will still<br />
be negligible compared to the momentum<br />
transfer of the jetting phase. In<br />
addition, the expelled jet rolls into an<br />
isolated vortex ring which inherently<br />
produces a large momentum transfer<br />
associated with the fluid impulse<br />
of the vortex ring itself. During ingestion,<br />
the small internal cavity severely<br />
limits vortex ring formation as well as<br />
momentum transfer associated with it.<br />
The thruster was tested over a wide<br />
range of stroke ratios <strong>and</strong> actuation frequencies.<br />
As can be seen in Figure 4,<br />
the thrust shows a square proportionality<br />
to frequency, within a certain frequency<br />
range. When producing jets<br />
with low stroke ratio, this range this<br />
is the entire sub-cavitation frequency<br />
range. However, when the jet stroke<br />
ratio goes above the formation number,<br />
the thruster exhibits a parabolic<br />
dependence on the actuation frequency<br />
after a short range of square dependence.<br />
To more clearly show this<br />
trend, we define a scale factor, which<br />
FIGURE 4<br />
Thrust versus frequency. Each set of markers<br />
shows thrust data for a different stroke ratio.<br />
The error bars plotted along with the thrust<br />
relationship represent a single st<strong>and</strong>ard deviation<br />
of the thrust data.<br />
is a measure of the accuracy of the<br />
slug model for various operating conditions<br />
α = T Exp / ‐ T (Krieg & Mohseni,<br />
2008) where ‐ T is the slug model predicted<br />
average thrust from equation (3).<br />
This scale factor is plotted with respect<br />
to frequency for stroke ratios below<br />
the formation number in Figure 5a<br />
<strong>and</strong> for stroke ratios above the formation<br />
number in Figure 5b. Note that<br />
the thruster of this study has a nozzle<br />
which is essentially a flat plate with a<br />
circular orifice in the middle. This<br />
type of nozzle produces a 2-D jet<br />
flow similar to a conical nozzle, meaning<br />
that the jet formation number is<br />
closer to 3 (Rosenfeld et al., 2009).<br />
For a more in depth analysis of the<br />
shift in formation number due to the<br />
2-D aspect of the jet, see Krieg <strong>and</strong><br />
Mohseni (2011).<br />
First, consider the thrust response<br />
of the actuator operating below the<br />
formation number (Figure 5a). In the<br />
low-frequency regime, the scale factor<br />
is higher than that predicted by the<br />
slug model due to the nozzle overpressure,<br />
reaching 1.4 times the predicted<br />
value. Krueger <strong>and</strong> Gharib (2003)<br />
showed that the pressure impulse can<br />
reach as much as 40% of the total<br />
156 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 5<br />
Scale factor (slug model accuracy) versus frequency for stroke ratios below (a) <strong>and</strong> above<br />
(b) the formation number. Error bars shown on data points indicate a st<strong>and</strong>ard deviation of<br />
measured values at that frequency (also taken in the scaled space).<br />
FIGURE 6<br />
Successive frames of jet flow showing the<br />
thruster re-ingesting wake flow.<br />
impulse for low stroke ratio jets corresponding<br />
to a total impulse 1.6 times<br />
the predicted slug model impulse.<br />
However, as the frequency increases,<br />
the total thrust settles on the value predicted<br />
by the slug model, meaning that<br />
in this frequency range the impulse<br />
due to overpressure during expulsion<br />
is equal to the impulse due to “underpressure”<br />
during refilling so that the<br />
net impulse transfer is that predicted<br />
by the slug model. Now consider<br />
the thruster response when operating<br />
above the formation number, shown<br />
in Figure 5b. Again the low-frequency<br />
ranges exhibit an added impulse due<br />
to the nozzle overpressure; albeit to a<br />
lower extent as observed by Krueger <strong>and</strong><br />
Gharib (2003). But the high-frequency<br />
range exhibits an added loss in thrust<br />
with respect to the slug model prediction.<br />
This relative loss is seen to<br />
increase monotonically with both actuation<br />
frequency <strong>and</strong> stroke ratio.<br />
This suggests that another assumption<br />
made in the slug model is no longer<br />
valid when operating above the formation<br />
number. We assume that this loss<br />
in model accuracy is tied into the assumption<br />
made that all fluid being<br />
ingested between pulsations is at rest<br />
outside of the thruster. When a jet is<br />
ejected with a stroke ratio above the<br />
formation number, some of the shear<br />
flow is left behind in the trailing<br />
wake of the leading vortex ring. The<br />
trailing wake has a lower momentum<br />
than the leading vortex ring <strong>and</strong> travels<br />
at a much lower induced velocity but<br />
still has a forward momentum substantially<br />
larger than the surrounding resting<br />
fluid. Therefore, the loss in slug<br />
model accuracy could be explained<br />
by the thruster ingesting some of<br />
the trailing wake during the refilling<br />
phase. Figure 6 shows successive<br />
frames from a video of the thruster’s<br />
forming jet (at a high stroke ratio)<br />
where some of the trailing wake is ingested<br />
back into the thruster.<br />
It should also be noted that the<br />
scale factor results are only presented<br />
above an actuation frequency of 4 Hz,<br />
because the thruster was designed to<br />
operate cyclically rather than generate<br />
individual pulsations. The 2-D nature<br />
of the jet created by this type of<br />
thruster (orifice nozzle) has an added<br />
effectonthenozzleoverpressurenot<br />
seen in the frequency ranges presented.<br />
This effect is fully explained (along<br />
with a more in depth description of<br />
impulse generation) for a single pulsation<br />
with constant jet velocity in Krieg<br />
<strong>and</strong> Mohseni (2011). Despite the<br />
relative magnitude of the overpressure<br />
impulse (with respect to the momentum<br />
impulse, I slug ), it is only observed<br />
in low actuation frequencies, which are<br />
coupled with low thrust output. Often<br />
vehicle mission scenarios will necessitate<br />
a large magnitude thrust output<br />
(higher actuation frequencies), where<br />
the overpressure is cancelled out, <strong>and</strong><br />
the slug model provides an accurate<br />
thrust measurement.<br />
The vortex ring formation phenomenon<br />
plays a key role in the jet<br />
locomotion process in squid as well.<br />
Bartol et al. (2009) observed that squid<br />
have two distinct swimming gaits. In<br />
the efficient cruising gate jets are expelled<br />
below the formation number,<br />
so that the majority of the jet rolls into<br />
the primary vortex ring. Alternatively,<br />
in threatening situations the squid employs<br />
a swimming technique referred<br />
to as escape jetting; which begins with<br />
the hyperinflation of the mantle followed<br />
by a fast contraction expelling a<br />
jet well above the formation number.<br />
July/August 2011 Volume 45 Number 4 157
Presumably this behavior indicates that<br />
this type of jet propulsion is most efficient<br />
when expelling jets below the formation<br />
number <strong>and</strong> that jetting above<br />
the formation number can achieve higher<br />
thrust at the expense of fluid losses.<br />
Therefore, all subsequent vehicle<br />
thrusters have been designed with a<br />
set diameter resulting in a stroke ratio<br />
near the formation number; to achieve<br />
a maximum level of thrust, while still<br />
being accurately described by the slug<br />
model. It should be noted that the formation<br />
number for a jet created on a<br />
moving vehicle will not be the same<br />
as the formation number of the jet in<br />
the static setup. Krueger et al. (2006)<br />
showed that vortex rings formed in<br />
the presence of a uniform background<br />
co-flow have a lower formation number<br />
than vortex rings formed in a resting<br />
fluid. Since a moving vehicle will<br />
inherently induce a co-flow with the<br />
jetting direction, this effect must be<br />
taken into consideration. However,<br />
the reduction in formation number<br />
is proportional to the ratio between<br />
co-flow velocity <strong>and</strong> jet velocity; therefore,<br />
the low-frequency (low jet velocity)<br />
pulsation will experience pinch off<br />
at an earlier formation time, but the<br />
lower-frequency pulsation is also less<br />
effected by the dynamics of cyclic vortex<br />
ring formation.<br />
One advantage of the VRT is that<br />
it produces a desired level of thrust<br />
almost instantaneously (Krieg &<br />
Mohseni, 2010). Propeller style thrusters<br />
suffer from a rise time associated<br />
with reaching the static thrust level<br />
after initiating rotation. This rise time<br />
is inversely proportional to the desired<br />
level of thrust <strong>and</strong> can be on the order<br />
of several seconds for low thrust levels<br />
(Fossen, 1991; Yoerger et al., 1990).<br />
VRTs also have a rise time associated<br />
with reaching the desired level of thrust,<br />
which is inversely proportional to the<br />
level of thrust. However, this rise time<br />
is an order of magnitude smaller for<br />
VRTs. The exact thrust program as a<br />
function of time is sinusoidal, due to<br />
the nature of the thruster; however,<br />
using several thrust data sets to average<br />
out the dynamic component <strong>and</strong> fitting<br />
the average thrust to a basic logarithmic<br />
curve the mean rise time can be observed.<br />
The fitted curves for several operational<br />
frequencies, <strong>and</strong> a stroke ratio<br />
of 4.3 is shown in Figure 7.<br />
FIGURE 7<br />
Mean thrust produced at various frequencies<br />
(static desired thrust level) versus time. Rise<br />
time inversely proportional to thrust level.<br />
Along with a minimal rise time, the<br />
VRT is also immune to a thrust lag seen<br />
in tunnel thrusters (Mclean, 1991),<br />
where thrust continues to be exerted<br />
on the vehicle after the thruster has<br />
been terminated.<br />
Thruster <strong>and</strong> CephaloBot<br />
Evolution<br />
The compact thrusters used in vehicle<br />
testbeds to provide validation of<br />
static testing have taken a wide variety<br />
of forms. The first generation utilized<br />
a solenoid driving mechanism <strong>and</strong> a<br />
flexing diaphragm to expel fluid (Figure<br />
8a). The solenoid driving mechanism<br />
suffered from reduced stroke<br />
at higher actuation frequencies as the<br />
solenoid stroke was load dependent.<br />
All of the subsequent iterations have<br />
used mechanical driving mechanisms<br />
for better flexibility in adjusting the operation<br />
conditions. Gen. 2 (Figure 8b)<br />
used a completely flexible cavity in a<br />
fitted mold, whereas Gen. 3 <strong>and</strong> 4<br />
switched to a semi-flexible cavity<br />
which was reinforced to ensure constant<br />
diameter but allow compression<br />
in height. Gen. 3 (Figure 8c) used a<br />
complicated encompassing cylindrical<br />
tube cam to drive cavity compression,<br />
but mechanical complexities <strong>and</strong> reliability<br />
issues caused us to simplify to<br />
a basic crank shaft design for Gen. 4<br />
(Figure 8d). In the figure, this thruster<br />
is shown with an acrylic casing to allow<br />
the components to be seen. The vehicle<br />
ready thruster uses an aluminum<br />
casing (Clark et al., 2009).<br />
Along with the thrusters themselves<br />
the vehicle testbeds housing<br />
the thrusters have evolved rapidly. Figure<br />
9 shows the evolution of vehicle<br />
testbeds used to demonstrate the feasibility<br />
of maneuvering using VRTs.<br />
Starting with the oldest vehicle at the<br />
bottom <strong>and</strong> successive generations<br />
upwards, the first vehicle only had<br />
fins to provide a st<strong>and</strong>ard for maneuvering<br />
capabilities, 2nd, 3rd <strong>and</strong> 4th<br />
generation vehicles contain 1st, 2nd<br />
<strong>and</strong> 3rd generation thrusters, respectively,<br />
<strong>and</strong> increasing levels of autonomy.<br />
The lessons learned from these<br />
vehicles directly lead to the development<br />
of the most recent hybrid class<br />
vehicle described in the following<br />
section.<br />
Hybrid Vehicle Description<br />
The newest generation of vehicle<br />
is intended to be used in autonomous<br />
sensor network applications. Therefore,<br />
the vehicle must be able to travel<br />
on long range missions collecting data<br />
158 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 8<br />
Successive generations of thrusters (a) utilize solenoid driver (7.5 cm/3-inch diameter), (b) utilize cavity <strong>and</strong> mold (10 cm/4-inch casing diameter),<br />
(c) have encompassing driving mechanism (12.5 cm/5-inch plate diameter) <strong>and</strong> (d) utilize simple crank shaft <strong>and</strong> semi-flexible ducting (10 cm/4-inch<br />
plate diameter).<br />
butmustalsobecapableofautonomously<br />
docking with permanent<br />
support structures. These structures<br />
will be responsible for downloading<br />
the vehicle’s mission data, changing<br />
mission objectives <strong>and</strong> recharging vehicle<br />
batteries.<br />
Autonomous docking with such a<br />
structure is a complicated problem,<br />
requiring not only high-accuracy maneuvering,<br />
but equally high-accuracy<br />
localization <strong>and</strong> attitude determination.<br />
In addition vehicles in this type of environment<br />
need to be capable of communicating<br />
with each other at short<br />
FIGURE 9<br />
The first four generations of our vehicle test<br />
beds. Oldest vehicle at bottom, <strong>and</strong> successive<br />
vehicles placed in ascending order.<br />
FIGURE 10<br />
Fifth generation hybrid vehicle CephaloBot.<br />
distances for coordinated missions. In<br />
general reduction in cost <strong>and</strong> internal<br />
structure are also desirable to allow<br />
for more vehicles in the network with<br />
a wider range of payload options.<br />
The base vehicle (Figure 10) is separated<br />
into three hull sections. The<br />
front <strong>and</strong> back sections house all of<br />
the actuators. Each section has two<br />
VRTs <strong>and</strong> an active buoyancy control<br />
device (BCD). Each thruster has an<br />
overall diameter of 7.6 cm (3 inches),<br />
a nozzle diameter of 1.8 cm (0.6 inch),<br />
has a stroke ratio of 4.5, <strong>and</strong> produces<br />
close to 2N of thrust at 30 Hz before<br />
cavitation starts to occur in the cavity.<br />
The back section also has the rear propeller<br />
motor. The primary batteries<br />
<strong>and</strong> all the electronics except motor<br />
control are housed in the center section.<br />
An additional payload section<br />
may be added between front <strong>and</strong> center<br />
to include additional sensors,<br />
devices, <strong>and</strong>/or batteries. Regulated<br />
power <strong>and</strong> digital communication<br />
lines interface the payload to the center<br />
section. When put together, the<br />
0.15 m (6 inches) diameter, 0.92 m<br />
(36 inches) long vehicle weights a neutrally<br />
buoyant 16 kg (36 lb). The<br />
BCDs can change the buoyancy by<br />
±1% to dive or surface, <strong>and</strong> 2 kg of<br />
internal ballast are adjustable to balance<br />
the pitch of the vehicle.<br />
Communication/<br />
Localization System<br />
Due to the physical properties of<br />
water, RF communication methods<br />
July/August 2011 Volume 45 Number 4 159
are not practical for small unmanned<br />
underwater vehicles. CephaloBot has<br />
a joint acoustic communication <strong>and</strong><br />
localization system which is ideally<br />
suited for underwater sensor network<br />
applications. The communication<br />
technique is based on binary frequency<br />
modulation whereas the localization<br />
methodology is based on a time delay<br />
of arrival technique. The system consists<br />
of a specialized hydrophone array<br />
(fabricated in house as described later<br />
in this section), which interpret acoustic<br />
signals for both information <strong>and</strong><br />
directional content. The receiving hydrophone<br />
array consists of three piezo<br />
electric ceramics spaced in a triangular<br />
arrangement parallel to the vehicle<br />
principle axis plane (Figure 11). Two<br />
piezo electric ceramics are placed on a<br />
line parallel to the pitching axis, <strong>and</strong><br />
the 3rd is extended along the roll axis<br />
from the midpoint of the other two.<br />
The entire array is encased in urethane<br />
rubber with an acoustic impedance<br />
similar to water. The transmitting<br />
node consists of a single high-power<br />
piezo electric ceramic encased in the<br />
same type of urethane rubber as the<br />
FIGURE 11<br />
Definition of principle submarine axes.<br />
hydrophone array. Each vehicle is<br />
equippedwithbothareceiving<strong>and</strong><br />
transmitting node on the underside<br />
of the vehicle (see Figure 12).<br />
FIGURE 12<br />
Transducer nodes on the vehicle. Receiving<br />
node on the right <strong>and</strong> transmitting node on<br />
the left.<br />
All three hydrophones in the receiving<br />
node receive a signal from a<br />
transmitting node (either on another<br />
vehicle or a docking station). The<br />
phase lag (ϕ) between the signal coming<br />
from the hydrophone on the roll<br />
axis <strong>and</strong> the signals coming from the<br />
two hydrophones on the pitching axis<br />
is measured <strong>and</strong> correlated with the<br />
signal frequency ( f ) to determine the<br />
time between when the hydrophones<br />
received the source signal Δt i = ϕ i /f.<br />
The time lag is then multiplied by<br />
the speed of sound in water to get<br />
relative source distances in the vehicle<br />
frame <strong>and</strong> transformed into the inertial<br />
frame to get the azimuth <strong>and</strong> elevation<br />
of the receiving node with respect to<br />
the source node.<br />
The hybrid localization/communication<br />
system is comprised of three<br />
main phases: data sending, data receiving<br />
<strong>and</strong> localization. Data receiving<br />
<strong>and</strong> localization both use the<br />
same incoming acoustic wave, the<br />
first 1000 cycles are dedicated to localization<br />
<strong>and</strong> the remaining cycles are<br />
frequency modulated. The acoustic<br />
wave propagates from a sending transducer<br />
located at some other node <strong>and</strong><br />
propagates towards the vehicle. Once<br />
this wave is received by the hydrophone<br />
array, it is conditioned by filtering<br />
<strong>and</strong> amplification circuitry.<br />
The conditioning circuitry contains a<br />
17-dB gain pre-amplifier followed by<br />
two stages of filtering <strong>and</strong> two stages<br />
of amplification to bring the signal<br />
voltage to 5 V. The electrical voltage<br />
is then passed through exclusive or<br />
gates with one of the other signals (in<br />
the case of hydrophone 0, the signal is<br />
also passed to demodulation circuitry).<br />
During the localization portion of receiving<br />
an onboard microcontroller<br />
reads <strong>and</strong> stores the output of each of<br />
the three exclusive-or gates <strong>and</strong> also<br />
determines which of the three signals<br />
arrives first. The duty cycle of the<br />
exclusive-or gates directly relates to<br />
the phase difference of the signals. The<br />
localization hardware calculates <strong>and</strong><br />
passes the azimuth <strong>and</strong> elevation angles<br />
(which defineaconeofpossiblevehicle<br />
locations with the source node at the<br />
vertex) to the main navigation processor<br />
which uses secondary positioning sensors<br />
(depth sensor <strong>and</strong> electronic compass)<br />
to determine a unique solution to<br />
the position of the receiving node with<br />
respect to the transmitting node.<br />
After the localization portion of<br />
receiving is complete the output of<br />
the exclusive-or gates are ignored <strong>and</strong><br />
the microcontroller reads <strong>and</strong> stores the<br />
DC voltage level from a frequency to<br />
voltage converter which directly represents<br />
the frequency of the incoming<br />
frequency modulated signal.<br />
In order for the localization methodology<br />
to function properly the hydrophones<br />
must be placed no more<br />
than λ/2 apart where the wavelength<br />
λ is equal to the speed of propagation<br />
divided by the frequency (λ =c/f ).<br />
This leads to a maximum spacing of<br />
3 cm for a frequency of 25 kHz <strong>and</strong><br />
1.875 cm for a frequency of 40 kHz.<br />
160 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
Miniature hydrophones which are approximately<br />
1 cm in diameter are commercially<br />
available from Reson but are<br />
on the order of $1000 per unit. Halfinch<br />
diameter cylindrical piezo electric<br />
ceramics (SMC14H12111) are available<br />
from Steminc for less than $10/<br />
each. Placement of the three halfinch<br />
diameter piezo electric ceramics<br />
in a triangle pattern yields a maximum<br />
navigational frequency of 25 kHz.<br />
The sending <strong>and</strong> receiving transducers<br />
were made in house using a method<br />
similar to that in Li et al. (2010).<br />
The sending piezoceramic was chosen<br />
primarily based on its resonant frequency<br />
of 22 kHz <strong>and</strong> cylindrical<br />
shape to provide an omni-directional<br />
signal. The receiving hydrophone<br />
array <strong>and</strong> transmitter are shown placed<br />
on the belly of the CephaloBot in<br />
Figure 12.<br />
Overall this customized communication<br />
localization system places a<br />
minimal load on the vehicle. The circuitry<br />
has a very small footprint (Figure<br />
13) compared to typical commercial<br />
acoustic modems. In a two-way communication<br />
mode, the power draw is<br />
less than 1 W, <strong>and</strong> in simple listening<br />
mode, the power draw is less than<br />
0.25 W. The transducers are fabricated<br />
in house so they can be customized to<br />
FIGURE 13<br />
Communication <strong>and</strong> localization hardware.<br />
any shape <strong>and</strong> placed at any location<br />
on the vehicle to improve vehicle drag<br />
characteristics. The entire system is<br />
fabricated for under $300 in materials,<br />
making it a suitable option for sensor<br />
networks requiring several low<br />
cost vehicles.<br />
FIGURE 14<br />
Embedded System<br />
The embedded system was custom<br />
designed for the CephaloBot. The vehicle<br />
will be used by researchers for<br />
various underwater sensor networking<br />
applications <strong>and</strong> multi-vehicle<br />
coordination. In order to enable this<br />
underwater network to interact with<br />
a potential aerial sensor network,<br />
CephaloBot is also equipped with RF<br />
communication capabilities, spare<br />
computation power, <strong>and</strong> the ability<br />
to quickly add new sensors. Each vehicle<br />
must be robust <strong>and</strong> easy to h<strong>and</strong>le<br />
<strong>and</strong> operate. The embedded system is<br />
separated into multiple printed circuit<br />
boards (PCB). A power distribution<br />
board h<strong>and</strong>les voltage regulation <strong>and</strong><br />
battery charging. An interface board<br />
connects the onboard devices <strong>and</strong> sensors<br />
with the power board <strong>and</strong> processing<br />
device. Smaller PCBs are located<br />
throughout the vehicle to provide<br />
specific functionality such as motor<br />
control, user interface, or simply wire<br />
routing. Figure 14 shows the electronics<br />
located in the center section, where<br />
everything except the motor controllers<br />
<strong>and</strong> user interface is located.<br />
Processing<br />
The primary processing device on<br />
the vehicle is a National Instruments<br />
Single-board RIO (sbRIO). It has an<br />
on-board 400 MHz processor running<br />
real-time LabVIEW software, 128 MB<br />
of RAM, 256 MB of flash, <strong>and</strong> a<br />
40-MHz 2 M gate FPGA (field programmable<br />
gate array), also programmed<br />
in LabVIEW, providing<br />
110 digital I/O pins. The combination<br />
of microprocessor <strong>and</strong> FPGA was<br />
proven effective in the 4th generation<br />
vehicle where a Compact RIO was<br />
used. All of the low-level communication,<br />
interface, <strong>and</strong> control tasks are<br />
h<strong>and</strong>led on the FPGA, leaving the<br />
microprocessor open to perform high<br />
level mission control. The core vehicle<br />
software is located on the FPGA. The<br />
primary benefit is that the mission critical<br />
software will always run at full<br />
speed <strong>and</strong> safety checks will be implemented<br />
that a vehicle user cannot easily<br />
override. In this way, even if the<br />
algorithms being tested fail, the vehicle<br />
will remain safe to itself <strong>and</strong> surroundings.<br />
The 400-MHz real-time processor<br />
is little used by the core system <strong>and</strong><br />
therefore provides significant computational<br />
power to the researcher. An<br />
Hybrid vehicle embedded system mounted on battery pack.<br />
July/August 2011 Volume 45 Number 4 161
additional 2 GB of flash storage is<br />
added with a serial data logger <strong>and</strong><br />
can be used for mission review.<br />
Batteries<br />
The power source is a four-cell lithium<br />
polymer pack. The nominal voltage<br />
of 14.8 V ranges from 10 to 16.8 V,<br />
depending on charge level. The lithium<br />
polymer chemistry was selected because<br />
of its high-power density <strong>and</strong><br />
excellent discharge characteristics. A<br />
single four-cell battery pack was chosen<br />
because, during most of its discharge<br />
cycle, it will have a voltage<br />
between 14 <strong>and</strong> 16 V, which minimizes<br />
the voltage change of the highcurrent<br />
devices. The availability of<br />
integrated circuits <strong>and</strong> components<br />
for four-cell batteries was found to be<br />
better than an eight-cell approach,<br />
which would have a maximum voltage<br />
of 33.6 V <strong>and</strong> require larger 35 V tolerant<br />
components. Overall current required<br />
dictates some trace thicknesses<br />
of approximately 8 mm. The capacity<br />
of each cell is 21 Ah, which provides a<br />
total of 310 Wh for the pack. A commercial<br />
frontend battery circuit board<br />
protects the batteries from overcharging,<br />
over discharge, <strong>and</strong> short<br />
circuit. It also balances the cells to increase<br />
overall service life. The circuit<br />
has a very low resistance to have minimal<br />
impact on the efficiency. The<br />
charger built into the power distribution<br />
can charge the batteries in approximately<br />
12 h. The high capacity of the<br />
batteries allows them to be discharged<br />
at a rate lower than 0.5 C, which increases<br />
battery runtime <strong>and</strong> overall<br />
lifetime (Murphy et al., 1990).<br />
Power Distribution<br />
The power distribution is the most<br />
complex custom-designed circuit board<br />
in the vehicle made into a four-layer<br />
10 × 17 cm PCB. It regulates <strong>and</strong><br />
monitors the voltages to power<br />
the rest of the electronics on the<br />
submarine.<br />
The power distribution module<br />
uses a Microchip PIC18F45K22 as<br />
a microcontroller supervisor. This<br />
microcontroller monitors <strong>and</strong> controls<br />
voltages <strong>and</strong> current draws <strong>and</strong> parts or<br />
the whole system can be shut off if<br />
excessive current is drawn or voltages<br />
are not in an acceptable range. The<br />
presence of power sources is also monitored,<br />
<strong>and</strong> the microcontroller correctly<br />
decides which one to use <strong>and</strong><br />
whether a battery requires charging.<br />
The sbRIO can signal the supervisor<br />
to enable or disable certain regulators<br />
on the vehicle to save power (i.e., wireless<br />
bridge is off when submerged). A<br />
function of the microcontroller also<br />
allows the vehicle to enter a “sleep”<br />
mode where everything except the supervisor<br />
microcontroller is turned off<br />
for a pre-determined period of time<br />
which reduces the total power usage<br />
to about 1 W. The microcontroller<br />
has an onboard analog-to-digital converter<br />
<strong>and</strong> a 16-channel multiplexer<br />
is used to read all of the required<br />
voltages <strong>and</strong> currents. Voltages are<br />
measured using a resistor-divider network<br />
to reduce the voltage into the<br />
multiplexer, <strong>and</strong> therefore the microcontroller<br />
to between 0 <strong>and</strong> 5 V with<br />
some error margin in case the voltage<br />
rises to more than intended. Allegro<br />
ACS714 Hall effect current sensors<br />
are used to provide very low loss method<br />
current measurements (0.06 W at 5 A).<br />
Sensors<br />
CephaloBot incorporates a minimum<br />
set of sensors needed to maintain<br />
a heading <strong>and</strong> depth underwater.<br />
Acoustics provide the vehicle with<br />
a relative position to a static pinger.<br />
Intervehicle communication prevents<br />
collisions between vehicles <strong>and</strong> the<br />
walls of the pool. The vehicle is robust<br />
enough to withst<strong>and</strong> a collision if it<br />
does occur. When the vehicles are<br />
deployed to an ocean environment,<br />
payload sensors may be added heterogeneously<br />
to the vehicles <strong>and</strong> shared so<br />
that each submarine has all required<br />
data to successfully navigate its environment.<br />
To simplify the design,<br />
CephaloBot uses an all-in-one IMU<br />
solution from VectorNav. The device<br />
has an onboard three-axis accelerometer,<br />
gyroscope, <strong>and</strong> magnetometer. It<br />
performs Kalman filtering <strong>and</strong> outputs<br />
quaternions, Euler angles, <strong>and</strong> the raw<br />
sensor data to the sbRIO FPGA. The<br />
onboard filtering eliminates the need<br />
to develop or perform the computations<br />
on the sbRIO. Stated accuracies<br />
are less than 2°, <strong>and</strong> because a magnetometer<br />
provides an absolute reference,<br />
this accuracy will not degrade<br />
with time. A Honeywell pressure sensor<br />
provides fine resolution (0.01 m)<br />
measurements to 10-m depths. The<br />
device outputs an analog voltage. The<br />
test pool for the vehicle is 5-m deep,<br />
<strong>and</strong> so a higher resolution device as<br />
chosen over one that may be used to<br />
a deeper depth.<br />
Conclusion<br />
The CephaloBot provides an ideal<br />
low cost option for underwater sensor<br />
networking <strong>and</strong> hybrid vehicle applications.<br />
The vehicle has maneuvering<br />
capabilities at zero forward velocity necessary<br />
for docking <strong>and</strong> high-resolution<br />
sensing. This capability is provided by<br />
an array of novel squid <strong>and</strong> jellyfish<br />
inspired thrusters. The thrusters are<br />
located internal to the hull with only<br />
asmallorifice exposed to the outer<br />
flow minimizing the effect on forward<br />
drag <strong>and</strong> allowing for efficient<br />
162 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
high-speed transit. Additionally these<br />
thrusters only require the single<br />
opening, allowing for greater system<br />
freedom internal to the vehicle inbetween<br />
thrusters.<br />
The embedded controller system<br />
designed for the vehicle has a compact<br />
modular design allowing for a wide<br />
variety of possible mission objectives.<br />
The microcontroller, which is operated<br />
on an easily adaptable LabVIEW<br />
platform, includes several open connections<br />
for future mission operations,<br />
on top of the base level vehicle operation<br />
input/outputs. The vehicle also has<br />
a wide variety of communication options<br />
including a low cost <strong>and</strong> in house<br />
developed acoustic communication/<br />
localization system for communication<br />
between underwater vehicles <strong>and</strong><br />
support structures. The vehicle has an<br />
RF system for communication with<br />
aerial vehicles while on the surface,<br />
<strong>and</strong> a WIFI bridge for communication<br />
with testers <strong>and</strong> data loggers while in<br />
controlled laboratory environments.<br />
Acknowledgments<br />
The authors would like to thank<br />
S. Lawrence-Simon, Tyler Thomas,<br />
Ryan Delgizzi, Dan Ambrosio, Colin<br />
Miller, Mikhail Kosna, <strong>and</strong> Matt<br />
Rhode for their hours of working on<br />
design <strong>and</strong> fabrication of the vehicle.<br />
We would also like to thank the Office<br />
of Naval Research (code 34) for funding<br />
this research project.<br />
Corresponding Author:<br />
Kamran Mohseni<br />
231 MAE-A, Department of<br />
Mechanical <strong>and</strong> Aerospace<br />
Engineering<br />
University of Florida, Gainesville, FL<br />
Email: mohseni@ufl.edu<br />
References<br />
Allen, D.G. 2010. James Rumsey: Steamboat<br />
inventor or pioneer in steam navigation.<br />
Magazine of the Jefferson County Historical<br />
<strong>Society</strong>. 96:48-54.<br />
Anderson, E., & Grosenbaugh, M. 2005.<br />
Jet flow in steadily swimming adult squid.<br />
J Exp Biol. 208:1125-46. doi: 10.1242/<br />
jeb.01507.<br />
Barrett, D., Triantafyllou, M.S., Yue, D.K.P.,<br />
Grosenbaugh, M.A., & Wolfgang, M.J. 1999.<br />
Drag reduction in fish-like locomotion.<br />
J Fluid Mech. 392:183-212. doi: 10.1017/<br />
S0022112099005455.<br />
Bartol, I.K., Krueger, P.S., Stewart, W.J., &<br />
Thompson, J.T. 2009. Hydrodynamics of<br />
pulsed jetting in juvenile <strong>and</strong> adult brief squid<br />
Lolliguncula brevis: Evidence of multiple jet<br />
‘modes’ <strong>and</strong> their implications for propulsive<br />
efficiency. J Exp Biol. 212:1889-903.<br />
doi: 10.1242/jeb.027771.<br />
Bartol, I.K., Krueger, P.S., Thompson, J.T.,<br />
& Stewart, W.J. 2008. Swimming dynamics<br />
<strong>and</strong> propulsive efficiency of squids throughout<br />
ontogeny. Integr Comp Biol. 48(6):720-33.<br />
Clark, T., Klein, P., Lake, G., Lawrence-<br />
Simon, S., Moore, J., Rhea-Carver, B., …<br />
Wu, A. 2009. Kraken: Kinematically roving<br />
autonomously kontrolled electro-nautic.<br />
In: Aerospace Sciences Meeting. Orl<strong>and</strong>o, FL:<br />
AIAA.<br />
Dabiri, J.O., Colin, S.P., Katija, K., &<br />
Costello, J.H. 2010. A wake based correlate of<br />
swimming performance <strong>and</strong> foraging behavior<br />
in seven co-occurring jellyfish species. J Exp<br />
Biol. 213:1217-75. doi: 10.1242/jeb.034660.<br />
Dunbabin, M., Roberts, J., Usher, K.,<br />
Winstanley, G., & Corke, P. 2005. A hybrid<br />
auv design for shallow water reef navigation.<br />
In: Proceedings of the International Conference<br />
on Robotics <strong>and</strong> Automation. pp. 2105-10.<br />
Barcelona, Spain: IEEE. doi: 10.1109/<br />
ROBOT.2005.1570424.<br />
Fossen, T.I. 1991. Nonlinear modeling <strong>and</strong><br />
control of underwater vehicles. Ph.D. thesis,<br />
Norwegian Inst. Technol., Trondheim,<br />
Norway.<br />
Gharib, M., Rambod, E., & Shariff, K. 1998.<br />
A universal time scale for vortex ring formation.<br />
J Fluid Mech. 360:121-40. doi: 10.1017/<br />
S0022112097008410.<br />
Glezer, A. 1988. The formation of vortex rings.<br />
Phys Fluids. 31(12):3532-42. doi: 10.1063/<br />
1.866920.<br />
Kato, N. 2011. Swimming <strong>and</strong> walking of<br />
an amphibious robot with fin actuators.<br />
Mar Technol Soc J. 45(4):181-97.<br />
Krieg, M., & Mohseni, K. 2008. Thrust<br />
characterization of pulsatile vortex ring<br />
generators for locomotion of underwater<br />
robots. IEEE J Oceanic Eng. 33(2):123-32.<br />
doi: 10.1109/JOE.2008.920171.<br />
Krieg, M., & Mohseni, K. 2010. Dynamic<br />
modeling <strong>and</strong> control of biologically inspired<br />
vortex ring thrusters for underwater robot<br />
locomotion. IEEE T Robot. 26(3):542-54.<br />
doi: 10.1109/TRO.2010.2046069.<br />
Krieg, M., & Mohseni, K. 2011. Thrust<br />
enhancement from radial velocity in squid<br />
inspired thrusters. OCEANS 2011, in press.<br />
Kona, HI: MTS/IEEE.<br />
Krueger, P.S., Dabiri, J., & Gharib, M. 2006.<br />
The formation number of vortex rings formed<br />
in a uniform background co-flow. J Fluid<br />
Mech. 556(1):147-66. doi: 10.1017/<br />
S0022112006009347.<br />
Krueger, P.S., & Gharib, M. 2003. The<br />
significance of vortex ring formation to the<br />
impulse <strong>and</strong> thrust of a starting jet. Phys<br />
Fluids. 15(5):1271-81. doi: 10.1063/<br />
1.1564600.<br />
Li, Y., Kastner, R., Faunce, B., Domond, K.,<br />
Kimball, D., Schurgers, C., & Benson, B.<br />
2010. Design of a low-cost underwater<br />
acoustic modem. IEEE ESL. 2:58-61.<br />
doi: 10.1109/LES.2010.2050191.<br />
Licht, S. 2008. Biomimetic oscillating foil<br />
propulsion to enhance underwater vehicle<br />
agility <strong>and</strong> maneuverability. Ph.D. thesis,<br />
Massachusetts Institute of <strong>Technology</strong>,<br />
Cambridge, Massachusetts.<br />
Licht, S., Polidoro, V., Flores, M., Hover, F.,<br />
& Triantafyllou, M. 2004. Design <strong>and</strong><br />
projected performance of a flapping foil<br />
July/August 2011 Volume 45 Number 4 163
auv. IEEE J Oceanic Eng. 29(3):786-68.<br />
doi: 10.1109/JOE.2004.833126.<br />
Lipinski, D., & Mohseni, K. 2009. A numerical<br />
investigation of flow structures <strong>and</strong><br />
fluid transport with applications to feeding<br />
for the hydromedusae Aequorea victoria <strong>and</strong><br />
Sarsiatubulosa. J Exp Biol. 212:2436-47.<br />
doi: 10.1242/jeb.026740.<br />
Mclean, M. 1991. Dynamic performance of<br />
small diameter tunnel thrusters. Ph.D. thesis,<br />
Naval Postgraduate School, Monterey,<br />
California.<br />
Murphy, T., Cason-Smith, D., James, S., &<br />
Smith, P. 1990. Characterization of aa size<br />
lithium rechargeable cells. In: Power Sources<br />
Symposium, Proceedings of the 34th International.<br />
pp. 176-80. Cherry Hill, NJ: IEEE.<br />
Mohseni, K. 2004. Zero-mass pulsatile jets<br />
for unmanned underwater vehicle maneuvering.<br />
In: 3rd “Unmanned Unlimited” Technical<br />
Conference, Workshop <strong>and</strong> Exhibit.<br />
pp. 2004-3686. Chicago, IL: AIAA.<br />
Mohseni, K. 2006. Pulsatile vortex generators<br />
for low-speed maneuvering of small underwater<br />
vehicles. Ocean Eng. 33(16):2209-23.<br />
Mohseni, K., & Gharib, M. 1998. A model<br />
for universal time scale of vortex ring formation.<br />
Phys Fluids. 10:2436-8. doi: 10.1063/1.869785.<br />
O’Dor, R.K., & Webber, D. 1991. Invertebrate<br />
athletes: Trade-offs between transport<br />
efficiency <strong>and</strong> power density in cephalopod<br />
evolution. J Exp Biol. 160:93-112.<br />
Sahin, M., & Mohseni, K. 2009. An arbitrary<br />
Lagrangian-Eulerian formulation for the numerical<br />
simulation of flow patterns generated<br />
by the hydromedusa Aequorea Victoria.<br />
J Comp Physiol. 228:4588-605. doi: 10.1016/<br />
j.jcp.2009.03.027.<br />
Sahin, M., Mohseni, K., & Colins, S. 2009.<br />
The numerical comparison of flow patterns<br />
<strong>and</strong> propulsive performances for the hydromedusae<br />
Sarsia tubulosa <strong>and</strong> Aequorea Victoria.<br />
J Exp Biol. 212:2656-67. doi: 10.1242/<br />
jeb.025536.<br />
Shariff, K., & Leonard, A. 1992. Vortex<br />
rings. Annu Rev Fluid Mech. 34:235-79.<br />
doi: 10.1146/annurev.fl.24.010192.001315.<br />
Torsiello, K. 1994. Acoustic positioning of<br />
the NPS autonomous underwater vehicle<br />
(AUV II) during hover conditions. Master’s<br />
thesis, Naval Postgraduate School, Monterey,<br />
California.<br />
Wilson, M.M., Peng, J., Dabiri, J.O., &<br />
Eldridge, J.D. 2009. Lagrangian coherent<br />
structures in low Reynolds number swimming.<br />
J Phys-Condens Mat. 21. Art. 204105.<br />
doi: 10.1088/0953-8984/21/20/204105.<br />
Yoerger, D., Cooke, J., & Slotine, J.J. 1990.<br />
The influence of thruster dynamics on underwater<br />
vehicle behavior <strong>and</strong> their incorporation<br />
into control system design. IEEE J Oceanic<br />
Eng. 15(3):167-78. doi: 10.1109/48.107145.<br />
Rosenfeld, M., Katija, K., & Dabiri, J.O.<br />
2009. Circulation generation <strong>and</strong> vortex ring<br />
formation by conical nozzles. J Fluids Eng.<br />
131, 091204:1-8.<br />
Rosenfeld, M., Rambod, E., & Gharib, M.<br />
1998. Circulation <strong>and</strong> formation number<br />
of laminar vortex rings. J Fluid Mech.<br />
376:297-318.<br />
Sahin, M., & Mohseni, K. 2008. The numerical<br />
simulation of flow patterns generated<br />
by the hydromedusa Aequorea victoria using<br />
an arbitrary Lagrangian- Eulerian formulation.<br />
In: 38th Fluid Dynamics Conference <strong>and</strong><br />
Exhibit. pp. 2008-3715. Seattle: AIAA.<br />
164 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
PAPER<br />
Modeling of Artificial Aurelia aurita<br />
Bell Deformation<br />
AUTHORS<br />
Keyur B. Joshi<br />
Alex Villanueva<br />
Colin F. Smith<br />
Shashank Priya<br />
Center for Energy Harvesting<br />
Materials <strong>and</strong> Systems,<br />
Center for Intelligent Material<br />
Systems <strong>and</strong> Structure,<br />
Virginia Polytechnic Institute<br />
<strong>and</strong> State University<br />
ABSTRACT<br />
Recently, there has been significant interest in developing underwater vehicles<br />
inspired by jellyfish. One of these notable efforts includes the artificial Aurelia aurita<br />
(Robojelly). The artificial A. aurita is able to swim with similar proficiency to the<br />
A. aurita species of jellyfish even though its deformation profile does not completely<br />
match the natural animal. In order to overcome this problem, we provide a systematic<br />
finite element model (FEM) to simulate the transient behavior of the artificial<br />
A. aurita vehicle utilizing bio-inspired shape memory alloy composite (BISMAC) actuators.<br />
The finite element simulation model accurately captures the hyperelastic<br />
