Biomimetics and Marine Technology - Marine Technology Society

Journal 

The International, Interdisciplinary Society Devoted to Ocean and Marine Engineering, Science, and Policy 

Volume 45 Number 4 July/August 2011 

Biomimetics and Marine Technology

INTRODUCING 

The Next Generation 

in Underwater Acoustic Modems 

The ATM-900 Series Modems 

from Teledyne Benthos 

• Advanced Data Logger Capabilities 

– Increased from 704K to 6MB (flash file format) 

– Searchable by record number and time 

• Dual Serial Port – Interface with more than one 

sensor at a time (optional) 

• Integrated real-time clock for data time 

stamping 

ATM-900 Series 

• New Command Structure & Modem Parameters 

• Pinger Mode 

• Arbitrary Waveform Generator (optional) 

Available April 2011 

See us at Ocean Business - Booth #S1. 

A Member of Teledyne Marine 

49 Edgerton Drive, North Falmouth, MA 02556 USA 

T: +1 508.563.1000 • F: +1 508.563.6444 

E: benthos@teledyne.com 

www.benthos.com

Volume 45, Number 4, July/August 2011 

Biomimetics and Marine Technology 

Guest Editors: Frank E. Fish and Donna M. Kocak 

Turn to page 7 for a key to the cover images. 

Text: SPi 

Cover and Graphics: 

Michele A. Danoff, Graphics By Design 

The Marine Technology Society Journal 

(ISSN 0025-3324) is published by the Marine Technology 

Society, Inc., 5565 Sterrett Place, Suite 108, Columbia, 

MD 21044. 

MTS members can purchase the printed Journal for $27 

domestic and $50 (plus $50 S&H) international. 

Non-members and library subscriptions are $420 online only, 

$124 print—domestic, $140 (plus $50 S&H) print— 

international, $435 print and online (worldwide); 

Single-issue (hardcopy) is $20 plus $7.50 S&H (domestic), 

$24.50 S&H (international); Pay-per-view (worldwide): 

$15/article. Postage for periodicals is paid at Columbia, MD, 

and additional mailing offices. 

POSTMASTER: 

Please send address changes to: 

Marine Technology Society Journal 

5565 Sterrett Place 

Suite 108 

Columbia, Maryland 21044 

Copyright © 2011 Marine Technology Society, Inc. 

In This Issue 

8 

Biomimetics and Marine Technology: 

An Introduction 

Frank E. Fish, Donna M. Kocak 

14 

Biomimicking Marine Mechanisms and 

Organizational Principles 

Commentary by Yoseph Bar-Cohen 

16 

Sink and Swim: Clues From Nature for 

Aquatic Robotics 

Commentary by Jeannette Yen 

19 

Developing Bioinspired Autonomous 

Systems 

Commentary by Thomas M. McKenna 

24 

GhostSwimmer AUV: Applying 

Biomimetics to Underwater Robotics for 

Achievement of Tactical Relevance 

Commentary by Michael Rufo, 

Mark Smithers 

31 

Autonomous Robotic Fish as Mobile 

Sensor Platforms: Challenges and 

Potential Solutions 

Xiaobo Tan 

41 

Robotic Models for Studying Undulatory 

Locomotion in Fishes 

George V. Lauder, Jeanette Lim, 

Ryan Shelton, Chuck Witt, Erik Anderson, 

James L. Tangorra 

56 

Thrust Production in Highly Flexible 

Pectoral Fins: A Computational Dissection 

Srinivas Ramakrishnan, Meliha Bozkurttas, 

Rajat Mittal, George V. Lauder 

65 

Learning From the Fins of Ray-Finned 

Fish for the Propulsors of Unmanned 

Undersea Vehicles 

James L. Tangorra, Timo Gericke, 

George V. Lauder 

74 

Bioinspired Design Process for an 

Underwater Flying and Hovering Vehicle 

Jason D. Geder, John S. Palmisano, 

Ravi Ramamurti, Marius Pruessner, 

Banahalli Ratna, William C. Sandberg 

83 

A Twistable Ionic Polymer-Metal 

Composite Artificial Muscle for 

Marine Applications 

Kwang J. Kim, David Pugal, 

Kam K. Leang 

99 

Batoid Fishes: Inspiration for the Next 

Generation of Underwater Robots 

Keith W. Moored, Frank E. Fish, 

Trevor H. Kemp, Hilary Bart-Smith 

110 

Bioinspired Propulsion Mechanisms 

Based on Manta Ray Locomotion 

Keith W. Moored, Peter A. Dewey, 

Megan C. Leftwich, Hilary Bart-Smith, 

Alexander J. Smits 

119 

Inspired by Sharks: A Biomimetic 

Skeleton for the Flapping, Propulsive 

Tail of an Aquatic Robot 

John H. Long, Jr., Tom Koob, 

Justin Schaefer, Adam Summers, 

Kurt Bantilan, Sindre Grotmol, 

Marianne Porter

In This Issue 

130 

Lateral-Line-Inspired Sensor Arrays for 

Navigation and Object Identification 

Vicente I. Fernandez, Audrey Maertens, 

Frank M. Yaul, Jason Dahl, 

Jeffrey H. Lang, Michael S. Triantafyllou 

147 

A Conserved Neural Circuit-Based 

Architecture for Ambulatory and 

Undulatory Biomimetic Robots 

Joseph Ayers, Anthony Westphal, 

Daniel Blustein 

153 

A Hybrid Class Underwater Vehicle: 

Bioinspired Propulsion, Embedded 

System, and Acoustic Communication 

and Localization System 

Michael Krieg, Peter Klein, 

Robert Hodgkinson, Kamran Mohseni 

165 

Modeling of Artificial Aurelia aurita 

Bell Deformation 

Keyur B. Joshi, Alex Villanueva, 

Colin F. Smith, Shashank Priya 

181 

Swimming and Walking of an 

Amphibious Robot With Fin Actuators 

Naomi Kato 

198 

Marine Applications of the Biomimetic 

Humpback Whale Flipper 

Frank E. Fish, Paul W. Weber, 

Mark M. Murray, Laurens E. Howle 

208 

Shark Skin Separation Control 

Mechanisms 

Amy Lang, Philip Motta, 

Maria Laura Habegger, Robert Hueter, 

Farhana Afroz 

216 

Can Biomimicry and Bioinspiration 

Provide Solutions for Fouling Control 

Emily Ralston, Geoffrey Swain 

228 

BOOK REVIEW: Sex, Drugs, and Sea 

Slime: The Oceans’ Oddest Creatures 

and Why They Matter 

by Ellen Prager 

Reviewed by Jason Goldberg

Deep Tracks! 

Teledyne RD Instruments’ new PAVS150 DVL keeps your 

vehicle on course up to 500 meters above the seafloor. 

The 150 kHz Phased Array Velocity Sensor (PAVS150) 

Doppler Velocity Log delivers over 2x the range of our 

industry-standard Workhorse Navigator. Now that’s 

deep tracks! 

The PAVS150’s 1000M depth rating, combined with its 

increased range allows you to submerge your manned 

or unmanned vehicle sooner and transit longer, saving 

you time, money, and peace of mind. 

Stay the course with PAVS! 

Full details at: 

http://www.rdinstruments.com/PAVS.aspx 

Poway, CA USA • Tel. +1-858-842-2600 

www.rdinstruments.com

Learn how MTS can help your company 

reach the technology leaders and innovators 

in the marine technology industry. 

Opportunities include advertising in Currents, the MTS Journal, 

sponsorships, online and custom-tailored programs. 

MTS — 

here for the marine technology community, 

here for you! 

Email Mary Beth Loutinsky at 

mbloutinsky@gmail.com or 

call 703-629-3810

Marine Technology Society Officers 

BOARD OF DIRECTORS 

President 

Jerry Boatman 

QinetiQ North America – Technology Solutions 

Group 

President-elect 

Drew Michel 

ROV Technologies, Inc. 

Immediate Past President 

Elizabeth Corbin 

VP—Section Affairs 

Lisa Medeiros 

OceanWorks International 

VP—Education and Research 

Jill Zande 

MATE Center 

VP—Industry and Technology 

Jerry C. Wilson 

Fugro Pelagos, Inc. 

VP—Publications 

Karin Lynn 

Treasurer and VP—Budget and Finance 

Debra Kill 

International Submarine Engineering 

VP—Government and Public Affairs 

Justin Manley 

Liquid Robotics 

SECTIONS 

Canadian Maritime 

Vacant 

Florida 

Vacant 

Gulf Coast 

Laurie Jugan 

Consultant 

Hampton Roads 

Raymond Toll 

SAIC 

Hawaii 

Stewart Burley 

Strategic Theories Unlimited 

Houston 

Robert Keith 

Phoenix International Holdings, Inc. 

Japan 

Prof. Toshitsugu Sakou 

Tokai University 

Monterey 


MATE Center 

New England 

Chris Jakubiak 

UMASS Dartmouth-SMAST 

Newfoundland and Labrador 

Bill O’Keefe 

Surmount Technologies, Inc. 