behavior of EcoFlex (Shore hardness-0010) room temperature vulcanizing silicone<br />
by invoking a three-parameter Mooney-Rivlin model. Furthermore, the FEM incorporates<br />
experimental temperature transformation curves of shape memory alloy<br />
wires by introducing negative thermal coefficient of expansion <strong>and</strong> considers the<br />
effect of gravity <strong>and</strong> fluid buoyancy forces to accurately predict the transient deformation<br />
of the vehicle. The actual power cycle used to drive artificial A. aurita vehicle<br />
was used in the model. The overall profile error between FEM <strong>and</strong> the vehicle profile<br />
is mainly due to the difference in initial relaxed profiles.<br />
Keywords: autonomous undersea vehicle, Aurelia aurita, BISMAC, finite element<br />
analysis, transient dynamics<br />
Introduction<br />
J<br />
ellyfish have been in existence<br />
for millions of years <strong>and</strong> are the earliestknownmetazoansthatusemuscles<br />
for swimming (Valentine, 2004).<br />
They are found at various ocean<br />
depths <strong>and</strong> possess the ability to survive<br />
under hostile ocean environment.<br />
They exhibit colonial behavior <strong>and</strong><br />
have the ability to maintain certain<br />
depth <strong>and</strong> certain distance from the<br />
ocean shore (Albert, 2009). Jellyfish<br />
have relatively simple biological form<br />
<strong>and</strong> muscle architecture, lacking advanced<br />
sensors <strong>and</strong> a complex neural<br />
network possessed by many oceanic<br />
creatures (Chapman, 1974; Gladfelter,<br />
1972, 1973). but they are still able<br />
to survive <strong>and</strong> adapt in hostile environments.<br />
It maintains territorial existence<br />
by swimming on minimal energy<br />
intake. These abilities have created<br />
tremendous interest in the scientific<br />
community to discover their structureproperty-performance<br />
relationships<br />
<strong>and</strong> apply the learning towards creating<br />
a jellyfish-inspired swimming vehicle<br />
to perform various surveillance<br />
<strong>and</strong> monitoring tasks.<br />
There have been various efforts<br />
in literature on developing jellyfishinspired<br />
robots by using smart materialbased<br />
actuators. Inspired by the jetter<br />
class of jellyfish that swim by creating<br />
a jet of water by forcing it out of the<br />
bell, Villanueva et al. (2009) developed<br />
the JETSUM. This prototype used<br />
shape memory alloy (SMA) wires <strong>and</strong><br />
created an actuating stroke of bell<br />
segments attached to a passive neutrally<br />
buoyant bell structure. Yang<br />
et al. (2007) used flappers with control<br />
surfaces made of ionomeric polymer<br />
metal composites (IPMC) to<br />
control directionality of the vehicle.<br />
Tadesse et al. (2010a) used polypyrrole–<br />
polyvinylidene difluoride composites<br />
to achieve bending actuation to create<br />
a jellyfish robot. Larger jellyfish typically<br />
use “rowing” locomotion (Colin<br />
& Costello, 2002) <strong>and</strong> are characterized<br />
by formation of counter rotating<br />
starting <strong>and</strong> stopping vortex rings. Interaction<br />
of these vortex rings reduces<br />
energy lost in the wake <strong>and</strong> lends<br />
rowerstheirsuperiorswimmingefficiencies<br />
relative to jetters (Colin &<br />
Costello, 2002). Thus, with regard to<br />
energy efficiency, rowers provide a better<br />
platform for larger vehicles. Yeom<br />
<strong>and</strong> Oh (2009) proposed an entire<br />
jellyfish made from IPMC actuators<br />
with segments cut such that, on<br />
contraction, all the segments close to<br />
form a contracted bell shape. Recently,<br />
asignificant breakthrough was made<br />
by Villanueva et al. (2010b), who proposed<br />
the high-energy density bioinspired<br />
shape memory alloy composite<br />
July/August 2011 Volume 45 Number 4 165
(BISMAC) actuator that opened the<br />
possibility of converting high-force generation<br />
capability of SMA wires into<br />
high displacements. Using BISMAC,<br />
the design <strong>and</strong> implementation of<br />
biomimetic rowers became feasible.<br />
In this study, we investigate the<br />
bell deformation of BISMAC-based<br />
jellyfish robots using a finite element<br />
model (FEM, conducted using<br />
ANSYS) <strong>and</strong> identify the correlation<br />
with the natural species. In order to<br />
do so, the first major challenge was a<br />
precise implementation of SMA in<br />
FEM due to their giant aspect ratio<br />
<strong>and</strong> hysteretic temperature transformation.Wewereabletosuccessfully<br />
demonstrate the deformation of SMA<br />
using ANSYS by optimizing the meshing<br />
technique, identifying the variability<br />
in thermal coefficient of expansion,<br />
<strong>and</strong> separating the total deformation<br />
cycle into individual heating <strong>and</strong> cooling<br />
curves. Building upon this success,<br />
we implemented the SMA in BISMAC<br />
structure <strong>and</strong> investigated its mechanics<br />
to optimize the BISMAC configuration<br />
for mimicking the jellyfish<br />
profile. The FEM model provides the<br />
underst<strong>and</strong>ing of mechanism for bending<br />
strain amplification in BISMAC<br />
actuators <strong>and</strong> clearly delineates the<br />
effect of structural <strong>and</strong> thermal variables.UsingtheFEMresults,we<br />
were able to identify the inaccuracies<br />
that can occur due to variability in<br />
prototyping of BISMAC. Furthermore,<br />
our results provide important<br />
insight towards the development of<br />
feedback controller based on resistance<br />
changes. Next, we introduce the design<br />
of bell geometry in FEM for artificial<br />
Aurelia aurita (later referred to as,<br />
A. aurita for convenience in this work)<br />
using radial arrangement of BISMAC<br />
actuators. The objective was to develop<br />
the proper joint geometry that allows<br />
BISMACs to provide maximum deformation.<br />
Next, we describe the method for fabricating artificial A. aurita <strong>and</strong><br />
experimental characterization. Lastly, using the FEM simulations, we present<br />
a comparative analysis between the biological A. aurita (swimming profile)<br />
<strong>and</strong> artificial A. aurita <strong>and</strong> FE simulation.<br />
BISMAC Actuator<br />
The BISMAC actuator is a composite of an incompressible flexible metal strip<br />
<strong>and</strong> SMA wires separated by distance d <strong>and</strong> embedded in silicone rubber. Thermal<br />
transformation from martensite phase into austenite upon Joule heating induces<br />
the contraction of SMA wires, which is resisted by the incompressible metal<br />
strip. This introduces a tensile force f in the SMA that opposes contraction <strong>and</strong><br />
reduces the strain in SMA wire by a small amount.<br />
Figure 1 helps in explaining the mechanics of BISMAC bending deformation<br />
(Figure 1(a)). Under Joule heating, as SMA transforms from martensite<br />
to austenite phase <strong>and</strong> tends to contract, but due to the structure of the BISMAC,<br />
the metal strip resists this contraction <strong>and</strong> induces tensile force f in the SMA wire<br />
<strong>and</strong> compressive force f of the same magnitude in the metal strip. This force<br />
couple being distance d apart generates effective moment in the BISMAC actuator<br />
causing it to bend. Figure 1(b) shows the deformed BISMAC geometry. The<br />
mechanics of BISMAC has been discussed in detail by Smith et al. (2011). By<br />
using constitutive relation for SMA, we can express the generated stress as<br />
σ σ 0 ¼ E SMA ðɛ ɛ 0 ÞþΩðζ ζ 0 ÞþΘðT T 0 Þ ð1Þ<br />
where σ is the stress, ɛ is the strain, E SMA is the effective Young’s modulus of<br />
SMA, Ω is the transformation coefficient, ζ is the martensite fraction, Θ is the<br />
thermoelastic coefficient, <strong>and</strong> T is the temperature. Subscript zero denotes the<br />
initial state. Since transformation here is from fully martensite to fully austenite<br />
phase, ζ = 0, <strong>and</strong> ζ 0 = 1. Pure thermoelastic expansion is negligible, <strong>and</strong> initial<br />
stress <strong>and</strong> strain states are zero. Using E SMA = E austenite , since SMA is completely<br />
in austenite phase, Eq. (1) transforms into<br />
σ ¼ E austenite ɛ Ω ð2Þ<br />
FIGURE 1<br />
(a) Schematic of the force <strong>and</strong> moment in BISMAC. (b) Geometry of beam curvature.<br />
166 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
If there was no BISMAC structure to resist the SMA contraction, σ = 0 <strong>and</strong> ɛ = ɛ l<br />
transformation strain, providing<br />
Ω ¼ E austenite ɛ l<br />
If there is a uniform tensile force applied onto the SMA wire while transformation<br />
takes place,<br />
σ ¼ E austenite ɛ þ E austenite ɛ l<br />
Since SMA wires are thin with negligible bending stiffness, it is safe to assume<br />
uniform distribution of axial stress. Thus, axial stress can be written as<br />
σ ¼<br />
f<br />
A SMA<br />
¼ ɛ f E austenite ⇒ ɛ f ¼ ɛ þ ɛ l<br />
where f is the force generated by the SMA wire, A SMA is the total area of SMA<br />
cross section, <strong>and</strong> ɛ f is the force induced tensile strain. If SMA length after contraction<br />
is reduced from L to L′, from kinematics consideration, we can write<br />
L 0<br />
L ¼ 1 þ ɛ ¼ R<br />
d<br />
R<br />
where R is the radius of curvature of the metal strip <strong>and</strong> d is the distance between<br />
SMA wire <strong>and</strong> the metal strip. Solving Eq.(5) <strong>and</strong> (6),<br />
1 þ ɛ f ɛ l ¼ 1<br />
R ¼<br />
d<br />
R<br />
d<br />
ɛ l ɛ f<br />
ð8Þ<br />
Using Euler beam theory,<br />
M<br />
I ¼ E R ⇒ fd I ¼ E ɛ l<br />
d<br />
ɛ f E austenite A SMA d<br />
I<br />
ɛ f ¼<br />
¼ E ɛ l<br />
d<br />
EI ɛ l<br />
E austenite A SMA d 2 þ EI<br />
ɛ f<br />
<br />
ɛ f<br />
<br />
⇒ ɛ f ¼ EI ɛ l ɛ f<br />
<br />
E austenite A SMA d 2<br />
ð3Þ<br />
ð4Þ<br />
ð5Þ<br />
ð6Þ<br />
ð7Þ<br />
ð9Þ<br />
ð10Þ<br />
ð11Þ<br />
where EI is the total bending stiffness of the composite beam. For the particular<br />
configuration of the BISMAC used by Smith et al. (2011), Figure 2(a) shows the<br />
relationship between the radius of<br />
curvature for the BISMAC <strong>and</strong> the<br />
distance between SMA wires <strong>and</strong> flexible<br />
metal strip. Figure 2(b) reveals<br />
that as the distance d decreases, the<br />
tensile force induced in SMA wires<br />
increases dramatically <strong>and</strong> attains a limiting<br />
value of 80 g force for 100-μmthick<br />
SMA wire (BioMetal Fiber,<br />
TOKI Corporation) beyond which<br />
high stresses cause austenite phase to<br />
transform into stress-induced martensite<br />
resulting in loss of transformation<br />
strain <strong>and</strong> thus loss of performance,<br />
which is not accounted by Eqs. (1)-(11).<br />
BISMAC Customization<br />
<strong>and</strong> Bell Geometry<br />
A. aurita curvature profiles in its<br />
relaxed <strong>and</strong> contracted states are<br />
showninFigure3(a)(Dabirietal.,<br />
2005). Before any actuation, the bell<br />
is fully exp<strong>and</strong>ed <strong>and</strong> said to be in<br />
the relaxed position. Fully contracted<br />
state refers to the state corresponding<br />
to complete contraction of subumbrellar<br />
muscles <strong>and</strong> minimum bell volume.<br />
After actuation, the bell passively<br />
regains its original (relaxed) position.<br />
In earlier work (Villanueva et al.,<br />
2010b), we have presented the methodology<br />
used to customize BISMAC<br />
configurations to mimic the natural<br />
A. aurita curvature profile. For a<br />
curved beam, we can write<br />
M ¼ EI ðÞ s<br />
dðΔθÞ<br />
ds<br />
ð12Þ<br />
where M is the moment, EI(s) isthe<br />
local bending stiffness at location s,<br />
Δθ(s) =θ(s) − θ 0 (s) is the change in<br />
slope at location s, <strong>and</strong>s is location<br />
on curved profile length. Since the<br />
force generated by SMA wires is uniform,weensureaconstantmoment<br />
M by maintaining the fixed distance<br />
July/August 2011 Volume 45 Number 4 167
FIGURE 2<br />
(a) Radius of curvature vs. distance d. (b) Tensile force in the SMA wires vs. distance d.<br />
FIGURE 3<br />
(a) A. aurita profile in relaxed <strong>and</strong> contracted conditions. (b) Curvature comparison of BISMAC<br />
muscle with A. aurita after customization. (c) Schematic of the BISMAC placement in the artificial<br />
A. aurita.<br />
between SMA wires <strong>and</strong> the metal<br />
strip. The bending stiffness was varied<br />
(i) to match jellyfish exumbrella <strong>and</strong><br />
subumbrella profiles in the relaxed<br />
state <strong>and</strong> (ii) to manipulate the bending<br />
stiffness EI(s) such that upon SMA<br />
actuation, dðΔθÞ<br />
ds<br />
matches that of real<br />
A. aurita in order to ensure that we<br />
have a good match between natural<br />
<strong>and</strong> artificial vehicles in the contracted<br />
state. The resultant BISMAC achieved<br />
close similarity with A. aurita profile as<br />
illustrated in Figure 3(b). The inflexion<br />
point in contracted A. aurita profile<br />
at ∼2.5 cm represents the fact that<br />
at the center bell thickens <strong>and</strong> the profile<br />
becomes a little convex near the<br />
center <strong>and</strong> changes to concave at inflexion<br />
point. Towards the end of the<br />
profile, the BISMAC shows reduction<br />
in curvature due to passive material at<br />
the tip end for protection from water.<br />
Similar reductions in contracted<br />
A. aurita curvature profile towards<br />
the bell margin lead to discovery of<br />
the passive flap in A. aurita near the<br />
bell margin. The passive flap was<br />
shown to improve the swimming performance<br />
of the artificial A. aurita significantly<br />
(Villanueva et al., 2010a).<br />
Figure 3(c) shows a schematic of artificial<br />
A. aurita consisting of eight<br />
BISMACs radially distributed around<br />
the bell, which is made of soft silicone.<br />
Artificial A. aurita has shown bell deformation<br />
<strong>and</strong> kinematics as well as a<br />
swimming performance comparable<br />
to that of natural animal (Villanueva<br />
et al., 2010a). Since most of the available<br />
engineering materials have lower<br />
compliance compared to that of natural<br />
animal, the artificial A. aurita design<br />
includes joint structures between<br />
BISMAC actuators to localize the material<br />
folding upon contraction. These<br />
joint structures modify the original<br />
axisymmetric A. aurita bell shape <strong>and</strong><br />
increases similarity to the Cyanea capillata<br />
bell shape. The joints are wedgelike<br />
cavities <strong>and</strong> are found on natural<br />
jellyfish. The joint structure is described<br />
by Smith <strong>and</strong> Priya (2010)<br />
<strong>and</strong> is copied from the Polyorchis montereyensis.<br />
Since the cross-sectional<br />
shape of the joints varies with bell<br />
height, we take the equation below to<br />
represent the profile at midbell. The<br />
joint structure can be described as a<br />
piecewise function representing two<br />
symmetric sides of a single function,<br />
mirrored about the y-axis,<br />
y ¼ δjk ð<br />
δk<br />
xÞ<br />
jx<br />
ð13Þ<br />
This function has been formed to describe<br />
the joint shape across a 2-D<br />
plane with height j, half-widthk,<br />
<strong>and</strong> curvature δ as parameters. In this<br />
way, the joint structure from a wide<br />
variety of species can be described<br />
with one basic equation by changing<br />
168 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 4<br />
(a) Original axisymmetric A. aurita bell (b). Schematic of joint geometry (Smith & Priya, 2010).<br />
(c) Top <strong>and</strong> side view showing location in bell for which joint calculations were made. (d) Final<br />
artificial A. aurita bell shape.<br />
−2.75. The final shape was evolved<br />
by sweeping circular section at bell<br />
tip into the curve at mid-bell height<br />
joint section for a smooth blending.<br />
Figure 4 shows the evolution of<br />
original axisymmetric A. aurita bell<br />
shape into the final bell design of artificial<br />
A. aurita. Figure 5 shows the details<br />
of the unigraphics model that was<br />
used in finite element (FE) simulation.<br />
the constants. Artificial A. aurita’s<br />
joint structure was produced by using<br />
a program that could manipulate the<br />
constants in Mathematica. The values<br />
to match the shape of joints in P. montereyensis<br />
were found to be height =<br />
2.15, half-width = 3, <strong>and</strong> curvature =<br />
FIGURE 5<br />
Unigraphics model of the artificial A. aurita.<br />
Experimental Setup<br />
Artificial A. aurita consists of a<br />
central mount that houses the electrical<br />
circuitry <strong>and</strong> clamps the BISMAC<br />
actuators together (see Figure 6). The<br />
radius of this hub was 25% of the bell<br />
diameter <strong>and</strong> covers a region where<br />
minimal deformation is expected to<br />
occur in both the natural <strong>and</strong> artificial<br />
A. aurita. Since the distance between<br />
the SMA wires <strong>and</strong> the metal strip is<br />
very crucial parameter in BISMAC design,<br />
the SMA wires <strong>and</strong> the metal<br />
strip are slide in position in small<br />
acrylic supports, designed to maintain<br />
this distance. The supports are then<br />
placed in the mold <strong>and</strong> silicone is<br />
poured in to settle (Villanueva et al.,<br />
2010a). The actual vehicle has a<br />
small portion (3-4 mm) of metal strip<br />
<strong>and</strong> SMA wires protruding out of the<br />
bell geometry. This is neglected in this<br />
work for model simplification. The<br />
simplification is justified by negligible<br />
bending moment contribution from<br />
the protruded part towards the bell<br />
deformation. The experimental data<br />
used for the deformation comparison<br />
were acquired by using the same artificial<br />
A. aurita <strong>and</strong> experimental setup as<br />
FIGURE 6<br />
Artificial A. aurita with uniform bell <strong>and</strong> flap,<br />
in the relaxed configuration.<br />
July/August 2011 Volume 45 Number 4 169
described in previous work (Villanueva<br />
et al., 2010a). Artificial A. aurita was<br />
submerged underwater <strong>and</strong> clamped<br />
down by supports pressing on the top<br />
<strong>and</strong> bottom of the bell. The contacting<br />
area between supports <strong>and</strong> bell covered<br />
the region of the internal central<br />
mount. This region is meant to undergo<br />
negligible deformation since<br />
BISMAC actuators do not directly deform<br />
the bell at that location. The bell<br />
was contracted by heating the SMA<br />
wires using a rapid heating control algorithm<br />
(Villanueva & Priya, 2010a).<br />
This controller uses SMA resistance<br />
feedback to monitor the bell state of<br />
deformation. It sends high-current<br />
pulses for rapid contraction <strong>and</strong> low<br />
current to maintain deformation<br />
allowing fast contraction while minimizing<br />
power consumption. The controller<br />
was developed in LabView<br />
(National Instruments), <strong>and</strong> the measurements<br />
were made using a NI<br />
cDAQ 9172 with NI-9215 <strong>and</strong><br />
NI-9263 analog input <strong>and</strong> output<br />
cards, respectively. A NF HAS 4052<br />
power amplifier was used to amplify<br />
the DAQ output <strong>and</strong> actuate the<br />
robot. The bell profile was recorded<br />
during the first actuation cycle. The<br />
deformation was captured using an<br />
IN250 high-speed camera from Fastec<br />
Imaging. The bell deformation was<br />
tracked by placing reflective beads<br />
along the profile <strong>and</strong> by processing<br />
the images manually using ImageJ.<br />
This process included an error on the<br />
order of ±1 cm.<br />
FEM Setup <strong>and</strong> Solution<br />
Material Properties<br />
To ensure accuracy of the model in<br />
adequately representing the behavior<br />
of various materials, we determined<br />
the properties experimentally. There<br />
are three materials critical to the simulation:<br />
silicone matrix, flexible but incompressible metal strip <strong>and</strong> SMA wires<br />
(BioMetal Fiber).<br />
Silicone Rubber<br />
Room temperature vulcanizing (RTV) silicone is a c<strong>and</strong>idate material for bell<br />
mesoglea. In addition to forming the main jellyfish body, it is also responsible for<br />
maintaining the required distance between spring steel <strong>and</strong> the SMA wires. We<br />
have tested several silicone rubbers with different shore hardness to evaluate their<br />
mechanical properties. Ecoflex TM (Smooth-On) with initial tensile Young’s modulus<br />
of the order of 10,580 Pa was selected to construct the Artificial A. aurita<br />
because it offered much less resistance to the BISMAC deflection. Figure 7(a)<br />
shows the tensile test, on a dog bone–shaped sample with gauge length of 7 mm<br />
<strong>and</strong> cross-sectional area of 2.45 × 2.8 mm 2 for several cycles at room temperature<br />
at 1 mm/s displacement rate up to 20 mm extension to ensure silicone properties<br />
do not change with multiple loading cycles. Hysteresis is evident in the figure, <strong>and</strong><br />
we used the average of the two curves to generate our model. Figure 7(b) represents<br />
completely defined stress-strain behavior of silicone including compression<br />
test data carried out on 16- mm-thick 25.4-mm diameter cylindrical specimen.<br />
We generated a three-parameter Mooney-Rivlin model for silicone from test<br />
data using ANSYS’s curve fitting tool, as a more conventional two-parameter<br />
Mooney-Rivlin model failed to provide a good fit to the experimental data.<br />
The Mooney-Rivlin model for Ecoflex TM is given by Eq. (14),<br />
W ¼ c 10 ðI 1 3Þþc 01 ðI 2 3Þþc 11 ðI 1 3ÞðI 2 3Þþ 1 d ðJ 1Þ2<br />
ð14Þ<br />
where W is the strain energy function, I 1 , I 2 , I 3 are the stretch invariants, J is the<br />
determinant of deformation gradient tensor, <strong>and</strong> c 10 = 2307.1 Pa, c 01 = −223.76 Pa,<br />
c 11 =142.83Pa,d =0Pa −1 (compressibility parameter). Silicone thermal conductivity<br />
was measured to be 0.22 W/m K. Silicone properties used in the<br />
FEM are tabulated in Table 1.<br />
SMA Wire<br />
We selected 100-μm diameter BioMetal fibers due to their superior performance<br />
(Tadesse et al., 2010b). The temperature transformation of the wires was<br />
measured experimentally as shown in Figure 8(a). Figure 8(b) represents the<br />
FIGURE 7<br />
(a) Tensile test for several cycles showing hysteresis. (b) Stress-strain curve used to model<br />
Ecoflex TM .<br />
170 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
TABLE 1<br />
Silicone properties used in the FEM.<br />
Property<br />
Value<br />
Thermal Conductivity<br />
0.22 W/m K<br />
Elastic modulus<br />
Mooney-Rivlin model<br />
Poisson’s ratio 0.49<br />
Density 982 kg/m 3<br />
Specific heat<br />
300 J/kg K<br />
TABLE 2<br />
SMA properties used in the FEM.<br />
Property Martensite Austenite<br />
Thermal Conductivity 8 W/m K 18 W/m K<br />
Elastic modulus 28 MPa 75 MPa<br />
Poisson’s ratio 0.33 0.33<br />
Density 6450 kg/m 3 6450 kg/m 3<br />
Specific heat 837.36 J/kg K 837.36 J/kg K<br />
stress-strain relationship as measured<br />
by tensile test to confirm the manufacturer’s<br />
claims (Toki Corporation) that<br />
the wires can easily take 400 MPa<br />
stress.<br />
Other properties that were used in<br />
the simulation are tabulated in Table 2.<br />
For temperatures other than transformation<br />
temperatures, the properties<br />
were interpolated according to martensite<br />
fraction in the SMA. Martensite<br />
fraction was calculated based on the<br />
temperature-strain curve.<br />
In order to model the temperature<br />
transformation hysteresis of SMA<br />
wires, we defined two separate material<br />
curves, one for the heating profile <strong>and</strong><br />
other for the cooling profile. Accordingly,<br />
the elastic modulus EX <strong>and</strong> thermal<br />
conductivity KXX also followed<br />
two separate profiles as depicted in Figures<br />
9(a) <strong>and</strong> 9(b). Figure 9(c) shows<br />
the artificially defined negative thermal<br />
coefficient of expansion to achieve the<br />
transformation strains with increase in<br />
temperature during heating <strong>and</strong> cooling,<br />
respectively.<br />
Metal Strip<br />
For choosing the metal strip, two<br />
criteria should be met: (i) incompressibility<br />
for BISMAC mechanics to work<br />
well <strong>and</strong> (ii) flexibility (low bending<br />
stiffness) to obtain maximum<br />
deformation with a small actuation<br />
moment. We selected st<strong>and</strong>ard spring<br />
steel (low carbon steel) as the suitable<br />
metal strip material with properties<br />
tabulated in Table 3.<br />
Transient Heat Transfer Model<br />
Meshing, Boundary Conditions,<br />
<strong>and</strong> Loads<br />
We chose ANSYS as our simulation<br />
package. To simulate transient<br />
heat transfer, we built the model<br />
by meshing the SMA, metal strip<br />
<strong>and</strong> one element thick silicone layer<br />
around them with SOLID70 (8-node<br />
brick element). The rest of the silicone<br />
matrix was meshed with SOLID87<br />
(10-node tetrahedrons) <strong>and</strong> SOLID90<br />
(20-node brick elements) was used as<br />
transition element between SOLID70<br />
FIGURE 8<br />
(a) BioMetal Fiber temperature transformation curve (A s = Austenite start temperature, A f = Austenite finish temperature, M s = Martensite start<br />
temperature, M f = Martensite finish temperature). (b) Stress-strain relationship of Martensite phase.<br />
July/August 2011 Volume 45 Number 4 171
FIGURE 9<br />
(a) Variation of SMA Young’s modulus with temperature. (b) Variation of SMA thermal conductivity<br />
with temperature. (c) Variation of thermal coefficient of expansion with temperature.<br />
<strong>and</strong> SOLID87 element meshes as shown<br />
in Figure 10(a). All these elements have<br />
TEMP degree of freedom.<br />
We took advantage of the circular<br />
symmetry <strong>and</strong> modeled only 1/8th of<br />
the bell segment with no loss of physics.<br />
To reduce the meshing efforts<br />
we chose to mesh only half of the<br />
1/8th segment of the bell, reflected<br />
the mesh about the plane of symmetry<br />
<strong>and</strong> finally merged the nodes to create<br />
the complete model. The boundary<br />
FIGURE 10<br />
(a) Mesh detail for transient thermal model. (b) Boundary conditions.<br />
conditions are depicted in Figure 10(b).<br />
Due to symmetry, the circular symmetric<br />
sides (Figure 10(b), violet colored<br />
faces) of the model do not have<br />
any temperature gradient; thus, they<br />
are modeled as insulated boundaries.<br />
Exumbrella <strong>and</strong> subumbrella surfaces<br />
(Figure 10(b), yellow colored faces)<br />
are in contact with surrounding water<br />
<strong>and</strong> conveys heat into the fluid, <strong>and</strong><br />
thus, were modeled as convective<br />
boundaries. The heat transfer coefficient<br />
of both these surfaces were<br />
taken to be 20 W/m 2 K (Baker, 1972),<br />
<strong>and</strong> the ambient temperature was<br />
fixed at 25°C. The thermal loading<br />
resulting from electrical heating was<br />
applied using Villanueva et al.’s resistance<br />
feedback control algorithm<br />
(Villanueva & Priya, 2010a) to reduce<br />
power requirement. The internal heating<br />
load was applied on the SMA wire<br />
as shown in Figure 11, which is consistently<br />
1/8th of total power consumed<br />
by the vehicle’s eight segments. It consists<br />
of rapid heating pulses of high<br />
TABLE 3<br />
Metal strip properties used in the FEM.<br />
Property<br />
Value<br />
Thermal Conductivity 47 W/m K<br />
Elastic modulus 210 GPa<br />
Poisson’s ratio 0.3<br />
Density 7860 kg/m 3<br />
Specific heat<br />
510 J/kg K<br />
172 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
current followed by a constant current<br />
regime <strong>and</strong> finally reducing the magnitude<br />
to idling minimum current governed<br />
by need for resistance feedback<br />
measurement. The three initial spikes<br />
seen in the curve correspond to the<br />
high-current impulse sent for rapid heating.<br />
Multiple pulses are usually needed<br />
during the first few actuation cycles<br />
since the low current is not enough to<br />
maintain the deformation. The material<br />
surrounding the SMA warms up <strong>and</strong><br />
eventually the low current input is<br />
enough to maintain deformation.<br />
Solution: Transient Thermal Analysis<br />
We first find the solution with<br />
SMA wires having material properties<br />
corresponding to heating curve. To<br />
capture initial sharp temperature rise<br />
accurately over the first 0.12 s, small<br />
time steps of 0.01 s were used to provide<br />
sufficient time resolution. The<br />
low current period from 0.12 to 0.70 s<br />
was solved in 100 uniform time steps.<br />
After solving two more transient current<br />
time steps at 0.71 <strong>and</strong> 0.72 s in<br />
one time step each, we switched the<br />
SMA wire material to the one with<br />
properties corresponding to cooling<br />
curve. Temperatures of SMA wires<br />
corresponding to t =0.72sbeingfar<br />
beyond A f ensures properties of both<br />
the curves are same at this point. Finally,<br />
relaxation phase of A. aurita<br />
jellyfish vehicle from 0.72 to 2.00 s<br />
was solved in 200 uniform time steps.<br />
FIGURE 11<br />
Variation of internal heating load on SMA wire with time.<br />
FIGURE 12<br />
Transient Structural<br />
Deformation Model<br />
Meshing, Boundary Conditions,<br />
<strong>and</strong> Loads<br />
To simulate transient structural<br />
deformation, we reused the mesh<br />
from transient heat transfer model<br />
for rapid model building. SOLID70<br />
(8-node brick element) of the SMA,<br />
metal strip <strong>and</strong> one element thick silicone<br />
layer around them were transformed<br />
into SOLID185 (8-node<br />
brick element) with structural degree<br />
of freedom UX, UY, UZ. Similarly,<br />
SOLID87 (10-node tetrahedrons)<br />
was transformed into SOLID187<br />
(10-node tetrahedrons) elements <strong>and</strong><br />
transition elements SOLID90 (20-<br />
node brick elements) were transformed<br />
into SOLID186 (20-node<br />
brick elements) as shown in Figure<br />
12(a). Figure 12(b) displays the<br />
displacement boundary condition for<br />
the model. The boundary condition<br />
was applied in cylindrical coordinate<br />
system defined using jellyfish vehicle<br />
centralaxisasthez-axis of the cylindrical<br />
coordinate system (CSYS = 5).<br />
(a) Transient structural deformation model mesh. (b) Displacement boundary conditions. (c) Temperature<br />
<strong>and</strong> acceleration boundary conditions.<br />
July/August 2011 Volume 45 Number 4 173
Both sides of 1/8th segment being circular symmetric were constrained from<br />
moving in hoop direction (UY = 0). The line on the axis of the vehicle was<br />
also constrained from moving in radial direction (UX = 0). Also, to constrain<br />
any rigid body translation, we chose to constrain one node on rigid central hub<br />
lying on the axis of the vehicle from moving in axial direction (UZ = 0).<br />
We accounted for gravitation <strong>and</strong> buoyancy force of the water by applying<br />
inertial acceleration corresponding to reduced gravitation under buoyancy from<br />
Eq. (15),<br />
<br />
ACEL ‐<br />
Y ¼ ΣN mat¼1 V matρ mat ρ water Σ N mat¼1 V <br />
mat<br />
Σ N mat¼1 V 9:81 m<br />
matρ mat<br />
s 2 ð15Þ<br />
where ACEL_Y is the inertial acceleration for the vehicle, mat is the material index,<br />
ρ is the density, <strong>and</strong> V is the volume. Earth’s gravitation acceleration was taken to<br />
be 9.81 m/s 2 . Temperatures were applied as body forces <strong>and</strong> were read from previously<br />
solved transient heat transfer model result. To account for the fluid drag<br />
we have introduced damping by defining BETAD = 0.03. This creates a damping<br />
matrix [C] =BETAD[K ], where [K ] = stiffness matrix of the FEM.<br />
The overall FE system equation can be written as<br />
½M<br />
ŠfÜ<br />
gþ ½CŠ<br />
:<br />
U þ ½K<br />
ŠfU<br />
g ¼ fFg<br />
ð16Þ<br />
Here, [M], [C]<strong>and</strong> [K ] are displacement dependent on the mass matrix, the<br />
damping matrix <strong>and</strong> the stiffness matrix, respectively, {U} is the displacement<br />
vector <strong>and</strong> {F } is effective force vector. The system defined by Eq. (16) is nonlinear<br />
due to variable [M ], [C], <strong>and</strong> [K ] matrices.<br />
Solution: Transient Structural Deformation Analysis<br />
As in transient heat transfer model, we start the modeling with SMA wires having<br />
properties corresponding to heating curve (martensite to austenite temperature<br />
transformation) for contraction phase of the cycle <strong>and</strong> during relaxation we<br />
use material corresponding to cooling curve (austenite to martensite temperature<br />
transformation). In simulation, this was achieved by defining two separate SMA<br />
material property curves corresponding to (martensite (M) to austenite (A) <strong>and</strong> austenite<br />
to martensite) <strong>and</strong> switching from material with M → A curve to material with<br />
A → M curve after completion of contraction phase. As SMA wire is well above austenite<br />
finish temperature A f , for which both the material curves (cooling <strong>and</strong> heating)<br />
have identical properties there’s no abrupt jump between these curves. We used the<br />
time intervals as used in transient heat transfer model.<br />
Results <strong>and</strong> Discussion<br />
Transient Heat Transfer Analysis<br />
Figure 13 summarizes all major results of transient heat transfer analysis. Figure<br />
13(a) shows overall temperature distribution at the end of heating phase. It<br />
suggests that practical region of interest is concentrated near SMA wire, <strong>and</strong> for<br />
most part, it does not change along SMA wire length. Near the edge of the bell,<br />
however, model predicts a little more temperature rise in the SMA wire compared<br />
to rest of the area due to very thin silicone<br />
layer as same amount of generated<br />
heat is absorbed by lesser silicone<br />
available. Since rate of heat conduction<br />
into silicone is higher compared to the<br />
rate of heat being convected away at<br />
the subumbrella surface, silicone temperature<br />
rises faster <strong>and</strong> temperature<br />
gradient between SMA wire <strong>and</strong> silicone<br />
reduces, inhibiting heat conduction<br />
from SMA wire to subumbrella<br />
surface. Figure 13(b) shows temperature<br />
history at a typical location at<br />
SMA center <strong>and</strong> in silicone 1 element<br />
away from SMA wire surface. It shows<br />
typical response of first order system to<br />
a given excitation. Initial high-current<br />
pulses till t = 0.12 s result in fast temperature<br />
rise at SMA center which<br />
keep increasing at logarithmic rate<br />
from t = 0.12 s to t = 0.70 s. Temperature<br />
at SMA center decreases exponentially<br />
from t = 0.70 s to t = 2.00 s<br />
but fails to return to starting temperature<br />
of 25°C at the end of the cycle<br />
(T =45°C at t = 2.00 s) suggesting<br />
net heat accumulation in the model.<br />
It also suggests that we may be adding<br />
heat unnecessarily at higher rate<br />
during low current constant heating<br />
cycle as the temperature at the end of<br />
high current pulsed heating is well<br />
beyond austenite finish temperature<br />
A f . This additional heating also aggravates<br />
heat accumulation problem that<br />
eventually results in partial loss of performance<br />
of the actuator as insufficient<br />
cooling results into incomplete transformation<br />
into martensite phase. Temperature<br />
in silicone 1 element away<br />
from SMA surface shows quite similar<br />
trend but has reduced temperature rise<br />
peaks during high-current pulse cycle<br />
as low conductivity dampens out the<br />
sharp peaks. Cooling cycle starts with<br />
significant temperature gradient between<br />
SMA center <strong>and</strong> the aforementioned<br />
silicone location but diminishes<br />
174 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 13<br />
(a) Overall temperature distribution at t = 0.70 s. (b) Temperature time history at SMA wire<br />
center <strong>and</strong> in silicone 1 element away from SMA surface. (c) Typical temperature distribution<br />
along SMA wire length.<br />
exponentially fast as is evident from Figure<br />
13(b). Figure 13(c) compares temperature<br />
distribution near SMA length<br />
at typical location for key time points<br />
t = 0.12 s (end of high-current pulse<br />
heating), t = 0.70 s (end of constant<br />
low current heating), <strong>and</strong> t = 2.00 s<br />
(end of cooling cycle).<br />
Transient Structural<br />
Deformation Analysis<br />
Overall Bell Deformation<br />
Since, gravity <strong>and</strong> buoyancy forces<br />
are accounted for, in the two equilibrium<br />
positions (relaxed <strong>and</strong> contracted),<br />
where fluid pressure differential does<br />
not exist across subumbrella <strong>and</strong><br />
exumbrella due to the FEM <strong>and</strong> the<br />
artificial A. aurita being held at the<br />
bell center; we can predict deformation<br />
at these positions accurately<br />
at this location without worrying<br />
about approximations involved in<br />
fluid drag. Figure 14(a) shows various<br />
views of FE simulation result<br />
for contracted state at t =0.12s.<br />
Original undeformed model is shown<br />
with only black edges superimposed<br />
on deformed model for comparison.<br />
Figure 14(a.1) represents bottom<br />
view of the contracted artificial<br />
A. aurita bell model, which qualitatively<br />
matches the contracted bell<br />
segments of C. capillata species (Figure<br />
14(d)) <strong>and</strong> experimental artificial<br />
FIGURE 14<br />
(a) FE result for artificial A. aurita bell contraction: (a.1) bottom view, (a.2) isometric view from bottom, (a.3) cross-sectional view at BISMAC<br />
location, <strong>and</strong> (a.4) side view. (b) Biological A. aurita contraction. (c) Artificial A. aurita experimental contraction. (d) C. capillata bell segment<br />
contraction bottom view. (e) Artificial A. aurita experimental contraction bottom view (Villanueva, submitted).<br />
July/August 2011 Volume 45 Number 4 175
A. aurita contraction (Figure 14(e))<br />
that we have intended to model in<br />
our FE simulation. Note that due to<br />
manufacturing imperfections not all<br />
the bell segments deform the same<br />
amount (Figure 14(e)). The comparison<br />
of contracted profile with that of<br />
C. capillata is justified as our artificial<br />
A. aurita possess joint structure <strong>and</strong><br />
radial muscles that were inspired by<br />
C. capillata possessing similar radial<br />
muscle arrangement (Gladfelter,<br />
1973). Figure 14(a.2) depicts isometric<br />
view of deformed model for<br />
better visualization. Figure 14(a.3)<br />
shows cross-sectional view taken along<br />
BISMAC center line. Figure 14(a.4)<br />
exhibits side view of the contracted<br />
FEM also matching well with natural<br />
A. aurita shown in Figure 14(b) <strong>and</strong><br />
experimental artificial A. aurita (Figure<br />
14(c)). It is evident that the bell<br />
curves inwards at the BISMAC locations<br />
radially <strong>and</strong> axially well <strong>and</strong><br />
confirms that the joint design provided<br />
by Smith <strong>and</strong> Priya (2010) is<br />
effective in assisting the bell deformation<br />
at the BISMAC locations.<br />
This behavior was also confirmed in<br />
experimental A. aurita deformation<br />
(Figure 14(c)).<br />
Comparison of Deformation<br />
at the BISMAC Location<br />
<strong>and</strong> at the Joint Location<br />
Figures 15(a)-15(f) reveal radial<br />
displacements in the bell at the<br />
BISMAC cross-section <strong>and</strong> in the joint<br />
cross-section, respectively, at the key<br />
time points (t = 0.12 s, t = 0.70 s <strong>and</strong><br />
t = 2.0 s); each group uses a common<br />
color legend for ease of comparison.<br />
It is obvious that the deformation is<br />
negligible near the centre of the jellyfish<br />
bell <strong>and</strong> maximum at the tip of<br />
the bell. At joint location, the bell is<br />
practically undeformed.<br />
Figure 16(b) plots the time history<br />
of radial <strong>and</strong> axial tip displacement at<br />
BISMAC location, which conforms<br />
to second order system response<br />
to step excitation. In fact, the highpower<br />
pulse heating is done at much<br />
higher frequency than the bell is able<br />
to respond, thus inertia of the bell<br />
acts as low-pass filter for the pulsed<br />
heating excitation. But as discussed<br />
in transient heat transfer results, SMA<br />
goes through complete martensite to<br />
austenite transformation in this period,<br />
contracting due to phase transformation<br />
strain <strong>and</strong> achieves maximum<br />
deformation of U radial =1.31mm<br />
<strong>and</strong> U axial =7.87mmatt =0.12s.<br />
At the maximum deformation point,<br />
the inertial forces <strong>and</strong> elastic restoring<br />
forces are not in equilibrium. The bell<br />
has a slight overshoot above equilibrium<br />
position due to inertia. During<br />
moderate constant current excitation<br />
from t = 0.12s to t =0.70sastemperature<br />
keeps rising, without any<br />
additional benefit toSMAtransformation<br />
strain the structure relaxes<br />
under stored strain energy during<br />
rapid contraction <strong>and</strong> after a small oscillation<br />
around equilibrium condition<br />
finally settles to U radial = 10.87 mm<br />
<strong>and</strong> U axial = 6.04 mm. During relaxation<br />
phase from t = 0.70 s to t = 2.0 s,<br />
SMA goes through austenite to<br />
martensite transition as temperature<br />
drops rapidly <strong>and</strong> returns to original<br />
configuration after little oscillations as<br />
FIGURE 15<br />
(a-c) Deformed artificial A. aurita bell at BISMAC location at different time t = 0.12 s, 0.70 s, <strong>and</strong> 2.0 s (d-f) Deformed artificial A. aurita bell at joint<br />
location at different time t = 0.12 s, 0.70 s <strong>and</strong> 2.0 s.<br />
176 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 16<br />
(a) Comparison of the FE results with artificial A. aurita bell deformation <strong>and</strong> natural A. aurita.<br />
(b) Time history of A. aurita tip displacement at BISMAC location.<br />
FIGURE 17<br />
is evident from Figures 16(b), 15(a)-<br />
15(b), <strong>and</strong> 17(a).<br />
Figure 16(a) compares the relaxed<br />
<strong>and</strong> contracted position of the<br />
simulated A. aurita bell with natural<br />
A. aurita <strong>and</strong> robotic A. aurita. The<br />
natural A. aurita with its muscle contracting<br />
in excess of 40% clearly<br />
achieves the highest deformation,<br />
not being fully matched by either<br />
artificial A. aurita or FE simulation.<br />
Experimental deformation on artificial<br />
jellyfish was measured through<br />
image processing thus only exumbrella<br />
points are available for comparison.<br />
Experimental model, however,<br />
has changed its form during manufacturing,<br />
which was designed to match<br />
FEM in relaxed state. This results in<br />
apparent mismatch in FE prediction<br />
from the experimental deformation.<br />
Inspite of this mismatch, the overall<br />
deformations are comparable. Figure<br />
16(b) shows comparison of time<br />
response of the transient tip displacements<br />
between FE simulation <strong>and</strong> artificial<br />
<strong>and</strong> natural A. aurita. It should<br />
be noted that the natural A. aurita<br />
Trace of tip displacement at BISMAC location by (a) FE simulation <strong>and</strong> (b) artificial A. aurita<br />
experiment <strong>and</strong> (c) natural A. aurita.<br />
swims freely in water moving forward,<br />
while artificial (experimental) A. aurita<br />
<strong>and</strong> the model are held at the bell center<br />
<strong>and</strong> does not move it water. Also,<br />
for the particular cycle for which data<br />
is obtained has cycle time of 1.7 s instead<br />
of 2 s. Natural A. aurita contracts<br />
<strong>and</strong> relaxes very smoothly <strong>and</strong><br />
does not have any steady equilibrium<br />
position. Artificial A. aurita achieves<br />
about twice as much radial displacement<br />
as predicted by simulation but<br />
follows similar trend of sharp rise<br />
(fall), overshooting above the equilibrium<br />
position beyond t =0.12s,relaxing<br />
to equilibrium position during<br />
t = 0.12 s to t = 0.70 s due to stored<br />
strain energy in the bell during contraction<br />
cycle. In relaxation cycle, as<br />
SMA cools down <strong>and</strong> undergoes austenite<br />
to martensite transition, the<br />
bell returns back to its original location<br />
slightly overshooting beyond<br />
equilibrium position before returning<br />
back to relaxed configuration. FE simulation<br />
captures this entire physics<br />
perfectly. Artificial A. aurita returns<br />
to relaxed position slower than predicted<br />
by model, suggesting that<br />
heat gets conducted away from SMA<br />
more slowly than we predict. However,<br />
artificial A. aurita axial displacement<br />
being too small; gets affected by<br />
finite resolution of image processing<br />
<strong>and</strong> shows inconclusive trend. FE<br />
simulation predicts axial tip displacement<br />
time history similar to that of<br />
radial tip displacement but reduced<br />
in magnitude.<br />
Figure 17(a) exhibits the trace of<br />
the bell tip at BISMAC location <strong>and</strong><br />
suggests that bell tip does not follow<br />
the same path while relaxing to its original<br />
configuration than it did during<br />
contraction. It also illustrates the<br />
overshoot around the equilibrium positions<br />
(relaxed <strong>and</strong> contracted). Tip<br />
displacement for the artificial A. aurita<br />
July/August 2011 Volume 45 Number 4 177
(Figure 17(b)) shows different trace patterns during contraction <strong>and</strong> relaxation as<br />
predicted by the model. Since artificial A. aurita begins with different relaxation<br />
configuration, this is expected. Interaction with surrounding fluid would cause<br />
the tip displacement trace to change which are not accommodated in the<br />
model. Oscillations in the trace of the experimental results are partly due to tracking<br />
error. The magnitude of the oscillations is close to the predicted error from the<br />
tracking method. Figure 17(c) represents trace of tip displacement of natural<br />
A. aurita scaled to the same bell diameter. Natural A. aurita is freely moving in<br />
water. The direction of contraction <strong>and</strong> relaxation is quite similar to the one predicted<br />
by the model (Figure 17(a)); however, relative location of the path during<br />
contraction <strong>and</strong> relaxation is reversed. We believe that this is caused by the fluid<br />
forces <strong>and</strong> the freely forward motion present in natural A. aurita.<br />
For a more quantitative comparison, we calculated the curvature of the profiles<br />
by calculating radius of circle passing through three consecutive points,<br />
RP ð 2 Þ ¼<br />
ρðP 2 Þ ¼ 1<br />
RP ð 2 Þ<br />
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi<br />
ðx 1 x 2 Þ 2 þðy 1 y 2 Þ 2 ðx 2 x 3 Þ 2 þðy 2 y 3 Þ 2 ðx 3 x 1 Þ 2 þðy 3 y 1 Þ 2<br />
2ðx 2 y 1 x 1 y 2 þ x 3 y 2 x 2 y 3 þ x 1 y 3 x 3 y 1 Þ<br />
ð17Þ<br />
ð18Þ<br />
where R(P 2 ) is the radius at point P 2 (x 2, y 2 ), having previous point P 1 (x 1 , y 1 )<br />
<strong>and</strong> next P 3 (x 3 , y 3 ). Except first <strong>and</strong> last point, radius <strong>and</strong> curvature of all points<br />
can be calculated by using Eq. (17) <strong>and</strong> (18). Equal interval of all the points is<br />
very essential for this method, thus additional or fewer points were generated<br />
from available FE nodes, experimental traces <strong>and</strong> A. aurita profile.<br />
Figures 18(a) <strong>and</strong> 18(b) compare the curvature of the profile along the<br />
length of the curved exumbrella surface in relaxed <strong>and</strong> contracted condition between<br />
FE simulation, experimental <strong>and</strong> natural A. aurita. In relaxed condition,<br />
FEM curvature matches that of A. aurita. Experimentally as evident from Figure<br />
16(a) <strong>and</strong> Figure 18(a) it does not match A. aurita profile in the relaxed state.<br />
Figure 18(b) emphasizes that curvatures of FEM, experimental as well as natural<br />
animal, increases as the fish contracts. Artificial A. aurita does not have the same<br />
relaxed state as the natural animal, but in contracted state it follows natural<br />
A. aurita exumbrella deformation. It outperforms the curvature of the natural<br />
A. aurita by achieving a maximum curvature of 67 m −1 at about 80% length.<br />
FEM <strong>and</strong> natural A. aurita achieve maximum curvature of 50 m −1 at 67% <strong>and</strong><br />
45 m −1 at 85%, respectively, <strong>and</strong> show sharp increase at the tip, suggesting dynamic<br />
overshoot of the passive flap that is hypothesized to help swimming performance.<br />
Figure 18(c) displays profile errors between various profiles <strong>and</strong><br />
conditions,<br />
ɛ FER ðÞ¼ s<br />
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi<br />
ðÞ s x ExpBR ðÞ s<br />
2<br />
þ yFEBR ðÞ s y ExpBR ðÞ s<br />
2<br />
x FEBR<br />
ð19Þ<br />
The profile error ɛ FER is defined as the<br />
distance between corresponding<br />
points (x(s), y(s)) (subscript BR corresponds<br />
to the BISMAC location <strong>and</strong><br />
relaxed condition) of FEM <strong>and</strong> the artificial<br />
A. aurita vehicle relaxed profiles<br />
at location s. This was calculated by<br />
Eq. (19), <strong>and</strong> we observe that profile<br />
errors between the FE simulation <strong>and</strong><br />
artificial A. aurita experiment in relaxed<br />
<strong>and</strong> contracted states, ɛ FER <strong>and</strong><br />
ɛ FEC , have similar trends (Figure 18(c)).<br />
This indicates that the difference between<br />
FE simulation <strong>and</strong> artificial<br />
A. aurita is mainly due to the initial difference<br />
in relaxed profiles. We hypothesize<br />
that if artificial A. aurita were<br />
re-designed to match the natual<br />
A. aurita’s relaxed state, the simulation<br />
would match it more closely. The profile<br />
error does not increase beyond<br />
5.8mm which corresponds to 22%<br />
of the initial profile error in relaxed<br />
configuration.<br />
The model developed in this study<br />
gives better underst<strong>and</strong>ing of the temperature<br />
rise in the SMA wires <strong>and</strong><br />
transient heat distribution in <strong>and</strong><br />
around it during entire cycle. It reveals<br />
that in spite of complex geometry, heat<br />
distribution is practically uniform<br />
along SMA length <strong>and</strong> can be effectively<br />
captured by 1st order unsteady heat<br />
conduction equation. It explains degradation<br />
of performance of artificial<br />
A. aurita, over number of cycle due<br />
to heat accumulation resulting from<br />
insufficient cooling <strong>and</strong> suggests<br />
opportunity of making the bioinspired<br />
A. aurita vehicle more energy<br />
efficient by reducing oversupply of<br />
heat to SMA wire <strong>and</strong> effectively controlling<br />
SMA temperature. The model<br />
predicts transient deformation behavior<br />
of the artificial A. aurita qualitatively<br />
at BISMAC <strong>and</strong> fold locations,<br />
confirming effectiveness of the join<br />
design. It reveals that the transient tip<br />
178 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 18<br />
Curvature comparison at BISMAC <strong>and</strong> fold location between ANSYS, experiment, <strong>and</strong> biological<br />
A. aurita (a) relaxed condition, (b) contracted condition, (c) profile errors at BISMAC location-<br />
FER: between FE <strong>and</strong> experiment relaxed profile, FEC:FE <strong>and</strong> experiment contracted profile, AEC:<br />
A. aurita <strong>and</strong> experimental contracted profile, FAC:FE <strong>and</strong> A. aurita contracted profile.<br />
displacement response can be modeled<br />
as a second-order system; however, due<br />
to mismatch in initial profile, it predicts<br />
the radial tip displacement response<br />
∼44% of the experimental curve. The<br />
model captures these physical aspects<br />
of bell deformation accurately <strong>and</strong> is a<br />
useful tool to evaluate effectiveness of<br />
design changes on performance of the<br />
vehicle without building one.<br />
Conclusion<br />
We introduce a customization procedure<br />
for BISMAC actuators to<br />
be used as radial muscles in artificial<br />
A. aurita such that we can match<br />
the relaxed <strong>and</strong> contracted profiles of<br />
A. aurita at the BISMAC locations.<br />
By accurately modeling the hyperelastic<br />
behavior of EcoFlex (Shore<br />
00-10) RTV silicone <strong>and</strong> using experimentally<br />
obtained temperature-strain<br />
transformation curves for SMA wires,<br />
we provide a high-fidelity model that<br />
captures most essential physics of<br />
artificial A. aurita deformation. The<br />
model suggests a unique approach<br />
to overcome ANSYS limitation in<br />
modeling SMA temperature transformation<br />
by using negative thermal coefficient<br />
of expansion <strong>and</strong> two separate<br />
material curves for heating <strong>and</strong> cooling.<br />
This approach is generic enough<br />
to be used in numerical modeling<br />
of the SMA temperature-dependent<br />
transformation in any design. The<br />
transient heat transfer model provides<br />
better underst<strong>and</strong>ing of temperature<br />
rise in SMA wire <strong>and</strong> heat distribution<br />
around it during an entire contractionrelaxation<br />
cycle. This information is<br />
crucial in designing <strong>and</strong> evaluating<br />
SMA heating algorithm <strong>and</strong> cannot<br />
be obtained from the prototype directly.<br />
The model also reveals that in<br />
spite of the complex geometry, the<br />
temperature distribution along the<br />
length of SMA wires is uniform <strong>and</strong><br />
we could use a simple first order heat<br />
transfer model to study temperature<br />
distribution in <strong>and</strong> around SMA<br />
wires. It also exposes the reason for<br />
degradation of the performance of artificial<br />
A. aurita over a number of cycles<br />
due to heat accumulation <strong>and</strong> reveals<br />
an opportunity to make the artificial<br />
A. aurita more energy-efficient by reducing<br />
the amount of heat supplied<br />
above the austenite’s finish temperature<br />
A f that does not contribute toward<br />
bell deformation. Transient structural<br />
model accounts for the gravity <strong>and</strong><br />
the buoyancy forces <strong>and</strong> thus in equilibrium<br />
conditions (fully contracted<br />
<strong>and</strong> fully relaxed states), where fluid<br />
pressure differentials across subumbrella<br />
<strong>and</strong> exumbrella do not<br />
exist, we can predict the bell deformations<br />
fairly accurately. Transient<br />
structural analysis captures all essential<br />
behavior of artificial A. aurita at<br />
BISMAC <strong>and</strong> fold locations <strong>and</strong> confirms<br />
effectiveness of the joint design,<br />
though an exact match was not<br />
achieved due to initial profile mismatch.<br />
The model reveals that the<br />
transient tip displacement response<br />
can be modeled as a second-order system;<br />
however, due to initial profile<br />
mismatch; the model predicts the radial<br />
tip displacement response at<br />
∼44% of the experimental curve. Profile<br />
error analysis suggests that initial<br />
error in simulation <strong>and</strong> artificial<br />
A. aurita relaxed profiles is a major<br />
cause for the mismatch. The profile<br />
error increases by a small amount<br />
with an average of 0.00015 m (0.58%<br />
of ɛ FER ) <strong>and</strong> reaches a maximum of<br />
0.0058 m (22% of ɛ FER ). The model<br />
captures these physical aspects of bell<br />
deformation physics accurately <strong>and</strong><br />
isausefultooltoevaluateeffectiveness<br />
of design changes on performance<br />
of the vehicle without the need for<br />
July/August 2011 Volume 45 Number 4 179
time-consuming physical construction.<br />
The modeling methodology is<br />
generic <strong>and</strong> could be used to model<br />
other bio-inspired robots using similar<br />
construction technique.<br />
Acknowledgment<br />
This research is sponsored by the<br />
Office of Naval Research through contract<br />
N00014-08-1-0654.<br />
Lead Authors:<br />
Keyur B. Joshi <strong>and</strong> Shashank Priya<br />
Center for Energy Harvesting<br />
Materials <strong>and</strong> Systems<br />
Center for Intelligent Material<br />
Systems <strong>and</strong> Structure<br />
Virginia Polytechnic Institute<br />
<strong>and</strong> State University<br />
310 Durham Hall, Blacksburg,<br />
VA 24061<br />
Email: key4josh@vt.edu;<br />
spriya@mse.vt.edu<br />
References<br />
Albert, D.J. 2009. Aurelia labiata (Scyphozoa)<br />
jellyfish in Roscoe Bay: Their spatial distribution<br />
varies with population size <strong>and</strong> their<br />
behaviour changes with water depth. J Sea<br />
Res. 61(3):140-3. doi: 10.1016/j.seares.<br />
2008.11.001.<br />
Baker, E. 1972. Liquid cooling of microelectronic<br />
devices by free <strong>and</strong> forced convection.<br />
Microelectron Reliab. 11(2):213-22.<br />
doi: 10.1016/0026-2714(72)90704-4.<br />
Chapman, D.M. 1974. Cnidarian histology.<br />
In Coelenterate Biology, eds. Muscatine, L.,<br />
& Lenhoff, H.M., 2-92. New York:<br />
Academic Press.<br />
Colin, S.P., & Costello, J.H. 2002. Morphology,<br />
swimming performance <strong>and</strong> propulsive<br />
mode of six co-occurring hydromedusae.<br />
J Exp Biol. 205:427.<br />
Dabiri, J.P., Colin, S.P., & Costello, J.H.<br />
2005. Flow patterns generated by oblate<br />
medusan jellyfish: Field measurements <strong>and</strong><br />
laboratory analyses. J Exp Biol. 208(7):1257.<br />
doi: 10.1242/jeb.01519.<br />
Gladfelter, W.G. 1972. Structure <strong>and</strong> function<br />
of the locomotory system of Polyorchis montereyensis<br />
(Cnidaria, Hydrozoa). Helgol<strong>and</strong> Wiss<br />
Meer. 23(1):38. doi: 10.1007/BF01616310.<br />
Gladfelter, W.G. 1973. Structure <strong>and</strong> function<br />
of the locomotory system of the Scyphomedusa<br />
Cyanea capillata. Mar Biol. 14(2):150.<br />
doi: 10.1007/BF00373214.<br />
ImageJ. Image Processing <strong>and</strong> Analysis in<br />
Java. http://rsbweb.nih.gov/ij/. (accessed 12<br />
July 2011).<br />
National Instruments. LabView. http://<br />
www.ni.com/labview/. (accessed 12 July 2011).<br />
Smith, C., Villanueva, A., Joshi, K., Tadesse,<br />
Y., & Priya, S. 2011. Working principle of<br />
bio-inspired shape memory alloy composite<br />
actuators. Smart Mater Struct. 20(1):012001.<br />
doi: 10.1088/0964-1726/20/1/012001.<br />
Smith, C.F., & Priya, S. 2010. Bio-inspired<br />
Unmanned Undersea Vehicle. Behavior <strong>and</strong><br />
Mechanics of Multifunctional Materials <strong>and</strong><br />
Composites 2010, March 8, 2010 - March<br />
11, 2010, The <strong>Society</strong> of Photo-Optical<br />
Instrumentation Engineers (SPIE); American<br />
<strong>Society</strong> of Mechanical Engineers. San Diego,<br />
CA: SPIE.<br />
Tadesse, Y., Brennan, J., Smith, C., Long,<br />
T.E., & Priya, S. 2010a. Synthesis <strong>and</strong><br />
characterization of polypyrrole composite<br />
actuator for jellyfish unmanned undersea<br />
vehicle. 764222-11. San Diego, CA: SPIE.<br />
Tadesse, Y., Thayer, N., & Priya, S. 2010b.<br />
Tailoring the response time of shape memory<br />
alloy wires through active cooling <strong>and</strong> prestress.<br />
J Intel Mat Syst Str. 21(1):19-40.<br />
doi: 10.1177/1045389X09352814.<br />
Toki Corp. Biometal Fiber. http://www.toki.<br />
co.jp/biometal/english/q<strong>and</strong>a.php. (accessed<br />
12 July 2011).<br />
Valentine, J.W. 2004. On the Origin of<br />
Phyla. Chicago: University of Chicago Press.<br />
Villanueva, A., Bresser, S., Chung, S.,<br />
Tadesse, Y., & Priya, S. 2009. Jellyfish<br />
inspired underwater unmanned vehicle.<br />
Proc SPIE. 7287:72871G. doi: 10.1117/<br />
12.815754.<br />
Villanueva, A., & Priya, S. 2010. BISMAC<br />
control using SMA resistance feedback.<br />
In: SPIE Conference Proceedings, SPIE<br />
NDE Conference. 76421Z-12. San Diego, CA:<br />
SPIE.<br />
Villanueva, A., Priya, S., Anna, C., & Smith,<br />
C. 2010a. Robojelly bell kinematics <strong>and</strong><br />
resistance feedback control. In: IEEE<br />
International Conference on Robotics <strong>and</strong><br />
<strong>Biomimetics</strong> (ROBIO). 1124-9. Tianjin,<br />
China: IEEE (Institute of Electrical <strong>and</strong><br />
Electronics Engineers).<br />
Villanueva, A., Smith, C., & Priya, S.<br />
submitted. Under review, Paper Ref no.<br />
“BB/378339/PAP/255146”. Biomimetic robotic<br />
jellyfish (Robojelly) using shape memory<br />
alloy. IOP Bioinspir Biomim.<br />
Villanueva, A.A., Joshi, K.B., Blottman, J.B.,<br />
& Priya, S. 2010b. A bio-inspired shape<br />
memory alloy composite (BISMAC) actuator.<br />
Smart Mater Struct. 19(2):025013.<br />
doi: 10.1088/0964-1726/19/2/025013.<br />
Yang, Y., Ye, X., & Guo, S. 2007. A new<br />
type of jellyfish-like microrobot. In: IEEE<br />
International Conference on Integration<br />
<strong>Technology</strong>. 673-8. Rourkela, India: IEEE<br />
(Institute of Electrical <strong>and</strong> Electronics<br />
Engineers). doi: 10.1109/ICITECHNOLOGY.<br />
2007.4290404.<br />
Yeom, S.-W., & Oh, I.-K. 2009. A biomimetic<br />
jellyfish robot based on ionic polymer<br />
metal composite actuators. Smart Mater<br />
Struct. 18(8):085002. doi: 10.1088/0964-<br />
1726/18/8/085002.<br />
180 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
PAPER<br />
Swimming <strong>and</strong> Walking of an Amphibious<br />
Robot With Fin Actuators<br />
AUTHOR<br />
Naomi Kato<br />
Graduate School of Engineering,<br />
Osaka University<br />
Introduction<br />
I<br />
t has recently been clarified that<br />
natural coastline areas <strong>and</strong> tidal flats<br />
play an important role in preserving<br />
ocean environments. To protect<br />
coastal environments, regular monitoring<br />
of these areas is important.<br />
Monitoring has previously been<br />
done by humans on foot or using<br />
boats, but this work can be dangerous<br />
because of breaking waves <strong>and</strong> rip<br />
currents. Monitoring on foot is limited<br />
because it cannot be carried out<br />
in deep waters, while monitoring by<br />
boat is limited to the areas accessible<br />
by water. Automatic monitoring by<br />
an amphibious robot is therefore expected<br />
to eliminate the safety threats<br />
to human monitors <strong>and</strong> improve operational<br />
efficiency. However, if an<br />
amphibious robot moves by gaining<br />
traction with screws <strong>and</strong> caterpillars,<br />
it will not be able to move about in<br />
areas such as marshes <strong>and</strong> will damage<br />
the environments of the areas in<br />
which it moves. An environmentally<br />
friendly amphibious robot is thus<br />
needed.<br />
Several studies have reported the<br />
development of amphibious robots.<br />
An amphibious snake-like robot, the<br />
ACM-R5 (Yamada et al., 2005), can<br />
operate both on ground <strong>and</strong> in water<br />
by undulating its long body. The<br />
ACM-R5 uses special paddles <strong>and</strong><br />
ABSTRACT<br />
With the goal of automatic monitoring of environments along natural coastal<br />
areas <strong>and</strong> tidal flats, researchers designed <strong>and</strong> developed an amphibious robot<br />
equipped with fin actuators called “RT-I” that mimics the locomotion of both a tortoise<br />
<strong>and</strong> a sea turtle. Experiments were carried out using a forearm with 4 degrees<br />
of freedom, which can reproduce the walking motions of tortoises <strong>and</strong> sea turtles on<br />
s<strong>and</strong>, to evaluate the walking performances of a robotic tortoise <strong>and</strong> a robotic sea<br />
turtle. It was clarified that the arm for a robotic tortoise is more suitable for use on<br />
soil compared with the arm for a robotic sea turtle. The advantages of both sea<br />
turtles <strong>and</strong> tortoises were adopted in a robotic turtle, namely, the lift-based swimming<br />
mode sea turtles use <strong>and</strong> the quadrupedal locomotion tortoises use. The present<br />
amphibious robot consists of four main components: (i) leg units, (ii) a control<br />
unit pressure hull, (iii) a buoyancy adjusting device, <strong>and</strong> (iv) a fairing cover. To realize<br />
not only swimming motion with the combination of flapping, rowing, <strong>and</strong><br />
feathering, but also tortoise-like walking motion, three motors were set up at the<br />
acromioclavicular joint using a differential gear mechanism to independently produce<br />
the three types of motion, <strong>and</strong> one motor was set up to produce elbow joint<br />
motion. A buoyancy-adjusting device was installed to realize walking on l<strong>and</strong> <strong>and</strong> in<br />
water as well as swimming in shallow water. The swimming <strong>and</strong> walking performances<br />
of the amphibious robot in water were evaluated by measuring the forward<br />
swimming speed, backward swimming speed, speed of turning, <strong>and</strong> speed of descending<br />
vertically as the indexes of the maneuverability of the robotic turtle, <strong>and</strong><br />
the walking speed <strong>and</strong> propulsive efficiency with the crawl gait for various walking<br />
patterns in still water <strong>and</strong> in waves.<br />
Keywords: locomotion, tortoise, sea turtle, experiment<br />
wheelsmountedarounditsbodyto<br />
propel itself through water <strong>and</strong> over<br />
ground in a snake-like fashion, generating<br />
propulsive force that allows it to<br />
glide freely in the tangential direction.<br />
The biomimetic amphibious soft cord<br />
robot (Wakimoto et al., 2006), which<br />
is made of Mckibben actuators <strong>and</strong><br />
plastic plates, can move both on the<br />
ground <strong>and</strong> in water, undulating its<br />
long body. A spinal cord model <strong>and</strong><br />
its implementation in an amphibious<br />
salam<strong>and</strong>er robot (Ijspeert et al.,<br />
2007) were studied to demonstrate<br />
how a primitive neural circuit for<br />
swimming can be extended using<br />
phylogenetically more recent limb oscillatory<br />
centers to explain the ability<br />
of salam<strong>and</strong>ers to switch between<br />
swimming <strong>and</strong> walking. The AQUA<br />
(Dudek et al., 2007), an amphibious<br />
robot, can swim <strong>and</strong> walk along the<br />
shore <strong>and</strong> on the bottom of the ocean<br />
by moving its fins. The AQUA uses<br />
six paddles, which act as control surfaces<br />
during swimming <strong>and</strong> as legs<br />
while walking. An amphibious walking<br />
robot was developed by Tanaka<br />
<strong>and</strong> Shirai (2006) to perform a shoreline<br />
survey. It has six legs, each of<br />
July/August 2011 Volume 45 Number 4 181
which has three joints. It was successful<br />
in obtaining the distribution of<br />
ground levels from l<strong>and</strong> to shallow<br />
water. An amphibious robotic turtle<br />
was built by Low et al. (2007) to imitate<br />
the locomotion of Cheloniidae,<br />
both in water <strong>and</strong> on l<strong>and</strong>, to perform<br />
various operations. The crawling <strong>and</strong><br />
lift-based swimming gaits were analyzed<br />
<strong>and</strong> implemented in the prototype.<br />
However, all of these robots are<br />
not operated in practice to monitor<br />
the coastal environment on l<strong>and</strong> <strong>and</strong><br />
in water by using the multiple functions<br />
of walking <strong>and</strong> swimming.<br />
For the field operation of an amphibious<br />
robot, a rigid fuselage with<br />
an adequate payload is necessary so<br />
that a control system <strong>and</strong> sensors for<br />
monitoring the environment can be<br />
installed. We have been studying a biomimetic<br />
underwater robot equipped<br />
with mechanical pectoral fins from<br />
the viewpoint of high maneuverability<br />
under disturbances such as waves <strong>and</strong><br />
water currents (Kato et al., 2006;<br />
Suzuki & Kato, 2005; Kato & Liu,<br />
2003). Taking the field operation<br />
<strong>and</strong> application of our experiences on<br />
the biomimetic underwater robot<br />
into account, this study focuses on an<br />
amphibious robotic turtle, which not<br />
only can swim in the sea but also<br />
walk on the l<strong>and</strong> to perform environmental<br />
monitoring of natural coast<br />
<strong>and</strong> tidal flat areas.<br />
Turtles that can walk <strong>and</strong> swim are<br />
generally categorized intoseaturtles<br />
<strong>and</strong> tortoises. Sea turtles have good<br />
swimming ability but poor walking<br />
ability because they drag their bodies<br />
on the l<strong>and</strong>, which causes friction<br />
against the s<strong>and</strong>. Tortoises, on the<br />
other h<strong>and</strong>, cannot swim smoothly,<br />
but they can walk better than sea<br />
turtles can because of their quadrupedal<br />
locomotion capability. In this<br />
study, we attempted to adopt the advantages<br />
of these two turtles into a robotic<br />
turtle.<br />
This paper presents (1) a description<br />
of the walking performance of<br />
an arm with 4 degrees of freedom<br />
(DOF) that can reproduce the walking<br />
motionsofseaturtles<strong>and</strong>tortoises<br />
from the viewpoints of mobility <strong>and</strong><br />
terrain trafficability, (2) details of the<br />
design <strong>and</strong> development of an amphibious<br />
robot with fin actuators,<br />
<strong>and</strong> (3) an evaluation of the walking<br />
<strong>and</strong> swimming performance of the<br />
robot in a laboratory environment.<br />
Here, mobility is defined as the walking<br />
performance of a vehicle depending<br />
on motor torque <strong>and</strong> vehicle<br />
configuration. Trafficability is the<br />
soil-bearing capacity of a vehicle.<br />
Locomotion of Tortoises<br />
<strong>and</strong> Sea Turtles<br />
Terrestrial Locomotion<br />
Walker (1971) studied the walking<br />
of a tortoise, Chrysemys picta, using<br />
cinephotography <strong>and</strong> X-rays <strong>and</strong><br />
clarified the relationship between the<br />
structure of the skeleton <strong>and</strong> the movements<br />
of the fore <strong>and</strong> hind limbs. The<br />
structure of the skeleton of the forelimb<br />
is the same as that of the hind<br />
limb. The forelimb consists of an<br />
acromioclavicular joint with 3 DOF,<br />
ahumerus,anelbowwith1DOF,a<br />
FIGURE 1<br />
Top view <strong>and</strong> side view of tortoise Chinemys reevesii.<br />
forearm, a wrist with 2 DOF, <strong>and</strong> a<br />
h<strong>and</strong>. Wyneken (1997) explained<br />
that in the locomotion of sea turtles<br />
on l<strong>and</strong>, clutching movements are<br />
seen in adult Chelonia mydas, Natator<br />
depressus, <strong>and</strong>Dermochelys coriacea,<br />
while adults of other cheloniid species<br />
employ quadrupedal gaits in which diagonally<br />
opposite feet move as a pair.<br />
Sea turtles support themselves on the<br />
carpus <strong>and</strong> the anterior edge of the<br />
h<strong>and</strong> rather than on the palmar<br />
surface.<br />
To design a robotic turtle, we analyzed<br />
quantitative information on the<br />
movement of the forelimb joints of a<br />
captured tortoise, Chinemys reevesii,<br />
with the following dimensions: length<br />
of shell × length of forelimb × length<br />
of hindlimb = 230 mm × 40 mm ×<br />
40 mm. The markers <strong>and</strong> body-fixed<br />
coordinates (x, y, z) on the tortoise<br />
were set up as shown in Figure 1.<br />
The origin of the body-fixed coordinates<br />
was set on the acromioclavicular<br />
joint. First, movies of the walking<br />
motions of the tortoise were taken<br />
using two CCD cameras. The movies<br />
from the top view <strong>and</strong> side view were<br />
taken at the same time. Second, the<br />
marked points were tracked using<br />
software that computed the twodimensional<br />
coordinates of the points.<br />
Third, the three-dimensional coordinates<br />
of the elbow <strong>and</strong> wrist on<br />
the body-fixed coordinates were<br />
182 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 2<br />
Trajectories of the elbow <strong>and</strong> the wrist in x-y plane.<br />
computed. Figures 2 <strong>and</strong> 3 show the<br />
two-dimensional tracks in the x-y<br />
plane <strong>and</strong> x-z plane of the motion of<br />
the wrist <strong>and</strong> the elbow, respectively,<br />
during walking in the case of a walking<br />
speed of 0.13 m/s, a period of walking<br />
of 1.3 s, <strong>and</strong> a stance phase of 0.78.<br />
Here, the stance phase denotes the fraction<br />
of time during which the forelimb<br />
is set on the ground during the walking<br />
period.<br />
If the forelimb is considered to be<br />
an arm, the joint angles of the forelimb<br />
FIGURE 3<br />
Trajectories of the elbow <strong>and</strong> the wrist in x-z plane.<br />
derived from inverse kinematics of the<br />
arm can be obtained. Here, an arm is<br />
assumed to consist of an acromioclavicular<br />
joint with 3 DOF, a humerus, an<br />
elbow with 1 DOF, a forearm, <strong>and</strong> a<br />
wrist. The (x, y, z) coordinates are<br />
fixed at the acromioclavicular joint, as<br />
shown in Figure 4. The rowing motion<br />
is defined as rotational motion around<br />
the z axis. The (x′, y′, z′) coordinates<br />
are defined as coordinates rotated<br />
around the z axis by the rowing motion.<br />
The feathering motion is defined<br />
as rotational motion around the y′ axis.<br />
The (x″, y″, z″) coordinates are defined<br />
as the coordinates rotated around the<br />
y′ axis by the feathering motion. The<br />
flapping motion is defined as rotational<br />
motion around the x″ axis.<br />
The coordinate system of the arm<br />
was set up as shown in Figure 4,<br />
where n, s, <strong>and</strong>a, denotingunitvectors<br />
fixed on the forearm, were set parallel<br />
to x, y, <strong>and</strong> z, respectively, when<br />
all of the joint angles were zero. The<br />
joint angles θ 1 , θ 2 , θ 3 , θ 4 are defined<br />
as the angle of rowing motion, angle<br />
of feathering motion, angle of flapping<br />
motion, <strong>and</strong> angle of bending of forearm,<br />
respectively. Lengths of the arms<br />
l 1 , l 2 are defined as the length of the<br />
humerus <strong>and</strong> the length of the forearm,<br />
respectively.<br />
Figure 5 shows time variations of<br />
thejointangles.Wecanseethatthe<br />
feathering angle varies from 5° to 49°<br />
during the power stroke from 0 to<br />
0.6 s, <strong>and</strong> the bending angle of the<br />
forelimb varies between −75° <strong>and</strong><br />
−115° during the power stroke,<br />
which indicates that the forelimb produces<br />
forward thrust by kicking the<br />
ground <strong>and</strong> positioning the forearm almost<br />
vertically on the ground. During<br />
the recovery stroke from 0.6 to 1.3 s,<br />
the bending angle of the forelimb<br />
reaches −20°, which indicates that the<br />
forearm is raised from the ground. The<br />
flapping angle does not vary much<br />
during the entire stroke.<br />
Aquatic Locomotion<br />
Sea turtles swim in water using<br />
their forelimbs to provide thrust. The<br />
synchronous sweeping of the flippers<br />
introduces a stable heading direction.<br />
Wyneken (1997) explains lift-based<br />
mechanisms of thrust production in<br />
which the locomotor apparatus of a<br />
sea turtle acts as a wing to generate<br />
July/August 2011 Volume 45 Number 4 183
FIGURE 4<br />
Coordinate system <strong>and</strong> definitions.<br />
minimize surface area as they are<br />
brought forward.<br />
Experiment on the<br />
Walking Performance<br />
of an Arm<br />
We constructed an arm that could<br />
reproduce the walking motion of a tortoise<br />
<strong>and</strong> a sea turtle to analyze the<br />
walking performance from the viewpoints<br />
of mobility <strong>and</strong> trafficability.<br />
lift forces during large portions of the<br />
powerstroke. Isobe et al. (2010) clarified<br />
that the feathering motion of<br />
fore flippers of a sea turtle influences<br />
the thrust production by measuring<br />
the 3-D motion of fore flippers of a<br />
sea turtle in a water circulating tank<br />
<strong>and</strong> conducting numerical simulations<br />
FIGURE 5<br />
Time variations of joint angles.<br />
based on quasi-steady wing element<br />
theory.<br />
Wyneken (1997) showed dragbased<br />
propulsion by a swimming semiaquatic<br />
turtle. Diagonally opposite<br />
limbs are protracted, then retracted together,<br />
<strong>and</strong> act as paddles. The distal<br />
elements of the limbs are flexed to<br />
Arm<br />
The robotic arm we developed<br />
makes the motions of rowing, feathering,<br />
<strong>and</strong> flapping, <strong>and</strong> the forearm can<br />
also be bent. The arm moves on rails to<br />
make these motions on s<strong>and</strong>. The angles<br />
of the motions of rowing, feathering,<br />
flapping, <strong>and</strong> bending were<br />
measured by four potentiometers, the<br />
distance of the movement along the<br />
rails was measured by a potentiometer,<br />
<strong>and</strong> the forces that were applied to the<br />
h<strong>and</strong> were measured by a six-axes force<br />
sensor, as shown in Figure 6. The humerus<br />
<strong>and</strong> forearm are each 150 mm<br />
long. The rowing angle, feathering<br />
angle, flapping angle, <strong>and</strong> bending<br />
angle of the forearm vary within the<br />
following ranges, respectively: ±70°,<br />
FIGURE 6<br />
Picture of manipulator.<br />
184 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
±70°, ±50°, <strong>and</strong> 0° to −110°. The rowing motion, feathering motion, flapping motion, <strong>and</strong> bending motion were produced by<br />
four motors independently. Because frictional force works between the rails <strong>and</strong> the arm, a force corresponding to the static<br />
frictional force was applied to the arm along the rail using a pulley <strong>and</strong> a weight. In the experiments on the walking performance<br />
of a sea turtle, an amount of force was subtracted from the static frictional force to simulate the friction a sea turtle<br />
generates on s<strong>and</strong>. The arm was connected to the weight by a string. Toyoura s<strong>and</strong>, which has almost constant particle<br />
diameter <strong>and</strong> known physical parameters, was used in the experiments.<br />
Figure 7 shows the h<strong>and</strong> shapes. The shape of the end of the h<strong>and</strong> of the model simulating a tortoise (T ) is a 70 mm ×<br />
70 mm square. That of the model simulating a sea turtle (S) is a 150 mm × 125 mm rectangle.<br />
Kinetic Relations Between External Force <strong>and</strong> Joint Torque<br />
To discuss the propulsive efficiency in terms of walking performance of the arm, we need to know the torque of each joint<br />
of the arm. We then consider the kinetic relations between external force <strong>and</strong> joint torque.<br />
First, we refer to the Jacobian matrix in the kinetics of the arm. We define θ(θ 1 , θ 2 , θ 3 , θ 4 ) T as the rotational angles of<br />
joints. The relations between the rotational velocities of the joints, θ : θ : 1 ; θ : 2 ; θ : 3 ; θ : T 4 , translational velocity of the end of the<br />
forearm (see Figure 6), ( P : : : :<br />
r ðx;<br />
y; zÞ<br />
T ), <strong>and</strong> rotational velocities of the end of the forearm, ( Φ : r ð ϕ : x ; ϕ : y ; ϕ : z Þ T ), are described as<br />
follows:<br />
: <br />
P : r<br />
¼ J θ<br />
: Φ r<br />
ð1Þ<br />
<br />
J ¼ s <br />
1 × ðP r P 1 Þ s 2 × ðP r P 2 Þ s 3 × ðP r P 3 Þ s 4 × ðP r P 4 Þ<br />
s 1 s 2 s 3 s 4<br />
ð2Þ<br />
where J, P i ,<strong>and</strong>s i denote the Jacobian matrix, position vector, <strong>and</strong> vector of rotational direction, respectively (refer to<br />
Figure 8).<br />
Next, we will explain the kinetic relations between the external force, F(F x , F y , F z ) T , moment, M(M x , M y ,M z ) T , <strong>and</strong><br />
joint torque, Q(q 1 , q 2 , q 3 , q 4 ). The moment, M i , that is applied to each joint is<br />
M i ¼ ðP r P i Þ×Fþ M ð3Þ<br />
<strong>and</strong> thus the torque of each joint is<br />
q i ¼ s i ⋅ M i<br />
¼ s i ⋅ðP r<br />
P i<br />
Þ× F þ s i ⋅M<br />
¼ fs i ×P ð r P i Þg⋅F þ s i ⋅M<br />
ð4Þ<br />
<strong>and</strong> the relation between eternal forces, moments, <strong>and</strong> joint torque is represented using a transposed Jacobian matrix as<br />
follows:<br />
2 3 2<br />
3<br />
q 1 s 1 × ðP r P 1 Þ s 1 <br />
q<br />
Q ¼ 6 2<br />
7<br />
4 q 3<br />
5 ¼ s 2 × ðP r P 2 Þ s 2<br />
6<br />
7<br />
4 s 3 × ðP r P 3 Þ s 3<br />
5 F <br />
¼ J T F<br />
M M<br />
q 4 s 4 × ðP r P 4 Þ s 4<br />
ð5Þ<br />
July/August 2011 Volume 45 Number 4 185
FIGURE 7<br />
H<strong>and</strong> models.<br />
In the present study, we discuss<br />
propulsive efficiency from the viewpoints<br />
of mobility <strong>and</strong> trafficability.<br />
We define propulsive efficiency,η, as<br />
follows:<br />
FIGURE 9<br />
Trajectories of wrist <strong>and</strong> elbow of type (a).<br />
motion of a robotic tortoise were configured<br />
with reference to the experimental<br />
motion analysis of a tortoise<br />
in locomotion of tortoises <strong>and</strong> sea<br />
turtles. Type (a) motion includes<br />
movement of the arm vertically during<br />
the power stroke. It has a symmetric<br />
trajectory in the anteroposterior direction<br />
(see Figure 9). Type (b) motion is<br />
based on the pulling motion during<br />
the power stroke. It has a larger anterior<br />
trajectory component within the<br />
whole trajectory, which is analogous<br />
to the observed trajectory of the tortoise.<br />
Type (c) motion is based on<br />
the kicking motion during the power<br />
stroke. It has a larger posterior trajectory<br />
component within the whole trajectory.<br />
Figures 10, 11, <strong>and</strong> 12 show<br />
the time variations of the joint angles<br />
of type (a), type (b), <strong>and</strong> type (c), respectively.<br />
The range of movement of<br />
η ¼<br />
∫ T F x vdt<br />
4 :<br />
∑ ∫ T q i θi dt<br />
i¼1<br />
ð6Þ<br />
Here, v denotes the translational speed<br />
along the x axis. The numerator is the<br />
value of the work that the arm applies<br />
to the s<strong>and</strong>. The denominator is the<br />
value of the total input work by the<br />
joints.<br />
Comparison of Walking<br />
Performance of an Arm<br />
Between a Robotic Tortoise<br />
<strong>and</strong> a Robotic Sea Turtle<br />
The walking performance of an<br />
arm for a robotic tortoise using the<br />
h<strong>and</strong> models (T ) was measured. The<br />
following three types of walking<br />
FIGURE 8<br />
Vectors of coordinates.<br />
186 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 10<br />
Time variations of joint angles of type (a).<br />
FIGURE 12<br />
Time variations of joint angles of type (c).<br />
FIGURE 11<br />
Time variations of joint angles of type (b).<br />
the wrist along the x axis <strong>and</strong> that along<br />
the z axis were taken as 162 <strong>and</strong><br />
70 mm, respectively. The range of<br />
movement of the wrist along the<br />
y axis was taken as 103 mm. The period<br />