Oregon 

TBD 

Puget Sound 

Fritz Stahr 

University of Washington 

San Diego 

Scott Mau 

Southwest Fisheries Sciences Center 

South Korea 

Dr. Seok Won Hong 

Maritime & Ocean Engineering Research Inst. 

(MOERI/KORDI) 

Washington, D.C. 

Brent Evers 

Hadal Technologies, Inc. 

PROFESSIONAL COMMITTEES 

Industry and Technology 

Buoy Technology 

Dr. Walter Paul 

Woods Hole Oceanographic Institution 

Cables and Connectors 

Helmut H. Portmann 

National Data Buoy Center 

Deepwater Field Development Technology 

Dr. Benton Baugh 

Radoil, Inc. 

Diving 

David C. Berry 

Subsea Construction and Diving Consultant 

Dynamic Positioning 

Howard Shatto 

Shatto Engineering 

Manned Underwater Vehicles 

William Kohnen 

SEAmagine Hydrospace Corporation 

Moorings 

Jack Rowley 

SAIC 

Oceanographic Instrumentation 

Dr. Jim Irish 

University of New Hampshire 

Offshore Structures 

Dr. Peter W. Marshall 

MHP Systems Engineering 

Remotely Operated Vehicles 

Drew Michel 

ROV Technologies, Inc. 

Renewable Energy 

Rich Chwaszczewski 

SAIC 

Ropes and Tension Members 

Evan Zimmerman 

Delmar Systems, Inc. 

Seafloor Engineering 

Herb Herrmann 

Naval Seafloor Cable Protection Office 

Underwater Imaging 

Dr. Fraser Dalgleish 

Harbor Branch Oceanographic Institute 

Unmanned Maritime Vehicles 

Rafael Mandujano 

Vehicle Control Technologies, Inc. 

Education and Research 

Marine Archaeology 

Dan Warren 

C & C Technologies 

Marine Education 

Erica Moulton 

MATE Center 

Marine Geodetic Information Systems 

Dave Zilkoski 

NOAA 

Marine Materials 

Vacant 

Ocean Exploration 

Guillermo Söhnlein 

OceanGate 

Physical Oceanography/Meteorology 

Dr. Richard L. Crout 

National Data Buoy Center 

Remote Sensing 

Herb Ripley 

Hyperspectral Imaging Limited 

Government and Public Affairs 

Marine Law and Policy 

Montserrat Gorina-Ysern 

Healthy Children–Healthy Oceans Foundation 

Marine Mineral Resources 

Dr. John C. Wiltshire 

University of Hawaii 

Marine Security 

Dallas Meggitt 

Sound & Sea Technology 

Ocean Economic Potential 

James Marsh 


Ocean Observing Systems 

Donna Kocak 

HARRIS CapRock Communications 

Ocean Pollution 

Jacob Sobin 

NOAA Coastal Services Center 

STUDENT SECTIONS 

Duke University 

Counselor: Douglas Nowacek, Ph.D. 

Florida Atlantic University 

Counselor: Douglas A. Briggs, Ph.D. 

Florida Institute of Technology 

Counselor: Stephen Wood, Ph.D., P.E. 

Long Beach City College 

Counselor: Scott Fraser 

Massachusetts Institute of Technology 

Counselor: Alexandra Techet, Ph.D. 

Monterey Peninsula College/Hartnell College 

Counselor: Jeremy R. Hertzberg 

Texas A&M University—College Station 

Counselor: Patrick Lynett 

Texas A&M—Corpus Christi 

Counselor: Lea-Der Chen, Ph.D. 

Texas A&M University—Galveston 

Counselor: Frank Warnakula, Ph.D. 

United States Naval Academy 

Counselors: Capt. Joseph T. Arcano 

(USN Ret), Ph.D. 

Cmdr. David J. Robillard 


Counselor: R. Cengiz Ertekin, Ph.D. 

University of Houston 

Counselors: Raresh Pascali, P.E., 

Chuck Richards 

University of North Carolina—Charlotte 

Counselor: James Conrad, Ph.D. 

University of Southern Mississippi 

Counselor: Stephen Howden, Ph.D. 

Webb Institute 

Counselor: Matthew Werner 

HONORARY MEMBERS 

†Robert B. Abel 

†Charles H. Bussmann 

John C. Calhoun, Jr. 

John P. Craven 

†Paul M. Fye 

David S. Potter 

†Athelstan Spilhaus 

†E. C. Stephan 

†Allyn C. Vine 

†James H. Wakelin, Jr. 

†deceased

Editorial Board 

Brian Bingham, Ph.D. 

Editor 

University of Hawaii at Manoa 

Corey Jaskolski 

Hydro Technologies 

Donna Kocak 


Scott Kraus, Ph.D. 

New England Aquarium 

Dhugal Lindsay, Ph.D. 

Japan Agency for Marine-Earth Science 

& Technology 

Justin Manley 

Liquid Robotics 

Stephanie Showalter 

National Sea Grant Law Center 

Jason Stanley 

Schilling Robotics 

Edith Widder, Ph.D. 

Ocean Research and Conservation 

Association 


MATE Center 

Editorial 

Karin Lynn 

VP of Publications 

Brian Bingham, Ph.D. 

Editor 

Amy Morgante 

Managing Editor 

Administration 

Jerry Boatman 

President 

Richard Lawson 

Executive Director 

Jeanne Glover 

Membership and Marketing Manager 

Michael Hall 

Member Groups Manager 

Chris Barrett 

Director of Professional Development 

and Meetings 

Suzanne Voelker 

Subscription Manager 

The Marine Technology Society is 

a not-for-profit, international professional 

society. Established in 1963, the Society’s 

mission is to promote the exchange of 

information in ocean and marine engineering, 

technology, science, and policy. 

Please send all correspondence to: 

The Marine Technology Society 

5565 Sterrett Place, Suite 108 

Columbia, MD 21044 

(410) 884-5330 Tel. 

(410) 884-9060 FAX 

MTS Journal: morganteeditorial@verizon.net 

Publications: publications@mtsociety.org 

Membership: Jeanne.Glover@mtsociety.org 

Programs: Michael.Hall@mtsociety.org 

Director: Rich.Lawson@mtsociety.org 

Online: www.mtsociety.org 

MEMBERSHIP INFORMATION 

may be obtained by contacting the Marine 

Technology Society. Benefits include: 

■ Free subscription to the online Marine 

Technology Society Journal, with highly 

reduced rates for the paper version 

■ Free subscription to the bimonthly news - 

letter, Currents, covering events, business 

news, science and technology, and 

people in marine technology 

■ Member discounts on all MTS publications 

■ Reduced registration rates to all MTS and 

MTS-sponsored conferences and workshops 

■ Member-only access to an expansive Job 

Bank and Member Directory 

■ Reduced advertising rates in MTS 

publications 

■ National recognition through our Awards 

Program 

Individual dues are $75 per year. Life 

membership is available for a one-time 

fee of $1,000. Patron, Student, Emeritus, 

Institutional, Business, and Corporate 

memberships are also available. 

ADVERTISING 

Advertising is accepted by the Marine 

Tech nology Society Journal. For more information 

on MTS advertising and policy, 

please contact Mary Beth Loutinsky, 

mbloutinsky@att.net. 

COPYRIGHT 

Copyright © 2011 by the Marine Technology 

Society, Inc. Authorization to photocopy 

items for internal or personal use, or the 

internal or personal use of specific clients, is 

granted by the Marine Technology Society, 

provided that the base fee of $1.00 per copy, 

plus .20 per page is paid directly to Copyright 

Clearance Center, 222 Rosewood Dr., 

Danvers, MA 01923. 

For those organizations that have been 

granted a photocopy license by CCC, a separate 

sys tem of payment has been arranged. 

The fee code for users of the Transactional 

Reporting Service is 0025-3324/89 $1.00 + 

.20. Papers by U.S Government employees 

are declared works of the U.S. Government 

and are therefore in the public domain. 

The Marine Technology Society cannot be 

held responsible for the opinions given and 

the statements made in any of the articles 

published. 

ABSTRACTS 

MTS Journal article abstracts, if available, 

can be accessed for free at http://www. 

ingentaconnect.com/content/mts/mtsj. 

Print and electronic abstracts of MTS Journal 

articles are also available through GeoRef , Aquatic Sciences and 

Fisheries Abstracts, published by Cambridge 

Scientific Abstracts , and Geobase’s 

Oceanbase published by Elsevier Science. 

CONTRIBUTORS 

Contributors can obtain an information and 

style sheet by contacting the managing editor. 

Sub missions that are relevant to the concerns 

of the Society are welcome. All papers are subjected 

to a stringent review procedure directed 

by the editor and the editorial board. The 

Journal focuses on technical material that may 

not otherwise be available, and thus technical 

papers and notes that have not been published 

previously are given priority. General commentaries 

are also accepted and are subject to review 

and approval by the editorial board. 

6 Marine Technology Society Journal

Key to Cover Images 

Front cover: GhostSwimmer AUVs from Boston 

Engineering Advanced Systems Group and Olin College 

Intelligent Vehicle Lab. Image courtesy of HARRIS CapRock 

Communication; underwater photo courtesy of Dr. Tamara Frank. 

Back cover: (Image courtesy of HARRIS CapRock Communications) 

1. Propeller with tubercles (propeller courtesy of Laurens Howls 

and morphing courtesy of John A. Lever) 

2. CephaloBot prototype hybrid vehicle 

3. Robotic Turtle, “RT-I” 

4. Mantabot 

5. Artificial Aurelia aurita bell deformation (photo courtesy of 

Alex Villanueva, CIMSS, Virginia Tech) 

6. Autonomous robotic fish (photo courtesy of Xiaobo Tan) 

7. Bluegill Sunfish (Lepomis macrochirus) with robotic pectoral 

fins (photo courtesy of George Lauder and James Tangorra) 

8. Four-fin vehicle based on Bird wrasse (Gomphosus varius) 

9. Tadro4 modeled after living electric ray Narcine 

(photo courtesy of Dr. Steve Kajiura) 

10. Lobster-based robot (robot photo courtesy of Brian Tucker, 

Bresnahan Photography; lobster photo courtesy of 

Dr. Tamara Frank)

INTRODUCTION 

Biomimetics and Marine Technology: 

An Introduction 

Frank E. Fish 

West Chester University 

Donna M. Kocak 


I 

nspiration for the development of new technologies is at the heart of the biomimetic 

approach. As there is a wide diversity of biological forms, particular attributes can be targeted 

that provide innovative solutions to engineering problems. Biology can provide new 

technological possibilities and enhance performance of existing technologies. Biomimicry is 

a tool for solving problems in the conceptual and embodiment phases of design (Reap 

et al., 2005). The goal of biomimetics is to use biological inspiration to engineer machines 

that emulate the performance of animals (Kumph & Triantafyllou, 1998; Taubes, 2000), 

particularly in instances where the animal’s performance exceeds current technology. The 

natural experimentation that has occurred through the evolutionary process has produced 

the plethora of organisms, both living and extinct. Within the phylogenetic lineages of 

these organisms, there has been essentially a “cost-benefit analysis” where particular designs 

for specific functions have been optimized to perform with respect to the rigors of their 

environment (Fish, 2006). 

Biologists are well acquainted with the specific adaptations present in animals, which may 

be of interest to engineers. For biologists, an adaptationist program has allowed for the identification 

of novel features of organisms based on engineering principles, whereas for engineers, 

identification of such novel features is necessary to exploit them for biomimetic 

development. This new synergy between biologists and engineers can be beneficial in advancing 

technology by looking to nature to provide solutions to current problems. 

For marine technologies, the biomimetic approach particularly holds the promise of enhanced 

performance and increased efficiency for operation in the aquatic realm. It was in the 

oceans that life first evolved and where complex animals have thrived for over 600 million 

years. There are marine representatives from every major animal phylum. Marine animals 

survive in environments as diverse as tropical coral reefs, polar ice-capped oceans, and the 

lightless abyssal depths. The diversity of habitats available in marine systems has led to a 

vast array of body designs and physiological and behavioral mechanisms. These adaptations 

that evolved in animals are used to overcome the biotic and abiotic challenges in the ocean. 

To deal with the rigors of the marine environment, animals have developed specialized 

sensory systems (e.g., echolocation, electroreception), mechanisms to deal with pressure 


(e.g., buoyancy control), strategies to economize on energy (e.g., fusiform body design, 

schooling, burst-and-glide swimming), armor (e.g., bony scales, mollusk shells), stability 

mechanisms (e.g., paired and median fins), maneuverability (e.g., flexible bodies, vectored 

thrust), speed (e.g., high-aspect-ratio oscillatory propulsors, jet propulsion), stealth (e.g., 

camouflage, low acoustic signature), and use of compliant materials (e.g., collagen, protein 

rubbers, mucous). In using such specializations to enhance their own survival, animals attempt 

to function in a manner to minimize their total energy budget while maximizing the 

performance of the specialization. Animals are doing the type of optimization that engineers 

seek to incorporate into designs (Vincent, 1990), and specifically, marine animals 

are dealing with the very problems that are of concern for marine engineers. 

Marine technology is well suited for the application of bioinspired design, as there is a need 

for exploitation of the oceans for new sources of food, energy, and minerals. Exploration of 

the world’s oceans is expensive. Ship time, support personnel, maintenance facilities, and 

energy costs can be prohibitively costly. Added to these costs are the expanse of the ocean 

surface (3.6 × 10 8 km 2 ) to be explored and the dangers associated with working in the 

deep-water environment (average depth = 3,650 m). The Challenger Deep (depth = 

10,902 m; pressure = 16,500 psi or 113,764 kPa) in the Mariana Trench was only visited 

once by a manned vehicle, Trieste, in 1960. Since that time, only two unmanned expeditions, 

robotic deep-sea probe Kaikō and HROV Nereus, have returned to the deepest surveyed point 

in the ocean. 

Historically, marine animals have served as the inspiration for technological design. During 

the Renaissance, animals were identified as streamlined bodies for drag reduction that 

could be applied to manufactured devices. Between 1505 and 1508, Leonardo da Vinci 

was particularly interested in flow in water, as revealed in his notebooks, Codex Leicester 

(Ball, 2009). Da Vinci wrote on the function of streamlined bodies in reducing drag and 

noted the streamlined shape of a fish (Anderson, 1998). He argued that the fish could 

move through the water with little resistance because its shape allowed the water to flow 

smoothly over the afterbody without prematurely separating. Da Vinci recognized and demonstrated 

a similar design with the hull shape of ships. 

Giovanni Borelli in 1680 made an examination of the swimming motions of animals with 

their application to submarine technology (Borelli, 1680). In his book De Motu Animalium 

(The Movement of Animals), Borelli likened swimming to flying in that both were accomplished 

by the displacement of fluids, although he noted the differences in density of air and 

water and their effects on stability and buoyancy. Borelli described the design of an early submarine 

that incorporated ideas based in part on animals for buoyancy regulation and propulsion. 

The submarine would submerge using a hydrostatic mechanism based on the swim 

bladder of a fish by filling goatskin bags, located inside the submarine, through holes in 

July/August 2011 Volume 45 Number 4 9

the sides of the boat. Propulsion would be accomplished by oars projecting through the hull 

and fitted with watertight seals. When the submarine was on the bottom, it was envisioned 

that the oars would push off the sandy substrate to move the boat along. In mid-water, the 

oars would paddle like the feet of frogs or geese. During the rearward power stroke, a flexible 

paddle at the end of the oar would expand to work on a large mass of fluid. During the forward 

recovery stroke, the paddle would fold passively to reduce the frontal area and drag on the oar. 

However, Borelli considered that propulsion of the boat would be easier if a flexible oar were 

positioned at the stern, emulating the motion of a fish tail. Despite the elaborate design for its 

time, it is doubtful if this early biomimetic experiment was successfully used. 

Cayley examined the streamlined body shapes of a trout and a dolphin in 1809 as solids of 

least resistance design (Gibbs-Smith, 1962). Cayley unsuccessfully attempted to apply these 

natural designs to the hull of a boat for moving on the water surface (Vogel, 1998). The 

rounded configuration of the hull was unstable with respect to roll, and low drag did not 

occur. The design of fish and dolphins is similar to the optimal shape for drag reduction 

of submerged bodies, such as modern submarines. These natural swimmers and submarines 

have fusiform body shapes with a rounded leading edge and slowly tapering tail. The forerunner 

for hulls used by modern nuclear submarines, the USS Albacore, was built in 1953 

with a fusiform shape. 

Both Cayley’s hull and the USS Albacore demonstrate limitations due to a misuse of the 

biomimetic approach. For Cayley, strict adherence to copying biological designs without 

proper insight into the function and limitations of those designs proved disastrous (Vogel, 

1998; Fish, 2006). While the shapes of fish and dolphins are appropriate for movement 

underwater, their shapes are not effective at the water surface. Hulls with broad beams and 

greater buoyancy, like those displayed by waterfowl, provide enhanced stability and opportunity 

for greater speed at the water surface (Aigeldinger & Fish, 1995). 

InthecaseoftheUSSAlbacore, the design is only analogous with fish and dolphins. 

Although the designs are convergent, there was no information exchange to determine design. 

The Albacore was likened to the shape of a fish, but the submarine’s design was not biologically 

inspired (Harris 1997; Largess & Mandelblatt, 1999). This was similar to the description 

of the fictional submarine, the Nautilus, in Jules Verne’s Twenty Thousand Leagues under 

the Sea: 

We were lying upon the back of a sort of submarine boat, which appeared (as far as I could 

judge) like a huge fish of steel. 

The hull shape of the Albacore was based on the “Lyon form” of airship models (Largess & 

Mandelblatt, 1999). Originally, submarine hulls were designed more as surface ships due to 


the limited amount of time that they could operate submerged. The streamlined hull of the 

Albacore made it the fastest and most maneuverable submarine of its time. 