of motion was changed from 7.5<br />
to 15.0 s with an interval of 2.5 s.<br />
The sinkage of the end of the forearm<br />
below the surface of the s<strong>and</strong> was also<br />
varied; values of 10, 15, <strong>and</strong> 20 mm<br />
were used. Although the sinkage of a<br />
robotic turtle changes with time according<br />
to the weight of the body,<br />
the type of arm motion <strong>and</strong> the condition<br />
of the soil, it was treated as an independent<br />
parameter affecting the<br />
walking performance of the arm in<br />
this study so we could evaluate it together<br />
with other parameters from<br />
the viewpoint of fundamental walking<br />
performance. It was controlled by<br />
a feedback controller located in the<br />
arm controller using the measured angles<br />
of the motions of rowing, feathering,<br />
flapping, <strong>and</strong> bending by four<br />
potentiometers, after the initial attitude<br />
of the arm was set up in relation<br />
to the surface of the s<strong>and</strong>.<br />
Figure 13 shows the averaged propulsive<br />
forces, Fx, in the direction of<br />
the x axis during one period for the<br />
three types of arm motion with the<br />
motion period of 10 s. Figure 14<br />
shows the averaged vertical forces, Fz,<br />
in the direction of the z axis during one<br />
period for the three types of arm motion<br />
with the motion period of 10 s.<br />
The fact that the averaged vertical<br />
forces, Fz, for type (b) motion are negative<br />
can be attributed to the posture of<br />
the forearm in the first half of the time<br />
period. Namely, the flat plate at the<br />
end of the forearm digs into the s<strong>and</strong><br />
by positioning the forearm anteriorly<br />
in the firsthalfofthetimeperiod.<br />
This action leads it to carry s<strong>and</strong> on<br />
the flat plate, which causes a negative<br />
vertical force during one cycle. On<br />
the other h<strong>and</strong>, the forearm in type<br />
(a) motion is positioned almost vertically<br />
on the s<strong>and</strong> during one period,<br />
<strong>and</strong> the forearm in type (c) motion is<br />
positioned posteriorly so as to kick<br />
s<strong>and</strong>. Although it is difficult to discriminate<br />
among the three motion<br />
types in terms of the propulsive efficiencies<br />
beyond the sinkage of 15 mm,<br />
type (a) shows the largest values for<br />
the averaged propulsive force. The<br />
July/August 2011 Volume 45 Number 4 187
FIGURE 13<br />
Averaged propulsive force Fx in the direction of the x-axis during one<br />
period for the three types of arm motions against sinkage of the end<br />
of the forearm below the surface of the s<strong>and</strong> using the h<strong>and</strong> model (T).<br />
FIGURE 15<br />
Time variations of joint angles of type (d).<br />
same relation among the three motion<br />
types can be applied to the averaged<br />
vertical forces. Type (a) arm motion is<br />
suitable not only from the viewpoint<br />
of mobility, but also trafficability.<br />
Because sea turtles use the anterior<br />
edge of the h<strong>and</strong> for locomotion on<br />
l<strong>and</strong>, we used type (d) of walking<br />
FIGURE 14<br />
motion, in which the flipper moves inclined<br />
at an angle of 45°, in the design<br />
of the robotic sea turtle. Figure 15<br />
shows the time variations of the joint<br />
angles of type (d) with the motion period<br />
of 10 s. The sinkage of the bottom<br />
of the flipper below the surface of the<br />
s<strong>and</strong> was varied among three different<br />
Averaged vertical force Fz in the direction of the z-axis during one period for the three types of arm<br />
motions against sinkage of the end of the forearm below the surface of the s<strong>and</strong> using the h<strong>and</strong><br />
model (T).<br />
levels:10mm,15mm,<strong>and</strong>20mm.<br />
The walking motion of a sea turtle generates<br />
friction on the s<strong>and</strong> because the<br />
turtle drags its body on the l<strong>and</strong> as it<br />
walks. To model this situation, friction<br />
on the carriage of the manipulator was<br />
added. We compared the walking performance<br />
of these two types of motion<br />
under the conditions including friction<br />
on the carriage of the manipulator<br />
set at the following levels: 2.94, 4.90,<br />
<strong>and</strong> 6.86 N.<br />
Figure 16 shows the averaged propulsive<br />
force, Fx, in the direction of the<br />
x axis during one period against sinkage<br />
of the end of the forearm below<br />
the surface of the s<strong>and</strong> <strong>and</strong> friction<br />
on the carriage of the manipulator<br />
(the key to the symbols expresses friction<br />
on the carriage of the manipulator).<br />
As the friction on the rail increases, the<br />
averaged propulsive force, Fx, also increases.<br />
Figure 17 shows the average<br />
vertical force during one period against<br />
sinkage. As the sinkage increases, the<br />
maximum vertical force, Fz, also<br />
increases. The dependency of the<br />
maximum vertical force, Fz, on the<br />
188 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 16<br />
Averaged propulsive force Fx in the direction of x-axis during one period<br />
for type (d) arm motion against sinkage of the end of the forearm<br />
below the surface of s<strong>and</strong> using the h<strong>and</strong> model (S).<br />
FIGURE 17<br />
Averaged vertical force Fz in the direction of z-axis during one period<br />
for type (d) of arm motion against sinkage of the end of the forearm<br />
from the surface of s<strong>and</strong> using the h<strong>and</strong> model (S).<br />
friction on the rail is clear. When we<br />
consider the walking motion of a robotic<br />
sea turtle on the beach, friction<br />
between the s<strong>and</strong> <strong>and</strong> the body depends<br />
on the vertical force subtracting<br />
the vertically upward force from the<br />
weight of the robotic sea turtle.<br />
When comparing type (a) walking<br />
motion of a robotic tortoise, which<br />
showed the best performance among<br />
the three types, with type (d) walking<br />
motion of a robotic sea turtle, it can be<br />
seen that the propulsive efficiency of<br />
type (d) is influenced by friction <strong>and</strong><br />
the averaged vertical force acting on<br />
the h<strong>and</strong> plate in the case of the arm<br />
for a robotic tortoise is larger than in<br />
the case of the arm for a robotic sea turtle.<br />
An arm for a robotic tortoise is suitable<br />
for moving heavy payloads over a<br />
variety of terrains. To discuss the trafficability<br />
of the arm for a sea turtle, we<br />
define M as the mass of the body, g as<br />
gravitational acceleration, k as the frictional<br />
coefficient (Setouchi & Shinjo,<br />
2001) between the soil <strong>and</strong> the ventral<br />
surface of the robotic sea turtle, Fxm as<br />
the mean value of propulsive force, <strong>and</strong><br />
Fzm as the mean value of vertical force<br />
against the soil. Assuming that a pair of<br />
fore flippers <strong>and</strong> a pair of rear flippers<br />
exert thrust forces simultaneously, we<br />
obtain the following equation of static<br />
equilibrium for walking on soil:<br />
ðM⋅g − 4⋅FzmÞ⋅k ¼ 4⋅Fxm ð7Þ<br />
IfweassumeFxm=Fzmforthe<br />
type (d) arm motion based on Figures<br />
16 <strong>and</strong> 17, we obtain the following<br />
relation:<br />
k<br />
Fzm ¼ M⋅g⋅ < M⋅g=4 ð8Þ<br />
41þ ð kÞ If we set the length of the humerus as l 1<br />
<strong>and</strong> the length between the elbow joint<br />
<strong>and</strong> center of the h<strong>and</strong> plate as l 2 using<br />
the arm structure shown in Figures 6<br />
<strong>and</strong> 8, the torque around the x axis<br />
<strong>and</strong> that around the z axis are Fzm∙<br />
(l 1 + l 2 ). For a robotic tortoise using a<br />
crawling gait with slow speed, we obtain<br />
Fzm = M · g/4 during the time when<br />
the 4 feet are placed on ground. If we<br />
assume that the robotic turtle uses<br />
type (a) arm motion <strong>and</strong> that Fxm is<br />
nearly equal to Fzm/25 based on Figures<br />
13 <strong>and</strong> 14, we obtain the torque<br />
around the x axis of Fzm∙l 1 <strong>and</strong> a<br />
torque around the z axis of Fzm · l 1 /25.<br />
The total torque on the arm of the<br />
sea turtle robot, Q S , is expressed as<br />
follows:<br />
k<br />
Q S ¼ 2⋅M⋅g⋅<br />
41þ ð kÞ ⋅ ð l 1 þ l 2 Þ: ð9Þ<br />
The total torque on the arm of the<br />
tortoise robot, Q L , is expressed as<br />
follows:<br />
Q L ¼ 26=25⋅M⋅g=4⋅l 1<br />
The ratio of Q S to Q L γ is<br />
γ ¼ 25 <br />
13 ⋅ k<br />
1 þ k ⋅ 1 þ l2 <br />
l1<br />
ð10Þ<br />
ð11Þ<br />
Because the friction ratio of s<strong>and</strong><br />
(Setouchi & Shinjo, 2001) is in the<br />
July/August 2011 Volume 45 Number 4 189
ange of 0.4-0.6 <strong>and</strong> l 2 /l 1 is larger than<br />
1.0 for a sea turtle, we find that γ tends<br />
to approach 1.0 if k becomes smaller,<br />
<strong>and</strong> that γ tends to become larger<br />
than 1.0 if k becomes larger. Because<br />
smaller torques acting around the<br />
joints are needed from the viewpoint<br />
of the mechanical design of a<br />
turtle robot, the arm for the robotic<br />
tortoise is suitable on soil from the<br />
viewpoint of trafficability.<br />
Design <strong>and</strong> Development<br />
of an Amphibious Robot<br />
With Fin Actuators<br />
A previously designed mechanical<br />
pectoral fin (Kato & Liu, 2003) has a<br />
drag-based swimming mode <strong>and</strong> a liftbased<br />
swimming mode. The former is<br />
characterized by the rowing action<br />
forming a high angle to the horizontal<br />
axis of the body, while the latter is<br />
characterized by the flapping action<br />
forming a small angle to the horizontal<br />
axis. It was revealed through the optimization<br />
of motion of the mechanical<br />
pectoral fin that the lift-based swimming<br />
mode rather than the dragbased<br />
swimming mode is suitable for<br />
generation of propulsive force in uniform<br />
flow, while the drag-based swimming<br />
mode rather than the lift-based<br />
swimming mode is suitable for generation<br />
of propulsive force in still water.<br />
To realize both the drag-based swimming<br />
mode <strong>and</strong> the lift-based swimming<br />
mode, a combination of rowing<br />
motion, flapping motion, <strong>and</strong> feathering<br />
motion is needed. The hydrodynamic<br />
characteristics of the drag-based<br />
swimming mode <strong>and</strong> the lift-based<br />
swimming mode are discussed in details<br />
elsewhere (Suzuki et al., 2007).<br />
Walking locomotion using fin actuators<br />
is of two types, imitating the motionofatortoise<strong>and</strong>thatofasea<br />
turtle. Sea turtle-like walking motion<br />
has the following characteristics:<br />
It is dynamically stable.<br />
Terrain condition strongly affects<br />
the attitude of the body.<br />
Friction between the soil <strong>and</strong> the<br />
body necessitates additional thrust<br />
force.<br />
Tortoise-like walking motion has the<br />
following characteristics:<br />
It is dynamically more unstable<br />
compared with the sea turtle-like<br />
walking motion.<br />
The attitude of the body has more<br />
DOF compared with the sea turtlelike<br />
walking motion.<br />
The body weight is vertically applied<br />
to the foot.<br />
In this study, we adopted the<br />
tortoise-like walking motion, which<br />
has better trafficability on soil than<br />
the sea turtle-like walking motion, as<br />
discussed in the previous section.<br />
FIGURE 18<br />
Components of Robotic Turtle (RT-I) <strong>and</strong> fin actuator with 4 DOF.<br />
The amphibious robot we designed<br />
consists of the following four main<br />
components (see Figure 18):<br />
(1) leg units,<br />
(2) a control unit in a pressure hull,<br />
(3) a buoyancy adjusting device in a<br />
pressure hull, <strong>and</strong><br />
(4) a fairing cover<br />
Each pressure hull was designed to resist<br />
the pressure at the water depth of<br />
10 m. We used “Solidworks” for 3-D<br />
CAD software to design the hardware<br />
<strong>and</strong> “LabVIEW” to develop<br />
the control software. Four DOFs are<br />
needed to realize not only three<br />
types of swimming motion but also<br />
the tortoise-like walking motion.<br />
Therefore, three motors were set up<br />
at the acromioclavicular joint using a<br />
differential gear mechanism to independently<br />
produce flapping motion,<br />
rowing motion, <strong>and</strong> feathering motion,<br />
<strong>and</strong> one motor was set up at<br />
190 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
the elbow joint to produce bending<br />
motion of the forearm (see Figure 18).<br />
Motors <strong>and</strong> reduction gears were<br />
selected according to the simulation<br />
results on walking. An open dynamic<br />
engine of the kinetics calculation library<br />
with open sources was used for<br />
the simulation. A fuselage with the dimensions<br />
(W × H × L = 0.50 × 0.20 ×<br />
0.80 m) was used. The length of the<br />
humerus was set at 0.25 m, <strong>and</strong> the<br />
length of the forearm was set at<br />
0.15 m. There are various walking<br />
gaits for quadrupedal locomotion.<br />
We selected a crawling gait for low<br />
speed where the legs are lifted up one<br />
by one. The fin actuators of this<br />
robot have two functions, walking <strong>and</strong><br />
swimming, so we had to design the<br />
shape of the fin withbothwalking<br />
<strong>and</strong> swimming in mind. For swimming,<br />
it is desirable to have a large<br />
fin area. However, the fin mayinterfere<br />
with the base of the leg unit<br />
during walking. The fin shapeisdesigned<br />
to minimize the interference.<br />
Figure 19 shows the form of the fin<br />
with the root chord of 0.195 m, the<br />
span of 0.15 m, <strong>and</strong> the maximum<br />
thickness of 0.32 m.<br />
Figure 20 shows the outline of<br />
the control unit’s electric circuit. The<br />
control unit consists of a CPU,<br />
motor drivers, an azimuth sensor,<br />
FIGURE 19<br />
Form of fin.<br />
FIGURE 20<br />
Electric circuit of the control unit.<br />
three-axes rate gyros, a pressure sensor,<br />
a GPS, <strong>and</strong> related minor parts. Batteries<br />
are used separately for motors<br />
<strong>and</strong> the CPU with sensors. Nickel<br />
hydride batteries with the capacity of<br />
13.2 V, 8.0 Ah are used for motors<br />
by arranging a series circuit of two<br />
<strong>and</strong> a parallel circuit of two. It is also<br />
possible to provide the control unit<br />
with electric power from outside.<br />
The amphibious robot has to realize<br />
walking in which the buoyancy is<br />
less than the weight <strong>and</strong> swimming<br />
in which the buoyancy is equal to or<br />
greater than the weight, because the<br />
mission of the robot includes travel<br />
in shallow water with breaking waves.<br />
To realize these capabilities, the robot<br />
must be equipped with a buoyancy<br />
adjusting device. The buoyancy adjusting<br />
device with the buoyancy capacity<br />
of ±0.35 kg was designed using<br />
a pair of pistons arranged symmetrically<br />
in the longitudinal direction so<br />
as not to have an effect on the attitude<br />
of the robot.<br />
It is desirable to cover the body<br />
of the robot with streamlined fairing.<br />
The general-purpose computer fluid<br />
dynamics software “FLUENT” was<br />
used to estimate the hydrodynamic<br />
drag on the fairing cover <strong>and</strong> to design<br />
a fairing cover with low hydrodynamic<br />
drag. The cover was made of glass<br />
fiber-reinforced plastics.<br />
Figure 21 shows a photograph of<br />
the amphibious robot equipped with<br />
fin actuators, named “RT-I.” The<br />
principal dimensions are shown in<br />
Table 1.<br />
July/August 2011 Volume 45 Number 4 191
FIGURE 21<br />
Photograph of the RT-I amphibious robot<br />
equipped with fin actuators.<br />
TABLE 1<br />
Specification of RT-I.<br />
Total mass [kg] (without buoyancy control device <strong>and</strong> battery) 90<br />
Depth of pressure resistant [m] 10<br />
Walking speed in the water [m/s] (experimental value) 0.025<br />
Swimming speed [m/s] (experimental value) 0.168<br />
Dimension [m] Body width 0.73<br />
Body height 0.55<br />
Body length 1.68<br />
Forearm length 0.26<br />
Humerus length 0.2<br />
Swimming <strong>and</strong> Walking<br />
Performance by the<br />
Robotic Turtle RT-I<br />
Swimming Performance<br />
A towing tank test was carried out<br />
to measure drag forces acting on the<br />
body of the robotic turtle in the towing<br />
tank of Osaka University. The forces<br />
applied to the body were measured<br />
by the force sensor at various towing<br />
speeds. From this experiment, the relationship<br />
between the drag forces <strong>and</strong><br />
the fluid velocities was obtained to estimate<br />
the thrust forces produced by<br />
the fins. The swimming speeds in<br />
free swimming condition of the robotic<br />
turtle were measured in the<br />
same towing tank for the estimation<br />
of the swimming performance. The<br />
lift-based swimming mode was<br />
adopted in the experiments. It uses<br />
the flapping motion <strong>and</strong> the feathering<br />
motion in horizontal plane of the<br />
robot; on the other h<strong>and</strong>, it uses the<br />
rowing motion <strong>and</strong> the feathering motion<br />
in vertical motion of the robot.<br />
The speeds of the towing carriage<br />
were determined by operating it parallel<br />
to the robot for several values of<br />
amplitudes of flapping motion <strong>and</strong><br />
feathering motion. From the experimental<br />
result of the swimming speeds,<br />
the thrust forces were estimated by<br />
using the relationship between drag<br />
forces <strong>and</strong> swimming speeds. Figure<br />
22 shows the thrust forces for the<br />
amplitudes of feathering motion of<br />
30° <strong>and</strong> 45° <strong>and</strong> the amplitudes of flapping<br />
motion of 20° <strong>and</strong> 30°. From this<br />
figure, we can see that the thrust force<br />
increases as the amplitude of the<br />
flapping motion increases. Figure 23<br />
shows the thrust forces for amplitudes<br />
of a flapping motion of 30° versus amplitudes<br />
of the feathering motion.<br />
From this figure, we can see that the<br />
thrust force becomes the largest at the<br />
FIGURE 22<br />
amplitude of the feathering motion of<br />
30°. Judging from these results, the<br />
largest thrust force is made at the amplitudes<br />
of flapping motion <strong>and</strong> feathering<br />
motion of 30°.<br />
The swimming speeds of forward<br />
motion, backward motion, turning<br />
motion, <strong>and</strong> vertically descending motion<br />
were measured as performances of<br />
the maneuverability of the robotic turtle.<br />
Table 2 shows the swimming<br />
speeds of these motions with fixed amplitudes<br />
of joint angle. Unlike the forward<br />
motion, the swimming speed in<br />
backward motion with the amplitudes<br />
of feathering motion of 45° is larger<br />
than the case of 30°.<br />
Thrust forces in swimming motion against amplitudes of flapping motion <strong>and</strong> feathering motion.<br />
192 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 23<br />
Thrust forces in swimming motion against amplitudes of feathering motion.<br />
Walking Performance<br />
in Still Water<br />
TABLE 2<br />
The swimming speeds for various swimming motions.<br />
The crawl gait was adopted as the<br />
walking gait of the robotic turtle because<br />
it has a high level of stability.<br />
During the walking motion using the<br />
crawl gait, the robot moves each arm<br />
individually. This means the robot<br />
supports the body with at least three<br />
arms at all times. In operation of the<br />
robot in sea water, it is supposed that<br />
the seabed is not flat <strong>and</strong> there exist<br />
disturbances such as waves <strong>and</strong> water<br />
currents. Therefore, the robotic turtle<br />
needs more stability under such conditions.<br />
For that reason, we introduced a<br />
modified walking motion by considering<br />
the movement of the center of<br />
gravity in the normal crawl gait, as<br />
showninFigure24.Themovement<br />
of the center of gravity makes the stability<br />
margin larger, which is defined as<br />
the distance from the side of the support<br />
polygon to the projection of the<br />
center of gravity on l<strong>and</strong>, as shown in<br />
Figure 25.<br />
Two types of walking patterns were<br />
made. The first walking pattern consists<br />
of four steps to move an arm, for<br />
a total of 16 steps to move four arms<br />
during a cycle. The trajectory of the<br />
end of an arm in this pattern forms a<br />
rectangle. The second walking pattern<br />
consists of three steps to move an arm,<br />
for a total of 12 steps to move four<br />
Swimming Pattern<br />
Flapping (Rowing)<br />
Amplitude (°)<br />
Feathering<br />
Amplitude (°)<br />
Swimming<br />
Speed<br />
Forward motion 30 30 0.168<br />
Forward motion 30 45 0.15<br />
Backward motion 30 30 0.088 m/s<br />
Backward motion 30 45 0.102 m/s<br />
Turning motion 30 30 9.33°/s<br />
Turning motion 30 45 10.30°/s<br />
Descending motion 30 (rowing) 45 0.05 m/s<br />
arms during a cycle (Figure 26). The<br />
trajectory of the end of the arm of<br />
the second walking pattern forms a<br />
right triangle. In this case, it is expected<br />
that walking will be faster because of<br />
the reduction in the number of walking<br />
steps <strong>and</strong> that the energy consumption<br />
will be smaller because of the<br />
reduction of total motor speed.<br />
The walking speed <strong>and</strong> the motor<br />
currents for each walking pattern<br />
were measured while the robot walked<br />
on the bottom of the pool (L × B × H =<br />
4.5 m × 2.0 m × 0.8 m) (see Figure 27)<br />
to evaluate the walking performances<br />
in still water. Here the length of the<br />
stride of each walking pattern was set<br />
at 0.2 m. The joint torque was derived<br />
from the value of the motor current<br />
measured by the experiment, the<br />
torque constant, the reduction ratio,<br />
<strong>and</strong> the transmission efficiency, as<br />
showninEq.(12).Theconsumption<br />
energy was derived from the<br />
value of the derived joint torque, as<br />
follows:<br />
q ij<br />
¼ I ij ⋅K t ⋅i⋅η′<br />
ð12Þ<br />
Here, I ij denotes the motor current of<br />
the joint P ij ,K t is the torque constant,<br />
i is the reduction ratio, <strong>and</strong> η′ is the<br />
transmission efficiency. Figure 28<br />
shows a comparison of walking speeds<br />
between measurement <strong>and</strong> simulation<br />
by the walking simulator for the two<br />
walking patterns. From this figure,<br />
we can see that the walking speed for<br />
the second walking pattern is greater.<br />
This is attributed to the small period<br />
of the second walking pattern for one<br />
cycle because of the reduction in walking<br />
steps. We can see that the experimental<br />
value of the walking speed for<br />
each walking pattern is smaller than<br />
the calculated value. This is because<br />
the end of the arm slips on the bottom<br />
July/August 2011 Volume 45 Number 4 193
FIGURE 24<br />
Schematic view of crawl gait <strong>and</strong> modified crawl gait.<br />
FIGURE 27<br />
Picture of Robotic Turtle in walking in a pool.<br />
FIGURE 25<br />
Definition of stability margin.<br />
FIGURE 26<br />
Trajectories of the arm end of the first <strong>and</strong> second walking patterns.<br />
of the pool when the robot walks in<br />
still water.<br />
Walking Performance in Waves<br />
We estimated the walking performance<br />
of the robotic turtle in waves<br />
through a series of experiments. Figure<br />
29 shows a schematic view of the<br />
experiment for estimating the walking<br />
performance in waves. Two types of<br />
experiments were carried out in the<br />
towing tank of Osaka University.<br />
The first type measured the walking<br />
performance by changing the walking<br />
patterns in waves <strong>and</strong> in still water.<br />
The second type measured the walking<br />
performances of the second walking<br />
pattern in waves by changing the<br />
wave conditions, namely, the wave<br />
height, the wavelength, <strong>and</strong> the<br />
depth of the virtual ground. For the<br />
first type of experiment, the walking<br />
speeds of the first <strong>and</strong> second normal<br />
walking patterns were measured for<br />
thestridelengths(L)of0.2m<strong>and</strong><br />
0.3 m under the wave height of 0.1 m,<br />
the wavelength of 2 m, <strong>and</strong> the water<br />
depth of 0.9 m. The walking speeds<br />
of the first <strong>and</strong> second modified walking<br />
patterns were also measured. Figure<br />
30 shows a comparison of the<br />
walking speeds for various walking patterns<br />
in waves <strong>and</strong> in still water. From<br />
this figure, we can see that the walking<br />
speeds for all cases decreased in waves<br />
compared to the performance in still<br />
water. The walking speed reaches its<br />
maximum value in the case of the<br />
length of the stride of 0.3 m <strong>and</strong> the<br />
second walking pattern, regardless of<br />
crawl gait. Figure 31 shows a comparison<br />
of the consumption energy for<br />
various walking patterns in still water<br />
<strong>and</strong> in waves. From this figure, we<br />
can see that the consumption energy<br />
of the first walking pattern in waves<br />
is larger than that in still water in the<br />
caseofthelengthofthestrideof<br />
194 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 28<br />
Comparison of walking speed in water.<br />
FIGURE 29<br />
Schematic view of experiment for walking performance in waves.<br />
FIGURE 30<br />
Comparison of walking speeds for various walking patterns in still water <strong>and</strong> in waves.<br />
0.3 m, although that of the second<br />
walking pattern in waves is smaller<br />
than that in still water. Regarding<br />
that the walking speeds of the first<br />
<strong>and</strong> second walking patters in waves<br />
become smaller than those in still<br />
water, we find that the propulsive efficiency<br />
of the first walking pattern in<br />
waves becomes worse than that of the<br />
second walking pattern in waves in<br />
the case of the length of the stride of<br />
0.3 m. This may be attributed to the<br />
interaction between the hydrodynamic<br />
forces acting on the arms <strong>and</strong> the paths<br />
of the arms, namely, larger energy is<br />
consumed in the first walking pattern<br />
during one cycle of the arm motion<br />
in waves by the hydrodynamic forces<br />
than in the second walking.<br />
Figure 32 shows the variation of<br />
walking speed against the wave heights<br />
from 0 m to 0.15 m with the wavelength<br />
of 2 m, the water depths of<br />
0.6m<strong>and</strong>0.9m,<strong>and</strong>normal<strong>and</strong><br />
crawl modified gaits for the second<br />
walking pattern with the length of<br />
the stride of 0.3 m. From this figure,<br />
we can see that the walking speed is decreased<br />
as the wave height increases<br />
<strong>and</strong> that the walking speed becomes<br />
smaller <strong>and</strong> the rates of the decrease become<br />
larger as the water depth decreases<br />
from 0.9 m to 0.6 m. Figure 33 shows<br />
the variation of consumption energy<br />
against the wave heights under the<br />
same condition for Figure 30. From<br />
this figure,wecanseethattheconsumption<br />
energy is almost constant<br />
against the wave height, although the<br />
consumption energies are smaller<br />
than those in still water, <strong>and</strong> that the<br />
propulsive efficiency in waves become<br />
worse if the water depth decreases because<br />
the rate of decrease of walking<br />
speed for d = 0.6m is larger than for<br />
d = 0.9 m. This is attributed to larger<br />
water current induced by waves in<br />
shallow water, which produces larger<br />
July/August 2011 Volume 45 Number 4 195
FIGURE 31<br />
Comparison of consumption energy for various walking patterns in still water <strong>and</strong> in waves.<br />
FIGURE 32<br />
Variation of walking speed against wave height.<br />
hydrodynamic drag force on the<br />
vehicle.<br />
Summary<br />
In this paper we discussed the<br />
mechanisms of swimming <strong>and</strong> walking<br />
of turtles <strong>and</strong> examined how those<br />
mechanisms could be applied to an<br />
amphibious robot, with the goal of<br />
designing a robot that can perform<br />
automatic monitoring of environments<br />
along natural coastal areas <strong>and</strong><br />
tidal flats. Using a manipulator with<br />
4 DOF, we discussed the characteristics<br />
of walking of sea turtles <strong>and</strong> tortoises<br />
from the viewpoints of mobility<br />
<strong>and</strong> trafficability. The advantages of<br />
sea turtles in swimming <strong>and</strong> tortoises<br />
in walking were adopted in the robotic<br />
turtle. The experiments in a water tank<br />
revealed that the mobility <strong>and</strong> the trafficability<br />
of the amphibious robot at<br />
the bottom of the water were greatly<br />
affected by waves in shallow water.<br />
The author will investigate the tracking<br />
performance of the amphibious<br />
robot on tidal flats, moving from<br />
l<strong>and</strong> to sea <strong>and</strong> vice versa.<br />
FIGURE 33<br />
Variation of consumption energy against wave height.<br />
Acknowledgments<br />
This research was funded for<br />
3 years beginning in 2008 by the Ministry<br />
of Education, Culture, Sports <strong>and</strong><br />
<strong>Technology</strong>, Japan (grant 20246130)<br />
under the project title “Establishment<br />
of basic technology of a biomimetic<br />
underwater vehicle <strong>and</strong> its applications.”<br />
Author:<br />
Naomi Kato<br />
Graduate School of<br />
Engineering, Osaka University<br />
Yamadaoka 2-1, Suita,<br />
Osaka 565-0871, Japan<br />
Email: kato@naoe.eng.osaka-u.ac.jp<br />
196 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
References<br />
Dudek, G., Giguere, P., Prahacs, C.,<br />
Saunderson, S., Sattar, J., Torres-Mendez,<br />
L.-A., … Georgiades, C. 2007. Aqua: An<br />
amphibious autonomous robot. IEEE<br />
Comput Mag. 40(1):46-53.<br />
Ijspeert, A.J., Crespi, A., Ryczko, D., &<br />
Cabelguen, J-M. 2007. From swimming to<br />
walking with a salam<strong>and</strong>er robot driven by<br />
a spinal cord model. Science. 315(5817):<br />
1416-20. doi: 10.1126/science.1138353.<br />
Isobe, Y., Kato, N., Suzuku, H., Senga, H.,<br />
Matsuda, Y., Kawamura, K., … Oshika, T.<br />
2010. Motion analysis of sea turtle with<br />
prosthetic flippers. In: The 20th (2010) Int.<br />
Offshore <strong>and</strong> Polar Eng. Conf. pp. 335-42.<br />
Beijing: International <strong>Society</strong> of Polar <strong>and</strong><br />
Offshore Eng.<br />
Kato, N., Ando, Y., Shigetomi, T., &<br />
Katayama, T. 2006. Biology-inspired precision<br />
maneuvering of underwater vehicles (Part 4).<br />
Int J Offshore Polar. 16(3):195-201.<br />
Kato, N., & Liu, H. 2003. Optimization<br />
of motion of a mechanical pectoral fin.<br />
JSME Int J C. 47(4):1356-62. doi: 10.1299/<br />
jsmec.46.1356.<br />
amphibious walking robot—The design<br />
concepts <strong>and</strong> the 1st field experiment.<br />
In: paper presented at the International<br />
Symposium on Automation <strong>and</strong> Robotics<br />
in Construction 2006 (ISARC 2006). 46-51.<br />
Tokyo: IEEE.<br />
Wakimoto, S., Suzumori, K., & K<strong>and</strong>a, T.<br />
2006. A bio-mimetic amphibious soft cord<br />
robot. JSME Int J C. 72(714):171-7.<br />
Walker, W.F., Jr. 1971. A structual <strong>and</strong><br />
functional analysis of walking in the turtle<br />
Chrysemys picta marginata. J Morphol.<br />
134:195-213. doi: 10.1002/jmor.1051340205.<br />
Wyneken, J. 1997. Sea turtle locomotion:<br />
Mechanisms, behavior, <strong>and</strong> energetics. In:<br />
The Biology of Sea Turtles, eds. Lutz, P.L.,<br />
& Musick, J.A., 165-98. New York: CRC<br />
Press.<br />
Yamada, H., Chigisaki, S., Mori, M., Takita,<br />
K., Ogami, K., & Hirose, S. 2005. Development<br />
of amphibious snake-like robot ACM-R5.<br />
In: paper presented at the 36th International<br />
Symposium on Robotics(ISR 2005). Tokyo:<br />
Japan Robot Association.<br />
Low, K.H., Zhou, C., Ong, T.W., & Yu, J.<br />
2007. Modular design <strong>and</strong> initial gait study<br />
of an amphibious robotic turtle. In: paper<br />
presented at the International Conference<br />
on Robotics <strong>and</strong> <strong>Biomimetics</strong>, 2007. 535-40.<br />
Sanya, China: IEEE.<br />
Setouchi, H., & Shinjo, T. 2001. The effect<br />
of water on frictional characteristics between<br />
calcareous s<strong>and</strong>s <strong>and</strong> a steel plate. In: Report<br />
of Faculty of Agriculture. University of the<br />
Ryukyus. 48:103-11.<br />
Suzuki, H., & Kato, N. 2005. A numerical<br />
study on unsteady flow around a mechanical<br />
pectoral fin. Int J Offshore Polar. 15(3):161-7.<br />
Suzuki, H., Kato, N., & Suzumori, K. 2007.<br />
Load characteristics of mechanical pectoral<br />
fin. Exp Fluids. 44(5):759-71. doi: 10.1007/<br />
s00348-007-0432-x.<br />
Tanaka, T., & Shirai, T. 2006. Development<br />
of automated shoreline surveying system using<br />
July/August 2011 Volume 45 Number 4 197
PAPER<br />
<strong>Marine</strong> Applications of the Biomimetic<br />
Humpback Whale Flipper<br />
AUTHORS<br />
Frank E. Fish<br />
West Chester University<br />
Paul W. Weber<br />
Applied Research Associates, Inc.<br />
Mark M. Murray<br />
United States Naval Academy<br />
Laurens E. Howle<br />
Duke University<br />
Introduction<br />
Life began with water. The highdensity<br />
<strong>and</strong> viscous nature of water<br />
has imposed a strong evolutionary selection<br />
pressure on the design of animals<br />
that move through this aqueous<br />
medium. Over the course of millions<br />
of years, different phylogenetic lines<br />
of animals have, in effect, experimented<br />
with various combinations<br />
of morphologies <strong>and</strong> behaviors to enhance<br />
locomotor performance. The<br />
great diversity of body shapes, surface<br />
textures, <strong>and</strong> propulsive mechanisms<br />
exhibited by aquatic animals has produced<br />
a variety of biomechanical solutions<br />
for the reduction of drag, increase<br />
in thrust production <strong>and</strong> efficiency,<br />
maintenance of stability, <strong>and</strong> enhancement<br />
of maneuverability. By emulating<br />
these biological characteristics in<br />
those instances where animal performance<br />
is superior to manufactured devices,theperformanceofengineered<br />
marine systems may be improved<br />
through the field of biomimetics.<br />
The cetaceans (whales, dolphins,<br />
porpoises) have been the focus of inspiration<br />
for technological development in<br />
ABSTRACT<br />
The biomimetic approach seeks technological advancement through a transfer<br />
of technology from natural technologies to engineered systems. The morphology of<br />
the wing-like flipper of the humpback whale has potential for marine applications.<br />
As opposed to the straight leading edge of conventional hydrofoils, the humpback<br />
whale flipper has a number of sinusoid-like rounded bumps, called tubercles, which<br />
are arranged periodically along the leading edge. The presence of the tubercles<br />
modifies the water flow over the wing-like surface, creating regions of vortex generation<br />
between the tubercles. These vortices interact with the flow over the tubercle<br />
<strong>and</strong> accelerate that flow, helping to maintain a partially attached boundary layer.<br />
This hydrodynamic effect can delay stall to higher angles of attack, increases lift,<br />
<strong>and</strong> reduces drag compared to the post-stall condition of conventional wings. As<br />
the humpback whale functions in the marine environment in a Reynolds regime<br />
similar to some engineered marine systems, the use of tubercles has the potential<br />
to enhance the performance of wing-like structures. Specific applications of the tubercles<br />
for marine technology include sailboat masts, fans, propellers, turbines, <strong>and</strong><br />
control surfaces, such as rudders, dive planes, stabilizers, spoilers, <strong>and</strong> keels.<br />
Keywords: tubercles, delayed stall, Megaptera novaeangliae, leading edge,<br />
bio-inspired design<br />
the marine environment. The cetacean<br />
lineage dates back 55 million years.<br />
The intense selection pressures for a<br />
fast swimming, maneuverable marine<br />
predator have culminated in a highly<br />
streamlined body with advanced sensory<br />
capabilities that is propelled by a<br />
highly efficient propulsion mechanism.<br />
There are a number of examples<br />
where cetaceans have been the inspiration<br />
for the development or improvements<br />
of marine technology. The<br />
cetacean body shape was used by<br />
Cayley (circa 1800) as a solid of leastresistance<br />
for the development of airplane<br />
fuselage <strong>and</strong> boat hull designs<br />
(Gibbs-Smith, 1962). The famous<br />
but erroneous “Gray’s paradox” led<br />
to examination of special drag reduction<br />
mechanisms (Gray, 1936; Fish<br />
& Rohr, 1999; Fish, 2006), including<br />
the biomimetic development of compliant<br />
coatings for viscous dampening<br />
(Kramer, 1960; Riley et al., 1988;<br />
Carpenter & Pedley, 2003). The compactness<br />
<strong>and</strong> high resolution of the<br />
echolocation system of dolphins provides<br />
a benchmark for the improvement<br />
of SONAR systems (Au, 1993).<br />
The thrust performance of oscillating,<br />
wing-like systems, such as the flukes of<br />
dolphins, has been considered superior<br />
to screw propellers (Peterson, 1925;<br />
Liu & Bose, 1993; Triantafyllou &<br />
Triantafyllou, 1995). The flexibility<br />
of the oscillatory flukes can allow operation<br />
at high efficiency over an extended<br />
speed range without cavitation<br />
(Fish & Lauder, 2006; Iosilevskii &<br />
Weihs, 2007).<br />
This report focuses on a unique morphology<br />
of a highly derived aquatic<br />
198 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
mammal that has general biomimetic<br />
applications for marine systems. The<br />
flipper of the humpback whale <strong>and</strong><br />
its bumpy leading edge provide a<br />
novel approach to enhance the hydrodynamic<br />
performance of wing-like<br />
structures for operation in water. A<br />
principal attribute of using the whale<br />
as a model to construct a biomimetic<br />
system is the scale. As the whale is of<br />
a large size <strong>and</strong> swims at speeds that<br />
compliment the scale <strong>and</strong> operation<br />
of engineered marine systems, application<br />
for marine technologies can be<br />
readily undertaken.<br />
Humpback Whale<br />
as Inspiration<br />
The humpback whale (Megaptera<br />
novaeangliae) hasthelongestflipper<br />
of any cetacean (i.e., whale, dolphin,<br />
porpoise), with regard to both absolute<br />
<strong>and</strong> relative size (Figure 1; Fish<br />
& Battle, 1995). The flippers are involved<br />
with the underwater maneuvers<br />
performed by the species that are associated<br />
with their mode of feeding<br />
(Friedlaender et al., 2009; Hazen<br />
et al., 2009). Humpback whales are<br />
the only baleen whales (e.g., blue<br />
whale, fin whale, minke whale, right<br />
whale) that rely on tight, rapid turns<br />
to capture prey (Fish & Battle,<br />
1995). The humpback whales use<br />
their flippers as biological hydroplanes<br />
to achieve tight circles to corral <strong>and</strong><br />
engulf prey (Fish et al., 2011).<br />
FIGURE 1<br />
Flippers on humpback whale (left), showing<br />
scalloped pattern of tubercles (center), <strong>and</strong> flipper<br />
cross-section (right) (from Fish & Lauder,<br />
2006).<br />
The humpback whale flippers are<br />
unique because of the presence of<br />
large tubercles along the leading edge,<br />
which gives this surface a scalloped<br />
appearance (Figure 1). The distances<br />
between tubercles decrease distally,<br />
although these distances remain relatively<br />
constant at 7-9% of span over<br />
the midspan of the flipper (Fish &<br />
Battle, 1995).<br />
The planform <strong>and</strong> cross-sectional<br />
views of the humpback whale flipper<br />
are shown in Figure 1. Whereas typical<br />
wing-like structures have a straight<br />
leading edge without the presence of<br />
irregularities or perturbations, the<br />
humpback flipper defies convention<br />
with prominent rounded bumps that<br />
are regularly spaced along the leading<br />
edge of its high-aspect ratio flipper<br />
(Fish et al., 2008). Humpback<br />
whale flippers closely resemble the<br />
21% thick, low drag NACA 63 4 -021<br />
foil in cross section (Abbott & von<br />
Doenhoff, 1959; Fish & Battle,<br />
1995). Furthermore, the flippers have<br />
high mobility (Edel & Winn, 1978).<br />
The elongate flippers function as<br />
wings to generate the forces necessary<br />
for turning maneuvers (Fish et al.,<br />
2011). Turning is important in the<br />
capture of elusive prey. Humpback<br />
whales feed on shoals of small fish<br />
<strong>and</strong> krill. The preys are forced into a<br />
tight ball by the whales striking the<br />
water surface with their flukes or circling<br />
the prey from underneath while<br />
emitting bubbles. The bubbles rise to<br />
corral the prey as a bubble net. In either<br />
case, the whale maneuvers under<br />
the prey for engulfment by executing<br />
a rapid turn. Lift generated by the flippers<br />
is used to produce a centripetal<br />
force for the turn. The tubercles produce<br />
vortical flows over the surface of<br />
the flipper <strong>and</strong> control lift characteristics<br />
at high angles of attack <strong>and</strong> delay<br />
stall (Fish & Battle, 1995; Miklosovic<br />
et al., 2004; Fish & Lauder, 2006; Fish<br />
et al., 2011).<br />
Hydrodynamic Effect<br />
of Tubercles<br />