The convergent designs of the Albacore and marine animals reflect selection, both artificial 

and natural, respectively, for optimizing similar performance parameters (i.e., speed, drag reduction, 

maneuverability). However, similarity of design does not necessarily always translate 

into identical performance and assumptions of performance expectations should be approached 

with caution. Similar shapes can have different functions or have limitations to a 

function when viewed in isolation without consideration of the whole system. Despite its 

enhanced maneuverability compared to other submarines, the Albacore’s rateofturnat 

2° s −1 is poor when compared with similarly shaped dolphins, which can turn at 453° s −1 

and in a confined space of 20% of body length (Fish, 2002). The rigid hull of the submarine 

in concert with yaw control mainly from an aft-positioned rudder limits agility and maneuverability 

compared to flexible-bodied dolphins with multiple fore and aft mobile control 

surfaces (Fish, 2002; Fish & Nicastro, 2003). 

The biomimetic approach demands first careful observation of the whole biological system 

to identify the principles and attributes of the system. Thus, major limitations and constraints 

of any biological design can be defined before translation to an engineered system. 

The association between biologists and engineers becomes paramount as biomimetic technologies 

are developed. 

This special issue of the Marine Technology Society Journal brings together both biologists 

and engineers who are currently involved with the development of biomimetic devices for 

applications in the marine environment. The intent is to assess the current state of technology 

and incite new applications that apply these and other innovative concepts as technology advances. 

The research presented in this issue mimics specific aspects of fish, skates, rays, sharks, 

lobsters, jellyfish, squids, and sea turtles. In many of these papers, the goal is focused on locomotion 

or propulsion in robotic counterparts so they can maneuver more efficiently in the 

animal’s designed-for environment. This may be acted out either individually or in multiples 

as schools (or swarms) of fish. Swimming and walking using amphibious designs are also considered. 

A recent student competition sponsored by the Office of Naval Research (ONR) was 

held for the first time (involving several of the authors in this issue) to evaluate the performance 

of “mantabots,” which aspire to perform as gracefully as the sleek and elegant manta 

rays they attempt to copy (Pennisi, 2011). A commentary by McKenna provides an overview 

of this and other biomimetic work sponsored by ONR. Other papers in this issue identify a 

single unique aspect and strive to achieve the same benefit in an engineered device. Simulating 

the lateral line system found in most aquatic vertebrates using pressure sensors is one example 

that can be used to supplement vision and sonar in turbid waters. Deriving a control 

mechanism based on shark skin properties to increase vehicle swimming speeds, imitating the 


tubercles of whale fins to enhance hydrodynamics, and engineering a bioinspired solution 

for natural antifouling mechanisms are other examples. Whether your interest is in underwater 

vehicles, imaging, dynamic positioning, marine materials, ocean pollution, remote 

sensing, oceanographic instrumentation, ocean observing, marine science, or education, 

this special issue should provide valuable content. 

References 

Aigeldinger, T.L., & Fish, F.E. 1995. Hydroplaning by ducklings: Overcoming limitations to swimming at the 

water surface. J Exp Biol. 198:1567-4. 

Anderson, J.D. 1998. A History of Aerodynamics. Cambridge: Cambridge University Press. 

Ball, P. 2009. Flow. Oxford: Oxford University Press. 

Borelli, G.A. 1680. De Motu Animalium Pars (The Movement of Animals, translated by P. Maquet (1989)). 

Berlin: Springer-Verlag. 

Fish, F.E. 2002. Balancing requirements for stability and maneuverability in cetaceans. Integr Comp Biol. 

42:85-93. doi: 10.1093/icb/42.1.85. 

Fish, F.E. 2006. Limits of nature and advances of technology in marine systems: What does biomimetics have 

to offer to aquatic robots Appl Bionics Biomech. 3:49-60. doi: 10.1533/abbi.2004.0028. 

Fish, F.E., & Nicastro, A.J. 2003. Aquatic turning performance by the whirligig beetle: constraints on 

maneuverability by a rigid biological system. J Exp Biol. 206:1649-56. doi: 10.1242/jeb.00305. 

Gibbs-Smith, C.H. 1962. Sir George Cayley’s Aeronautics 1796-1855. London: Her Majesty’s Stationery 

Office. 

Harris, B. 1997. The Navy Times Book of Submarines: A Political, Social, and Military History. New York: 

Berkley. 

Kumph, J.M., & Triantafyllou, M.S. 1998. A fast-starting and maneuvering vehicle, the ROBOPIKE. 

In: Proceedings of the International Symposium on Seawater Drag Reduction, ed. Meng, J.C.S., 

pp. 485-90. Newport, Rhode Island. 

Largess, R.P., & Mandelblatt, J.L. 1999. U.S.S. Albacore: Forerunner of the Future. Portsmouth, NH: 

The Portsmouth Marine Society. 

Pennisi, E. 2011. Manta Machines, 332:28-9, Science, 27 May 2001, www.sciencemag.org. Retrieved on 

June 14, 2011. 

Reap, J., Baumeister, D., & Bras, B. 2005. Holism, biomimicry and sustainable engineering. In: Proceedings 

of IMECE2005. November 5-11, 2005. Orlando, FL. 


Taubes, G. 2000. Biologists and engineers create a new generation of robots that imitate life. Science. 288:80-3. 

doi: 10.1126/science.288.5463.80. 

Vincent, J. 1990. Structural Biomaterials. Princeton: Princeton Univ. Press. 

Vogel, S. 1998. Cat’s Paws and Catapults. New York: W. W. Norton. 


COMMENTARY 

Biomimicking Marine Mechanisms 

and Organizational Principles 

AUTHOR 

Yoseph Bar-Cohen 

Jet Propulsion Laboratory, 

California Institute of Technology 

Nature is effectively a giant laboratory 

where trial-and-error evolutionary 

experiments are taking place. As 

nature performs its experiments, all 

the fields of science and engineering 

are employed, including physics, 

chemistry, mechanical engineering, 

and materials science. The processes 

range in scale from nano and micro 

(e.g., viruses and bacteria) to macro 

and mega (e.g., our life scale, elephants, 

and whales). To address the 

many survival challenges, biological 

systems came up with superb solutions. 

The constraints in addressing 

these challenges are similar to those 

that human engineers are facing, including 

the need to maximize the 

functionality of their design and produce 

systems that use minimal resources 

(e.g., materials, energy, cost, 

etc.). Humans have always made efforts 

to use nature as a model for 

inspiring innovation and problem 

solving. However, biological and 

botanical systems have superior 

capabilities, including producing 

materials—they use their body temperature, 

the materials are recyclable, 

and the process does not involve 

pollution. 

To take advantage of the capabilities 

of nature, the field of biomimetics 

involves seeking to understand and 

use of the capabilities as a model for 

copying, adapting, and inspiring concepts 

and designs (Bar-Cohen, 2005; 

Bar-Cohen, 2011; Benyus, 1998; 

Vincent, 2001). For this purpose, 

scientists are seeking rules, concepts, 

mechanisms, and principles to inspire 

new possibilities. Some of the benefits 

that resulted from biomimetic approaches 

have led to improved structures, 

actuators, sensors, interfaces, 

control algorithms, software, drugs, 

defense, and intelligence, and may 

help to improve our ability to recycle 

materials and protect the environment. 

Some of the biomimetic characteristics 

that are being developed include shape 

morphing, self-repair, self-replication, 

and self-reconfiguration (Bar-Cohen 

& Breazeal, 2003; Bar-Cohen, 2011). 

Increasingly, researchers are working 

towards adapting the capabilities of 

many creatures to perform tasks in 

hard-to-reach areas and in conditions 

that are too harsh or dangerous for 

humans. 

Marine biosystems are quite rich 

in capabilities that have and can further 

benefit humans from biomimicking. 

There are many examples that 

one can list, including one as simple 

as the fins that are used by swimmers 

and divers that significantly enhance 

their performance. While it may be 

arguable that the fins were a biologically 

inspired invention, one can state 

that it is common knowledge that 

swimming creatures (e.g., geese, 

swans, seagulls, seals, and frogs) have 

feet with membranes that help them 

swim. The stability, maneuvering, 

and swimming performance of underwater 

animals are determined by the 

morphology, position, and mobility 

of their control surfaces. For cetaceans 

(i.e., whales, dolphins, and porpoises) 

the pectoral flippers are mobile hydrofoils 

that generate lift similar to engineered 

hydrofoils. The flippers have 

various shapes and are used to perform 

lateral turning, dive, surface, 

brake, and other mobility-related 

functions. Studies of the 3-D geometry 

and hydrodynamic performance 

of cetacean flippers with various 

morphologies help provide insight 

into the maneuverability, drag and 

lift performance at high Reynolds 

numbers (Fish et al., 2011). Wagging 

the body and tail is the main propulsion 

method of marine swimmers 

(e.g., billfish and sailfish), allowing 

them to reach significant speeds of 

over 75 km/h. Submarines that 

could perform efficiently as marine 

swimmers using a flexible body 

would be an important exploration 

tool for scientists and would potentially 

have many military applications. 

While significant advances were 

reached via mimicking and the inspiration 

of marine biology, there are many 

capabilities in nature that are still far 

superior to engineered capabilities; 

examples include the following: 

■ The sonar of marine animals is far 

superior to any existing marine 

sonar (Muller & Hallam, 2004). 

■ Sea shells and skeletons of marine 

invertebratesarefarstrongerand 

lighter than human-made materials, 

and their fabrication does not create 

pollution concerns. 


■ Muscles stick to rocks even though 

the adhesion is done in water, and 

they sustain sticksion in spite of 

the strong impacts of ocean waves. 

On the other hand, most humanmade 

adhesives fail when the adhesion 

is done in on a wet surface. 

■ The chiton, which is a diminutive 

mollusk, has very strong teeth that 

it uses to munch on rocks and extract 

food (Weaver et al., 2010). 

This capability may be used to inspire 

the development of effective 

lightweight bits for in situ planetary 

exploration sampling drills 

(Figure 1). 

One hopes that engineers will be 

able to rapidly prototype biological 

capabilities as fast as it is now possible 

to graphically edit photos of biological 

systems, as illustrated in Figure 2. In 

this figure, a photo of a shark was edited 

to create an imaginary image of a 

U.S. Navy vessel that is shark-like. 

While we are somewhat far from this 

capability, significant advances have 

been made by scientists and engineers 

whoseektomimicmarinebiology. 

This special issue is dedicated to describing 

and discussing the latest advances 

that were inspired by marine 

mechanisms and their organizational 

principles. 

FIGURE 2 

An example is shown where software was 

used to turn a photo of a natural shark into 

a vessel-like naval system. The photograph 

(a) was taken by the author; the image 

below (b) shows the modification to a Navy 

marine vehicle and is courtesy of David 

Hanson, Hanson Robotics LLC, TX. 

Acknowledgment 

Some of the research reported in 

this article was conducted at the Jet 

Propulsion Laboratory, California Institute 

of Technology, under a contract 

with the National Aeronautics 

and Space Administration. 

Author: 

Yoseph Bar-Cohen 

Jet Propulsion Laboratory, 

California Institute of Technology 

4800 Oak Grove Drive, 

Pasadena, CA 91109-8099 

Email: yosi@jpl.nasa.gov 

Benyus, J.M. 1998. Biomimicry: Innovation 

Inspired by Nature. New York, NY: Harper- 

Collins Publishers. pp. 1-302. 

Fish, F.E., Smits, A.J., Haj-Hariri, H., 

Iwasaki, T., & Bart-Smith, H. 2011. 

Biomimetic swimmer inspired by the manta 

ray, Chapter 17. In Biomimetics: Nature- 

Based Innovation, ed. Bar-Cohen, Y. Boca 

Raton, FL: CRC Press, Taylor & Francis 

Group. 

Muller, R., & Hallam, J.C.T. 2004. From 

bat pinnae to sonar antennae: augmented 

obliquely truncated horns as a novel parametric 

shape model. In: Proceeding of the 8th 

International Conference on the Simulation 

of Adaptive Behavior, SAB’04. Massachusetts, 

USA: The International Society for Adaptive 

Behavior. 

Vincent, J.F.V. 2001. Stealing ideas from 

nature, Chapter 3. In: Deployable Structures, 

ed. Pellegrino, S. Vienna, Austria: Springer- 

Verlag. pp. 51-58. 

Weaver, J.C., Wang, Q., Miserez, A., 

Tantuccio, A., Stromberg, R., Bozhilov, 

K.N.P., … Kisailus, D. 2010. Analysis of an 

ultra hard magnetic biomineral in chiton 

radular teeth. Mater Today, 13(1-2): 

pp. 42-52. 

FIGURE 1 

A front view of the Eudoxochiton nobilis 

(“Noble” chiton). Courtesy of Iain Anderson. 

References 

Bar-Cohen, Y. (Ed.). 2005. Biomimetics— 

Biologically Inspired Technologies. Boca 

Raton, FL: CRC Press. pp. 1-527. 

Bar-Cohen, Y. (Ed.). 2011. Biomimetics: 

Nature-Based Innovation. Boca Raton, FL: 

CRC Press, Taylor & Francis Group. 

pp. 1-788. 

Bar-Cohen, Y., & Breazeal, C. (Eds.). 2003. 

Biologically Inspired Intelligent Robots. 

Bellingham, WA: SPIE Press. pp. 1-393. 

Vol. PM122. 


COMMENTARY 

Sink and Swim: Clues From Nature 

for Aquatic Robotics 

AUTHOR 

Jeannette Yen 

Georgia Institute of Technology 

There are many ways to travel by 

water. Watercraft can use propellers 

and sails while living organisms exhibit 

quite a variety of modes of transport. 

They sink and swim, flap and 

glide, stroke and jet. From this, we 

see a distinction between the limited 

human solutions and the diverse natural 

solutions. This natural diversity 

continues to inspire inventions in robotics. 

Leonardo da Vinci envisioned 

walking on water (Figure 1), something 

water striders could do millions 

of years ago (Hu et al., 2007). Bathtub 

toys such as the TwiddleFish, 

designed by Chuck Pell of Duke 

University (Figure 2), taught us the 

importance of stiffness in how fish 

FIGURE 1 

Leonardo da Vinci: Shoes for walking on water 

(image from http://en.wikipedia.org/wiki/ 

Walking_on_water). 

swim so well. Pell commented, “In 

the hands of kids…they can really 

feel what’s goingon.” To the delight 

of pre-K–12 children and their parents, 

there are propulsive toys on the 

market with remarkably functional 

body parts, serving as inspiration and 

outreach to youngsters. 

Recent efforts have focused on 

flapping by fish and mantas (e.g., 

Tangorra et al., 2011; Fish et al., 

2011a) or jetting by jellyfish and 

squid (e.g., Moslemi & Krueger, 

2011). By replicating nature via robotics, 

we understand the significance 

of the number of joints (Dean et al., 

2009), the shape of the fin (Curet 

et al., 2011), and the direction of a 

tail swish (Long et al., 2006) for controlling 

movement. Using these robots 

to repeatedly vary the timing of 

undulations of a fish or jellyfish, we 

discover that the frequency of the 

real organism is optimized to capture 

FIGURE 2 

TwiddleFish by Charles Pell, Bio-Design Studio, 

Duke University (cited by Guterl, 1996). Photo 

courtesy of F. Fish. 

the energy shed in the vortices left by 

the previous pulse (Triantafyllou & 

Triantafyllou, 1995; Fish, 2006; 

Ruiz et al., 2010). This is not something 

you could ask a fish or a jellyfish 

to do over and over again (though as 

biologists, we have patiently waited 

and recorded our aquatic creatures 

performing these behaviors over and 

over and over again; see Catton et al., 

2011). We identify the body parts that 

are important for propulsion, and we 

figure out how many propulsors are 

needed, along with their placement. 

Control systems coordinate the multiple 

fins. Sensors integrating different 

sensory modalities are sought to develop 

navigation systems that reliably 

guide the robot to reach its destination. 

We test out new actuators that 

accelerate unsteadily to increase thrust. 

We use new materials for joints or 

bodies with a compliance to achieve 

the flexibility needed for realistic undulations 

and squeezes (Lauder et al., 

2007; Cutkosky & Kim, 2009; Ruiz 

et al., 2010). By matching nature as 

closely as possible, we gain a greater 

understanding of how nature works. 

But is matching reality necessary 

What if instead of understanding 

how shape and structure are optimized 

for best propulsion, we determine 

how to achieve stealth by 

designing robots that blend in with 

the background turbulence or match 

the disturbance made by the other 

members of the school A robot like 

that would enable us to spy on 

schools from the inside out, by becoming 

a schoolmate. Can we study 


how the sensing system is integrated 

with the propulsors to enhance the 

acuity of information capture What 

principles can we abstract from natural 

aquatic propulsion to improve 

how we save energy or save materials 

or improve performance, as all surviving 

species do in a variety of ways to 

suit the constraints of the environment 

in which they have evolved 

Perhaps we can take advantage of the 

environment and glide and surf where 

possible, using the free energy of the 

sea. Oceanographic gliders like the 

Slocum glider Scarlet 27 (Figure 3; 

Schofield et al., 2010) are beautifully 

designed, wherein one version varies its 

ballast by a phase change in response to 

ambient ocean temperature (Webb et al., 

2001). By using whale- or copepodlike 

buoyancy control (Clarke, 1978; 

Pond & Tarling, 2011) and harvesting 

environmental energy in the ocean 

temperature gradient, this glider traveled 

great distances with much less 

energy than other forms of propulsion. 