The prominence of the tubercles<br />
on the leading edge of the humpback<br />
whale flippers <strong>and</strong> the swimming pattern<br />
of the whale suggests that these<br />
novel structures have a distinct hydrodynamic<br />
effect (Figure 2). In addition,<br />
the pattern of barnacle distribution on<br />
the flippers (i.e., barnacles are confined<br />
to the peak of the tubercle <strong>and</strong> do<br />
not occur between tubercles; Fish &<br />
Battle, 1995) indicates that the flow<br />
over the flipper is affected by tubercles.<br />
This section reviews potential hydrodynamic<br />
advantages, flow control,<br />
<strong>and</strong> limitations of the tubercles on<br />
wing-like structures.<br />
Hydrodynamic Advantages<br />
The presence of leading-edge tubercles<br />
on a wing-like structure can have<br />
a positive influence on the hydrodynamic<br />
performance. Wind tunnel<br />
tests showed that wings with tubercles<br />
improved maximum lift by over 6%,<br />
increased the ultimate stall angle by<br />
40%, <strong>and</strong> decreased drag by as much<br />
as 32% (Figure 3; Miklosovic et al.,<br />
2004). The tubercles, when facing<br />
FIGURE 2<br />
Idealized humpback flipper with leading-edge<br />
tubercles.<br />
July/August 2011 Volume 45 Number 4 199
FIGURE 3<br />
Humpback whale flipper models <strong>and</strong> results of wind tunnel experiments. The models (left) with<br />
<strong>and</strong> without tubercles were machined from clear polycarbonate, based on a symmetrical NACA<br />
0020 foil section. Lift <strong>and</strong> drag data (right) for the flipper models were obtained from tests in a<br />
wind tunnel. The solid lines in a, b, <strong>and</strong> c show the average of the data for the flipper model without<br />
tubercles, <strong>and</strong> open triangles are for the model with tubercles. Lift coefficient C L (a), drag coefficient<br />
C D (b), <strong>and</strong> aerodynamic efficiency L/D (c) are plotted against angle of attack, α (from<br />
Miklosovic et al., 2004).<br />
intothefreestreamflow, alter the<br />
fluid flow over wing-like structures<br />
(Bushnell & Moore, 1991; Fish &<br />
Battle, 1995; Fish et al., 2011). Furthermore,<br />
the lift to drag (L/D) ratio,<br />
which represents the aerodynamic efficiency,<br />
displayed a greater peak L/D<br />
for a wing geometry with tubercles<br />
(Miklosovic et al., 2004, 2007;<br />
Hansen et al., 2009).<br />
The position <strong>and</strong> number of tubercles<br />
on the flipper suggest analogues<br />
with specialized leading edge control<br />
devices that improve the hydrodynamic<br />
performance of wings. The<br />
occurrence of “morphological complexities”<br />
on a lifting body could reduce<br />
or use pressure variation at the tip to decrease<br />
drag <strong>and</strong> improve lift to prevent<br />
tip stall (Bushnell & Moore, 1991). Alternatively,<br />
various biological wings<br />
utilize leading-edge control devices to<br />
maintain lift <strong>and</strong> avoid stall at high attack<br />
angles <strong>and</strong> low speeds.<br />
The function of the tubercles may<br />
be analogous to strakes used on aircraft.<br />
Strakes are large vortex generators<br />
that change the stall characteristics<br />
of a wing (Hoerner, 1965;<br />
Shevell, 1986; Bertin & Smith,<br />
1998). Stall is postponed because the<br />
vortices exchange momentum within<br />
the boundary layer to keep it partially<br />
attached over the wing surface. Lift is<br />
thus maintained at higher angles of attack<br />
with strakes compared to wings<br />
without strakes, although maximum<br />
lift is not increased by strakes (Shevell,<br />
1986). Another leading-edge device are<br />
slots or slats that delay stall or move<br />
the angle of attack of maximum lift<br />
to a lower value (Wegener, 1991;<br />
Bertin & Smith, 1998). The moveable<br />
slats create a space anterior of the fixed<br />
wing to allow higher-pressure fluid to<br />
rise from the underside of the wing.<br />
The movement of fluid from the<br />
high-pressure side to the low-pressure<br />
side of the wing improves mixing <strong>and</strong><br />
helps to maintain the boundary layer<br />
<strong>and</strong> delay stall (Wegener, 1991). However,<br />
the tubercles have distinct advantages<br />
over slats. Tubercles are<br />
passive structures that have no drag<br />
penalty when designed onto wings<br />
(Miklosovicetal.,2004),whereas<br />
slats are actively deployed <strong>and</strong> incur increased<br />
drag (Hoerner, 1965).<br />
Vortex Generation<br />
The mechanism for enhanced hydrodynamic<br />
performance due to the<br />
presence of tubercles appears due to<br />
the specific pattern of vortex generation<br />
over the surface of the flipper.<br />
Flow visualization experiments on<br />
wavy bluff bodies showed periodic<br />
variation in the wake width across the<br />
span (Owen et al., 2000). A wide wake<br />
with two simultaneous vortices occurred<br />
where the body protruded<br />
downstream <strong>and</strong> a narrow wake occurred<br />
where the body protruded upstream.<br />
The flowinthewakeofthe<br />
wavy body was different from the<br />
wake of a straight cylinder, which exhibited<br />
a typical von Karman Vortex<br />
Street of alternating vortex pairs. A<br />
bluff body with a spanwise sinusoidal<br />
form could reduce drag by at least<br />
30%, compared to equivalent straight<br />
bodies (Bearman & Owen, 1998).<br />
The vortices produced by a wing<br />
section with tubercles are shown in<br />
Figure 4. The tubercles generate<br />
separated, chordwise vortices in the<br />
troughs at high angles of attack.<br />
These vortices are formed as the flow<br />
strikes the leading edge of the trough.<br />
As the flow does not strike the leading<br />
edge normally, the flow is sheared into<br />
the trough’s center to generate the<br />
vortices. These vortices are convected<br />
along the chord. The spanwise arrangement<br />
of the vortices is in a pair<br />
on each side of the tubercle crest with<br />
opposite spins (Hansen et al., 2010).<br />
The flow directly over the tubercle interacts<br />
with the vortices located downstream<br />
<strong>and</strong> lateral to the tubercle crest.<br />
200 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 4<br />
Pressure contours <strong>and</strong> streamlines at α = 10° for NACA 63-021 with straight leading edge (left) <strong>and</strong><br />
with tubercles (right). An unsteady Reynolds-averaged Navier-Stokes (RANS) simulation was<br />
used. A separation line is shown on the wing section without tubercles. For the wing with tubercles,<br />
large vortices are formed posterior of the troughs along the leading edge <strong>and</strong> flow posterior of<br />
the tubercles is shown as straight streamlines without separation. Images courtesy of E. Paterson.<br />
The tangential velocities of the inward<br />
facing flows of the pair of vortices are<br />
directed toward the trailing edge of<br />
the wing section. The flow from the<br />
tubercle peak is accelerated posteriorly<br />
due to the interaction with the vortex<br />
pair. These effects prevent the local<br />
boundary layer downstream of the tubercles<br />
from separating <strong>and</strong> push the<br />
stall line further posterior toward the<br />
trailing edge. When integrated over<br />
the entire structure, the wing with tubercles<br />
will stall at a higher angle of<br />
attack than a wing without tubercles.<br />
Flow experiments conducted on a<br />
model wing section with leading-edge<br />
tubercles at low speeds showed flow<br />
separation from the troughs between<br />
adjacent tubercles but attached flow<br />
on the tubercles (Johari et al., 2007).<br />
Flow separation pattern <strong>and</strong> surface<br />
pressure was dramatically altered by<br />
the tubercles. For regions downstream<br />
of tubercle crest, separation was delayed<br />
almost to the trailing edge. Although<br />
this flow pattern did not<br />
result in improved lift generation,<br />
drag reduction or delay in the stall<br />
angle of attack, the post-stall characteristics<br />
were greatly smoothed (Johari<br />
et al., 2007; Saadat et al., 2010),<br />
which could provide important performance<br />
benefits for systems that routinely<br />
operate beyond the stall point.<br />
The vortices produced from the<br />
tubercles re-energize the boundary<br />
layer by carrying high-momentum<br />
flow close to the flipper’s surface<br />
(Figure 5; Wu et al., 1991; Pedro &<br />
FIGURE 5<br />
Kobayashi, 2008; Hansen et al.,<br />
2010). The flow dynamics are improved<br />
also by confining separation<br />
to the tip region. Tubercles delay stall<br />
by causing a greater portion of the flow<br />
to remain attached on a wing, with the<br />
attached flow localized behind the tubercle<br />
crests (Weber et al., 2011).<br />
Thesize<strong>and</strong>frequencyofthetubercles<br />
along the leading edge influences<br />
the performance of a wing.<br />
Small amplitude tubercles show the<br />
best performance with regard to lift<br />
<strong>and</strong> stall characteristics (Johari et al.,<br />
2007; Hansen et al., 2009, 2011).<br />
The wavelength <strong>and</strong> thus frequency<br />
of tubercles was found, however, to<br />
have little effect on performance<br />
(Johari et al., 2007).<br />
The tubercle effect is further enhanced<br />
when sweepback is added to<br />
the wings (Murray et al., 2005).<br />
Wings with sweep angles of 15° <strong>and</strong><br />
30°requiredhigheranglesofattack<br />
to achieve stall than nonswept wings<br />
<strong>and</strong> showed superior drag performance<br />
over most of the range of α compared<br />
to models without tubercles (Murray<br />
et al., 2005). Flow tests on delta<br />
Vorticity computed from detached eddy simulation (DES) for flippers with (a) <strong>and</strong> without (b) tubercles<br />
at an angle of attack of 15°. Vortices re-energize boundary layer to delay separation <strong>and</strong><br />
stall. Images courtesy of H. T. C. Pedro <strong>and</strong> M. H. Kobayashi.<br />
July/August 2011 Volume 45 Number 4 201
wings with a sweep of 50° showed that<br />
at high angles of attack large-scale<br />
three-dimensional separation occurred<br />
for the wing with a straight<br />
leading edge (Goruney & Rockwell,<br />
2009). However when tubercles are<br />
added, the flow is radically transformed.<br />
Tubercles with amplitude of<br />
4% of wing chord can completely<br />
eradicate the negative effect of the<br />
separation <strong>and</strong> foster re-attachment.<br />
Experiments performed on flapping<br />
wings with tubercles (Figure 6)<br />
showed an affect on the spanwise<br />
flow (Ozen & Rockwell, 2010). Typically<br />
a straight wing, whether flapping<br />
FIGURE 6<br />
or static, will develop a spanwise flow<br />
due to the pressure differential that develops<br />
between the upper <strong>and</strong> lower<br />
surfaces. This spanwise flow becomes<br />
manifest as a wing tip vortex, which increases<br />
the drag <strong>and</strong> reduces the efficiency<br />
of a wing. A flapping wing<br />
with tubercles does not produce a pronounced<br />
region of spanwise flow, but<br />
the wing tip vortex generation is unaffected<br />
(Ozen & Rockwell, 2010).<br />
Limitations of Tubercles<br />
Enhanced performance due to the<br />
presence of the tubercles is not universal.<br />
There exist limitations to the<br />
Comparison of flapping plate at Reynolds number of 1300 at angle of attack of 8° with <strong>and</strong> without<br />
tubercles from Ozen <strong>and</strong> Rockwell (2010). The pattern of the flow structure for the plate with tubercles<br />
produces a series of spanwise vortices that limit spanwise flow compared with the flapping<br />
plate without tubercles. Image courtesy of D. Rockwell.<br />
advantages of the tubercles. Tubercles<br />
improve performance when in concert<br />
with wing geometries that are characterized<br />
by the combination of finite<br />
span, swept wing, tapered planform,<br />
<strong>and</strong> thick foil.<br />
Foil sections with no wing tip that<br />
emulate infinite wings do not demonstrate<br />
reduced drag <strong>and</strong> increased lift, although<br />
stall is still delayed (Miklosovic<br />
et al., 2004; Johari et al., 2007; Van<br />
Nierop et al., 2008). Tip effects occur<br />
as a consequence of lift generation<br />
when a fully three-dimensional wing<br />
is canted at an angle of attack to an incident<br />
flow. Induced drag is produced<br />
in lift generation from kinetic energy<br />
imparted to the fluid from pressure differences<br />
between the two surfaces of<br />
the wing as there is leakage of fluid<br />
from high pressure to low pressure<br />
around the distal tip of a lifting surface<br />
resulting in spanwise flow <strong>and</strong> the formation<br />
of tip vortices (Vogel, 1981).<br />
The flow pattern set up by the tubercles<br />
helps to maintain a chordwise<br />
flow <strong>and</strong> reduce the induced drag due<br />
to tip vortices. This effect is only realized<br />
for finite wings.<br />
The tubercles require wings with<br />
thick sectional geometries to function.<br />
The section must have a prominent<br />
nose radius. In part, this is due to the<br />
necessity to contour the tubercles threedimensionally<br />
into the leading edge<br />
to avoid any flat surfaces. Leadingedge<br />
tubercles were tested on foils<br />
based on NACA 0021, 63 4 -021 <strong>and</strong><br />
65-021 sections (Miklosovic et al.,<br />
2004; Johari et al., 2007; Custodio<br />
et al., 2010; Hansen et al., 2009,<br />
2011). These designs approach the<br />
cross-sectional geometry of the humpback<br />
whale flipper (Fish & Battle,<br />
1995).<br />
It is necessary to have a relatively<br />
steady flow to maintain the pattern of<br />
thevortices<strong>and</strong>incurthehydrodynamic<br />
202 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
advantages (Stanway, 2008). Foils<br />
with tubercles that were oscillated in<br />
roll <strong>and</strong> pitch demonstrated that the<br />
tubercles did not improve hydrodynamic<br />
performance (Stanway, 2008).<br />
Flapping degraded performance by<br />
the redirection of energy to tuberclegenerated<br />
vortices from the vortices<br />
of the wake, which are necessary for<br />
thrust production during flapping.<br />
Furthermore, if the period of oscillations<br />
is too rapid, there may be<br />
insufficient time to allow the full development<br />
of the vortices over a wing. It is<br />
necessary to have a relatively steady<br />
flow to maintain the pattern of the<br />
vortices <strong>and</strong> incur the hydrodynamic<br />
advantages (Stanway, 2008).<br />
Tubercle Technologies<br />
Application of natural technologies<br />
into biomimetic-engineered systems<br />
has a number of problems that are inherent<br />
due to differences in biological<br />
<strong>and</strong> engineered systems (Fish, 2006).<br />
Engineered systems are relatively large<br />
in size, are composed of dry rigid<br />
materials, including metals <strong>and</strong> ceramics,<br />
use rotation motors, <strong>and</strong> are<br />
controlled by computational systems<br />
with limited sensory feedback. Biological<br />
structures associated with animal<br />
systems are relatively small in size, are<br />
composed of wet compliant materials,<br />
including composites of ceramics,<br />
polysaccharides <strong>and</strong> proteins, move<br />
by translational displacements generated<br />
by muscles, <strong>and</strong> are controlled<br />
by complex neural networks with multiple<br />
sensory inputs <strong>and</strong> fine scale<br />
motor outputs.<br />
The humpback whale tubercles<br />
provide an ideal solution for application<br />
to engineered marine systems.<br />
The size of the whale <strong>and</strong> its flippers<br />
operate near or at the same scale as<br />
some marine vehicles. The mature<br />
whales have a maximum length of<br />
17 m <strong>and</strong> weigh 40,000 kg (Clapham<br />
& Mead, 1999). Flipper length is approximately<br />
one-third the length of<br />
the animal <strong>and</strong> can be over 5 m. The<br />
whale cruises between 1.1 <strong>and</strong> 4.0 m/s<br />
<strong>and</strong> is able to burst to a speed of 7.5 m/s<br />
(Fish & Rohr, 1999). The flow experienced<br />
by <strong>and</strong> modified by the tubercles<br />
is within the same Reynolds regime<br />
that coincides with a large array of<br />
engineered applications. The Reynolds<br />
number of a flipper is 1.6 × 10 6 when<br />
the whale is lunge feeding (2.6 m/s).<br />
Thus, the flipper <strong>and</strong> tubercles are operating<br />
in a turbulent flow regime,<br />
which is the st<strong>and</strong>ard operating condition<br />
for most marine systems. Furthermore,<br />
the tubercle functions passively<br />
to modify flow <strong>and</strong> maintain favorable<br />
hydrodynamic conditions. Therefore,<br />
control systems can be simplified.<br />
Control Surfaces<br />
There are few other passive means<br />
of altering fluid flow around a winglike<br />
structure that can delay stall <strong>and</strong><br />
both increase lift <strong>and</strong> reduce drag at<br />
the same time. As a result, the application<br />
of leading-edge tubercles for passive<br />
flow control has potential in the<br />
design of marine technologies. The<br />
ubiquity of wing-like structures with<br />
marine applications for stability <strong>and</strong><br />
maneuverability presents an opportunity<br />
to enlist tubercles to improve performance.<br />
Included in such structures<br />
are fixed surfaces, such as keels, fins<br />
<strong>and</strong> skegs, <strong>and</strong> mobile control surfaces,<br />
such as rudders <strong>and</strong> dive planes.<br />
Delay of stall by tubercles on both<br />
fixed <strong>and</strong> mobile control surfaces<br />
provides a benefit in tight turning situations.<br />
To produce the required centripetal<br />
force to effect a turn, a lift force<br />
that is directed toward the center of the<br />
turn is generated by the control surface.<br />
The magnitude of the lift is<br />
directly dependent on the angle of attack<br />
with a higher angle providing a<br />
greater centripetal force. As stall can<br />
be delayed with tubercles, the control<br />
surface can operate at higher angles of<br />
attack producing a tighter turn radius<br />
with more control. If stall were to<br />
occur, the control surface would not<br />
be able to generate the centripetal<br />
force to maintain the turn. In effect,<br />
itwouldbelikedrivingacaralonga<br />
curved road <strong>and</strong> slipping on a patch<br />
of ice. The reduced friction <strong>and</strong> centripetal<br />
force between the ice <strong>and</strong> the<br />
tires would cause the car to drive off<br />
the road tangential to the curve, rather<br />
than following the original curved<br />
trajectory.<br />
A low-aspect ratio rudder with<br />
tubercles (Figure 7) <strong>and</strong> an unswept<br />
leading edge generated more lift at angles<br />
of attack above 22° compared to a<br />
smooth rudder at a Reynolds number<br />
of 200,000 (Weber et al., 2010). At<br />
higher Reynolds numbers, this effect<br />
diminishes <strong>and</strong> the tubercles accelerate<br />
the onset of cavitation. A humanpowered<br />
submarine, Umpty Squash,<br />
utilized tubercled dive planes <strong>and</strong> rudders<br />
(Figure 8). Students of the Sussex<br />
FIGURE 7<br />
Rudder with leading-edge tubercles.<br />
July/August 2011 Volume 45 Number 4 203
FIGURE 8<br />
Human-powered submarine (left) with rudder <strong>and</strong> dive planes with tubercles (right). Courtesy of<br />
Chris L<strong>and</strong> <strong>and</strong> the Sussex County Technical School.<br />
FIGURE 11<br />
Iceboat with leading-edge tubercles on the<br />
mast supporting a sail.<br />
County Technical High School,<br />
Sparta, NJ, constructed the submarine.<br />
In 2005, the submarine competed<br />
in the International Submarine<br />
RacesheldattheDavidTaylor<br />
Model Basin in Bethesda, MD. The<br />
submarine was capable of making a<br />
90° turn within 25 feet. The designers<br />
at Feadship De Voogt developed<br />
Breathe, a concept superyacht that<br />
incorporates biomimicry into the<br />
design (www.feadship.nl). The<br />
stabilizers <strong>and</strong> steering fins were<br />
based on the humpback whale flipper<br />
with tubercles (Figure 9).<br />
Commercially, the company Fluid<br />
Earth markets a surfboard skeg with<br />
leading-edge tubercles (Figure 10;<br />
Anders, 2009). The addition of tubercles<br />
would provide enhanced control<br />
during a cutback, when a surfer rapidly<br />
changes direction by 180° to maneuver<br />
the surfboard opposite to the direction<br />
of the wave’s braking motion.<br />
Application of tubercles to the mast<br />
of a sailboat (Figure 11) could be<br />
useful during close reach maneuvers.<br />
A close reach would have the sail set<br />
FIGURE 10<br />
Surfboard skeg with leading-edge tubercles.<br />
with a high angle of attack to the<br />
apparent wind. The lack of a thick<br />
cross-section by the sail itself may preclude<br />
any advantage in lift <strong>and</strong> stall.<br />
However, the presence of tubercles<br />
on the mast, representing a bluff<br />
body, could have advantages in terms<br />
of drag. Bluff bodies, like cylinders,<br />
FIGURE 9<br />
Conceptual yacht incorporating biomimetic<br />
structures. Stabilizers are shaped like the flippers<br />
of the humpback whale. Image courtesy<br />
of Feadship.<br />
204 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 12<br />
Two views of marine propeller designs with<br />
tubercles.<br />
can experience lowered drag when the<br />
leading edge has a sinusoidal design<br />
(Bearman & Owen, 1998).<br />
Propellers <strong>and</strong> Turbines<br />
The use of tubercles can effectively<br />
be employed in the generation of<br />
power by wing-like structures. Propellers<br />
operating in a marine system have<br />
the potential to be improved by the<br />
addition of tubercles (Figure 12).<br />
The effective angle of attack of a propeller<br />
blade can be increased by increasing<br />
the blade angle (Larrabee,<br />
1980). A higher angle of attack can<br />
produce more lift to derive greater<br />
thrust <strong>and</strong> increase the effective pitch<br />
of the propeller. Prevention of stall<br />
by the tubercle effect would reduce<br />
flow-induced vibrations. Furthermore,<br />
suppression of tonal noise is possible<br />
by the addition of tubercles to a propeller<br />
(Hansen et al., 2010). Tonal<br />
noise is most effectively reduced by<br />
large amplitude <strong>and</strong> smaller wavelength<br />
tubercles. As propellers produce<br />
a particular noise signature, reduction<br />
of noise would be advantageous for<br />
stealth in naval operations. Similarly,<br />
a reduction of noise pollution by commercial<br />
marine traffic could be beneficial<br />
to marine organisms, although the<br />
use of radiated noise by whales to avoid<br />
collisions with ships may be negatively<br />
impacted.<br />
Tubercle modified blades were also<br />
found to be effective in power generationofamarinetidalturbineatlow<br />
flow speeds (Murray et al., 2010). Tubercles<br />
were placed on the distal 40%<br />
of the three turbine blades. Compared<br />
to blades with smooth leading edges,<br />
blades with leading-edge tubercles<br />
demonstrated enhanced performance.<br />
The marine tidal turbine is analogous<br />
to wind turbines. A variable pitch<br />
wind turbine with retrofitted blades<br />
with tubercles demonstrated increased<br />
electrical generation at moderate wind<br />
speeds compared to unmodified blades<br />
(Howle, 2009; Wind Energy Institute<br />
of Canada, 2008).<br />
Conclusions<br />
A passive means of altering fluid<br />
flow around a wing-like structure that<br />
can delay stall <strong>and</strong> both increase lift<br />
<strong>and</strong> reduce drag simultaneously is<br />
highly novel. This performance by<br />
leading-edge tubercles, therefore, has<br />
potential application for passive flow<br />
control in the design of various marine<br />
technologies. The flow is modified by<br />
the formation of paired vortices in the<br />
troughs between tubercles. These vortices<br />
interact with the flow over the tubercles<br />
to keep the flow attached to the<br />
wing surface <strong>and</strong> delay stall. The tubercles<br />
perform best when designed into<br />
tapered wings with finite span, swept<br />
planform, <strong>and</strong> with a thick foil section<br />
that have limited oscillatory movement.<br />
Such applications for marine technology<br />
include fins, rudders, dive planes,<br />
water turbines <strong>and</strong> propellers. The fusion<br />
of these marine devices <strong>and</strong> tubercles<br />
can produce biomimetic designs<br />
that can exhibit superior performance<br />
to conventional engineered systems.<br />
Acknowledgments<br />
We thank the technical support<br />
staff of the United States Naval<br />
Academy. This work was supported<br />
by the National Science Foundation<br />
(IOS-0640185) to FEF <strong>and</strong> the National<br />
Defense Science <strong>and</strong> Engineering<br />
Graduate (NDSEG) Fellowship<br />
to PWW.<br />
Lead Author:<br />
Frank E. Fish<br />
Department of Biology<br />
West Chester University<br />
West Chester, PA 19383<br />
Email: ffish@wcupa.edu<br />
References<br />
Abbott, I.H., & von Doenhoff, A.E. 1959.<br />
Theory of Wing Sections. New York: Dover.<br />
Anders, M. 2009. Fins I like. Surfer’s Path,<br />
April/May 2009:116.<br />
Au, W.W.L. 1993. The Sonar of Dolphins.<br />
New York: Springer-Verlag.<br />
Bearman, P.W., & Owen, J.C. 1998.<br />
Reduction of bluff body drag <strong>and</strong> suppression<br />
of vortex shedding by the introduction of<br />
wavy separation lines. J Fluid Struct.<br />
12:123-30. doi: 10.1006/jfls.1997.0128.<br />
Bertin, J.J., & Smith, M.L. 1998. Aerodynamics<br />
for Engineers. Upper Saddle River,<br />
NJ: Prentice Hall.<br />
Bushnell, D.M., & Moore, K.J. 1991. Drag<br />
reduction in nature. Annu Rev Fluid Mech.<br />
23:65-79. doi: 10.1146/annurev.<br />
fl.23.010191.000433.<br />
Carpenter, P.W., & Pedley, T.J. 2003.<br />
IUTAM Symposium on Flow Past Highly<br />
Compliant Boundaries <strong>and</strong> in Collapsible<br />
Tubes. Dordrecht: Kluwer.<br />
Clapham, P.J., & Mead, J.G. 1999.<br />
Megaptera novaeangliae. Mamm Spec.<br />
604:1-9.<br />
Custodio, D., Henoch, C., & Johari, H.<br />
2010. The performance of finite-span<br />
hydrofoils with humpback whale-like leading<br />
edge protuberances. 63 rd Ann Mtg APS Div<br />
July/August 2011 Volume 45 Number 4 205
Fluid Dyn, 55 (16), Nov. 21, 2010, Long<br />
Beach, CA.<br />
Edel, R.K., & Winn, H.E. 1978. Observations<br />
on underwater locomotion <strong>and</strong> flipper<br />
movement of the humpback whale Megaptera<br />
novaeangliae. Mar Biol. 48:279-87.<br />
doi: 10.1007/BF00397155.<br />
Fish, F.E. 2006. The myth <strong>and</strong> reality of<br />
Gray’s paradox: Implication of dolphin drag<br />
reduction for technology. Bioinspir Biomim.<br />
1:R17-25. doi: 10.1088/1748-3182/1/2/R01.<br />
Fish, F.E., & Battle, J.M. 1995. Hydrodynamic<br />
design of the humpback whale flipper.<br />
J Morphol. 225:51-60. doi: 10.1002/<br />
jmor.1052250105.<br />
Fish, F.E., Howle, L.E., & Murray, M.M.<br />
2008. Hydrodynamic flow control in marine<br />
mammals. Integr Comp Biol. 211:1859-67.<br />
Fish, F.E., & Lauder, G.V. 2006. Passive<br />
<strong>and</strong> active flow control by swimming fishes<br />
<strong>and</strong> mammals. Annu Rev Fluid Mech.<br />
38:193-224. doi: 10.1146/annurev.<br />
fluid.38.050304.092201.<br />
Fish, F.E., & Rohr, J. 1999. Review of<br />
Dolphin Hydrodynamics <strong>and</strong> Swimming<br />
Performance. San Diego, CA: SPAWARS<br />
TR 1801.<br />
Fish, F.E., Weber, P.W., Murray, M.M., &<br />
Howle, L.E. 2011. The humpback whale’s<br />
flipper: Application of bio-inspired tubercle<br />
technology. Integr Comp Biol. 51:203-13.<br />
doi: 10.1093/icb/icr016.<br />
Friedlaender, A.S., Hazen, E.L., Nowacek,<br />
D.P., Halpin, P.N., Ware, C., Weinrich,<br />
M.T., … Wiley, D. 2009. Diel changes in<br />
humpback whale Megaptera novaeangliae<br />
feeding behavior in response to s<strong>and</strong> lance<br />
Ammodytes spp. behavior <strong>and</strong> distribution.<br />
Mar Ecol Prog Ser. 395:91-100. doi:<br />
10.3354/meps08003.<br />
Gibbs-Smith, C.H. 1962. Sir George<br />
Cayley’s Aeronautics 1796-1855. London:<br />
Her Majesty’s Stationery Office.<br />
Goruney, T., & Rockwell, D. 2009.<br />
Flow past a delta wing with a sinusoidal<br />
leading edge: Near-surface topology <strong>and</strong> flow<br />
structure. Exp Fluids. 47:321-31.<br />
doi: 10.1007/s00348-009-0666-x.<br />
Gray, J. 1936. Studies in animal locomotion<br />
VI. The propulsive powers of the dolphin.<br />
J Exp Biol. 13:192-9.<br />
Hansen, K.L., Kelso, R.M., & Dally, B.B.<br />
2009. The effect of leading edge tubercle<br />
geometry on the performance of different<br />
airfoils. In: 7th World Conf Exp Heat Transfer,<br />
Fluid Mechanics <strong>and</strong> Thermodynamics.<br />
28 June-03 July 2009. Krakow, Pol<strong>and</strong>.<br />
Hansen, K.L., Kelso, R.M., & Dally, B.B.<br />
2011. Performance variations of leading-edge<br />
tubercles for distinct airfoil profiles. AIAA<br />
J Aircraft. 49:185-94.<br />
Hansen, K.L., Kelso, R.M., & Doolan, C.J.<br />
2010. Reduction of flow induced tonal noise<br />
through leading edge tubercle modifications.<br />
In: 16th AIAA/CEAS Aeroacoustics Conf.<br />
7-9 June 2010. Stockholm, Sweden.<br />
Hazen, E.L., Friedlaender, A.S., Thompson,<br />
M.A., Ware, C.R., Weinrich, M.T., Halpin,<br />
P.N., & Wiley, D.N. 2009. Fine-scale<br />
prey aggregations <strong>and</strong> foraging ecology of<br />
humpback whales Megaptera novaeangliae.<br />
Mar Ecol Prog Ser. 395:75-89.<br />
doi: 10.3354/meps08108.<br />
Hoerner, S.F. 1965. Fluid-dynamic Drag.<br />
Published by author. Brick Town, NJ.<br />
Howle, L.E. 2009. WhalePower Wenvor<br />
blade. A report in the efficiency of a WhalePower<br />
Corp. 5 meter prototype wind turbine blade.<br />
BelleQuant Engineering, PLLC.<br />
Iosilevskii, G., & Weihs, D. 2007. Speed<br />
limits on swimming of fishes <strong>and</strong> cetaceans.<br />
J R Soc Interface. 5:329-38. doi: 10.1098/<br />
rsif.2007.1073.<br />
Johari, H., Henoch, C., Custodio, D., &<br />
Levshin, A. 2007. Effects of leading-edge<br />
protuberances on airfoil performance. AIAA<br />
J. 45:2634-42. doi: 10.2514/1.28497.<br />
Kramer, M.O. 1960. The dolphin’s secret.<br />
New Sci. 7:1118-20.<br />
Larrabee, E.E. 1980. The screw propeller.<br />
Sci Am. 243:134-48.<br />
Liu, P., & Bose, N. 1993. Propulsive performance<br />
of three naturally occurring oscillating<br />
propeller planforms. Ocean Eng. 20:57-75.<br />
doi: 10.1016/0029-8018(93)90046-K.<br />
Miklosovic, D.S., Murray, M.M., &<br />
Howle, L.E. 2007. Experimental evaluation of<br />
sinusoidal leading edges. J Aircraft. 44:1404-7.<br />
doi: 10.2514/1.30303.<br />
Miklosovic, D.S., Murray, M.M., Howle,<br />
L.E., & Fish, F.E. 2004. Leading-edge<br />
tubercles delay stall on humpback whale<br />
(Megaptera novaeangliae) flippers. Phys<br />
Fluids. 16:L39-42. doi: 10.1063/1.1688341.<br />
Murray, M.M., Fish, F.E., Howle, L.E., &<br />
Miklosovic, D.S. 2005. Stall delay by leading<br />
edge tubercles on humpback whale flipper<br />
at various sweep angles. In 14 th Int Symp<br />
Unmanned Untethered Submersible Tech.<br />
Durham New Hampshire: Autonomous<br />
Undersea Systems Insitute.<br />
Murray, M., Gruber, T., & Fredriksson, D.<br />
2010. Effect of leading edge tubercles on<br />
marine tidal turbine blades. 63 rd Ann Mtg<br />
APS Div Fluid Dyn, 55(16), Nov. 21, 2010,<br />
Long Beach, CA.<br />
Owen, J.C., Szewczyk, A.A., & Bearman,<br />
P.W. 2000. Suppression of Karman vortex<br />
shedding. Phys Fluids. 12:S9.<br />
Ozen, C.A., & Rockwell, D. 2010. Control<br />
of vortical structures on a flapping wing<br />
via a sinusoidal leading-edge. Phys Fluids.<br />
22:021701. doi: 10.1063/1.3304539.<br />
Pedro, H.T.C., & Kobayashi, M.H.<br />
2008. Numerical study of stall delay on<br />
humpback whale flippers. In: 46th AIAA<br />
Aerospace Mtg Exhibit. 7-10 Jan. 2008.<br />
Reno, Nevada.<br />
Peterson, C.J.G. 1925. The motion of<br />
whales during swimming. Nature. 116:327-9.<br />
doi: 10.1038/116327a0.<br />
Riley, J.J., Gad-el-Hak, M., & Metcalfe,<br />
R.W. 1988. Compliant coatings. Annu Rev<br />
Fluid Mech. 20:393-420. doi: 10.1146/<br />
annurev.fl.20.010188.002141.<br />
Saadat, M., Haj-Hariri, H., & Fish, F. 2010.<br />
Explanation of the effects of leading-edge<br />
206 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
tubercles on the aerodynamics of airfoils<br />
<strong>and</strong> finite wings. 63 rd Ann Mtg APS Div<br />
Fluid Dyn, 55(16), Nov 21, 2010, Long<br />
Beach, CA.<br />
Shevell, R.S. 1986. Aerodynamic anomalies:<br />
Can CFD prevent or correct them J Aircraft.<br />
23:641-9. doi: 10.2514/3.45356.<br />
Stanway, M.J. 2008. Hydrodynamic effects<br />
of leading-edge tubercles on control surfaces<br />
<strong>and</strong> in flapping foil propulsion. M.S. thesis,<br />
Massachusetts Institute of <strong>Technology</strong>.<br />
Triantafyllou, G.S., & Triantafyllou, M.S.<br />
1995. An efficient swimming machine.<br />
Sci Amer. 272:64-70. doi: 10.1038/<br />
scientificamerican0395-64.<br />
Van Nierop, E.A., Alben, S., & Brenner,<br />
M.P. 2008. How bumps on whale flippers<br />
delay stall: An aerodynamic model. Phys<br />
Rev Lett. 100:054502. doi: 10.1103/<br />
PhysRevLett.100.054502.<br />
Vogel, S. 1981. Life in Moving Fluids.<br />
Boston: Willard Grant Press.<br />
Weber, P.W., Howle, L.E., & Murray, M.M.<br />
2010. Lift, drag, <strong>and</strong> caviation onset on<br />
rudders with leading-edge tubercles. <strong>Marine</strong><br />
Technol Sname N. 47(1):26.<br />
Weber, P.W., Howle, L.E., Murray, M.M.,<br />
& Miklosovic, D.S. 2011. Computational<br />
evaluation of the performance of lifting<br />
surfaces with leading edge protuberances.<br />
J Aircraft. 48:591-600. doi: 10.2514/<br />
1.C031163.<br />
Wegener, P.P. 1991. What Makes Airplanes<br />
Fly New York: Springer-Verlag.<br />
Wind Energy Institute of Canada. 2008.<br />
WhalePower Tubercle Blade Power Performance<br />
Test Report. North Cape, PE:<br />
Wind Energy Institute of Canada.<br />
Wu, J.Z., Vakili, A.D., & Wu, J.M. 1991.<br />
Review of the physics of enhancing<br />
vortex lift by unsteady excitation.<br />
Prog Aerosp Sci. 28:73-131. doi: 10.1016/<br />
0376-0421(91)90001-K.<br />
July/August 2011 Volume 45 Number 4 207
PAPER<br />
Shark Skin Separation Control Mechanisms<br />
AUTHORS<br />
Amy Lang<br />
University of Alabama<br />
Philip Motta<br />
Maria Laura Habegger<br />
University of South Florida<br />
Robert Hueter<br />
Mote <strong>Marine</strong> Laboratory<br />
Farhana Afroz<br />
University of Alabama<br />
Introduction<br />
Natural selection, operating for<br />
hundreds of millions of years, has<br />
honed fast swimming marine organism<br />
forms to reduce energy expenditure<br />
through streamlining. Among<br />
the fastest swimming fishes, certain<br />
species of sharks have converged on<br />
a suite of adaptations to reduce drag.<br />
Consequently, fast swimming sharks<br />
have been studied for insight into potential<br />
drag-reducing mechanisms for<br />
well over three decades. Drag on a<br />
submerged, swimming body consists<br />
of three sources. These include (i) form<br />
drag due to a difference in pressure<br />
around the body, (ii) drag due to lift,<br />
<strong>and</strong> (iii) skin friction due to boundary<br />
layer formation (Bushnell & Moore,<br />
1991). At low Reynolds numbers<br />
(Re) skin friction predominates, while<br />
at higher Re pressure drag can dominate<br />
if not minimized. It is not surprising<br />
then to consider the fact that<br />
aquatic organisms have evolved to<br />
minimize drag (Fish, 1998), with<br />
the primary decrease coming from<br />
a streamlined body shape to reduce<br />
flow separation <strong>and</strong> thus form<br />
drag. Aquatic organisms that swim at<br />
ABSTRACT<br />
Drag reduction by marine organisms has undergone millions of years of natural<br />
selection, <strong>and</strong> from these organisms biomimetic studies can derive new technologies.<br />
The shortfin mako (Isurus oxyrinchus), considered to be one of the fastest <strong>and</strong><br />
most agile marine predators, is known to have highly flexible scales on certain locations<br />
of its body. This scale flexibility is theorized to provide a passive, flow-actuated<br />
mechanism for controlling flow separation <strong>and</strong> thereby decreasing drag. Recent biological<br />
observations have found that the shortfin mako has highly flexible scales,<br />
bristling to angles in excess of 50°, particularly on the sides of the body downstream<br />
of the gills. High “contragility,” which is explicitly definedhereastheabilityto<br />
change or move in a new or opposing direction while already in a turn, would<br />
occur if form drag were minimized. This would thus indicate the potential control<br />
of flow separation on body regions aft of the point of maximum girth or in regions of<br />
adverse pressure gradient. Thus results are consistent with the hypothesis that<br />
scale bristling controls flow separation. This scale flexibility appears to be a result<br />
of a reduction in the relative size of the base of the scales as well as a reorganization<br />
of the base shape as evidenced by histological examination of the<br />
skin <strong>and</strong> scales. Probable mechanisms leading to separation control are discussed.<br />
Keywords: shark skin, flow separation, drag reduction<br />
high Re (>10 3 )haveavarietyof<br />
shapes <strong>and</strong> structures to reduce drag,<br />
which we often attempt to duplicate<br />
(Vogel, 2003; Fish & Lauder, 2006).<br />
It has been deduced that a crescent tail<br />
design could decrease induced drag on<br />
the order of 8%, <strong>and</strong> not surprisingly<br />
this lunate tail design is found on<br />
many fast swimming marine animals<br />
(Fish, 1998; Donley et al., 2004).<br />
Several researchers (e.g., Anderson<br />
et al., 2001) have observed that swimming<br />
fish experience more friction<br />
drag than the same rigid body<br />
towed. This higher friction is attributed<br />
to motion of the body as it<br />
swims to produce thrust. The undulating<br />
body motion can result in measurably<br />
thinner boundary layers <strong>and</strong><br />
thus higher skin friction (Fish, 2006).<br />
The argument has been made that<br />
because of this higher power output<br />
requirement, to overcome drag <strong>and</strong><br />
maintain a certain speed, swimming<br />
drag reduction due to various morphological<br />
mechanisms is extremely<br />
probable (Schultz & Webb, 2002).<br />
Separation of the boundary layer<br />
from a body typically occurs in vicinities<br />
where the flow is decelerating<br />
along a curved body after the point of<br />
maximum thickness, resulting in an<br />
adverse pressure gradient. As a result<br />
separation typically occurs in areas<br />
posterior of the maximum body thickness.<br />
Incipient separation is characterized<br />
by regions of decreasing skin<br />
friction approaching zero, <strong>and</strong> consequent<br />
reversal of the flow at the surface<br />
(Doligalskietal.,1994).Swimming<br />
kinematics in thunniform fish such as<br />
tunas <strong>and</strong> mako sharks are characterized<br />
by cyclically repeating motions <strong>and</strong><br />
small linear <strong>and</strong> angular accelerations<br />
(Blake, 2004). Most fast swimming<br />
sharks, such as the shortfin mako Isurus<br />
208 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
oxyrinchus, are thunniform swimmers<br />
where oscillations are for the most<br />
part limited to the posterior end of<br />
the body. It has been reported that incipient<br />
separation (inflected boundary<br />
layer profile) is often observed during<br />
swimming movements, <strong>and</strong> this motion<br />
may be tuned by the fish to take<br />
advantage of the lower shear stress in<br />
a nearly separating boundary layer;<br />
yet separation must be avoided to reduce<br />
form drag <strong>and</strong> increase thrust<br />
produced by the caudal fin (Anderson<br />
et al., 2001).<br />
The skin of sharks is covered by<br />
minute scales, called dermal denticles<br />
or placoid scales, which originally<br />
evolved as a hard, protective covering<br />
to the animal (Raschi & Tabit,<br />
1992). The bases of these hard scales<br />
are embedded in the superficial collagenous<br />
layer of the skin (dermis) termed<br />
the stratum laxum, with the crowns<br />
of the scales exposed to the water. It is<br />
the unique geometry observed on fast<br />
swimming sharks for these tooth-like<br />
scales that is of interest from a drag reduction<br />
st<strong>and</strong>point. Beginning in the<br />
1970s researchers became interested<br />
in the small, streamwise ridges,<br />
or keels, located on the top of each<br />
crown (see scanning electron microscopy<br />
(SEM) pictures of shark scales<br />
shown in Figure 1). Now labeled as<br />
riblets, these ridges were found to result<br />
in a reduction in turbulent skin<br />
friction drag of up to 9.9% when<br />
sized correctly (Bechert et al., 1997).<br />
Even as early as the 1980s, riblets<br />
were utilized on boat hulls competing<br />
in the Olympics <strong>and</strong> America’s Cup<br />
but later were banned (Gad-el Hak,<br />
2000).<br />
However, the scales on some fast<br />
swimming sharks exhibit a different<br />
property that is flexibility or capability<br />
to bristle, <strong>and</strong> this is the focus of our<br />
current work. Previous work (Bechert<br />
FIGURE 1<br />
The average scale erection angles for 16 regions on the shortfin mako shark (Isurus oxyrinchus)<br />
together with scanning electron micrographs (top row) <strong>and</strong> histological sections (middle row) of<br />
the skin for three regions on the body. The flank region, including B2, has the most flexible scales,<br />
which are characterized by long backwardly projecting crowns <strong>and</strong> relatively short bases. The scale<br />
bases are embedded in the superficial part of the dermis. The surface of the scales has three riblets<br />
visible on the scanning electron micrographs. The scales in any region are oriented along the longitudinal<br />
axis of the shark, <strong>and</strong> anterior is to the left. The erection angles noted on the figure are the<br />
angles to which the scales in that region can be manually manipulated without damage to their<br />
attachment <strong>and</strong> which remain at that angle after release by the needle used to erect them.<br />
et al., 2000) investigating this aspect of<br />
the shark skin found no advantage<br />
from a skin friction reduction st<strong>and</strong>point,<br />
<strong>and</strong> only greatly increased friction<br />
drag if the scales were allowed to<br />
remain bristled. Thus, a new, passive<br />
flow-actuated separation control<br />
mechanism is proposed that is inspired<br />
by the scale flexibility found on the<br />
shortfin mako shark.<br />
This investigation chose to focus<br />
on the skin of the shortfin mako<br />
based on several factors. First, of fast<br />
swimming pelagic sharks, the shortfin<br />
mako is considered by most to be the<br />
fastest <strong>and</strong> most agile (Stevens, 2009).<br />
It is also one of the more derived<br />
species of shark <strong>and</strong> is recognized as<br />
making its appearance roughly 55 million<br />
years ago in the line of shark<br />
evolution dating back more than<br />
400 million years (Naylor et al.,<br />
1997). Next, it is one of two species<br />
of shark previously reported in literature,<br />
the other being the smooth hammerhead<br />
Sphyrna zygaena, ashaving<br />
July/August 2011 Volume 45 Number 4 209
flexible scales over large portions of its<br />
body (Bruse et al., 1993). Finally, the<br />
shortfin mako is readily obtainable off<br />
the Atlantic coast of the United States<br />
<strong>and</strong> is not currently listed as overfished<br />
or otherwise in a vulnerable state<br />
for conservation purposes (NMFS,<br />
2010).<br />
Materials <strong>and</strong> Methods<br />
To investigate scale structure <strong>and</strong><br />
erection, we acquired two subadult<br />
shortfin makos (female: total length,<br />
192 cm; fork length, 171.5 cm; male:<br />
total length, 158 cm; fork length,<br />
150 cm) from commercial <strong>and</strong> recreational<br />
fishers in the coastal waters off<br />
Montauk, New York. The frozen<br />
specimens were shipped to the University<br />
of South Florida in Tampa. There,<br />
following Reif (1985), scales at 16 regions<br />
along the body were marked in<br />
order to sample the scales under a<br />
variety of flow regimes (Figure 1).<br />
Three 1 cm 2 samples from each location<br />
were removed, two for histological<br />
analysis <strong>and</strong> one for SEM. Because<br />
swimming sharks have superambient<br />
subcutaneous pressure ranging as<br />
high as 100-200 kPa increasing skin<br />
stiffness (Wainwright et al., 1978;<br />
Martinez et al., 2002), scale erection<br />
angles were recorded with <strong>and</strong> without<br />
subcutaneous pressure, as follows.<br />
We measured scales erection angle<br />
under a dissecting microscope at a<br />
magnification of 135×. In seven of<br />
the 16 regions (B1, B2, B4, B5, A1,<br />
A2, A3; Figure 1), an aneroid sphygmomanometer<br />
was placed under the<br />
skin <strong>and</strong> underlying muscle <strong>and</strong> the<br />
pressure elevated <strong>and</strong> held at 15 psi<br />
(103 kPa), the maximum possible<br />
without damaging the underlying<br />
muscle tissue. Scale erection was not<br />
noted with the increase in subcutaneous<br />
pressure, leading us to suspect a<br />
passive mechanism. While the skin<br />
was pressurized, we used a fine acupuncture<br />
needle to gently manipulate<br />
five haphazardly selected scales to<br />
their maximum erected position without<br />
tearing them from the skin. Releasing<br />
the scales, we allowed them to<br />
settle at an erected angle, which we<br />
then measured by calculating the<br />
change in length of the scale crown<br />
when viewed from above. The inverse<br />
cosine of the apparent crown length divided<br />
by the resting or true crown<br />
length provided the angle of erection.<br />
Because the individual scales could actually<br />
be easily erected past this resting<br />
angle, we were in essence calculating a<br />
minimal erection angle. The pressure<br />
was then released <strong>and</strong> the angle similarly<br />
calculated on the flaccid skin.<br />
Our aprioritest determined that<br />
stretching the skin in this manner did<br />
not affect the non-pressurized erection<br />
angle.<br />
At the other regions (H2, B3, B6,<br />
P1, P2, P3, C1, C2, C3; Figure 1)<br />
where subcutaneous pressure was not<br />
possible, such as in the fins <strong>and</strong> over<br />
the body cavity, the erection angles<br />
were measured by gently erecting the<br />
scales (if possible) with the acupuncture<br />
needle <strong>and</strong> measuring their<br />
pre- <strong>and</strong> post-erection crown length.<br />
Finally, to get a more global picture<br />
of scale flexibility over the entire<br />
body, 35 equidistant sampling locations<br />
encompassing the entire dorsal,<br />
left lateral, <strong>and</strong> ventral surfaces of<br />
each shark were marked <strong>and</strong> scales in<br />
each area manually erected as before<br />
without subcutaneous pressure. The<br />
erected scales were simply recorded as<br />
greater or less than 50°, an angle that<br />
appeared to the approximate maximum.<br />
We also measured the crown<br />
<strong>and</strong> base length of scales in select<br />
areas, as well as the spacing of the riblets<br />
on the scales.<br />
To underst<strong>and</strong> the attachment of<br />
the scales to the skin, we prepared histological<br />
sections of the skin <strong>and</strong> scales,<br />
decalcified the scales, stained the samples<br />
to reveal the fibrous attachment,<br />
<strong>and</strong> examined the sections at 20× <strong>and</strong><br />
40× with a compound microscope. Finally,<br />
surface pictures of the scales, at<br />
all studied regions, were prepared by<br />
examining the skin at 100× <strong>and</strong> 200×<br />
under a SEM.<br />
Findings<br />
The placoid scales of sharks have<br />
a pulp cavity <strong>and</strong> a hard enameloid<br />
covering over dentine <strong>and</strong> are anchoredatthebaseofthescaletothe<br />
stratum laxum collagenous layer of<br />
the dermis (Figures 1 <strong>and</strong> 2). The exposed<br />
crowns overlap each other on<br />
the shark’s surface, <strong>and</strong> the majority<br />
of scales have small riblets or keels on<br />
their surface, oriented in the streamwise<br />
direction of the flow (Figures 1<br />
<strong>and</strong> 2). On the fast swimming shortfin<br />
mako, the flank scales (e.g., areas B2,<br />
B5, <strong>and</strong> A2) have a crown length of approximately<br />
0.18 mm, <strong>and</strong> each crown<br />
typically has three keels each having a<br />
height of 0.012 mm <strong>and</strong> a spacing of<br />
0.041 mm. This differs from other<br />
slower swimming sharks such as the<br />
blacktip shark (Carcharhinus limbatus),<br />
whereby preliminary data indicate the<br />
flank scales are typically 0.32 mm<br />
in length with each crown typically<br />
having five keels with a height of<br />
0.029 mm <strong>and</strong> a spacing of 0.065 mm.<br />
When considering shark species as a<br />
whole, length of the scales is typically<br />
fixed for specific regionsofthebody<br />
within a species but differs among regions<br />
<strong>and</strong> species. Similarly, the number<br />
of keels per scale is also consistent<br />
per body location for a species.<br />
Scale flexibility on the shortfin<br />
mako varies considerably across the<br />
210 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
FIGURE 2<br />
Scanning electron micrographs of the scales <strong>and</strong> histological sections of the embedded scales<br />
from three regions of the pectoral fin, P1, P2, <strong>and</strong> P3 of the shortfin mako (Isurus oxyrinchus).<br />
The scales on the leading edge of the fin cannot erect <strong>and</strong> lack riblets, whereas those on the trailing<br />
edge can erect to greater angles than either the scales on the leading edge or the midregion of the<br />
fin. The bases of the scales (B) are anchored in the dermis (D), <strong>and</strong> the thin epidermis (E) is visible<br />
between the scales. The pulp cavity (PC) is visible in some of the scales, <strong>and</strong> not all scales are<br />
sectioned through their center, resulting in some crown (CR) lengths appearing shorter than<br />
others. Ceratotrichia or fin rays (CE) are visible in region P3 because this part of the fin is very<br />
thin. Anterior is to the left.<br />
body <strong>and</strong> fins, with average erection<br />
angles varying from 0° on the leading<br />
edge of the pectoral fin to approximately<br />
50° on the widest part of the<br />
body just behind the gill region (Figure<br />
3). However, only certain portions<br />
ofthebodyhaveveryflexible scales.<br />
The most flexible scales are found on<br />
the flank of the body extending behind<br />
the gills to the tail; here scales<br />
are found to be easily erected with slight<br />
manipulation on dead specimens to angles<br />
of approximately 50° or greater<br />
(Figure 3). The lateral scales (B2, B5,<br />
A2) had significantly greater erection<br />
angles (mean angle = 44° ± 1°) than<br />
both the dorsal (mean angle = 26° ±<br />
1° SE) <strong>and</strong> ventral regions (mean<br />
angle = 25° ± 2° SE). Erection angles<br />
for the dorsal region did not differ<br />
from that of the ventral region<br />
(Kruskal-Wallis one-way ANOVA,<br />
Tukey’s pairwise test; H = 53.173,<br />
FIGURE 3<br />
Outline of a representative shortfinmako,Isurus<br />
oxyrinchus, showing the approximate region<br />
(lines) on the flank with the most flexible scales<br />
capable of erection to at least 50°. This region<br />
approximates the region of flow separation.<br />
df =2,P ≤ 0.001). Highly flexible<br />
scales are also found at the trailing<br />
edge of the pectoral fins. The scales<br />
on the trailing edge of the pectoral<br />
(P3) had the highest erection angles<br />
compared to the leading edge (P1)<br />
<strong>and</strong> the central region of the fin (P2),<br />
<strong>and</strong> all the regions were significantly<br />
different from each other (Kruskal-<br />
Wallis one-way ANOVA, Tukey’s<br />
pairwise test; H = 26.289, df =2,<br />
P ≤ 0.001). Conversely, for the caudal<br />
fin the mean angles were not significantly<br />
different among the three regions<br />
(ANOVA; F = 0.0614, df =2,<br />
P = 0.941).<br />
At least two factors appear to control<br />
scale flexibility on the body. The<br />
first is a reduction in the length of<br />
the base relative to the length of the<br />
crown over certain regions of the<br />
body such as the flank in the shortfin<br />
mako (Table 1). The flank scales<br />
have a greater ratio of crown length<br />
to base length compared to the dorsal<br />
scales (Kruskal-Wallis one-way<br />
ANOVA, Tukey’s pairwisetest;H =<br />
18.104, df =2,P < 0.001) due to a significantly<br />
shorter base on the flank<br />
scales (ANOVA, Holm-Sidak method;<br />
F = 45.967, df =2,P =0.001).The<br />
flank scales have a relatively wide base<br />
compared to its length, whereas the<br />
base of the dorsal scales is more uniform<br />
in shape (Table 1, BL/BW ratio;<br />
Figure 4). The ventral scales of the<br />
shortfin mako are smaller than the<br />
dorsal <strong>and</strong> flank scales as they are<br />
shorter in crown (ANOVA, Holm-<br />
Sidak method; F =45.967,df =2,<br />
P < 0.001) <strong>and</strong> base length (ANOVA,<br />
Holm-Sidak method; F = 42.916,<br />
df =2,P < 0.001) overall. Secondly, relative<br />
changes in the length of the leading<br />
<strong>and</strong> trailing edges of the scale base<br />
affect its anchoring in the dermis, similar<br />
to the root system of a tree. More<br />
firmly anchored scales have a more<br />
July/August 2011 Volume 45 Number 4 211
TABLE 1<br />
Means <strong>and</strong> st<strong>and</strong>ard errors for scale crown length (CL), base length (BL), base width (BW), ratio of CL/BL, <strong>and</strong> ratio BL/BW for three body areas of<br />
shortfin mako (Isurus oxyrinchus).<br />
Body Position CL (mm) BL (mm) BW (mm) CL/BL BL/BW<br />
B1, B4, A1 Dorsum 0.173 ± 0.004 0.145 ± 0.005 0.161 ± 0.005 1.201 ± 0.029 0.938<br />
B2, B5, A2 Flank 0.179 ± 0.004 0.104 ± 0.003 0.161 ± 0.005 1.745 ± 0.076 0.625<br />
B3, B6, A3 Ventrum 0.128 ± 0.004 0.091 ± 0.005 0.125 ± 0.003 1.472 ± 0.102 0.692<br />
evenly distributed base like a tree with<br />
a broad but evenly distributed shallow<br />
root system (Figure 4B). The more<br />
erectable scales on the flank have a narrow<br />
base on the trailing edge (Figure 4A).<br />
In this manner, the scales can pivot up,<br />
analogous to a tree with a narrow root<br />
system on one side that is blown over<br />
away from this side. The scales return<br />
to their resting position with the help<br />
of elastic fibers that anchor its base<br />
(stained black in the histology sections).<br />
A sideview picture of the skin with the<br />
scales in the foreground manually<br />
bristled is shown in Figure 5.<br />
Effects on Fluid<br />
Flow Patterns<br />
The findings of scale flexibility <strong>and</strong><br />
angle of erection on the shortfin mako<br />
FIGURE 4<br />
Representative scales of the shortfin mako<br />
(Isurus oxyrinchus) from(A)theflexible<br />
flankareaB5,<strong>and</strong>(B)thelessflexible area<br />
B4. The scales in (A) have a relatively short<br />
but wide base compared to those of (B) with<br />
a more uniformly shaped base.<br />
have led to a working hypothesis currently<br />
being investigated through<br />
hydrodynamic testing. Our discovery<br />
of highly flexible scales on the flank<br />
<strong>and</strong> trailing edges of the pectoral fin<br />
fits with the hypothesis that the scales<br />
act as a means of controlling flow separation.<br />
First, these results indicate that<br />
the erection of the scales is most likely<br />
initiated by a passive, flow-actuated<br />
mechanism, i.e., in a region of turbulent<br />
flow separation the flow consists of<br />
moments when the flow close to the<br />
wall is both in the main direction as<br />
well as reversed. Pressurizing the skin<br />
had no effect on scale erection. Flow<br />
reversal over a large region indicates<br />
FIGURE 5<br />
Coronal section through the shortfin mako<br />
skin showing the scales in the foreground<br />
that have been manually erected from location<br />
B2. Not all scales are erected to the same<br />
degree because of the individual manual erection.<br />
Flow would normally pass over the skin<br />
from left to right <strong>and</strong> reversed flow, as occurs<br />
during separation, is believed to cause bristling<br />
as shown.<br />
that global separation from the surface<br />
(body) occurs, resulting in increased<br />
pressure drag. In the case of the shark<br />
this would not only increase drag but<br />
also inhibit contragility or the ability<br />
to change direction quickly <strong>and</strong> to a<br />
large degree. As the shark swims <strong>and</strong><br />
turns, the body is bent laterally with<br />
regions of greatest curvature occurring<br />
on the flank. From the nose to<br />
the point of maximum girth (around<br />
the location of the gills), the flow will<br />
experience a favorable pressure gradient,<br />
<strong>and</strong> unfavorable pressure gradient<br />
regions will be located downstream of<br />
this point. Thus, the findings that the<br />
most flexible scales are found on the<br />
flank of the body <strong>and</strong> downstream of<br />
the gills corroborate the hypothesis<br />
that these scales are working to control<br />
flow separation. Likewise, the delay of<br />
flow separation over the pectoral fin<br />
can lead to increased performance in<br />
their ability to act as lifting surfaces<br />
during high-speed swimming maneuvers.<br />
The generation of high lift<br />
by the pectoral fins is important for<br />
quick upward maneuvers while attacking<br />
prey, as has been observed in video<br />
evidence of a shortfin mako in pursuit<br />
of a towed baitfish. This same video<br />
evidence shows the shark’s ability to<br />
turn in one direction <strong>and</strong> then change<br />
direction before the body completes<br />
the initial turn. This type of turning<br />
behavior, also defined as contragility,<br />
requires not only large muscular effort<br />
but also low form drag (Frank Fish,<br />
212 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
personal communication, February<br />
18, 2008).<br />
Flow visualization images (Figure<br />
6) of particles illuminated with a<br />
laser sheet show characteristic flow scenarios<br />
found in a turbulent boundary<br />
layer undergoing separation. In this<br />
case, separation is induced on a flat<br />
surface via the presence of a rotating<br />
cylinder located above the wall on<br />
which a turbulent boundary layer is<br />
formed. The free stream flow moves<br />
FIGURE 6<br />
Flow visualization generated by particle streaking<br />
in a plane parallel to the free stream flow<br />
(16.5 cm/s) illuminated by a laser sheet. A rotating<br />
(40 RPM) cylinder (5.1 cm diameter) is<br />
located above the image with its center, 6.4 cm<br />
above the wall, to induce boundary layer separation.<br />
Boundary layer thickness prior to the<br />
test area was approximately 1 cm <strong>and</strong> an area<br />
of approximately 3 cm × 1.5 cm is imaged.<br />
Main flow moves left to right. Distinguishing<br />
moments in time for this unsteady flow are<br />
shown when (a) the flow is attached, (b) the<br />
separation process is initiated with reversed<br />
flow close to wall, <strong>and</strong> (c) separated flow characterizedbyvortexburstingawayfromthe<br />
wall is observed, with a large region of reversed<br />
flow near the wall.<br />
at 17 cm/s, giving a local Reynolds<br />
number, based on distance from the<br />
leading edge of 0.3 m, in the boundary<br />
layer of 5 × 10 4 . In comparison, a<br />
shark’s boundary layer will have a Re of<br />
∼10 6 -10 7 when swimming at 10 m/s.<br />
However, the characteristics of the turbulent<br />
flow will be similar; the flow on<br />
a shark will be faster <strong>and</strong> the boundary<br />
layer thinner resulting in shorter time<br />
<strong>and</strong> length scales than the water tunnel<br />
experiments. Under these conditions,<br />
there are moments when the flow is<br />
for the most part attached (Figure 6a),<br />
developing into a large region of separation<br />
with reversed flow near the wall<br />
(Figure 6b), <strong>and</strong> finally times when<br />
large vortex bursting occurs as the separated<br />
region becomes unstable resulting<br />
in a shedding of vortices (Figure 6c).<br />
A time trace of the velocity measured at<br />
a location adjacent to the wall where<br />
the flow is reversed about 50% of the<br />
timeisshowninFigure7.Here,the<br />
cyclic nature of a separating turbulent<br />
boundary is clearly evident. Regions in<br />
FIGURE 7<br />
theplotwherethevelocityisnegativeindicate<br />
moments of reversed flow,<br />
which would actuate scale bristling<br />
on the shark. It is transitions in the<br />
time trace of the velocity where the<br />
flow moves (Figure 7) from positive<br />
(u > 0) to negative (u < 0) (or from<br />
point a to point b labeled in Figure 7)<br />
when scale actuation would be initiated<br />
<strong>and</strong> potentially disrupt the evolving<br />
flow that leads to the formation<br />
of a separation bubble as shown in<br />
Figures 6b <strong>and</strong> 6c.<br />
Cassel et al. (1996) provide a description<br />
of the process leading to<br />
flow separation. Because of the suction<br />
pressure upstream (adverse pressure<br />
gradient) the region closest to<br />
the wall, where the flow has the lowest<br />
momentum, is where flow reversal<br />
is first initiated. This patch of fluid<br />
moves upstream <strong>and</strong> thickens, ultimately<br />
leading to large scale flow separation<br />
from the surface. It is our<br />
hypothesis that on the shark’s body<br />
in the region close to the wall where<br />
Velocity (u) measured as a function of time at a point close to the wall corresponding to approximately<br />
two-thirds the distance downstream in Figure 6. Measurements were made using timeresolved<br />
digital particle image velocimetry (TR-DPIV), which sampled the flow at 1 kHz. At this<br />
location for about 50% of the time, flow reversal is occurring, as can be seen by the regions where<br />
u < 0. This is a point in the vicinity where flow reversal is initiated in the turbulent boundary layer<br />
<strong>and</strong> develops into a large-scale region of reversed flow due to the presence of the adverse pressure<br />
gradient induced by the rotating cylinder above. Points labeled (a), (b), <strong>and</strong> (c) are<br />
moments in the flow typified correspondingly to those shown in Figure 6.<br />
July/August 2011 Volume 45 Number 4 213
flow reversal begins to occur, the scales<br />
are actuated by the flow to erect, thereby<br />
disrupting the unsteady flow separation<br />
process. Future experiments<br />
will investigate this hypothesis through<br />
hydrodynamic testing using models<br />
<strong>and</strong> real specimens of shark skin.<br />
This aspect of shark skin resulting in<br />
asurfacewithapreferredflow direction<br />
is likely key to its ability to control<br />
flow separation.<br />
Previousexperimentsoverabristled<br />
shark skin model confirmed the<br />
presence of embedded vortices forming<br />
between replicas of the scales<br />
(Lang et al., 2008). Thus, if flow is<br />
induced to form between the scales<br />
when bristled, there are two additional<br />
mechanisms that may aid to control<br />
the flow. The formation of embedded<br />
vortices, similar as occurs with dimples<br />
on a golf ball, would allow the flow to<br />
pass over the skin with a resulting partial<br />
slip condition, thereby leading to<br />
higher momentum adjacent to the surface.<br />
Secondly, with a turbulent flow<br />
forming in the boundary layer above<br />
the cavities, there may be additional<br />
momentum exchange whereby high<br />
momentum fluid typically located<br />
away from the surface is induced at a<br />
greater rate to move towards the surface<br />
<strong>and</strong> into the cavities. This latter<br />
mechanism, resulting in turbulence<br />
augmentation (Gad el-Hak, 2000), is<br />
another potential means to increase<br />
the momentum overall in the flow adjacent<br />
to the wall. These three mechanisms<br />
may be working in conjunction<br />
to inhibit flow separation over the<br />
surface of the shark.<br />
This method has advantages, above<br />
<strong>and</strong> beyond other methods currently<br />
in use to control flow separation, in<br />
that it is passive with no energy input<br />
required. Also it causes no additional<br />
drag penalty when not in use in that<br />
scale bristling would be controlled by<br />
on-dem<strong>and</strong> erection of the scales induced<br />
by regions of flow reversal as<br />
occurs under conditions of incipient<br />
flow separation. This obviates the use<br />
of bristled scales to act as vortex generators<br />
as a means of separation control,<br />
as previously theorized by Bechert<br />
et al. (2000). Vortex generators, which<br />
consist of small, typically V-shaped<br />
protrusions, require careful placement<br />
<strong>and</strong> protrusion into the boundary layer<br />
flow upstream of the point of flow separation<br />
<strong>and</strong> work by mixing higher<br />
momentum flow down towards the<br />
surface (Lin, 2002). Vortex generators<br />
also result in a drag penalty due to their<br />
protrusion into the flow (Gad-el-Hak,<br />
2000). Our findings suggest that the<br />
scales are bristled passively <strong>and</strong> are activated<br />
in a region of flow reversal that<br />
occurs downstream of the point of separation<br />
<strong>and</strong> is thus a different methodology<br />
from that of vortex generators<br />
currently in use today. Finally, this<br />
new passive, flow-actuated mechanism<br />
may in fact go to the initial root cause<br />
of the separation, that of flow reversal<br />
adjacent to the wall, <strong>and</strong> disrupt it<br />
prior to growth into a fully separated<br />
region. Our ultimate aim is to gain a<br />
fundamental underst<strong>and</strong>ing of how<br />
the flow-actuated bristling of shark<br />
skin scales can control flow separation<br />
so that bio-inspired surfaces can be<br />
engineered for greater flow control in<br />
marine <strong>and</strong> other applications.<br />
Summary<br />
Fast swimming shortfin mako<br />
sharks Isurus oxyrinchus have highscale<br />
flexibility on the flank <strong>and</strong><br />
trailing edges of the pectoral fin, with<br />
bristling angles up to a range of approximately<br />
50° on the flank. These regions<br />
correspond to those on the body<br />
where flow separation control is likely<br />
to be most beneficial. In the case of the<br />
flank region, high curvature of the<br />
body will result from the shark’s lateral<br />
swimming motion, <strong>and</strong> this is also<br />
in the vicinity of the point of maximum<br />
girth; both conditions indicate<br />
the presence of an adverse pressure<br />
gradient. In the case of the pectoral<br />
fin, it is hypothesized that the flexible<br />
scales can lead to the control of flow<br />
separation which indicates high drag<br />
<strong>and</strong> loss of lift. Control of the flow in<br />
both regions will lead to increased swimming<br />
speeds with high contragility for<br />
the shark. Increased scale flexibility in<br />
this shark appears to be due to a reduction<br />
in the relative size of the scale base<br />
compared to the crown <strong>and</strong> changes<br />
in the shape of the base where it is anchored<br />
into the dermis. Future work<br />
will lead to the manufacturing of bioinspired<br />
surfaces based on shark skin<br />
microgeometry whereby a passive,<br />
flow-actuated surface patterning can<br />
be used for applications where flow<br />
separation control is required.<br />
Acknowledgments<br />
Funding for this work received<br />
through collaborative NSF grants<br />
(0932352, 0744670, <strong>and</strong> 0931787)<br />
to A. Lang, P. Motta, <strong>and</strong> R. Hueter<br />
to support both the engineering <strong>and</strong><br />
biological work is gratefully acknowledged.<br />
We also thank Jessica Davis<br />
for assisting in the shark measurements<br />
<strong>and</strong> C<strong>and</strong>y Mir<strong>and</strong>a for preparing the<br />
histological samples. Finally, we wish<br />
to express our gratitude to Paul <strong>and</strong><br />
Jane Majeski <strong>and</strong> crew, Captain Mark<br />
Sampson, Captain Al VanWormer,<br />
Philip Pegley, Lisa Natanson, Jack<br />
Morris, <strong>and</strong> Mote <strong>Marine</strong> Laboratory<br />
for assistance in obtaining shark<br />
specimens.<br />
214 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
Lead Author:<br />
Amy Lang<br />
Department of Aerospace<br />
Engineering <strong>and</strong> Mechanics<br />
University of Alabama<br />
Box 870280<br />
Tuscaloosa, AL 35487<br />
Email: alang@eng.ua.edu<br />
References<br />
Anderson, E., McGillis, W., & Grosenbaugh,<br />
M. 2001. The boundary layer of swimming<br />
fish. J Exp Biol. 204:81-102.<br />
Bechert, D.W., Bruse, M., Hage, W., &<br />
Meyer, R. 2000. Fluid mechanics of biological<br />
surfaces <strong>and</strong> their technological application.<br />
Naturwissenschaften. 80:157-71.<br />
Bechert, D.W., Bruse, M., Hage, W.,<br />
Van der Hoeven, J., & Hoppe, G. 1997.<br />
Experiments on drag-reducing surfaces <strong>and</strong><br />
their optimization with an adjustable<br />
geometry. J Fluid Mech. 338:59-87.<br />
Blake, R. 2004. Fish functional design <strong>and</strong><br />
swimming performance. J Fish Biol. 65:1193-222.<br />
Bruse, M., Bechert, D., van der Hoeven, J.,<br />
Hage, W., & Hoppe, G. 1993. Experiments<br />
with conventional <strong>and</strong> with novel adjustable<br />
drag-reducing surfaces. In: Proc. of the Int.<br />
Cong. On Near-Wall Turbulent Flows.<br />
719-738. Tempe, AZ: Elsevier Science<br />
Publishers.<br />
Bushnell, D., & Moore, K. 1991. Drag<br />
reduction in nature. Annu Rev Fluid Mech.<br />
23:65-79.<br />
Cassel, K., Smith, F., & Walker, J. 1996. The<br />
onset of instability in unsteady boundary-layer<br />
separation. J Fluid Mech. 315:223-56.<br />
Doligalski, T., Smith, C., & Walker, J. 1994.<br />
Vortex interactions with walls. Annu Rev<br />
Fluid Mech. 26:573-616.<br />
Donley, J.M., Sepulveda, C.A., Konstantinidis,<br />
P., Gemballa, S., & Shadwick, R.E. 2004.<br />
Convergent evolution in mechanical design of<br />
lamnid sharks <strong>and</strong> tunas. Nature. 429:61-65.<br />
Fish, F.E. 1998. Imaginative solutions by<br />
marine organisms for drag reduction. In:<br />
Proceedings of the International Symposium<br />
on Seawater Drag Reduction, ed. Meng, J.C.S.,<br />
443-50. Newport, Rhode Isl<strong>and</strong>.<br />
Fish, F. 2006. The myth <strong>and</strong> reality of<br />
Gray’s paradox: Implication of dolphin drag<br />
reduction for technology. Bioinspir Biomim.<br />
1:17-25.<br />
Fish, F., & Lauder, G. 2006. Passive <strong>and</strong><br />
active flow control by swimming fishes <strong>and</strong><br />
mammals. Annu Rev Fluid Mech. 38:193-224.<br />
Gad-el-Hak, M. 2000. Flow Control: Passive,<br />
Active <strong>and</strong> Reactive Flow Management.<br />
Cambridge, UK: Cambridge University<br />
Press. 421 pp.<br />
Lang, A., Hidalgo, P., Motta, P., & Westcott,<br />
M. 2008. Bristled shark skin: A microgeometry<br />
for boundary layer control<br />
Bioinspir Biomim. 3:046005.<br />
Lin, J. 2002. Review of research on lowprofile<br />
vortex generators to control boundarylayer<br />
separation. Prog Aerosp Sci. 38:389-420.<br />
Martinez, G., Drucker, E., & Summers, A.<br />
2002. Under pressure to swim fast. Integr<br />
Comp Biol. 42(6):1273-4.<br />
Naylor, G., Martin, A., Mattison, E., & Brown,<br />
W. 1997. The inter-relationships of lamniform<br />
sharks: Testing phylogenetic hypotheses with<br />
sequence data. In: Molecular Systematics of<br />
Fishes, eds. Kocher, T.D., & Stepien, C.,<br />
199-217. New York: Academic Press.<br />
NMFS. 2010. Final Amendment 3 to the<br />
Consolidated Atlantic Highly Migratory<br />
Species Fishery Management Plan. Silver<br />
Spring, MD: National Oceanic <strong>and</strong> Atmospheric<br />
Administration, National <strong>Marine</strong><br />
Fisheries Service, Office of Sustainable Fisheries,<br />
Highly Migratory Species Management<br />
Division. 632 pp. Public Document.<br />
Raschi, W., & Tabit, C. 1992. Functional<br />
aspects of placoid scales: A review <strong>and</strong> update.<br />
Aust J Mar Fresh Res. 43:123-147.<br />
Reif, W. 1985. Squamation <strong>and</strong> ecology of<br />
sharks, no. 78, Courier Forschungs-Institut<br />
Senckenberg, Frankfurt am Main.<br />
Schultz, W., & Webb, P. 2002. Power<br />
requirements of swimming: Do new methods<br />
resolve old questions Integr Comp Biol.<br />
42:1018-25.<br />
Stevens, J. 2009. The biology <strong>and</strong> ecology of<br />
the shortfin mako shark, Isurus oxyrinchus.<br />
In: Sharks of the Open Ocean: Biology,<br />
Fisheries <strong>and</strong> Conservation, eds. Camhi,<br />
M.D., Pikitch, E.K., Babcock, E.A., 87-94.<br />
Oxford, UK: Blackwell Publishing Ltd.<br />
Vogel, S. 2003. Comparative Biomechanics:<br />
Life’s Physical World. Princeton: Princeton<br />
University Press.<br />
Wainwright, S., Vosburgh, F., & Hebrank, J.<br />
1978. Shark skin: A function in locomotion.<br />
Science. 202:747-9.<br />
July/August 2011 Volume 45 Number 4 215
PAPER<br />
Can Biomimicry <strong>and</strong> Bioinspiration Provide<br />
Solutions for Fouling Control<br />
AUTHORS<br />
Emily Ralston<br />
Geoffrey Swain<br />
Florida Institute of <strong>Technology</strong><br />
Introduction<br />
Antifouling is changing. The ban<br />
on tributyltin (TBT), arguably the<br />
most effective <strong>and</strong> yet environmentally<br />
damaging antifouling precipitated an<br />
increase in research to develop new<br />
technology (Swain, 1999; Omae,<br />
2003). The immediate response was<br />
to return to copper as the antifouling<br />
of choice. However, high levels of copper<br />
in many ports <strong>and</strong> harbors has led<br />
to concern <strong>and</strong> even the banning of<br />
copper based antifouling (i.e., San<br />
Diego, Washington State; Carson<br />
et al., 2009; Nehring, 2001; Qian<br />
et al., 2010; Thomas et al., 2001).<br />
The use of toxic coatings may also be<br />
linked to the transport of biocide tolerant<br />
non-indigenous species (Dafforn<br />
et al., 2008). Some users have switched<br />
to biocide-free silicone antifouling<br />
coatings for improved hydrodynamic<br />
<strong>and</strong> environmental performance. However,<br />
these coatings may foul, which<br />
necessitates cleaning, <strong>and</strong> there may<br />
be an increased risk of transporting<br />
non-native species. They are mechanically<br />
weak <strong>and</strong> can be damaged more<br />
easily than traditional coatings<br />
(Chambers et al., 2006; Nehring, 2001;<br />
Swain, 2010; Swain et al., 2007; Yebra<br />
et al., 2004). An ideal coating will be<br />
hydraulically smooth <strong>and</strong> control fouling<br />
for the lifetime of the vessel, be<br />
ABSTRACT<br />
Biomimicry, modeling biological systems to find engineering methods, <strong>and</strong> bioinspiration,<br />
improving upon or repurposing the biological model, may provide direction<br />
for the development of new antifouling solutions. Despite being subject to constant<br />
pressure from foulers, many organisms maintain a clean surface. The challenge lies<br />
in selecting the most effective <strong>and</strong> reproducible antifouling mechanisms from<br />
nature <strong>and</strong> mimicking or modifying them to provide a realistic engineered solution.<br />
Keywords: natural antifouling, marine coatings, biomimicry, bioinspiration, antifouling,<br />
foul release<br />
environmentally compliant, control<br />
invasive species, be easily applied, repaired<br />
<strong>and</strong> maintained, be compatible<br />
with materials <strong>and</strong> methods of hull<br />
construction <strong>and</strong> decommissioning,<br />
<strong>and</strong> be cost effective (Swain, 2010).<br />
In the decades since a ban on TBT<br />
was proposed, research has been directed<br />
towards new solutions to the<br />
fouling problem. Some of the discoveries<br />
have been in the form of new<br />
booster biocides to improve the performance<br />
of copper based antifouling<br />
against biofilms (slimes). These include<br />
Irgarol 1051, Seanine 211, diuron,<br />
pyrithiones, <strong>and</strong> many others.<br />
Unfortunately, some of these boosters<br />
may parallel TBT in terms of environmental<br />
impact. For example, Irgarol<br />
1051 is a photosynthesis inhibitor<br />
that reduces plants’ ability to create<br />
energy causing growth to slow, reproduction<br />
to stop <strong>and</strong> eventually death of<br />
the plant. It has a long half-life <strong>and</strong><br />
does not partition into sediment so it<br />
is found primarily in the water column.<br />
It has been found to occur in estuaries,<br />
ports <strong>and</strong> harbors at elevated<br />
levels <strong>and</strong> is highly toxic to non-target<br />
marine plants (Hall et al., 1994;<br />
Thomas & Brooks, 2010; Thomas<br />
et al., 2001; Yebra et al., 2004). Recently,<br />
it has been detected in water<br />
samples collected from the Caribbean,<br />
Bermuda, Florida <strong>and</strong> Australia in the<br />
vicinity of seagrass beds <strong>and</strong> coral reefs<br />
where the effects could be devastating<br />
(Carbery et al., 2006; Owen et al.,<br />
2002; Scarlett et al., 1999). Researchers<br />
must be careful to avoid any chemistry<br />
that may impact the environment<br />
<strong>and</strong> non-target organisms. One source<br />
of new technology is to underst<strong>and</strong><br />
how organisms prevent fouling <strong>and</strong><br />
then mimic or draw inspiration from<br />
these biological models to create an<br />
engineering solution.<br />
Organisms that are unfouled or<br />
only lightly fouled provide insights<br />
into the mechanisms that have evolved<br />
to prevent surface colonization or epibiosis<br />
(Wahl, 1989). These “clean” organisms<br />
are the focus of research into<br />
biomimetic <strong>and</strong> bioinspired solutions<br />
to fouling on human structures. Biomimicry<br />
refers to the study of the<br />
structure <strong>and</strong> function of biological<br />
systems as models for the design of<br />
engineering solutions (dictionary.com)<br />
while bioinspiration exp<strong>and</strong>s on biomimicry<br />
by not only copying or imitating<br />
nature but also improving them or<br />
216 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
epurposing the biological model for<br />
an idealized engineering solution.<br />
This review will discuss the methods<br />
by which nature controls fouling<br />
<strong>and</strong> characterize natural antifouling<br />
<strong>and</strong> engineered solutions in terms of<br />
chemical, physical, mechanical, behavioral<br />
<strong>and</strong> combined mechanisms.<br />
Natural Antifouling<br />
Organisms have evolved several<br />
different strategies to prevent fouling.<br />
These natural antifouling methods include<br />
chemical, physical, mechanical,<br />
behavioral <strong>and</strong> a combination of<br />
more than one of the others (Bers &<br />
Wahl, 2004; Pawlik, 1992; Wahl,<br />
1989).<br />
Chemical<br />
Chemical antifouling has a long<br />
history of research <strong>and</strong> has been the<br />
subject of many reviews (Armstrong<br />
et al., 2000; Clare, 1996; Fusetani,<br />
2004; Omae, 2006; Pawlik, 1992;<br />
Qian et al., 2010; Raveendran &<br />
Mol, 2009; Rittschoff, 2000; <strong>and</strong><br />
others). To date, thous<strong>and</strong>s of active<br />
natural products have been identified<br />
(Pawlik, 1992). The activity of natural<br />
chemistries includes low pH, deterrents,<br />
anesthetics, attachment <strong>and</strong><br />
metamorphosis inhibitors, or toxic<br />
chemicals. The chemicals may be surface<br />
bound or water soluble (Omae,<br />
2006; Rittschof, 2000; Wahl, 1989).<br />
Despite issues with the ecological relevance<br />
of some of the chemicals, active<br />
natural products have been isolated<br />
from an algae, sponges, soft corals<br />
<strong>and</strong> a limpet that are either available<br />
at the surface or released into the<br />
water column when the organism<br />
is disturbed (de Nys & Steinberg,<br />
2002; Fusetani, 2004; Hay, 1996;<br />
Hellio et al., 2002). The mucus of dolphins,<br />
echinoderms, fish <strong>and</strong> corals<br />
contain antifouling chemicals that<br />
have been found to dissolve glue, prevent<br />
attachment or act as antimicrobial<br />
toxins (Baum et al., 2002; Bavington<br />
et al., 2004; Ebran et al., 2000; Ritchie,<br />
2006; Shephard, 1994; Videler et al.,<br />
1999). Many tunicates have acidic<br />
body pH <strong>and</strong> low epibiosis, especially<br />
in areas where density of acidic<br />
vacuoles is high (Hirose et al., 2001;<br />
Stoeker, 1980). Additionally, when<br />
looking at whole animal extracts,<br />
some tunicates contain chemicals that<br />
are cytotoxic, antimicrobial <strong>and</strong> antiviral<br />
(Davis & Wright, 1989). The<br />
eggs of many organisms, including<br />
fish <strong>and</strong> coral, are well protected with<br />
antimicrobial chemistries (Marquis<br />
et al., 2005, Ramasamy & Marugan,<br />
2007). Terrestrial plants have also<br />
yielded interesting chemistries such<br />
as tannins, pyrethroids <strong>and</strong> capsaicin<br />
(Feng et al., 2009; Perez et al., 2007;<br />
Xu et al., 2005).<br />
More recently, attention has turned<br />
to microorganisms. Antifouling metabolites<br />
that had been attributed in the<br />
past to algae, sponges, corals, etc., have<br />
been found on closer study, to be produced<br />
by surface associated bacteria<br />
<strong>and</strong> cyanobacteria (Armstrong et al.,<br />
2000; Clare, 1996; Krug, 2006).<br />
These surface associated microorganisms<br />
are distinct from the communities<br />
in the water column, often found in<br />
higher densities than the water column<br />
community <strong>and</strong> are often highly<br />
pigmented (Dobretsov et al., 2005;<br />
Faimalietal.,2004;Holmstrom<br />
et al., 1992). Deterrent biofilms preventfoulingbytoxicordeterrent<br />
chemistries (Dobretsov et al., 2006;<br />
Pawlik, 1992).<br />
Physical<br />
The two primary physical means<br />
identified to prevent fouling are surface<br />
energy <strong>and</strong> surface texture. A surface<br />
energy range of 20-30 dynes/cm<br />
(Baier, 1972; Dexter, 1979) has been<br />
shown to minimize adhesion <strong>and</strong><br />
favor the removal of epibionts. Such<br />
surface energy values have been measured<br />
on the surface of killer whales<br />
(Baier & Meyer, 1986), gorgonians<br />
(Vrolijk et al., 1990) <strong>and</strong> healthy<br />
teeth (Baier & Meyer, 1986; Glantz<br />
et al., 1991). Surface energy also effects<br />
settlement of some organisms, albeit<br />
in a species specific manner (Anderson<br />
et al., 2003; Meyer et al., 1988; Molino<br />
& Weatherbee, 2008; Rittschof &<br />
Costlow, 1989). For example, it has<br />
been found that barnacles <strong>and</strong> bryozoans<br />
prefer to settle on different surface<br />