Simple modifications of shape on 

key control surfaces can lead to large 

variations in the balance of lift and 

drag essential for tight maneuvering, 

as exemplified by the tubercles of 

whale fins that now enable wind turbines 

to capture energy at lower wind 

speeds (Fish et al., 2011b). Can we further 

achieve an economy of materials 

by adopting the streamlined form of 

the shark and applying the hierarchical 

structure of its denticles to fine 

tune that drag reduction (Dean & 

Bhushan, 2010) 

Indeed, looking at the familiar 

players in the sea points out many 

cases of convergent evolution in 

terms of undulatory or jet-like propulsion. 

A closer look reveals other 

unusual adaptations for aquatic mobility 

such as the parachutes of pteropods 

or the floats of siphonophores or 

FIGURE 3 

(A) Autonomous robots that follow the routes of swimming penguins are collecting information 

that could help scientists understand why the birds’ populations are dropping rapidly. The underwater 

robots, called gliders, are programmed to record ocean conditions as they follow the 

tracks of Adelie penguins swimming in the Southern Ocean surrounding Antarctica. (B) Diagram 

of Teledyne-Webb Corporation’s Slocum Glider (coastal model). The Front Main Housing 

Section glider’s ballast, and consequently its flight, is controlled by moving water into or out 

of the Fore Wet Section. From Kahl et al. (2010). (C) The 221-day path taken by Scarlet Knight 

(adapted from Schofield et al., 2010; Kahl et al., 2010; images courtesy of O. Schofield: http:// 

rucool.marine.rutgers.edu). 

the multiple oars of copepods, krill, 

and polychaetes. Analyses of their 

transport mechanisms may again 

open our eyes to novel designs of future 

underwater vehicles that can 

steadily hover, smoothly cruise, rapidly 

escape, quietly sink or perform any 

gait as needed in tight quarters. With 


the rapid evolution occurring in materials 

science research and in control 

systems, we may find ourselves traveling 

through fluids in unexpected 

ways. 

Author: 

Jeannette Yen 

School of Biology, 

Center for Biologically Inspired Design 

Georgia Institute of Technology, 

Atlanta, GA. 30332-0230 

Email: jeannette.yen@biology.gatech. 

edu 

References 

Catton, K.B., Webster, D.R., Kawaguchi, S., 

& Yen, J. 2011. The hydrodynamic disturbances 

of two species of krill: Implications for aggregation 

structure. J Exp Biol. 214:1845-56. 

Clarke, M.R. 1978. Buoyancy control as a 

function of the spermaceti organ in the sperm 

whale. J Mar Biol Assoc UK. 58:27-71. 

doi: 10.1017/S0025315400024395. 

Curet, O.M., Patankar, N.A., Lauder, G.V., 

& MacIver, M.A. 2011. Aquatic manoeuvering 

with counter-propagating waves: A novel 

locomotive strategy. J Roy Soc Interface. 

8(60):1041-50. doi: 10.1098/rsif.2010.0493. 

Cutkosky, M.R., & Kim, S. 2009. Design 

and fabrication of multi-material structures for 

bioinspired robots. Philos T R Soc A. 

367:1799-813. doi: 10.1098/rsta.2009.0013. 

Dean, B., & Bhushan, B. 2010. Shark-skin 

surfaces for fluid-drag reduction in turbulent 

flow: A review. Philos T R Soc A. 368(1929): 

4775-806. doi: 10.1098/rsta.2010.0201. 

Dean, M.N., Swanson, B.O., & Summers, 

A.P. 2009. Biomaterials: Properties, variation 

and evolution. Integr Comp Biol. 49(1): 

15-20. doi: 10.1093/icb/icp012. 

Fish, F.E. 2006. Limits of nature and advances 

of technology: What does biomimetics have to 

offer to aquatic robots Appl Bionics Biomech. 

3(1):49-60. doi: 10.1533/abbi.2004.0028. 

Fish, F.E., Nichols, R.H., Dudas, M.A., 

Moored, K.W., & Bart-Smith, H. 2011a. 

Kinematics of swimming in the manta ray 

(Manta birostris): 3D analysis of open 

water maneuverability. Integr Comp Biol. 

51(Suppl. 1):E42. 

Fish, F.E., Weber, P.W., Murray, M.M., & 

Howle, L.E. 2011b. The tubercles on 

humpback whale’s flipper: Application of 

bio-inspired technology. Integr Comp Biol. 

51(1):203-13. doi: 10.1093/icb/icr016. 

Guterl, F. 1996. Playthings of Science. 

http://discovermagazine.com/1996/dec/ 

playthingsofscie946. (accessed 15 July 2011). 

Hu, D., Chan, B., & Bush, J.W.M. 2007. 

Water-walking devices. Exp Fluids. 43: 

769-78. doi: 10.1007/s00348-007-0339-6. 

Kahl, L., Oscar Schofield, A., & Fraser, W.R. 

2010. Autonomous gliders reveal features 

of the water column associated with foraging 

by adelie penguins. Integr Comp Biol. 

50(6):1041-50. doi: 10.1093/icb/icq098. 

Lauder, G.V., Anderson, E.J., Tangorra, J., & 

Madden, P.G. 2007. Fish biorobotics: Kinematics 

and hydrodynamics of self-propulsion. 

J Exp Biol. 210:2767-80. doi: 10.1242/ 

jeb.000265. 

Long, J.H.J., Koob, T.J., Irving, K., Combie, 

K., Engel, V., Livingston, N., … Schumacher, 

J. 2006. Biomimetic evolutionary analysis: 

Testing the adaptive value of vertebrate tail 

stiffness in autonomous swimming robots. 

J Exp Biol. 209:4732-46. doi: 10.1242/ 

jeb.02559. 

Moslemi, A.A., & Krueger, P.S. 2011. The 

effect of Reynolds number on the propulsive 

efficiency of a biomorphic pulsed-jet 

underwater vehicle. Bioinspir Biomim. 

6(2):026001. doi: 10.1088/1748- 

3182/6/2/026001. 

Pond, D.W., & Tarling, G.A. 2011. Phase 

transitions of wax esters adjust buoyancy 

in diapausing Calanoides acutus. Limnol 

Oceanogr. 56(4):1310-8. 

Ruiz, L.A., Whittlesey, R.W., & Dabiri, J.O. 

2010. Vortex-enhanced propulsion. 

J Fluid Mech. 668:5-32. doi: 10.1017/ 

S0022112010004908. 

Schofield, O., Kohut, J., Glenn, S., Morell, 

J., Capella, J., Corredor, J., … Boicourt, W. 

2010. A regional Slocum glider network in the 

Mid-Atlantic coastal waters leverages broad 

community engagement. Mar Technol Soc J. 

44(6):64-74. 

Tangorra, J., Phelan, C., Esposito, C., & 

Lauder, G.V. 2011. Use of biorobotic models 

of highly deformable fins for studying the 

mechanics and control of fin forces in fishes. 

Integr Comp Biol. 51(1):176-89. doi: 

10.1093/icb/icr036. 

Triantafyllou, M.S., & Triantafyllou, G.S. 

1995. An efficient swimming machine. 

Sci Am. 272:64-70. doi: 10.1038/ 

scientificamerican0395-64. 

Webb, D.C., Simonetti, P.J., & Jones, C.P. 

2001. SLOCUM: An underwater glider 

propelled by environmental energy. 

IEEE J Oceanic Eng. 26(4):447-52. 

doi: 10.1109/48.972077. 


COMMENTARY 

Developing Bioinspired Autonomous Systems 

AUTHOR 

Thomas M. McKenna 

Office of Naval Research 

Research that seeks to identify the 

principles, strategies, and mechanisms 

used by motile aquatic animals offers 

opportunities to develop undersea 

vehicles that exceed current capabilities 

and enable the Navy to expand 

the operational envelope of autonomous 

undersea vehicles. Autonomous 

undersea vehicles serve in a number of 

important current and emerging roles 

in Navy missions, including surveillance 

for anti-submarine warfare, 

mine countermeasures, Improvised 

Explosive Device (IED) detection 

and localization, force protection (e.g., 

counter-diver missions) in harbors, and 

riverine exploration and characterization. 

However, current unmanned undersea 

vehicles (UUVs) have technical 

gaps in areas such as mission duration 

(due to power constraints), excessive 

noise generation, speed, limited maneuverability, 

lack of tight integration 

of sensing and maneuver, propulsion 

and maneuver dead zones at low 

speeds, and inability to operate in the 

surf zone. Considering the capabilities 

of sea creatures for efficient propulsion 

over a large range of speeds, their ability 

to operate in or even exploit energetic 

ocean environments like the surf 

zone, extraordinary maneuverability, 

and the special sensing evolved for predation, 

schooling and navigation, 

there are many lessons for technologists. 