energies (Dahlstrom et al., 2004;<br />
Rittschof & Costlow, 1989). Diatoms<br />
<strong>and</strong> the green algae Ulva have different<br />
adhesion strengths on surfaces with<br />
different wettability. Diatoms are more<br />
easily removed from hydrophilic surfaces,<br />
whereas Ulva releases more easily<br />
from hydrophobic surfaces (Finlay<br />
et al., 2002; Kirshnan et al., 2006).<br />
Low adhesion surfaces in nature are associated<br />
with waxes, oils, surfactants,<br />
mucuses or fluorinated or methylated<br />
compounds (Baum et al., 2003; Krug,<br />
2006; Shephard, 1994; Wahl, 1989).<br />
The antifouling properties of surface<br />
topography have received extensive<br />
attention <strong>and</strong> review (Scardino<br />
& de Nys, 2011; Scardino et al., 2008).<br />
The effectiveness of topography as<br />
antifouling appears to be the relationship<br />
of scale between the texture <strong>and</strong><br />
the settling organism. An ultra-smooth<br />
surface offers no refuge from predation<br />
or hydrodynamic stresses <strong>and</strong> is therefore<br />
unattractive (Kohler et al., 1999;<br />
Walters & Wethey, 1996). Surfaces<br />
with textures that are smaller than<br />
the settler reduce settlement <strong>and</strong>/or<br />
attachment strength, the “attachment<br />
point theory” (Scardino et al., 2008).<br />
Textures that are the same size or<br />
July/August 2011 Volume 45 Number 4 217
slightly larger than the propagules offer<br />
the greatest number of attachment<br />
points, the strongest attachment <strong>and</strong><br />
the best protection to settling organisms<br />
(Callow et al., 2002; Scardino<br />
et al., 2006). Organisms that have<br />
only one attachment point (barnacles,<br />
arborescent bryozoans, etc.) are more<br />
specific in searching for high quality<br />
pits than colonial organisms with multiple<br />
attachment points, likely because<br />
the colonial organisms outgrow the<br />
“refuge” of the pit quickly <strong>and</strong> can<br />
survive partial mortality (Walters &<br />
Wethey, 1996). The use of hairs in<br />
mussel spat (Dixon et al., 1995), spicules<br />
in gorgonian coral (Scardino &<br />
de Nys, 2011; Vrolijk et al., 1990)<br />
<strong>and</strong> spines in some colonial organisms<br />
(Dyrynda, 1986; Wahl, 1989) has<br />
been shown to prevent fouling <strong>and</strong><br />
overgrowth of fouling organisms. It<br />
has been suggested that many other organisms<br />
including crabs, brittle stars,<br />
molluscs, marine mammals <strong>and</strong> sharks<br />
(adult skin <strong>and</strong> dogfish egg cases), use<br />
textured surfaces with one or more<br />
scales of complexity to prevent micro<strong>and</strong><br />
macro-fouling (Baum et al.,<br />
2002; Bers & Wahl, 2004; Scardino<br />
& de Nys, 2004; Scardino & de Nys,<br />
2011).<br />
Mechanical<br />
Grooming is a common antifouling<br />
mechanism found in nature.<br />
Grooming involves specialized structuresthateitherpickorsweepan<br />
animals surface clean (Wahl, 1989).<br />
Decapod crustaceans have highly<br />
evolved brushes used to remove epibionts<br />
<strong>and</strong> parasites from specific<br />
parts of their bodies like the gills <strong>and</strong><br />
carapace (Acosta & Poirrier, 1992;<br />
Batang & Suzuki, 2003; Bauer, 1981).<br />
Historically, echinoderms <strong>and</strong> bryozoans<br />
were thought to use specialized<br />
structures to clean their surfaces<br />
(Campbell & Rainbow, 1977; Dyrynda,<br />
1986), but recently this has been called<br />
into question as the pedicellaria of the<br />
crown of thorns starfish (Acanthaster<br />
planci) were found to be too unresponsive<br />
<strong>and</strong> widely placed to be effective in<br />
keeping their surfaces clean (Guenther<br />
et al., 2007). Many organisms use<br />
ciliary cleaning in conjunction with<br />
mucus to keep surfaces clean (Wahl<br />
et al., 1998). Symbiotic or mutualistic<br />
relationships such as fish visiting<br />
“cleaning stations” (Poulin & Grutter,<br />
1996), mutualistic grazing of snails<br />
within populations (Wahl & Sonnichsen,<br />
1992; Wahl et al., 1998) <strong>and</strong> branchiobdellid<br />
annelids that feed on epibionts<br />
in the gill chamber of crayfish<br />
(Brown et al., 2002) are examples of<br />
beneficial relationships that may prevent<br />
fouling.<br />
The other mechanical method of<br />
antifouling is surface renewal via shedding<br />
or molting of outer layers. Crustaceans,<br />
stone fish <strong>and</strong> algae all molt,<br />
either the entire surface simultaneously<br />
or in patches, which removes<br />
all attached fouling (Bakus et al.,<br />
1986; Keats et al., 1997; Wahl, 1989).<br />
Additionally, many organisms use<br />
mucus as membrane to separate themselves<br />
from their environment. The<br />
mucus sloughs off removing foulers,<br />
makes adhesion difficult <strong>and</strong> fouls<br />
sensory <strong>and</strong> attachment apparatus of<br />
epibionts (Brown & Bythell, 2005;<br />
Davies & Hawkins, 1998; Denny,<br />
1989; Dyrynda, 1986; Shephard,<br />
1994; Wahl, 1989; Wahl et al., 1998).<br />
Behavioral<br />
Behavioral antifouling is the direct<br />
or indirect active avoidance of fouling<br />
organisms (Becker & Wahl, 1996).<br />
Burrowing into sediment, moving<br />
into the air, between fresh <strong>and</strong> salt<br />
water or into areas with very different<br />
oxygen contents <strong>and</strong> nocturnal activity<br />
or hiding in crevices are all mechanisms<br />
that remove less tolerant epibiotic<br />
organisms (Becker & Wahl, 1996;<br />
Wahl, 1989; Wahl et al., 1998). Organisms<br />
with a similar range of tolerances<br />
to their hosts will be unaffected<br />
by these behavioral methods (Brock<br />
et al., 1999).<br />
Combination<br />
Most organisms that are well studied<br />
use a combination of methods to<br />
prevent surface fouling. This was highlighted<br />
in reviews by Ralston <strong>and</strong><br />
Swain (2009) <strong>and</strong> Scardino <strong>and</strong><br />
de Nys (2011). Crustaceans groom<br />
<strong>and</strong> shed their shells <strong>and</strong> use behavioral<br />
mechanisms like burrowing <strong>and</strong> moving<br />
among habitats to ensure clean surfaces<br />
(Becker & Wahl, 1996; Wahl<br />
et al., 1998). Echinoderms groom,<br />
slough, excrete anti-adhesive mucus,<br />
have chemical antifoulants <strong>and</strong> may<br />
even use a strong negatively charged<br />
cuticle to prevent surface colonization<br />
(Bakus et al., 1986; Bavington et al.,<br />
2004; Bryan et al., 1996; McKenzie<br />
& Grigolava, 1996). Corals use antibacterial<br />
mucus, select specific microbial<br />
colonists which in turn protect<br />
them from other microbes, slough<br />
mucus <strong>and</strong> surface layers <strong>and</strong> secrete<br />
secondary metabolites to keep their<br />
surfaces clean (Brown & Bythell,<br />
2005; Ritchie, 2006; Targett et al.,<br />
1983). Dolphins <strong>and</strong> whales are hypothesized<br />
to use microtopography<br />
in conjunction with an enzymatically<br />
active zymogel to prevent attachment<br />
of macrofoulers (Baum et al., 2003;<br />
Meyer & Seegers, 2004) <strong>and</strong> surface<br />
sloughing, skin compliance <strong>and</strong> a critical<br />
surface tension in the preferred<br />
range for minimal adhesion combined<br />
with breaching may remove fouling at<br />
an early stage (Baum et al., 2003; Fish<br />
& Rohr, 1999; Scardino & de Nys,<br />
2011). Algae have provided many<br />
218 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
new chemical metabolites that prevent<br />
fouling, shed their outer layers <strong>and</strong><br />
may remove settled epibionts by flexing<br />
beyond what their epibionts can<br />
withst<strong>and</strong> (Nylund & Pavia, 2005;<br />
Scardino & de Nys, 2011; Walters<br />
et al., 2003; Wikstrom & Pavia, 2004).<br />
Biomimetic <strong>and</strong> Bioinspired<br />
Engineering Solutions<br />
Chemical<br />
Due to our vast experience with incorporating<br />
chemicals into coatings, it<br />
is not surprising that natural products<br />
are the most investigated biological<br />
antifouling. SeaNine 211 is a booster<br />
biocide added to copper coatings to<br />
boost efficacy against fouling plants.<br />
It degrades quickly in water <strong>and</strong> sediment,<br />
binds strongly to sediment, has<br />
low environmental toxicity <strong>and</strong> has an<br />
excellent performance record from lab<br />
<strong>and</strong> field tests <strong>and</strong> ship trials (Thomas<br />
& Brooks, 2010; Yebra et al., 2004).<br />
SeaNine 211 is based on the natural<br />
product isothiozolone originally isolated<br />
in the 1980s from the soft coral<br />
Eunicea (Raveendran & Mol, 2009).<br />
Econea is a halogenated pyrrol that is<br />
the active ingredient in copper-free<br />
antifouling paints from manufacturers<br />
such as Petit, Interlux, Sea Hawk <strong>and</strong><br />
others. Halogenated pyrrols are common<br />
secondary metabolites in bacteria<br />
<strong>and</strong> sponges (or possibly surface bacteria<br />
associated with sponges) <strong>and</strong> are<br />
potent settlement <strong>and</strong> metamorphosis<br />
inhibitors for barnacles <strong>and</strong> other<br />
animal foulers (Dahms et al., 2006;<br />
Omae, 2006). Because of its specificity<br />
against animal fouling, Econea is often<br />
combined with a booster like SeaNine<br />
to prevent plant fouling as well.<br />
Perhaps the most studied natural<br />
chemistry is the halogenated furanone<br />
originally isolated from the red algae<br />
Delisea pulchra. The chemical is present<br />
on the surface of the plant in concentrations<br />
that prevent fouling <strong>and</strong><br />
the coverage of epibionts corresponds<br />
to concentrations of the furanone<br />
(de Nys & Steinberg, 2002). It has<br />
not yet been successfully incorporated<br />
into a long-lasting ship hull coating<br />
(Chambers et al., 2006); however,<br />
some have reported that it is available<br />
products called “Netsafe” <strong>and</strong> “Pearlsafe”<br />
marketed in Australia for use in<br />
commercial aquaculture (Raveendran<br />
& Mol, 2009). We were unable to<br />
find any record of these products for<br />
sale at this time so they may no longer<br />
be available. Many other natural<br />
chemicals from marine macroorganisms<br />
show promise for non-toxic or<br />
low-toxicity antifouling paints <strong>and</strong><br />
are being investigated (see reviews by<br />
Armstrong et al., 2000; Fusetani,<br />
2004; Omae, 2006; Qian et al.,<br />
2010; Raveendran & Mol, 2009).<br />
Terrestrial plants have also yielded<br />
promising chemistries for antifouling.<br />
These include products like tannins,<br />
pyrethroids <strong>and</strong> capsaicin (Feng et al.,<br />
2009; Perez et al., 2007; Thomas &<br />
Brooks, 2010; Xu et al., 2005). Pyrethroids,<br />
synthetic analogs of pyrethrin<br />
from chrysanthemum flowers, are<br />
of particular interest because they are<br />
already approved for use as environmentally<br />
safe insecticides. These insecticides<br />
have low toxicity to mammals,<br />
do not persist, do not bioaccumulate<br />
<strong>and</strong> are available in industrial quantities<br />
(Feng et al., 2009). Tannin is present<br />
in terrestrial plants, mangroves <strong>and</strong><br />
in some marine algae, primarily as an<br />
anti-herbivory chemical. However,<br />
some have found it to have antifouling<br />
properties as well (Brock et al., 2007;<br />
Lau & Qian, 1997; Perez et al., 2007;<br />
Wikstrom & Pavia, 2004). The longterm<br />
efficacy of tannin isolated from<br />
the quebracho tree was improved by<br />
precipitating it with aluminum forming<br />
a salt which increased the life<br />
span of the coating to 1 month in the<br />
field (Perez et al., 2007).<br />
Another avenue of research is isolating<br />
chemicals from microorganisms<br />
<strong>and</strong> using the microorganisms themselves.<br />
There are many benefits to<br />
this strategy including culturability,<br />
abundance <strong>and</strong> ability to trick or stress<br />
the organisms into producing large<br />
quantities of the necessary chemical<br />
(Dobretsov et al., 2006; Holmstrom<br />
& Kjelleberg, 1994). Holmstrom et al.<br />
(2000) were able to keep bacteria alive<br />
in a coating for 14 days in the laboratory.<br />
Microencapsulation is another<br />
strategy being investigated, not just<br />
for microorganisms but for all natural<br />
products, as a way to increase length<br />
of efficacy. Coatings with microencapsulated<br />
living bacteria were able<br />
to prevent fouling up to 7 weeks in<br />
field trials (Chambers et al., 2006; Yee<br />
et al., 2007).<br />
The use of enzymes <strong>and</strong> hormones<br />
that are commercially available is another<br />
strategy for chemical antifouling.<br />
Many patents have been awarded <strong>and</strong><br />
there is an enzymatic coating available<br />
on the Danish yacht market, although<br />
little scientific evidence of effectiveness<br />
exists (Olsen et al., 2007). Enzymes<br />
may act directly by dissolving<br />
glues, lysing cells or decomposing<br />
exoskeletons of barnacles (Abarzua &<br />
Jakubowski, 1995; Evans & Clarkson,<br />
1993; Olsen et al., 2007). They may<br />
also act indirectly by increasing the<br />
effectiveness of an antifoulant or by<br />
acting on the coating to improve release<br />
or polishing rates (Olsen et al.,<br />
2007). Hormones, such as noradrenaline,<br />
may also be used as a non-toxic<br />
deterrent in antifouling coatings<br />
(Gohad et al., 2010). However, both<br />
enzymes <strong>and</strong> hormones have several<br />
drawbacks to their widespread use<br />
July/August 2011 Volume 45 Number 4 219
including expense, instability, potentially<br />
specific response <strong>and</strong> need for<br />
environmental approval (Gohad et al.,<br />
2010; Olsen et al., 2007; Rittschof,<br />
2000).<br />
There are several challenges in getting<br />
new chemicals approved for use in<br />
antifouling paints. The environmental<br />
problems associated with TBT have<br />
increased awareness of the potential<br />
risks associated with introducing new<br />
chemistries <strong>and</strong> attention must be<br />
given to proving environmental compliance.<br />
This greatly increases the<br />
time <strong>and</strong> the cost to go from the identification<br />
<strong>and</strong> isolation of a chemical<br />
to commercialization, which can cost<br />
millions of dollars <strong>and</strong> take over<br />
10 years to get approval. Furthermore,<br />
many natural products are structurally<br />
complex <strong>and</strong> only available in small<br />
amounts in the organism so there are<br />
issues with obtaining or synthesizing<br />
the active chemicals (Fusetani, 2004;<br />
Rittschof, 2000; Rittschof, 2001).<br />
Natural products tend to have a short<br />
life span as they cannot be toxic to the<br />
organism.Thisleadstoissueswhen<br />
incorporated into a coating as coatings<br />
need to maintain efficacy for<br />
3-12 years of service life (de Nys &<br />
Steinberg, 2002; Fusetani, 2004;<br />
Ingle, 2007; Marechal & Hellio,<br />
2009; Rittschof, 2000; Rittschof,<br />
2001). Conversely, the short life span<br />
is beneficial for environmental compliance<br />
as one of the characteristics of an<br />
ideal chemical antifoulant is short half<br />
life (Clare, 1996). Other factors for an<br />
ideal chemical antifoulant include<br />
non-toxicity, activity against a wide<br />
variety of fouling organisms, easy incorporation<br />
into a controlled release<br />
coating <strong>and</strong> should come from a culturable<br />
organism or have an active<br />
chemistry that can be industrially synthesized<br />
(Clare, 1996; Hellio et al.,<br />
2002; Marechal & Hellio, 2009;<br />
Raveendran & Mol, 2009; Rittschof,<br />
2000; Yebra et al., 2004).<br />
Physical<br />
Slippery coatings are not new<br />
technology. Commercially available<br />
fouling release coatings have been<br />
available since the mid-1970s <strong>and</strong> are<br />
proving to be an increasingly successful<br />
method for fouling control. These<br />
coatings combine polydimethylsiloxane<br />
silicone with low surface energy,<br />
oils <strong>and</strong> compliance to reduce the<br />
adhesion strength of organisms to a<br />
surface. Fouling is removed by hydrodynamic<br />
shear forces or with gentle<br />
cleaning (Anderson et al., 2003).<br />
There are several lanolin based waxes<br />
on the market for use on ship hulls<br />
<strong>and</strong> propellers. While the wax may<br />
provide short-term antifouling, the<br />
main purpose is to lessen attachment<br />
strength <strong>and</strong> make cleaning easier.<br />
These products are often used over a<br />
tough epoxy, however the duration of<br />
effect is unknown <strong>and</strong> we were unable<br />
to find any published data in a peer reviewed<br />
journal to back the claims of<br />
manufacturers. Results obtained from<br />
these coatings may vary depending<br />
on the fouling communities. Effects<br />
of surface energy <strong>and</strong> wettability on<br />
surface colonization are species specific<br />
with some responding favorably to hydrophobic<br />
surfaces <strong>and</strong> some to hydrophilic<br />
or intermediate surfaces (Callow<br />
et al., 2002, Dahlstrom et al., 2004).<br />
Additionally primary colonizers often<br />
change the surface energy of surfaces<br />
which will change the effect on subsequent<br />
settlers (Scardino & de Nys,<br />
2011).<br />
Mimicking the surface texture of<br />
marine organisms has been investigated<br />
as an environmentally friendly<br />
antifoulant. “Sealcoat” is a commercially<br />
available antifouling coating<br />
that is flocked with fibers mimicking<br />
the fur of a seal. According to the company’s<br />
website, fouling is prevented for<br />
up to 5 years. However, no scientific<br />
data exists to back this claim. Other<br />
studies looking at flocked or furred<br />
coatings found mixed responses with<br />
green <strong>and</strong> brown algae, encrusting<br />
bryozoans <strong>and</strong> barnacles deterred, red<br />
algae <strong>and</strong> hydroids unaffected <strong>and</strong><br />
solitary tunicates <strong>and</strong> tube worms<br />
increased by these coatings (Phillippi<br />
et al., 2001). It must also be remembered<br />
that seals do not only depend<br />
on their fur to keep them fouling free<br />
but also groom <strong>and</strong> spend large amounts<br />
of time out of the water.<br />
Mimics of topographies from other<br />
organisms such as crustose coralline<br />
algae, molluscs, crabs, brittle stars,<br />
soft corals <strong>and</strong> dogfish egg cases, have<br />
been investigated for antifouling activity<br />
<strong>and</strong> have shown short-term efficacy<br />
in laboratory assays. Additionally,<br />
topographies from pilot whale <strong>and</strong><br />
shark skins have been characterized<br />
<strong>and</strong> had an antifouling activity attributed<br />
to the microstructures. These active<br />
topographies range in scale from 1<br />
to 300 μm with multiple length scales<br />
occurring on natural surfaces (Baum<br />
et al., 2002; Bers & Wahl, 2004;<br />
Scardino & de Nys, 2004; Scardino<br />
& de Nys, 2011). The “Sharklet” is<br />
an example of a biomimetic texture<br />
used as an engineered surface to prevent<br />
fouling. It has performed well in<br />
laboratory assays against Ulva spores<br />
<strong>and</strong> Balanus amphitrite cyprids (Carman<br />
et al., 2006; Schumacher et al., 2007).<br />
In order to improve the effect of this<br />
<strong>and</strong> other topographies, a mathematical<br />
model was created called<br />
the “Engineered Roughness Index”;<br />
this index can also be used to predict<br />
settlement of marine organisms on<br />
the engineered topographies (Long<br />
et al., 2010). Engineered surfaces<br />
with hierarchically wrinkled surfaces<br />
220 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
have shown promising results in field<br />
trials, especially against barnacles<br />
(Efimenko et al., 2009; Scardino & de<br />
Nys, 2011).<br />
Sound has been suggested as an<br />
antifouling method. However, there<br />
are no published field test data that scientifically<br />
prove that it can provide<br />
long-term antifouling. This is not a<br />
truly biomimetic method as it is not reported<br />
as a natural antifouling mechanism.<br />
Sound is used by competent<br />
larvae of fish <strong>and</strong> invertebrates like<br />
crabs to navigate to appropriate settlement<br />
sites (Radford et al., 2010;<br />
Simpson et al., 2008; Stanley et al.,<br />
2010). Specific habitats have different<br />
auditory signatures <strong>and</strong> larvae can use<br />
these to differentiate <strong>and</strong> pilot to their<br />
adult habitats (Radford et al., 2010).<br />
Both high- <strong>and</strong> low-frequency sound<br />
waves have been shown to be effective<br />
at inhibiting settlement of barnacles<br />
<strong>and</strong> mussels (Branscomb & Rittschof,<br />
1984; Donskoy & Ludyanskiy, 1995;<br />
Guo et al., 2011). Additionally, ultrasound<br />
waves have been used to destroy<br />
barnacle larvae via cavitation for ballast<br />
water treatment (Seth et al., 2010).<br />
The use of low-frequency sound is limited<br />
because it is audible to humans<br />
<strong>and</strong> other organisms <strong>and</strong> therefore<br />
noise pollution is an issue. Several<br />
companies worldwide (Ultrasonic Antifouling,<br />
ASM, Sonihull <strong>and</strong> others)<br />
offer ultrasonic units that can be installed<br />
that are purported to prevent<br />
fouling or conversely to kill settling<br />
fouling organisms thereby making<br />
them easy to remove. However, this<br />
method is variable in effect, with settlement<br />
rates ranging from 1% up to<br />
55% for low <strong>and</strong> high frequency, respectively<br />
(Branscomb & Rittschof,<br />
1984; Guo et al., 2011). Guo <strong>and</strong> colleagues<br />
(2011) reported a settlement<br />
rate for barnacle cyprids in the laboratory<br />
of about 20% for their best ultrasonic<br />
treatment compared to a rate of<br />
around 70% for the control so the<br />
method is not perfect. Additionally,<br />
Sonihull reports changes in fish behavior<br />
when their ultrasonic units are<br />
in use so there are noise pollution concerns<br />
with high-frequency sound as<br />
well.<br />
Physical methods of antifouling are<br />
often inferred but seldom proved due<br />
to challenges associated with testing<br />
living materials. Surface topography<br />
effects are scale dependent (Scardino<br />
et al., 2006; Schumacher et al., 2007)<br />
<strong>and</strong> effectiveness in fouling prevention<br />
may vary geographically (Bers et al.,<br />
2010). Finding a universal physical<br />
antifoulant may be difficult <strong>and</strong> the<br />
results are often short lived, lasting a<br />
monthorlessinfield testing (Holm<br />
et al., 1997).<br />
Mechanical<br />
Mechanical cleaning is performed<br />
on ships, aquaculture nets, instruments<br />
<strong>and</strong> other marine structures<br />
when they become fouled, either because<br />
an antifouling coating was not<br />
used or if that coating becomes fouled.<br />
The U.S. Navy cleans their vessels<br />
when a set level of fouling is reached<br />
as set out in the Naval Ships’ Technical<br />
Manual (Cologer, 1984; NSTM,<br />
2006). Cleaning is reactive <strong>and</strong> has<br />
been shown to speed the rate of recolonization<br />
<strong>and</strong> increase the risk of<br />
transport of nonindigenous species<br />
(Floerl et al., 2005). Additionally,<br />
commercially available brush cleaning<br />
devices (i.e., SCAMP, Mini-Pamper,<br />
etc.) are harsh <strong>and</strong> may damage the<br />
antifouling coating. A new direction<br />
for mechanical antifouling is to mimic<br />
natural grooming. This is the idea behind<br />
the HullBUG (Hull Bioinspired<br />
Underwater Grooming), an autonomous<br />
robot that will proactively pass<br />
over a hull while a ship is in port. Its<br />
mode of action is a gentle wiping or<br />
brushing of the surface on a frequent<br />
schedule sufficient to remove fouling<br />
at its earliest stages before it can become<br />
established (Borchardt, 2010;<br />
Tribou & Swain, 2010). Results<br />
from field testing of fouling release<br />
<strong>and</strong> copper-coated panels subjected<br />
to grooming are so promising that further<br />
experiments <strong>and</strong> scale up on this<br />
method are being investigated.<br />
Ecospeed is a commercially available<br />
hull coating system. It is a tough<br />
glass flake reinforced vinyl ester<br />
coating. When combined with hull<br />
cleaning, this non-toxic coating is<br />
purported to maintain a fouling free<br />
surface with no repainting for up to<br />
25 years. Additionally, the coating<br />
smooths during cleaning, decreasing<br />
drag. Again, no scientifically published<br />
data exists to back the claims made<br />
by the manufacturer.<br />
Surface renewal has been attempted<br />
as an antifoulant for ship<br />
hull coatings. Polymers were developed<br />
that hydrolyze in seawater leaving<br />
a clean surface as they dissolve<br />
(C<strong>and</strong>ries et al., 2000). To date, however,<br />
these coatings have only been<br />
successful when combined with biocides<br />
as the rate of dissolution <strong>and</strong><br />
thickness of the coating required<br />
would be too great without the help<br />
of toxic chemicals.<br />
Mechanical antifouling has proven<br />
to be an effective but imperfect method<br />
of keeping submerged surfaces clean.<br />
Cleaning requires the deployment of<br />
equipment <strong>and</strong> usually divers, which<br />
increases both the expense <strong>and</strong> human<br />
risk factor of this antifouling<br />
mechanism. When applied to toxic<br />
coatings, cleaning may increase the release<br />
of biocides, at least in the short<br />
term (Schiff et al., 2004). Cleaning<br />
may cause damage to coatings which<br />
increases the rate of re-colonization<br />
July/August 2011 Volume 45 Number 4 221
<strong>and</strong> may increase the risk of transport<br />
of invasive species (Floerl et al., 2005;<br />
Piola & Johnston, 2008). Mechanical<br />
antifouling works better when combined<br />
with another antifouling method<br />
such as a biocidal coating or a fouling<br />
release surface. Grooming, however,<br />
is proactive <strong>and</strong> more closely matches<br />
many of the behavioral activities<br />
found in nature. Many organisms<br />
benefit from self or mutual grooming<br />
to maintain their surfaces free of<br />
fouling.<br />
Behavioral<br />
Behavioral methods include removing<br />
a vessel from the water when<br />
not in use or moving between fresh<br />
<strong>and</strong> salt water. The former is commonly<br />
practiced by recreational boat<br />
owners; however, removing a large<br />
vessel from the water is impractical, especially<br />
if that ship is frequently used.<br />
Moving vessels between fresh <strong>and</strong><br />
salt water, by traversing through the<br />
Panama Canal, for instance, is performed<br />
occasionally <strong>and</strong> has been<br />
credited with preventing the unobstructed<br />
movement of Caribbean <strong>and</strong><br />
Pacific species between the two bodies<br />
of water. Brock <strong>and</strong> colleagues (1999)<br />
found that moving a ship into fresh<br />
waterfor9dayswassufficient to remove<br />
90% of fouling from the hull.<br />
However, tolerant fouling organisms<br />
will not be affected by this antifouling<br />
method as shown by the survival <strong>and</strong><br />
subsequent introduction of the mussel<br />
Mytilus galloprovincialis to Oahu,<br />
Hawaii, from Washington. The mussel<br />
was one of the 10% of fouling organisms<br />
remaining on the USS Missouri<br />
after its Pacific transit <strong>and</strong> was seen<br />
spawning shortly after arrival in Pearl<br />
Harbor <strong>and</strong> later found colonizing<br />
the ballast tanks of a submarine (Apte<br />
et al., 2000).<br />
Combined<br />
It is unlikely that any one antifouling<br />
mechanism will be sufficient to<br />
prevent all fouling in all situations<br />
that may be encountered by submerged<br />
structures. Indeed, every<br />
organism that is well studied with<br />
regards to natural antifouling uses a<br />
combination of strategies to maintain<br />
a clean surface. The most effective<br />
coatings in use today also use more<br />
than one antifouling mechanism; antifouling<br />
coatings use a biocide combined<br />
with self-polishing or ablative<br />
mechanism to keep an active layer at<br />
the surface. Fouling release coatings<br />
combine low surface energy, oils <strong>and</strong><br />
compliance to maximize self-cleaning.<br />
To date, most researchers investigating<br />
a biomimetic solution to antifouling<br />
have focused on only one method.<br />
However, that is beginning to change<br />
with the recent publication of reviews<br />
focusing on combined antifouling<br />
mechanisms (Ralston & Swain, 2009;<br />
Scardino & de Nys, 2011).<br />
Bioinspired Approaches<br />
The examples highlighted above<br />
represent biomimetic solutions for<br />
biofouling control. Very little research<br />
exists that takes lessons from nature<br />
<strong>and</strong> adapts or alters them for a true<br />
bioinspired solution. It has only been<br />
recently that novel uses have been proposed<br />
from biological models. For example,<br />
the dopamine based adhesive<br />
system in mussels, a common fouling<br />
organism, has been investigated as a way<br />
to obtain better adhesion of non-stick<br />
coatings to a substrate. The dopamine<br />
adhesive allows testing on polyethylene<br />
glycol (PEG) <strong>and</strong> other slippery<br />
polymers where before it was not possible<br />
because the polymers would not<br />
sticktoanything.Thosecoatings<br />
using the bioinspired PEG-DOPA system<br />
outperformed traditional silicone<br />
fouling release coatings in laboratory<br />
assays, comparing both the settlement<br />
<strong>and</strong> adhesion of a common fouling<br />
diatom <strong>and</strong> alga (Statz et al., 2006).<br />
Conclusions<br />
<strong>Marine</strong> organisms can achieve<br />
long-term protection from fouling<br />
using short-lived renewable mechanisms.<br />
This is attributed to using a<br />
combination of chemical, physical,<br />
mechanical <strong>and</strong> behavioral mechanisms.<br />
Much research has been published<br />
investigating the specific ways<br />
that organisms maintain a clean surface<br />
but frequently focus on only one<br />
mechanism without considering the efficacy<br />
of a holistic combined method.<br />
The challenge, for us, is to identify <strong>and</strong><br />
select the best natural systems to solve<br />
the problem of biofouling. Through<br />
improved knowledge of natural systems,wewillbebetterabletoboth<br />
mimic <strong>and</strong> innovate using biological<br />
models to find engineering solutions.<br />
Results so far have been promising<br />
but better interactions between biologists,<br />
ecologists, engineers, chemists<br />
<strong>and</strong> materials scientists are needed<br />
<strong>and</strong> publishing of results is vitally important.<br />
Despite some issues, biomimetics<br />
<strong>and</strong> bioinspiration hold great<br />
promise for new antifouling solutions.<br />
For example, the HullBUG grooming<br />
method now being developed by the<br />
Office of Naval Research demonstrates<br />
how a proactive grooming method will<br />
enhance the long-term effectiveness<br />
of the presently available commercial<br />
antifouling or fouling release surfaces.<br />
Acknowledgments<br />
The authors would like to thank<br />
the Office of Naval Research (grants<br />
N000140210217, N000140810034<br />
<strong>and</strong> N000140910843), who has<br />
222 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
funded much of this work <strong>and</strong> continues<br />
to support research directed<br />
towards discovering improved antifouling<br />
technology <strong>and</strong> the participants<br />
in the ONR Coatings Research<br />
Group.<br />
Authors:<br />
Emily Ralston <strong>and</strong> Geoffrey Swain<br />
Florida Institute of <strong>Technology</strong><br />
150 W. University Blvd.,<br />
Melbourne, FL 32901<br />
Emails: eralston@fit.edu;<br />
swain@fit.edu<br />
References<br />
Abarzua, S., & Jakubowski, S. 1995. Biotechnological<br />
investigation for the prevention<br />
of biofouling. I. Biological <strong>and</strong> biochemical<br />
principles for the prevention of biofouling.<br />
Mar Ecol Prog Ser. 123:301-12. doi: 10.3354/<br />
meps123301.<br />
Acosta, C., & Poirrier, M. 1992. Grooming<br />
behavior <strong>and</strong> associated structures of the<br />
mysid Mysidopsis bahia. J Crustacean Biol.<br />
12:383-91. doi: 10.2307/1549032.<br />
Anderson, C., Atlar, M., Callow, M., C<strong>and</strong>ries,<br />
M., Milne, A., & Townsin, R. 2003. The<br />
development of foul-release coatings for<br />
seagoing vessels. J Mar Des Operations.<br />
B4:11-23.<br />
Apte, S., Holl<strong>and</strong>, B., Godwin, L., & Gardner,<br />
J. 2000. Jumping ship: A stepping stone<br />
event mediated transfer of a non-indigenous<br />
species via a potentially unsuitable environment.<br />
Biol Invasions. 2:75-79. doi: 10.1023/<br />
A:1010024818644.<br />
Armstrong, E., Boyd, K., & Burgess, J. 2000.<br />
Prevention of marine biofouling using natural<br />
compounds from marine organisms. Biotech<br />
Ann Rev. 6:221-41. doi: 10.1016/S1387-<br />
2656(00)06024-5.<br />
Baier, R. 1972. Influence of the initial surface<br />
condition of materials on bioadhesion.<br />
In: Proc. 3rd Int. Congr. Mar. Corr. Foul.<br />
Maryl<strong>and</strong>: Northwestern University Press.<br />
Baier, R., & Meyer, A. 1986. Biosurface<br />
chemistry for fun <strong>and</strong> profit. Chemtech.<br />
16:178-85.<br />
Bakus, G., Targett, N., & Schulte, B. 1986.<br />
Chemical ecology of marine organisms.<br />
J Chem Ecol. 12:951-87. doi: 10.1007/<br />
BF01638991.<br />
Batang, Z., & Suzuki, H. 2003. Gill-cleaning<br />
mechanisms of the burrowing thalassinidean<br />
shrimps Nihonotrypaea japonica <strong>and</strong> Upogebia<br />
major (Crustacea: Decapoda). J Zool. 261:<br />
69-77. doi: 10.1017/S0952836903003959.<br />
Bauer, R. 1981. Grooming behavior <strong>and</strong><br />
morphology in the decapod crustacea.<br />
J Crustacean Biol. 1:153-73. doi: 10.2307/<br />
1548154.<br />
Baum, C., Meyer, W., Stelzer, R., &<br />
Fleischer, L-G. 2002. Average nanorough<br />
skin surface of the pilot whale (Globicephala<br />
melas, Delphinidae): Considerations on the<br />
self-cleaning abilities based on nanoroughness.<br />
Mar Biol. 140:653-7. doi: 10.1007/s00227-<br />
001-0710-8.<br />
Baum, C., Simon, F., Meyer, W., Fleischer,<br />
L-G., Siebers, D., Kacza, J., & Seeger, J. 2003.<br />
Surface properties of the skin of the pilot whale<br />
Globicephala melas. Biofouling. 19:181-6.<br />
doi: 10.1080/0892701031000061769.<br />
Bavington, C., Lever, R., Mulloy, B.,<br />
Grundy, M., Page, C., Richardson, N., &<br />
McKenzie, J. 2004. Anti-adhesive glycoproteins<br />
in echinoderm mucus secretions.<br />
Comp Biochem Physiol B. 139:607-17.<br />
doi: 10.1016/j.cbpc.2004.07.008.<br />
Becker, K., & Wahl, M. 1996. Behaviour<br />
patterns as natural antifouling mechanisms<br />
of tropical marine crabs. J Exp Mar Biol<br />
Ecol. 203:245-58. doi: 10.1016/0022-<br />
0981(96)02575-0.<br />
Bers, A., Diaz, E., Da Gama, B., Vieira-Silva,<br />
F., Dobretsov, S., Valdivia, N., … Wahl, M.<br />
2010. Relevance of mytilid shell microtopographies<br />
for fouling defence—A global<br />
comparison. Biofouling. 26:367-77.<br />
doi: 10.1080/08927011003605888.<br />
Bers, A., & Wahl, M. 2004. The influence of<br />
natural surface microtopographies on fouling.<br />
Biofouling. 20:43-51. doi: 10.1080/<br />
08927010410001655533.<br />
Borchardt, J. 2010. Grooming the fleet.<br />
Mech Eng. 132:33-35.<br />
Branscomb, E., & Rittschof, D. 1984. An<br />
investigation of low frequency sound waves<br />
as a means of inhibiting barnacle settlement.<br />
J Exp Mar Biol Ecol. 79:149-54. doi: 10.1016/<br />
0022-0981(84)90215-6.<br />
Brock, E., Nylund, G., & Pavia, H. 2007.<br />
Chemical inhibition of barnacle larval settlement<br />
by the brown alga Fucus vesiculosus.<br />
Mar Ecol Prog Ser. 337:165-74. doi: 10.3354/<br />
meps337165.<br />
Brock, R., Bailey-Brock, J., & Goody, J.<br />
1999. A case study of the efficacy of freshwater<br />
immersion in controlling introduction of<br />
alien marine fouling communities: The USS<br />
Missouri. Pac Sci. 53:223-31.<br />
Brown, B., & Bythell, J. 2005. Perspectives on<br />
mucus secretion in reef corals. Mar Ecol Prog<br />
Ser. 296:291-309. doi: 10.3354/meps296291.<br />
Brown, B., Creed, R., & Dobson, W. 2002.<br />
Branchiobdellid annelids <strong>and</strong> their crayfish<br />
hosts: Are they engaged in a cleaning symbiosis<br />
Oecologia. 132:250-5. doi: 10.1007/s00442-<br />
002-0961-1.<br />
Bryan, P., Rittschof, D., & McClintock, J.<br />
1996. Bioactivity of echinoderm ethanolic<br />
body-wall extracts: An assessment of marine<br />
bacterial attachment <strong>and</strong> macroinvertebrate<br />
larval settlement. J Exp Mar Biol Ecol. 196:<br />
79-96. doi: 10.1016/0022-0981(95)00124-7.<br />
Callow, M., Jennings, A., Brennan, A.,<br />
Seegert, C., Gibson, A., Wilson, L., …<br />
Callow, J. 2002. Microtopographic cues for<br />
settlement of zoospores of the green fouling<br />
alga Enteromorpha. Biofouling. 18:237-45.<br />
doi: 10.1080/08927010290014908.<br />
Campbell, A., & Rainbow, P. 1977. The<br />
role of pedicellariae in preventing barnacle<br />
settlement on the sea-urchin test. Mar Behav<br />
Physiol. 4:253-60. doi: 10.1080/<br />
10236247709386957.<br />
C<strong>and</strong>ries, M., Atlar, M., & Anderson, C.<br />
2000. Considering the use of alternative<br />
July/August 2011 Volume 45 Number 4 223
antifoulings: The advantages of foul-release<br />
systems. In: Conference Proceedings, ENSUS<br />
2000. pp. 88-95. Newcastle, UK: University<br />
of Newcastle-upon-Tyne.<br />
Carbery, K., Owen, R., Frickers, T., Otero,<br />
E., & Readman, J. 2006. Contamination of<br />
Caribbean coastal waters by the antifouling<br />
herbicide Irgarol 1051. Mar Pollut Bull.<br />
52:635-44. doi: 10.1016/j.marpolbul.2005.<br />
10.013.<br />
Carman, M., Estes, T., Feinberg, A.,<br />
Schumacher, J., Wilkerson, W., Wilson, L., …<br />
Brennan, A. 2006. Engineered antifouling<br />
microtopographies—correlating wettability<br />
with cell attachment. Biofouling. 22:11-21.<br />
doi: 10.1080/08927010500484854.<br />
Carson, R., Damon, M., Johnson, L., &<br />
Gonzalez, J. 2009. Conceptual issues in<br />
designing a policy to phase out metal-based<br />
antifouling paints on recreational boats in<br />
San Diego Bay. J Environ Manag. 90:2460-8.<br />
doi: 10.1016/j.jenvman.2008.12.016.<br />
Chambers, L., Stokes, K., Walsh, F., &<br />
Wood, R. 2006. Modern approaches to<br />
marine antifouling coatings. Surf Coat Tech.<br />
201:3642-52. doi: 10.1016/j.surfcoat.2006.<br />
08.129.<br />
Clare, A. 1996. <strong>Marine</strong> natural product<br />
antifoulants: Status <strong>and</strong> potential. Biofouling.<br />
9:211-29. doi: 10.1080/08927019609378304.<br />
Cologer, C. 1984. Six year interaction of<br />
underwater cleaning with copper based antifouling<br />
paints on Navy surface ships. Nav Eng J.<br />
96:200-8. doi: 10.1111/j.1559-3584.1984.<br />
tb01829.x.<br />
Dafforn, K., Glasby, T., & Johnston, E.<br />
2008. Differential effects of tributyltin <strong>and</strong><br />
copper antifoulants on recruitment of nonindigenous<br />
species. Biofouling. 24:23-33.<br />
doi: 10.1080/08927010701730329.<br />
Dahlstrom, M., Jonsson, H., Jonsson, P.,<br />
& Elwing, H. 2004. Surface wettability as a<br />
determinant in the settlement of the barnacle<br />
Balanus improvisus (Darwin). J Exp Mar Biol<br />
Ecol. 305: 223-32. doi: 10.1016/j.jembe.<br />
2003.12.013.<br />
Dahms, H., Ying, X., & Pfeiffer, C. 2006.<br />
Antifouling potential of cyanobacteria:<br />
A mini-review. Biofouling. 22:317-27.<br />
doi: 10.1080/08927010600967261.<br />
Davis, A., & Wright, A. 1989. Interspecific<br />
differences in fouling of two congeneric ascidians<br />
(Eudistoma olivaceum <strong>and</strong> E. capsulatum): Is<br />
surface acidity an effective defense Mar Biol.<br />
102:491-7. doi: 10.1007/BF00438350.<br />
Davies, M., & Hawkins, S. 1998. Mucus<br />
from marine molluscs. In: Advances in <strong>Marine</strong><br />
Biology, eds. Blaxter, J., Southward, A., &<br />
Tyler, P., 1-71. San Diego: Academic Press.<br />
Denny, M. 1989. Invertebrate mucous secretions:<br />
Functional alternatives to vertebrate<br />
paradigms. Symp Soc Exp Biol. 43:337-66.<br />
de Nys, R., & Steinberg, P. 2002. Linking<br />
marine biology <strong>and</strong> biotechnology. Curr Opin<br />
Biotech. 13:244-8. doi: 10.1016/S0958-<br />
1669(02)00311-7.<br />
Dexter, S. 1979. Influence of substratum<br />
critical surface tension on bacterial adhesion—<br />
in situ studies. J Colloid Interf Sci. 70:<br />
346-54. doi: 10.1016/0021-9797(79)90038-9.<br />
Dixon, D., Sole-Cava, A., Pascoe, P., &<br />
Holl<strong>and</strong>, P. 1995. Periostracal adventitious<br />
hairs on spat of the mussel Mytilus edulis.<br />
J Mar Biol Assoc UK. 75:363-72.<br />
doi: 10.1017/S0025315400018233.<br />
Dobretsov, S., Dahms, H., & Qian, P. 2006.<br />
Inhibition of biofouling by marine microorganisms<br />
<strong>and</strong> their metabolites. Biofouling.<br />
22:43-54. doi: 10.1080/08927010500504784.<br />
Dobretsov, S., Dahms, H-U., Tsoi, M., &<br />
Qian, P-Y. 2005. Chemical control of epibiosis<br />
by Hong Kong sponges: The effect of sponge<br />
extracts on micro- <strong>and</strong> macrofouling communities.<br />
Mar Ecol Prog Ser. 297:119-29.<br />
doi: 10.3354/meps297119.<br />
Donskoy, D., & Ludyanskiy, M. 1995. Low<br />