The basic research programs in 

Bio-Inspired Autonomous Systems at 

the Office of Naval Research (ONR) 

have supported research using four 

approaches: (1) the identification, 

modeling, and emulation of the biomechanics 

and fluid mechanics of 

underwater propulsion and control 

in swimming organisms, (2) identification 

and exploration of the 

sensorimotor control of animals and 

integrated closed loop control using 

biosensing (e.g., biosonar, electrosense, 

lateral line sensors, optic flow, 

magnetic sense), (3) the development 

of muscle-like actuators and fin designs 

that exploit these materials, 

and (4) design and development of 

swimming prototype Autonomous 

Underwater Vehicle (AUVs) as proof 

of principle and for performance 

evaluation. These studies take their 

inspiration from diverse sea creatures 

and include high-performance swimmers 

like bluefin tuna, squid, rays 

and seals; animals that thrive in the 

surf zone like lobsters and crabs; and 

animals with special senses like 

biosonar (i.e., dolphins) and electrosense 

(i.e., sharks and ribbonfish). 

More recently, opportunities have 

emerged for microrobotic underwater 

systems capable of sensing, reporting, 

and remediation of environmental 

chemicals that exploit the convergence 

of synthetic biology, nanotechnology, 

and electro-optic systems. 

This has prompted renewed interest 

in the propulsion biology of organisms 

that use cilia and flagella for 

locomotion. 

Key Science and 

Technology Issues 

There are a number of key issues 

for the development of bioinspired 

autonomous systems. 

1. Developing high-efficiency 

propulsion that exceeds the capability 

of propellers. This is particularly 

important for achieving long-duration 

missions. Although there is a sizable 

Navy effort to develop new energy 

sources for underwater vehicles, introduction 

of more efficient propulsion 

systems would reduce the power requirements 

and thereby lengthen mission 

durations. Early efforts to mimic 

the caudal fins of high-performance 

swimmers like tuna did not produce 

performance that exceeds propellers 

(but recent work on the Ghostswimmer 

TM , discussed in Rufo and Smithers’ 

commentary in this issue, looks promising), 

and Bandyopadhyay (2005) has 

produced a meta-analysis showing that 

animals do not have an advantage over 

man-made systems in cruise, but animals 

do show greater maneuverability 

relative to man-made underwater 

vehicles. One promising new form of 

propulsion is the use of bioinspired 

high-lift foil propulsors. High-lift propulsion 

was introduced in the biological 

context in the analysis of fly wing 

lift (Bandyopadhyay, 2009; Ellington, 

1984; Dickinson et al., 1999). Fly 

wings, which pitch and heave with a 

90° phase difference, can achieve lift 

coefficients substantially greater than 

rigid foils that have a constant angle 

of attack, and high-lift foil propulsors 

should be capable of an order of magnitude 

greater lift than traditional 

naval propellers. High-lift foils also 

produce substantially less noise than 

traditional propellers. Bandyopadhyay 

et al. (2008) have produced a series of 

rigid high-lift foils (roughly analogous 

to penguin pectoral fins), characterized 


their propulsion properties, and 

mounted them on an undersea vehicle. 

This first version was called bioinspired 

autonomous undersea vehicle (BAUV) 

and a second, lower-diameter version 

was called self-propelled line array 

(SPLINE) (Figures 1 and 2). High-lift 

flapping foils are efficient, and combining 

six multiple foils on a vehicle can 

achieve extraordinary maneuverability 

and hover and exhibit low-noise emission, 

but the flapping foil configuration 

is not consistent with producing high 

speeds. Recently, Bandyopadhay has 

designed a new propulsor called 

“Slosher” that has multiple foils with 

variable angles of attack arranged radially 

that can operate as a high-lift propulsor 

at low speeds or, with the foils 

locked, as a traditional propeller at 

high speeds. This avoids the deadband 

of controllability at low speeds. 

Moreover, hybrid vehicles, such as the 

RAZOR (Figure 3), have been developed 

that combine four high-lift 

foils and two props. At low speeds, 

the high-lift foils provide high maneuverability 

and hover, but when the 

props provide higher speeds for transit 

of the vehicle, the foils become control 

surfaces. 

FIGURE 1 

BAUV developed by Dr. Promode Bandyopadyay 

at the Naval Undersea Warfare Center, Newport, 

RI (NUWC-NPT). The BAUV has six high-lift 

foils and a controller based on a model of the 

olivo-cerebellar circuits of mammals. 

FIGURE 2 

SPLINE. This is a refinement of the BAUV, designed 

for long-duration testing of its ability 

to pull a load, maintain accurate position, and 

keep a line taut that has been fixed at the distal 

end while maneuvering (NUWC-NPT). 

FIGURE 3 

RAZOR vehicle. This vehicle was developed by 

Richard Berube and Promode Bandyopadyay 

at NUWC-NPT. It has four high-left foils and 

two rotating propulsors. At low speeds, it 

uses the foils for high maneuverability. 

2. Developing adaptive controllers 

for high-degree-of-freedom 

bioinspired propulsors. Many of 

the bio-inspired propulsion systems 

exhibit high degrees of freedom. For 

the case of the high-lift flapping 

foils,roll,pitch,andfrequencyare 

motion parameters to be specified to 

achieve efficient fin kinematics, and 

for multiple fin vehicles, the phasing 

of the fins is critical. In addition to 

commanding these parameters, for a 

given vehicle maneuver, the fin controller 

must respond to perturbations. 

Fortunately, there are bioinspired solutions 

to such control issues. One of 

the key components of the mammalian 

motor control system is the olivocerebellar 

system. Coupled neurons 

in the inferior olive generate a 10-Hz 

rhythm, and animal muscle contractions 

are initiated from particular 

phases of this rhythm. The inferior 

olive has multiple oscillating domains, 

with phase shifts between these domains, 

and these domains are under 

the control of the cerebellar cortex, 

which is a recipient of sensory inputs 

from many modalities. Llinas et al. 

(2004) developed a model whereby 

sequences of motor commands can 

be generated by the olivo-cerebellar 

system. The biophysics of coupling 

of these neurons also promotes a 

phase reset property for rapid synchronization 

(Kazantsev et al., 

2004). Bandyopadhyay (2008) has 

implemented this model in analog circuitsandhasshownittobeaneffective 

controller of power and thrust in 

thehigh-liftfoilsandtheabilityto 

rapidly return to normal following 

perturbations. This model is able to 

synchronize across multiple fins in 

order to achieve optimum gaits that 

minimize pitching and rolling of the 

BAUV.Thiscontrollerenabledthe 

BAUV to precisely hold a line load 

for 20 days. Based on these laboratory 

results, the BAUV vehicle is predicted 

to support a mission duration of 

3 weeks with current battery technology, 

although this assumes most of 

that time is spent in hover. 

For fish and eels that use whole 

body or caudal fin propulsion,the 

oscillations are generated by central 

pattern generators. However, the modulation 

of these neural generators 

during locomotion to achieve desired 

thrust and vectors has not been fully 

characterized for animals, and using 

this approach in artificial systems requires 

learning or genetic algorithms 

to tune them to particular maneuvers. 


3. Exploitation of fish swimming 

modes for underwater vehicle propulsion 

and maneuvers. Fish have a 

number of swimming modes that are 

worthy of consideration for emulation. 

Fish swimming types can be 

classified as either body and/or caudal 

fin movements vs. those that use median 

and/or paired fin propulsion. 

One review of fish swimming modes 

(Sfakiotakis et al., 1999) singled out 

lunate tail propulsion (e.g., tuna), undulating 

fins (e.g., some rays), and 

labriform (oscillatory pectoral fin) 

swimming mechanisms as having the 

greatest potential for exploitation in 

artificial systems. ONR is currently 

supporting research to exploit all 

three of these mechanisms in addition 

to the gymnotiform mode that involves 

undulations of a long ventral 

median fin. Batoid fishes utilize one 

of two modes of locomotion, employing 

either undulatory (passing multiple 

waves down the fin orbody) 

or oscillatory (flapping) kinematics 

(Rosenberger, 2001). An ambitious 

effort to build batoid vehicles, using 

predominately oscillating kinematics, 

is being undertaken by a team led 

by Hillary Bart-Smith (described in 

Moored, Fish, Kemp, & Bart-Smith, 

this issue) supported by an ONR program 

managed by Robert Brizzolara. 

Two prototype manta vehicles were 

recently demonstrated in a student 

competition (Pennisi, 2011). 

Exploitation of fish locomotion 

using lunate tail fins began with 

the seminal studies of Triantafyllou 

(Barrett et al., 1996; Anderson et al., 

1998) analyzing how bio-inspired 

foils create and exploit vortex structures. 

His laboratory also developed 

the interesting Robotuna (based on 

the bluefin tuna) and Robopike prototypes, 

which were propelled using 

caudal fins. Triantafyllou’s group 

demonstrated very high-propulsion 

efficiencies (up to 91%) for the 

Robotuna; however, the Robotuna 

did not achieve very high speeds. 

One of the creators of Robotuna, 

David Barrett at Olin College, is now 

teamed with Boston Engineering to 

develop the Ghostswimmer vehicle 

(described in Rufo’s commentary 

in this issue). The objective of the 

Ghostswimmer project is to exceed 

the performance of current similar 

size UUVs on speed, maneuver, mission 

duration, noise, rapid response 

and cost. The goal is to demonstrate 

the capability to conduct fully autonomous 

missions. 