frequency sound as a control measure for zebra<br />
mussel fouling. In: Proceedings of the 5th<br />
International Zebra Mussel <strong>and</strong> Other Aquatic<br />
Nuisance Organisms Conference. pp. 103-112.<br />
Pembroke, Ontario: Professional Edge.<br />
Dyrynda, P. 1986. Defensive strategies of<br />
modular organisms. Philos T Roy Soc Lon B.<br />
313:227-43. doi: 10.1098/rstb.1986.0035.<br />
Ebran, N., Julien, S., Orange, N., Auperin,<br />
B., & Molle, G. 2000. Isolation <strong>and</strong> characterization<br />
of novel glycoproteins from fish<br />
epidermal mucus: Correlation between their<br />
pore-forming properties <strong>and</strong> their antibacterial<br />
activities. Biochim Biophys Acta. 1467:271-80.<br />
doi: 10.1016/S0005-2736(00)00225-X.<br />
Efimenko, K., Finlay, J., Callow, M., Callow,<br />
J., & Genzer, J. 2009. Development <strong>and</strong><br />
testing of hierarchically wrinkled coatings<br />
for marine antifouling. ACS Appl Mat Int.<br />
1:1031-40. doi: 10.1021/am9000562.<br />
Evans, L., & Clarkson, N. 1993. Antifouling<br />
strategies in the marine environment. J Appl<br />
Bacteriol Symp Suppl. 74:119S-24S.<br />
Faimali, M., Garaventa, F., Terlizzi, A.,<br />
Chiantore, M., & Cattaneo-Vietti, R. 2004.<br />
The interplay of substrate nature <strong>and</strong> biofilm<br />
formation in regulating Balanus amphitrite<br />
Darwin, 1854 larval settlement. J Exp Mar<br />
Biol Ecol. 306:37-50. doi: 10.1016/j.jembe.<br />
2003.12.019.<br />
Feng, D., Ke, C., Li, S., Lu, C., & Guo, F.<br />
2009. Pyrethroids as promising marine antifoulants:<br />
Laboratory <strong>and</strong> field studies. Mar<br />
Biotechnol. 11:153-60. doi: 10.1007/s10126-<br />
008-9130-9.<br />
Finlay, J., Callow, M., Ista, L., Lopez, G.,<br />
& Callow, J. 2002. The influence of surface<br />
wettability on the adhesion strength <strong>and</strong><br />
settled spores of the green algae Enteromorpha<br />
<strong>and</strong> the diatom Amphora. Integr Comp Biol.<br />
42:1116-22. doi: 10.1093/icb/42.6.1116.<br />
Fish, F., & Rohr, J. 1999. Review of<br />
dolphin hydrodynamics <strong>and</strong> swimming<br />
performance. Technical Report no. 1801. SSC.<br />
136 pp.<br />
Floerl, O., Inglis, G., & Marsh, H. 2005.<br />
Selectivity in vector management <strong>and</strong> investigation<br />
of the effectiveness of measures used to<br />
prevent transport of non-indigenous species.<br />
Biol Invasions. 7:459-75. doi: 10.1007/<br />
s10530-004-4863-5.<br />
224 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
Fusetani, N. 2004. Biofouling <strong>and</strong> antifouling.<br />
Nat Prod Res. 21:94-104. doi: 10.1039/<br />
b302231p.<br />
Glantz, P., Baier, R., Attstrom, R., Meyer, A.,<br />
& Gucinski, H. 1991. Comparative clinical<br />
wettability of teeth <strong>and</strong> intraoral mucosa.<br />
J Adhes Sci Technol. 5:401-8. doi: 10.1163/<br />
156856191X00459.<br />
Gohad, N., Shah, N., Metters, A., & Mount,<br />
A. 2010. Noradrenaline deters marine invertebrate<br />
biofouling when covalently bound<br />
in polymeric coatings. J Exp Mar Biol Ecol.<br />
394:63-73. doi: 10.1016/j.jembe.2010.07.014.<br />
Guenther, J., Heimann, K., & de Nys, R.<br />
2007. Pedicellariae of the crown-of-thorns sea<br />
star Acanthaster planci are not an effective<br />
defence against fouling. Mar Ecol Prog Ser.<br />
340:101-108. doi: 10.3354/meps340101.<br />
Guo, S., Lee, H., Chaw, K., Miklas, J., Teo,<br />
S., Dickinson, G., … Khoo, B. 2011. Effect<br />
of ultrasound on cyprids <strong>and</strong> juvenile barnacles.<br />
Biofouling. 27:185-92. doi: 10.1080/<br />
08927014.2010.551535.<br />
Hall, J., Giddings, J., Solomon, K., & Balcomb,<br />
R. 1994. An ecological risk assessment for the<br />
use of Irgarol 1051 as an algaecide for antifouling<br />
paints. Crit Rev Toxicol. 29:367-437.<br />
Hay, M. 1996. <strong>Marine</strong> chemical ecology:<br />
What’s known <strong>and</strong> what’s next J Exp Mar<br />
Biol Ecol. 200:103-34. doi: 10.1016/S0022-<br />
0981(96)02659-7.<br />
Hellio, C., Berge, J., Beaupoil, C., Le Gal, Y.,<br />
& Bourgougnon, N. 2002. Screening of<br />
marine algal extracts for anti-settlement<br />
activities against microalgae <strong>and</strong> macroalgae.<br />
Biofouling. 18:205-15. doi: 10.1080/<br />
08927010290010137.<br />
Hirose, E., Yamashiro, H., & Mori, Y. 2001.<br />
Properties of tunic acid in the ascidian Phallusia<br />
nigra (Ascidiidae, Phlebobranchia). Zool Sci.<br />
18:309-14. doi: 10.2108/zsj.18.309.<br />
Holm, E., Cannon, G., Roberts, D., Schmidt,<br />
A., Sutherl<strong>and</strong>, J., & Rittschof, D. 1997.<br />
The influence of initial surface chemistry on<br />
development of the fouling community at<br />
Beaufort, North Carolina. J Exp Mar Biol<br />
Ecol. 215:189-203. doi: 10.1016/S0022-<br />
0981(97)00040-3.<br />
Holmstrom, C., & Kjelleberg, S. 1994. The<br />
effect of external biological factors on settlement<br />
of marine invertebrate <strong>and</strong> new antifouling<br />
technology. Biofouling. 8:147-60.<br />
doi: 10.1080/08927019409378269.<br />
Holmstrom, C., Rittschof, D., & Kjelleberg,<br />
S. 1992. Inhibition of settlement by larvae<br />
of Balanus amphitrite <strong>and</strong> Ciona intestinalis<br />
by a surface-colonizing marine bacterium.<br />
Appl Environ Microb. 58:2111-5.<br />
Holmstrom, C., Steinberg, P., Christov, V.,<br />
Christie, G., & Kjelleberg, S. 2000. Bacteria<br />
immobilised in gels: Improved methodologies<br />
for antifouling <strong>and</strong> biocontrol applications.<br />
Biofouling. 15:109-17. doi: 10.1080/<br />
08927010009386302.<br />
Ingle, M. 2007. Naval Sea Systems Comm<strong>and</strong><br />
Antifouling Program. Presented at the National<br />
Paint & Coatings Association International<br />
<strong>Marine</strong> <strong>and</strong> Offshore Coatings Expo. Virginia<br />
Beach, VA.<br />
Keats, D., Knight, M., & Pueschel, C. 1997.<br />
Antifouling effects of epithallial shedding in<br />
three crustose coralline algae (Rhodophyta,<br />
Coralinales) on a coral reef. J Exp Mar Biol<br />
Ecol. 213:281-93. doi: 10.1016/S0022-<br />
0981(96)02771-2.<br />
Kohler, J., Hansen, P., & Wahl, M. 1999.<br />
Colonization patterns at the substratum-water<br />
interface: How does surface microtopography<br />
influence recruitment patterns of sessile<br />
organisms Biofouling. 14:237-48. doi: 10.1080/<br />
08927019909378415.<br />
Krishnan, S., Wang, N., Ober, C., Finlay, J.,<br />
Callow, M., Callow, J., … Fischer, D. 2006.<br />
Comparison of the fouling release properties<br />
of hydrophobic fluorinated <strong>and</strong> hydrophilic<br />
PEGylated block copolymer surfaces: Attachment<br />
strength of the diatom Navicula <strong>and</strong><br />
the green algae Ulva. Biomacromolecules.<br />
7:1449-62. doi: 10.1021/bm0509826.<br />
Krug, P. 2006. Defense of benthic invertebrates<br />
against surface colonization by larvae: A<br />
chemical arms race. In: Progress in Molecular<br />
<strong>and</strong> Subcellular Biology Subseries <strong>Marine</strong><br />
Molecular Biotechnology, eds. Fusetani, N.,<br />
& Clare, A., 1-53. Berlin Heidelberg: Springer-<br />
Verlag.<br />
Lau, S., & Qian, P. 1997. Phlorotannins<br />
<strong>and</strong> related compounds as larval settlement<br />
inhibitors of the tube-building polychaete<br />
Hydroides elegans. Mar Ecol Prog Ser.<br />
159:219-27. doi: 10.3354/meps159219.<br />
Long, C., Schumacher, J., Robinson, P., III,<br />
Finlay, J., Callow, M., Callow, J., & Brennan,<br />
A. 2010. A model that predicts the attachment<br />
behavior of Ulva linza zoospores on surface<br />
topography. Biofouling. 26:411-9. doi: 10.1080/<br />
08927011003628849.<br />
Marechal, J-P., & Hellio, C. 2009. Challenges<br />
for the development of new non-toxic antifouling<br />
solutions. Int J Mol Sci. 10:4623-37.<br />
doi: 10.3390/ijms10114623.<br />
Marquis, C., Baird, A., de Nys, R., Holmstrom,<br />
C., & Koziumi, N. 2005. An evaluation<br />
of the antimicrobial properties of the eggs<br />
of 11 species of scleractinian corals. Coral<br />
Reefs. 24:248-53. doi: 10.1007/s00338-005-<br />
0473-7.<br />
McKenzie, J., & Grigolava, I. 1996. The<br />
echinoderm surface <strong>and</strong> its role in preventing<br />
microfouling. Biofouling. 10:261-72.<br />
doi: 10.1080/08927019609386285.<br />
Meyer, A., Baier, R., & King, R. 1988. Initial<br />
fouling of nontoxic coatings in fresh, brackish,<br />
<strong>and</strong> sea water. Can J Chem Eng. 66:55-62.<br />
doi: 10.1002/cjce.5450660108.<br />
Meyer, W., & Seegers, U. 2004. A preliminary<br />
approach to epidermal antimicrobial<br />
defense in the Delphinidae. Mar Biol.<br />
144:841-4. doi: 10.1007/s00227-003-<br />
1256-8.<br />
Molino, P., & Weatherbee, R. 2008. The<br />
biology of biofouling diatoms <strong>and</strong> their role<br />
in the development of microbial slimes.<br />
Biofouling. 24:365-79. doi: 10.1080/<br />
08927010802254583.<br />
Naval Ships’ Technical Manual. 2006.<br />
Chapter 081—Waterborne underwater hull<br />
cleaning of Navy ships. Publication # S9086-<br />
CQ-STM-010/CH-081 Revision 5. Washington,<br />
DC: Naval Sea Systems Comm<strong>and</strong>.<br />
July/August 2011 Volume 45 Number 4 225
Nehring, S. 2001. After the TBT era: Alternative<br />
anti-fouling paints <strong>and</strong> their ecological<br />
risks. Senck Marit. 31:341-51.<br />
Nylund, G., & Pavia, H. 2005. Chemical<br />
versus mechanical inhibition of fouling in the<br />
red alga Dilsea carnosa. Mar Ecol Prog Ser.<br />
299:111-21. doi: 10.3354/meps299111.<br />
Olsen, S., Pedersen, L., Laursen, M., Kiil, S.,<br />
& Dam-Johansen, K. 2007. Enzyme-based<br />
antifouling coatings: A review. Biofouling.<br />
23:369-83. doi: 10.1080/08927010701566384.<br />
Omae, I. 2003. Organotin antifouling paints<br />
<strong>and</strong> their alternatives. Appl Organometal<br />
Chem. 17:81-105. doi: 10.1002/aoc.396.<br />
Omae, I. 2006. General aspects of natural<br />
products antifoulants in the environment.<br />
H<strong>and</strong>book Env Chem. 5:227-62. doi: 10.1007/<br />
698_5_057.<br />
Owen, R., Knap, A., Toaspern, M., & Carbery,<br />
K. 2002. Inhibition of coral photosynthesis<br />
by the antifouling herbicide Irgarol 1051. Mar<br />
Pollut Bull. 44:623-32. doi: 10.1016/S0025-<br />
326X(01)00303-4.<br />
Pawlik, J. 1992. Chemical ecology of the<br />
settlement of benthic marine invertebrates.<br />
Oceanogr Mar Biol Annu Rev. 30:273-335.<br />
Perez, M., Garcia, M., Blustein, G., & Stupak,<br />
M. 2007. Tannin <strong>and</strong> tannate from the<br />
quebracho tree: An eco-friendly alternative<br />
for controlling marine biofouling. Biofouling.<br />
23:151-9. doi: 10.1080/08927010701189484.<br />
Phillippi, A., O’Connor, N., Lewis, A., &<br />
Kim, Y. 2001. Surface flocking as a possible<br />
anti-biofoulant. Aquaculture. 195:225-38.<br />
doi: 10.1016/S0044-8486(00)00556-1.<br />
Piola, R., & Johnston, E. 2008. The potential<br />
for translocation of marine species via small<br />
scale disruptions to antifouling surfaces.<br />
Biofouling. 24:145-55. doi: 10.1080/<br />
08927010801930480.<br />
Poulin, R., & Grutter, A. 1996. Cleaning<br />
symbioses: Proximate <strong>and</strong> adaptive explanations.<br />
Bioscience. 46:512-7. doi: 10.2307/1312929.<br />
Qian, P., Xu, Y., & Fusetani, N. 2010.<br />
Natural products as antifouling compounds:<br />
Recent progress <strong>and</strong> future perspectives.<br />
Biofouling. 26:223-34. doi: 10.1080/<br />
08927010903470815.<br />
Radford, C., Stanley, J., Tindle, C.,<br />
Montgomery, J., & Jeffs, A. 2010. Localised<br />
coastal habitats have distinct underwater<br />
sound signatures. Mar Ecol Prog Ser.<br />
401:21-9. doi: 10.3354/meps08451.<br />
Ralston, E., & Swain, G. 2009. Bioinspiration—the<br />
solution for biofouling control<br />
Bioinspir Biomim. 4:015007. doi: 10.1088/<br />
1748-3182/4/1/015007.<br />
Ramasamy, M., & Murugan, A. 2007. Fouling<br />
deterrent chemical defense in three muricid<br />
gastropod egg masses from the Southeast<br />
coast of India. Biofouling. 23:259-65.<br />
doi: 10.1080/08927010701278857.<br />
Raveendran, T., & Mol, V. 2009. Natural<br />
products antifoulants. Curr Sci. 97:508-20.<br />
Ritchie, K. 2006. Regulation of microbial<br />
populations by coral surface mucus <strong>and</strong><br />
mucus-associated bacteria. Mar Ecol Prog Ser.<br />
322:1-14. doi: 10.3354/meps322001.<br />
Rittschof, D. 2000. Natural product antifoulants:<br />
One perspective on the challenges<br />
related to coating development. Biofouling.<br />
15:119-27. doi: 10.1080/08927010009386303.<br />
Rittschof, D. 2001. Natural product antifoulants<br />
<strong>and</strong> coatings development. In:<br />
<strong>Marine</strong> Chemical Ecology, eds. McClintock,<br />
J., & Baker, B., 543-66. Boca Raton, Florida:<br />
CRC Press.<br />
Rittschof, D., & Costlow, J., Jr. 1989.<br />
Bryozoan <strong>and</strong> barnacle settlement in relation<br />
to initial surface wettability: A comparison<br />
of laboratory <strong>and</strong> field studies. Topics Mar<br />
Biol. 53:411-6.<br />
Scardino, A., & de Nys, R. 2004. Fouling<br />
deterrence on the bivalve shell Mytilus galloprovincialis:<br />
A physical phenomenon. Biofouling.<br />
20:249-57. doi: 10.1080/08927010400016608.<br />
Scardino, A., & de Nys, R. 2011. Minireview:<br />
Biomimetic models <strong>and</strong> bioinspired<br />
surfaces for fouling control. Biofouling. 27:<br />
73-86. doi: 10.1080/08927014.2010.536837.<br />
Scardino, A., Guenther, J., & de Nys, R.<br />
2008. Attachment point theory revisited: The<br />
fouling response to a microtextured matrix.<br />
Biofouling. 24:45-53. doi: 10.1080/<br />
08927010701784391.<br />
Scardino, A., Harvey, E., & de Nys, R. 2006.<br />
Testing attachment point theory: Diatom<br />
attachment on microtextured polyamide biomimetics.<br />
Biofouling. 22:55-60. doi: 10.1080/<br />
08927010500506094.<br />
Scarlett, A., Donkin, P., Fileman, T., &<br />
Morris, R. 1999. Occurrence of the antifouling<br />
herbicide, Irgarol 1051, within coastalwater<br />
seagrasses from Queensl<strong>and</strong>, Australia.<br />
Mar Pollut Bull. 38:687-91. doi: 10.1016/<br />
S0025-326X(99)00003-X.<br />
Schiff, K., Diehl, D., & Valkirs, A. 2004.<br />
Copper emissions from antifouling paint<br />
on recreational vessels. Mar Pollut Bull.<br />
48:371-7. doi: 10.1016/j.marpolbul.2003.<br />
08.016.<br />
Schumacher, J., Aldred, N., Callow, M.,<br />
Finlay, J., Callow, J., Clare, A., & Brennan, A.<br />
2007. Species-specific engineered antifouling<br />
topographies: Correlations between the settlement<br />
of algal zoospores <strong>and</strong> barnacle cyprids.<br />
Biofouling. 23:307-17. doi: 10.1080/<br />
08927010701393276.<br />
Seth, N., Chakravarty, P., Kh<strong>and</strong>eparker, L.,<br />
Anil, A., & P<strong>and</strong>it, A. 2010. Quantification<br />
of the energy required for the<br />
destruction of Balanus Amphitrite larva<br />
by ultrasonic treatment. J Mar Biol Assoc<br />
UK. 90:1475-82. doi: 10.1017/<br />
S0025315409991548.<br />
Shephard, K. 1994. Functions for fish<br />
mucus. Rev Fish Biol Fisher. 4:401-29.<br />
doi: 10.1007/BF00042888.<br />
Simpson, S., Meekan, M., Jeffs, A.,<br />
Montgomery, J., & McCauley, R. 2008.<br />
Settlement-stage coral reef fish prefer the<br />
higher-frequency invertebrate-generated<br />
audible component of reef noise. Anim<br />
Behav. 75:1861-8. doi: 10.1016/j.anbehav.<br />
2007.11.004.<br />
Stanley, J., Radford, C., & Jeffs, A. 2010.<br />
Induction of settlement in crab megalopae by<br />
ambient underwater reef sound. Behav Ecol.<br />
21:113-20. doi: 10.1093/beheco/arp159.<br />
226 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
Statz, A., Finlay, J., Dalsin, J., Callow, M.,<br />
Callow, J., & Messersmith, P. 2006. Algal<br />
antifouling <strong>and</strong> fouling release properties of<br />
metal surfaces coated with a polymer inspired<br />
by marine mussels. Biofouling. 22:391-9.<br />
doi: 10.1080/08927010601004890.<br />
Stoeker, D. 1980. Relationships between<br />
chemical defense <strong>and</strong> ecology in benthic<br />
ascidians. Mar Ecol Prog Ser. 3:257-65.<br />
doi: 10.3354/meps003257.<br />
Swain, G. 1999. Redefining antifouling<br />
coatings. J Prot Coat Linings. 16:26-35.<br />
Swain, G. 2010. The importance of ship<br />
hull coatings <strong>and</strong> maintenance as drivers for<br />
environmental sustainability. In: Ship Design<br />
<strong>and</strong> Operation for Environmental Sustainability.<br />
pp 8. London: Royal Institute of<br />
Naval Architects.<br />
Swain, G., Kovach, B., Touzot, A., Casse, F.,<br />
& Kavanagh, C. 2007. Measuring the<br />
performance of today’s antifouling coatings.<br />
J Ship Prot. 23:164-170.<br />
Targett, N., Bishop, S., O’Connell, O., &<br />
Yoder, J. 1983. Antifouling agents against the<br />
benthic marine diatom, Navicula salinicola<br />
Homarine from the gorgonians Leptogorgia<br />
virgulata <strong>and</strong> L. setacea <strong>and</strong> analogs. J Chem<br />
Ecol. 9:817-29. doi: 10.1007/BF00987807.<br />
Thomas, K., & Brooks, S. 2010. The environmental<br />
fate <strong>and</strong> effects of antifouling paint<br />
biocides. Biofouling. 26:73-88. doi: 10.1080/<br />
08927010903216564.<br />
Thomas, K., Fileman, T., Readman, J., &<br />
Waldock, M. 2001. Antifouling paint booster<br />
biocides in the UK coastal environment <strong>and</strong><br />
potential risks of biological effects. Mar<br />
Pollut Bull. 42:677-88. doi: 10.1016/S0025-<br />
326X(00)00216-2.<br />
Tribou, M., & Swain, G. 2010. The use of<br />
proactive in-water grooming to improve the<br />
performance of ship hull antifouling coatings.<br />
Biofouling. 26:47-56. doi: 10.1080/<br />
08927010903290973.<br />
Videler, H., Geertjes, G., & Videler, J. 1999.<br />
Biochemical characteristics <strong>and</strong> antibiotic<br />
properties of the mucus envelope of the<br />
queen parrotfish. J Fish Biol. 54:1124-7.<br />
doi: 10.1111/j.1095-8649.1999.tb00864.x.<br />
Vrolijk, N., Targett, N., Baier, R., & Meyer,<br />
A. 1990. Surface characterisation of two<br />
gorgonian coral species: Implications for a<br />
natural antifouling defence. Biofouling.<br />
2:39-54. doi: 10.1080/08927019009378128.<br />
Wahl, M. 1989. <strong>Marine</strong> epibiosis. I. Fouling<br />
<strong>and</strong> antifouling: Some basic aspects. Mar<br />
Ecol Prog Ser. 58:175-89. doi: 10.3354/<br />
meps058175.<br />
Wahl, M., Kroger, K., & Lenz, M. 1998.<br />
Non-toxic protection against epibiosis.<br />
Biofouling. 12:205-26. doi: 10.1080/<br />
08927019809378355.<br />
Wahl, M., & Sonnichsen, H. 1992. <strong>Marine</strong><br />
epibiosis. IV. The periwinkle Littorina littorea<br />
lacks typical antifouling defences—why are<br />
some populations so little fouled Mar Ecol<br />
Prog Ser. 88:225-35. doi: 10.3354/meps088225.<br />
Walters, L., Smith, C., & Hadfield, M. 2003.<br />
Recruitment of sessile marine invertebrates<br />
on Hawaiian macrophytes: Do pre-settlement<br />
or post-settlement processes keep plants free<br />
from fouling Bull Mar Sci. 72:813-39.<br />
Walters, L., & Wethey, D. 1996. Settlement<br />
<strong>and</strong> early post-settlement survival of sessile<br />
marine invertebrates on topographically<br />
complex surfaces: The importance of refuge<br />
dimensions <strong>and</strong> adult morphology. Mar<br />
Ecol Prog Ser. 137:161-71. doi: 10.3354/<br />
meps137161.<br />
Wikstrom, S., & Pavia, H. 2004. Chemical<br />
settlement inhibition versus post-settlement<br />
mortality as an explanation for differential<br />
fouling of two congeneric seaweeds. Oecologia.<br />
138:223-30. doi: 10.1007/s00442-003-1427-9.<br />
Xu, Q., Barrios, C., Cutright, T., & Newby,<br />
B. 2005. Assessment of antifouling effectiveness<br />
of two natural product antifoulants<br />
by attachment study with freshwater bacteria.<br />
Environ Sci Pollut Res. 12:278-84.<br />
doi: 10.1065/espr2005.04.244.<br />
Yebra, D., Kiil, S., & Dam-Johansen, K. 2004.<br />
Antifouling technology—past, present <strong>and</strong><br />
future steps towards efficient <strong>and</strong> environmentally<br />
friendly antifouling coatings. Prog Org<br />
Coat. 50: 75-104. doi: 10.1016/j.porgcoat.<br />
2003.06.001.<br />
Yee, L., Holmstrom, C., Fuary, E., Lewin, N.,<br />
Kjelleberg, S., & Steinberg, P. 2007. Inhibition<br />
of fouling by marine bacteria immobilised<br />
in K-carrageenan beads. Biofouling. 23:<br />
287-94. doi: 10.1080/08927010701366280.<br />
July/August 2011 Volume 45 Number 4 227
BOOK REVIEW<br />
Sex, Drugs, <strong>and</strong> Sea Slime: The Oceans’<br />
Oddest Creatures <strong>and</strong> Why They Matter<br />
By Ellen Prager<br />
University of Chicago Press, April 15, 2011 (International Publication: May 15, 2011)<br />
184 pp., $26.00 (Hardcover)<br />
Reviewed by Jason Goldberg<br />
U.S. Fish <strong>and</strong> Wildlife Service<br />
When it comes to reviewing this book,<br />
please let me be c<strong>and</strong>id about a personal<br />
bias: as a volunteer at the Smithsonian’s<br />
National Museum of Natural History,<br />
I’ve been known to give tours focused<br />
on the Sant Ocean Hall’s oddities, such<br />
as the two-horned narwhal skull, giant<br />
squid, <strong>and</strong> the seadevil. There’s a method<br />
to such madness: if you can capture a lay<br />
audience’s attention with the exciting <strong>and</strong><br />
unusual, they may be more receptive to<br />
discussing more serious messages about<br />
the importance of our oceans. With Sex,<br />
Drugs, <strong>and</strong> Sea Slime, Dr. Ellen Prager has<br />
unquestionably written one of the more<br />
bizarre <strong>and</strong> fascinating books in ocean literature.<br />
Combine Tina Fey with Carl<br />
Sagan or perhaps Mel Brooks <strong>and</strong> Rachel<br />
Carson <strong>and</strong> you have this book. Prager’s<br />
writing is reminiscent of the works of<br />
Mary Roach, Anthony Aveni, <strong>and</strong><br />
Michael Shermer, all of whom also have<br />
a talent for writing about science’s odder<br />
side. To the best of my knowledge,<br />
however, none has written as eloquently<br />
about, as Prager calls it, the lobster’s<br />
“Super Soaker Pee Blaster.” As she writes,<br />
the purpose of her book is to be “a brief<br />
<strong>and</strong> entertaining look at some of the<br />
oceans’ most fascinating creatures, their<br />
unusual tactics for survival, <strong>and</strong> their invaluable<br />
links to humankind. The end<br />
goal is to showcase the importance of<br />
the great diversity of life in the sea, why<br />
it is at risk, <strong>and</strong> why we should all care.”<br />
With this book, she has succeeded exceptionally<br />
well in achieving her goals.<br />
Everyone in the <strong>Marine</strong> <strong>Technology</strong><br />
<strong>Society</strong> likely has some favorite story of a<br />
bizarre creature or other factoid about<br />
the oceans they learned while in school<br />
or on the job. Prager has thoroughly researched<br />
<strong>and</strong> captured the best of these<br />
tales. More importantly, what she has<br />
really done is write entertainingly about<br />
why the ocean is relevant regardless of<br />
where you might live. She has an ability<br />
to wax poetic about the ocean’s strangest<br />
denizens, whether it is the unassuming<br />
dinoflagellate or the majestic humpback<br />
whale. The range of material she covers<br />
in such a slender volume is really quite astonishing.<br />
I often found myself wondering<br />
as I read the book whether she would cover<br />
this species or that, <strong>and</strong> inevitably she did.<br />
Plankton, hagfishes, corals, eels, parrotfish,<br />
conesnails, cephalopods, kelp, <strong>and</strong> more—<br />
they’re all in here, along with all the<br />
(copious) mucus <strong>and</strong> other excretions<br />
they produce. As the book’s title indicates,<br />
there’s also plenty of sex, as well as a few sex<br />
changes. For those of you who might be<br />
wondering, yes, she included the pearlfish,<br />
a species that provides clear proof that<br />
evolution has a sense of humor.<br />
Some chapters focus on specific species,<br />
such as those on plankton or the<br />
denizens of a coral reef, while others target<br />
specialized functions, such as species that<br />
might compete in the “X-Games” or live<br />
in extreme environments. The descriptions<br />
of the species <strong>and</strong> their unusual habits are<br />
always entertaining <strong>and</strong> sometimes laughout-loud<br />
funny. She then turns serious at<br />
the end of each chapter when she covers<br />
why these species matter. I was very<br />
pleased to see that Prager doesn’t just<br />
cover the usual reasons, such as food,<br />
drugs, <strong>and</strong> recreation, but that she highlights<br />
things many people don’t realize,<br />
such as the value of corals in mitigating<br />
storm damages. She concludes the book<br />
onamoreseriousbutoptimisticnote,<br />
highlighting the dangers that still lurk in<br />
ocean conservation <strong>and</strong> suggesting actions<br />
we can all take so we can continue to enjoy<br />
the ocean for generations to come.<br />
Overall, Prager’s work makes for fascinating<br />
reading for anyone interested in<br />
marine science. While it does sometimes<br />
get a little technical, it is certainly appropriate<br />
for lay audiences. It’s also an invaluable<br />
resource for anyone who talks<br />
about the oceans <strong>and</strong> needs some good references<br />
to spice up their talk, although the<br />
inclusion of an index would have been<br />
beneficial for such purposes. If you want<br />
to get another sense of her book, you can<br />
download a free National Public Radio<br />
podcast from http://www.npr.org/<br />
2011/04/07/135043954/under-the-seasex-is-slimy-business.<br />
Thesamecombination<br />
of titillating humor <strong>and</strong> practical<br />
discussion of the ocean’s valueisdemonstrated<br />
in Prager’s interview.<br />
Our livelihoods depend on the ocean,<br />
<strong>and</strong> the talent that has helped develop <strong>and</strong><br />
effectively use technology is extraordinary.<br />
It is therefore incumbent upon each of us<br />
to be able to talk in some way with the<br />
public <strong>and</strong> decision-makers about how<br />
228 <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Journal
the well-being of everyone living on Terra<br />
Firma relies on blue, brown, or white<br />
water. Books such as Prager’s offer an outline<br />
that many of us can use to foster a discussion<br />
about our own respective fields.<br />
Yes, the book is about weird science. The<br />
giggle factor is unmistakable. Even so, each<br />
time it gets weird, it casts light on the wonders<br />
of the ocean <strong>and</strong> makes that weirdness<br />
important <strong>and</strong> meaningful. Perhaps for<br />
that reason, more than any other, the<br />
book is highly recommended.<br />
July/August 2011 Volume 45 Number 4 229
UPCOMING MTS JOURNAL ISSUES<br />
September/October 2011<br />
General Issue<br />
November/December 2011<br />
Legacy Underwater Munitions: Assessment,<br />
Evaluation of Impacts, <strong>and</strong> Potential<br />
Response Technologies<br />
Guest Editors: Geoffrey Carton <strong>and</strong> Terrance Long<br />
CALL FOR PAPERS<br />
The MTS Journal<br />
includes technical<br />
papers, notes <strong>and</strong><br />
commentaries<br />
of general interest in<br />
most issues.<br />
Contact Managing Editor<br />
Amy Morgante<br />
(morganteeditorial@verizon.net)<br />
if you have material you<br />
would like considered.<br />
Specifications for<br />
submitting a manuscript<br />
are on the MTS website<br />
under the Publications menu.<br />
www.mtsociety.org<br />
E-mail:<br />
morganteeditorial@verizon.net<br />
Or visit our homepage at<br />
Check the <strong>Society</strong> website for future Journal topics.<br />
www.mtsociety.org<br />
www.mtsociety.org
Open Positions<br />
The MTS Career Center is an<br />
online resource to help employers<br />
<strong>and</strong> job seekers make career<br />
connections in the marine<br />
technology industry. Career<br />
professionals bring fresh ideas<br />
<strong>and</strong> new energy, <strong>and</strong> can keep<br />
your company current with modern<br />
trends <strong>and</strong> tactics.<br />
MTS Career Center<br />
Visit www.mtsociety.org/careers<br />
The MTS Career Center Provides:<br />
• Unmatched, Targeted Exposure for Job Postings<br />
MTS represents the largest audience of qualied marine technology<br />
industry professionals.<br />
• Easy Online Job & Applicant Management<br />
You can enter job descriptions, check the status of postings, renew or<br />
discontinue postings <strong>and</strong> even make payments online. Plus, you can<br />
easily manage your applicants through automatic email noti cations<br />
<strong>and</strong> applicant history access.<br />
• Resume Database Access<br />
With a paid job posting, you can search the resume database <strong>and</strong><br />
use an automatic notication system to receive email notications<br />
when new resumes match your criteria.<br />
• Employment Br<strong>and</strong>ing Opportunities<br />
Along with each job posting, you can include information about your<br />
company <strong>and</strong> a link to your website.<br />
International <strong>Marine</strong> Forensics Symposium<br />
April 3 – 5, 2012 • National Harbor, MD<br />
The Symposium is sponsored by the<br />
<strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong>,<br />
American <strong>Society</strong> of Naval Engineers,<br />
<strong>Society</strong> of Naval Architects & <strong>Marine</strong> Engineers.<br />
This symposium will bring together marine professionals<br />
<strong>and</strong> historians to exchange information on historic marine<br />
losses, on marine forensic investigation processes <strong>and</strong><br />
tools, <strong>and</strong> on case studies where causes of failures <strong>and</strong><br />
losses have been determined or are under continued<br />
study. The Symposium will report on the latest research<br />
<strong>and</strong> underst<strong>and</strong>ing of the Titanic, Lusitania, Edmund<br />
Fitzgerald, the Monitor <strong>and</strong> Passaic, HMS Prince of<br />
Wales, Bismarck, HMS Hood, <strong>and</strong> the Andrea Doria.<br />
Keynote speaker James<br />
Cameron (conrmed).<br />
His undersea documentaries<br />
include Expedition Bismarck,<br />
Ghosts of the Abyss,<br />
Volcanoes of the Deep Sea,<br />
Aliens of the Deep <strong>and</strong><br />
Last Mysteries of the Titanic.<br />
The International <strong>Marine</strong> Forensics Symposium will be held at the Gaylord Hotel, National Harbor, MD.<br />
Exhibition space is available at $650 for a 10 x 10-foot space. Freeman Exhibit Services will h<strong>and</strong>le the<br />
show decorating. A number of Sponsorship Opportunities are also available to maximize your impact<br />
with Symposium attendees. Contact Mary Beth Loutinsky at mbloutinsky@gmail.com for details.
Oceans of Opportunity:<br />
International Cooperation <strong>and</strong> Partnerships Across the Pacic<br />
September 19–22, 2011<br />
Kona, Hawaii<br />
Register Today for Outst<strong>and</strong>ing Topics in Oceanology,<br />
the Latest <strong>Technology</strong> in the Exhibition Hall, <strong>and</strong><br />
Excellent Networking Opportunities!<br />
go to:<br />
http://www.oceans11mtsieeekona.org/main.cfm/CID/16/Registration/<br />
Of cial Notice of <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Annual Meeting<br />
2011 Ofcer Elections <strong>and</strong> Notice of Annual Membership Meeting<br />
The terms of the individuals holding the following<br />
<strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> (“MTS”) ofces<br />
end on December 31, 2011: Vice President of<br />
Publications, Vice President of Industry <strong>and</strong><br />
<strong>Technology</strong>, Vice President of Education <strong>and</strong><br />
Research, <strong>and</strong> Vice President of Government <strong>and</strong><br />
Public Affairs. Elections for these ofces will<br />
be conducted at the MTS Annual Membership<br />
Meeting <strong>and</strong> Awards Luncheon held in conjunction<br />
with the OCEANS’11 MTS/IEEE Kona<br />
Conference, September 21, 2011, at the Hilton<br />
Waikoloa Village, Kona, Hawaii. The meeting will<br />
begin at noon PST.<br />
Article IV, Section 1.6 of the MTS Bylaws denes<br />
the nominations <strong>and</strong> election procedures for MTS<br />
ofcer elections. On June 20, 2011, the MTS Nominating<br />
Committee presented MTS’s President-<br />
Elect with at least two (2) nominees for each of the<br />
ofces to be elected at this year’s Annual Meeting.<br />
After reviewing these nominations, the President-<br />
Elect directed MTS staff to prepare the ballot for<br />
this year’s election <strong>and</strong> Member Proxy, both of<br />
which contain the names of the nominees for each<br />
ofce to be elected at this year’s election. The<br />
MTS Bylaws establish August 1, 2011 as the record<br />
date for determining the MTS members entitled to<br />
notice of the Annual Meeting <strong>and</strong> eligible to vote in<br />
this year’s election. If you are an active member of<br />
MTS as of August 1, 2011 (membership dues paid),<br />
you are a member in good st<strong>and</strong>ing <strong>and</strong> therefore<br />
entitled to vote in this year’s ofcer elections.<br />
Pursuant to MTS’s Bylaws, a member entitled to<br />
vote may do so in person at the Annual Meeting<br />
or by appointing a proxy to vote on the member’s<br />
behalf. If you wish to appoint a proxy to vote on<br />
your behalf at the MTS Annual Membership<br />
Meeting, please complete the online Member<br />
Proxy (http://www.votenet.com).<br />
This Proxy form must be completed no later than<br />
5:00 p.m. (Eastern St<strong>and</strong>ard Time) on September<br />
11, 2011 (“Proxy Deadline”). Proxies that are<br />
incomplete or received after the Proxy Deadline<br />
will not be accepted or counted. A hardcopy proxy<br />
form may be requested by calling the MTS ofce at<br />
410.884.5330.
<strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> Member Organizations<br />
CORPORATE MEMBERS<br />
ABCO Subsea<br />
Houston, Texas<br />
AMETEK Sea Connect Products, Inc.<br />
Westerly, Rhode Isl<strong>and</strong><br />
C & C Technologies, Inc.<br />
Lafayette, Louisiana<br />
C-MAR Group<br />
Houston, Texas<br />
Compass Publications, Inc.<br />
Arlington, Virginia<br />
Converteam<br />
Houston, Texas<br />
Delcor USA<br />
Houston, Texas<br />
DOF Subsea USA<br />
Houston, Texas<br />
Dynacon, Inc.<br />
Bryan, Texas<br />
E.H. Wachs Company<br />
Houston, Texas<br />
Fluor Offshore Solutions<br />
Sugar L<strong>and</strong>, Texas<br />
Fugro Chance, Inc.<br />
Lafayette, Louisiana<br />
Fugro Geoservices, Inc.<br />
Houston, Texas<br />
Fugro-McClell<strong>and</strong> <strong>Marine</strong> Geosciences<br />
Houston, Texas<br />
Fugro Pelagos, Inc.<br />
San Diego, California<br />
Geospace Offshore Cables<br />
Houston, Texas<br />
GS-Hydro US<br />
Houston, Texas<br />
HARRIS CapRock Communications<br />
Melbourne, Florida<br />
Hydroid, LLC<br />
Pocasset, Massachusetts<br />
Innerspace Corporation<br />
Covina, California<br />
INTECSEA<br />
Houston, Texas<br />
InterMoor, Inc.<br />
Houston, Texas<br />
iRobot Corporation<br />
Durham, North Carolina<br />
J P Kenny, Inc.<br />
Houston, Texas<br />
Kongsberg Maritime, Inc.<br />
Houston, Texas<br />
L-3 Communications<br />
Houston, Texas<br />
L-3 MariPro<br />
Goleta, California<br />
Lockheed Martin Sippican<br />
Marion, Massachusetts<br />
<strong>Marine</strong> Cybernetics AS<br />
Trondheim, Norway<br />
Mitsui Engineering <strong>and</strong> Shipbuilding Co. Ltd.<br />
Tokyo, Japan<br />
Mohr Engineering & Testing<br />
Houston, Texas<br />
Oceaneering Advanced Technologies<br />
Hanover, Maryl<strong>and</strong><br />
Oceaneering International, Inc.<br />
Houston, Texas<br />
Odyssey <strong>Marine</strong> Exploration<br />
Tampa, Florida<br />
Oil States Industries, Inc.<br />
Arlington, Texas<br />
Perry Slingsby Systems, Inc.<br />
Houston, Texas<br />
Phoenix International Holdings, Inc.<br />
Largo, Maryl<strong>and</strong><br />
QinetiQ North America – <strong>Technology</strong> Solutions<br />
Group<br />
Slidel, Louisiana<br />
Raytheon Technical Service Company, LLC<br />
Saipem America, Inc.<br />
Houston, Texas<br />
Schilling Robotics, LLC<br />
Davis, California<br />
SEA CON Brantner <strong>and</strong> Associates, Inc.<br />
El Cajon, California<br />
SonTek/YSI, Inc.<br />
San Diego, California<br />
South Bay Cable Corp.<br />
Idyllwild, California<br />
Subconn, Inc.<br />
Burwell, Nebraska<br />
Subsea 7 (US), LLC<br />
Houston, Texas<br />
Technip<br />
Houston, Texas<br />
Teledyne Geophysical Instruments<br />
Houston, Texas<br />
Teledyne Oil & Gas<br />
Daytona Beach, Florida<br />
Teledyne RD Instruments, Inc.<br />
Poway, California<br />
Tyco Electronics Subsea Communications<br />
Morristown, New Jersey<br />
UniversalPegasus International, Inc.<br />
Houston, Texas<br />
BUSINESS MEMBERS<br />
A<strong>and</strong>eraa Data Instruments, Inc.<br />
Attleboro, Massachusetts<br />
Ashtead <strong>Technology</strong>, Inc.<br />
Houston, Texas<br />
Bastion Technologies, Inc.<br />
Houston, Texas<br />
Bennex Subsea, Houston<br />
Houston, Texas<br />
BioSonics, Inc.<br />
Seattle, Washington<br />
BIRNS, Inc.<br />
Oxnard, California<br />
C.A. Richards <strong>and</strong> Associates, Inc.<br />
Houston, Texas<br />
Cochrane Technologies, Inc.<br />
Lafayette, Louisiana<br />
Compass Personnel Services, Inc.<br />
Katy, Texas<br />
Contros Systems & Solutions GmbH<br />
Kiel, Germany<br />
DeepSea Power <strong>and</strong> Light<br />
San Diego, California<br />
Deepwater Rental <strong>and</strong> Sypply<br />
New Iberia, Louisiana<br />
DOER <strong>Marine</strong><br />
Alameda, California<br />
DPS Offshore Inc.<br />
Houston, Texas<br />
DTC International<br />
Houston, Texas<br />
Energy Sales, Inc.<br />
Redmond, Washington<br />
Environ-Tech Diving<br />
Stanwood, Washington<br />
Falmat, Inc.<br />
San Marcos, California<br />
Fugro Atlantic<br />
Norfolk, Virginia<br />
Fugro GeoSurveys, Inc.<br />
St. John’s, Newfoundl<strong>and</strong> <strong>and</strong> Labrador,<br />
Canada<br />
Global Industries Offshore, LLC<br />
Houston, Texas<br />
Horizon <strong>Marine</strong>, Inc.<br />
Marion, Massachusetts<br />
ICAN<br />
Mt. Pearl, Newfoundl<strong>and</strong> <strong>and</strong> Labrador, Canada<br />
Intrepid Global, Inc.<br />
Houston, Texas<br />
IPOZ Systems, LLC<br />
Katy, Texas<br />
IVS 3D<br />
Portsmouth, New Hampshire<br />
iXBlue, Inc.<br />
Cambridge, Massachusetts<br />
KDU Worldwide Technical Services<br />
Sarjah, United Arab Emirates<br />
KnightHawk Engineering<br />
Houston, Texas<br />
Liquid Robotics, Inc.<br />
Palo Alto, California<br />
Makai Ocean Engineering, Inc.<br />
Kailua, Hawaii<br />
Matthews-Daniel Company<br />
Houston, Texas<br />
Oceanic Imaging Consultants, Inc.<br />
Honolulu, Hawaii<br />
The <strong>Marine</strong> <strong>Technology</strong> <strong>Society</strong> gratefully acknowledges the critical support of the Corporate, Business, <strong>and</strong> Institutional members listed.<br />
Member organizations have aided the <strong>Society</strong> substantially in attaining its objectives since its inception in 1963.<br />
OceanWorks International<br />
Houston, Texas<br />
Poseidon Offshore Mining<br />
Oslo, Norway<br />
Quest Offshore Resources<br />
Sugar L<strong>and</strong>, Texas<br />
Remote Ocean Systems, Inc.<br />
San Diego, California<br />
RRC Robotica Submarina<br />
Macaé, Brazil<br />
SeaBotix<br />
San Diego, California<br />
SeaL<strong>and</strong>Aire Technologies, Inc.<br />
Jackson, Mississippi<br />
SeaView Systems, Inc.<br />
Dexter, Michigan<br />
Sonardyne, Inc.<br />
Houston, Texas<br />
SIMCorp <strong>Marine</strong> Environmental, Inc.<br />
St. Stephens, Canada<br />
Sound Ocean Systems, Inc.<br />
Redmond, Washington<br />
Stress Subsea, Inc.<br />
Houston, Texas<br />
Subsea Riser Products, Inc.<br />
Houston, Texas<br />
SURF Subsea, Inc.<br />
Magnolia, Texas<br />
Team Trident<br />
Cypress, Texas<br />
<strong>Technology</strong> Systems Corporation<br />
Palm City, Florida<br />
Teledyne Impulse<br />
San Diego, California<br />
Tension Member <strong>Technology</strong><br />
Huntington Beach, California<br />
VideoRay, LLC<br />
Phoenixville, Pennsylvania<br />
WET Labs, Inc.<br />
Philomath, Oregon<br />
Xodus Group<br />
Houston, Texas<br />
INSTITUTIONAL MEMBERS<br />
Associacio Institut Ictineu Centre, Catala De<br />
Recerca Submarina<br />
Barcelona, Spain<br />
Canadian Coast Guard<br />
St. John’s, Newfoundl<strong>and</strong> <strong>and</strong> Labrador, Canada<br />
City of St. John’s<br />
Newfoundl<strong>and</strong> <strong>and</strong> Labrador, Canada<br />
CLS America, Inc.<br />
Largo, Maryl<strong>and</strong><br />
Consortium for Ocean Leadership<br />
Washington, DC<br />
Department of Innovation, Trade <strong>and</strong> Rural<br />
Development<br />
St. John’s, Newfoundl<strong>and</strong> <strong>and</strong> Labrador, Canada<br />
Fundação Homem do Mar<br />
Rio de Janeiro, Brazil<br />
Harbor Branch Oceanographic Institute<br />
Fort Pierce, Florida<br />
International Seabed Authority<br />
Kingston, Jamaica<br />
<strong>Marine</strong> Applied Research & Exploration<br />
Richmond, California<br />
<strong>Marine</strong> Institute<br />
Newfoundl<strong>and</strong> <strong>and</strong> Labrador, Canada<br />
Monterey Bay Aquarium Research Institute<br />
Moss L<strong>and</strong>ing, California<br />
National Research Council Institute for Ocean<br />
<strong>Technology</strong><br />
St. John’s, Newfoundl<strong>and</strong> <strong>and</strong> Labrador,<br />
Canada<br />
Naval Facilities Engineering Service Center<br />
Port Hueneme, California<br />
NOAA/PMEL<br />
Seattle, Washington<br />
Noblis<br />
Falls Church, Virginia<br />
OceanGate, LLC<br />
Everett, Washington<br />
Oregon State University College of Oceanic <strong>and</strong><br />
Atmospheric Sciences<br />
Corvalis, Oregon<br />
<strong>Society</strong> of Ieodo Research<br />
Jeju-City, South Korea
5565 Sterrett Place, Suite 108<br />
Columbia, Maryl<strong>and</strong> 21044<br />
Postage for periodicals<br />
is paid at Columbia, MD,<br />
<strong>and</strong> additional mailing offices.