Another active research area seeks 

to exploit the principles of fish pectoral 

fins for propulsion and maneuver. 

Fish pectoral fins are highly flexible 

and enable a high degree of precise 

maneuver and station keeping in 

currents. Sunfish fins have flexible 

rays, and attached membranes exhibit 

a cupping motion on the forward 

stroke that produces upper and lower 

leading edge vortices. These fins generate 

positive thrust throughout the fin 

beat, and turning involves asymmetric 

use of these fins. Small prototype fins 

with this ray and membrane structure 

have been shown in the laboratory, 

but no free swimming vehicles have 

yet been developed (Gottlieb et al., 

2010; Tangorra & Lauder, 2011, this 

issue). Rigid pectoral fins have been 

implemented on a number of swimming 

vehicles, but these are mainly 

used as control surfaces. 

Aquatic animals that are propelled 

by jetting can also provide inspiration 

for novel propulsion mechanisms. 

Mohseni (Krieg & Mohseni, 2010; 

Krieg et al., this issue) has characterized 

the biomechanics and hydrodynamics 

of squid propulsion and 

developed new bioinspired thrusters, 

and Priya and his colleagues (see 

Joshi et al., this issue) have built prototype 

jellyfish that closely mimic 

their swimming modes. 

4. Development of muscle-like 

actuators. To fully exploit animallike 

locomotion and sensorimotor 

control mechanisms, it is essential to 

develop muscle-like actuators, linear 

actuators, adaptive compliant structural 

materials, and elastic skins with 

properties much closer to biological 

systems than current technology. 

There are substantial inefficiencies 

with implementing bio-inspired locomotion 

using rotary motors and complex 

power transmission systems. Two 

recent efforts to develop fins based on 

ionic polymer-metal composites are 

described in the papers by Kim et al. 

and Tan in this issue. 

5. Closed-loop control of bioinspired 

underwater vehicles. In 

crustaceans, there are detailed accounts 

of sensorimotor reflexes for control of 

tail and legs, and robotic lobsters have 

been built with many levels of bioinspiration 

(Ayers & Crisman, 1992; 

Ayers et al., this issue). However, for 

fish, the neural mechanisms by which 

the sensory afferent information on 

motion, flows, and hydroacoustic 

pressure is processed, integrated and 

relayed to motor neurons is not 

known. One reason for this is that 

electrophysiological recording in 

swimming fish is technically challenging. 

One open question is whether fish 

can sense vortices and maneuver or 

time fin movements to exploit them 

or avoid them. The ONR is currently 

supporting a number of efforts in 

closed-loop control and bionavigation. 

These projects include identifying the 

role of hydroacoustic receptors in the 

lateral line and fin andbodymechano-reception 

in flow and vortex 

sensing and tracing the connectivity 


of these receptors into spinal sensorimotor 

circuits involved in locomotion 

and fin control (Green et al., 2011; 

Green & Hale, 2011). This project is 

also developing robotic fish prototypes 

(see Tangorra et al., this issue; Lauder 

et al., this issue) with detailed pectoral 

fins. This work entails some basic research 

on fish neurophysiology, since 

circuit-level neurophysiology is much 

more challenging in fish than in terrestrial 

vertebrates, and many of the 

spinal motor reflexes and circuits involved 

in limb control that were well 

established for mammals by the 

1980s are still in nascent stage for the 

pectoral fins of fish. 

6. Exploitation of special senses 

of aquatic animals. Anumberof 

fish species and sharks are able to navigate 

and search for prey using electroreception; 

some animals also exploit 

geomagnetic signals for navigation, 

and the biosonar of dolphins supports 

a range of behaviors. The ribbon fish 

can sense prey in murky waters using 

electrosense and then use its ventral 

ribbon fin to maneuver to this prey. 

Ongoing ONR projects are performing 

system identification of visual 

and eletroreceptive feedback control of 

locomotion in the ribbon fish (Roth 

et al., 2011; Mitchell et al., 2011) to 

build a working electrosense and electromagetosense 

module for navigation 

and target localization and to build a 

prototype ribbon fish (“Ghostbot”) 

(Curet et al., 2011). A vehicle designed 

to sense electric and magnetic anomalies 

could enable new mine countermeasures 

systems or enable navigation 

that does not depend on GPS or 

expensive Inertial Measurement Units 

(IMUs). 

Sharks exhibit exquisite sensitivity 

to perturbations of electric fields and 

one current project seeks to develop 

new highly sensitive electromagnetic 

and hydroacoustic sensors based on 

shark sensor biophysics (Kalmijn, 

Scripps). 

Dolphins have extraordinary abilities 

to recognize objects using biosonar. 

ONR has supported research characterizing 

this ability in terms of psychoacoustics, 

and several dolphin-inspired 

sonars have been developed and demonstrated. 

Recently, Forsythe et al. 

(2008) demonstrated closed-loop 

control of his bio-inspired autonomous 

underwater vehicle using a simple 

dolphin-inspired biosonar in the exploration 

of acoustic targets. Dolphins, 

however, have the disadvantage that 

one cannot analyze directly the neural 

circuits involved in their biosonar. 

Hence, from the earliest days of its existence, 

ONR has supported the study 

of bat biosonar. The neural circuits of 

bat biosonar have been well characterized. 

Recently research has focused on 

how bats use multistatic biosonar for 

obstacle avoidance. Bats not only have 

an extraordinary ability to navigate 

around obstacles and strike prey 

using their own sonar, but they can 

also perform these feats using the returns 

from calls emitted by other bats 

in a swarm. This has implications for 

the cooperative behavior of multiple 

AUVs. 

7. Group behaviors. The ONR 

Science of Autonomy program is supporting 

efforts to identify the principles 

of fish schooling and applying them to 

multiple AUVs conducting ocean surveys. 

Such studies have addressed the 

minimal sensing requirements of fish 

used in schooling, including vision. 

Additionally, for AUVs, one could conceivably 

use man-made communications 

like acoustic modems or lasers to 

enhance coordinated actions, but successful 

emulation of sense modalities 

used in nature (i.e., fish hydroacoustic 

sensing) might greatly simplify coordination 

of multiple vehicles—in particular 

when a large number of vehicles 

is operating at close quarters. There is 

also an effort to emulate the group 

hunting behaviors of dolphins in intelligent 

control systems for cooperative 

autonomous vehicles. 

8. The Navy is interested in distributed, 

persistent sensing systems 

for tasks such as anti-submarine warfare. 

One bioinspired approach to 

this is to develop small sensor platforms 

that float in the water column 

with limited mobility, similar to jellyfish. 

An ambitious effort to exploit 

multifunctional materials and build 

such jellyfish colonies (Priya) is underway. 

Additionally, the emerging convergence 

of nanotechnology, synthetic 

biology and micro-optoelectronics has 

fostered the nascent development of 

very small autonomous systems. Some 

of the new microrobots being designed 

have components consisting of colonies 

of micro-organisms and hybrid 

biomolecules, with motility being 

provided by a multitude of cilia or 

flagella. These new systems could 

function as distributed sensors of hazardous 

materials, but with integrated 

mobility, mitigation and reporting 

capabilities. 

In summary, there are a number of 

promising opportunities for the Navy 

to extend the capabilities of autonomous 

undersea platforms using bioinspired 

technologies in propulsion, 

control and sensing. 

Author: 

Dr. Thomas McKenna 

Office of Naval Research 

Division of Human and 

Bioengineered Systems 

875 N. Randolph St., Suite 1425 

Arlington, VA 22203-1995 

Email: tom.mckenna@navy.mil 


References 

Anderson, J.M., Streitlien, K., Barrett, D.S., 

& Triantafyllou, M.S. 1998. Oscillating 

foils of high propulsive efficiency. J Fluid 

Mech. 360:41-72. doi: 10.1017/ 

S0022112097008392. 

Ayers, J., & Crisman, J. 1992. The lobster as 

a model for an omnidirectional robotic 

ambulation control architecture. In: Biological 

Neural Networks in Invertebrate Neuroethology 

and Robotics, eds. Beer, R., Ritzmann, R., 

& McKenna, T., 287-316. San Diego: 

Academic Press. 

Bandyopadhyay, P.R. 2005. Trends in biorobotic 

autonomous undersea vehicles. IEEE J 

Oceanic Eng. 30(1):109-39. doi: 10.1109/ 

JOE.2005.843748. 

Bandyopadhyay, P.R. 2009. Swimming and 

flying in nature—The route toward applications: 

The Freeman Scholar lecture. ASME J Fluid 

Eng. 131:031801-29. doi: 10.1115/1.3063687. 

Bandyopadhyay, P., Beal, D.N., & Menozzi, 

A. 2008. Biorobotic insights into how animals 

swim. J Exp Biol. 211:206-14. doi: 10.1242/ 

jeb.012161. 

Bandyopadhyay, P.R. 2008. Synchronization 

of animal-inspired multiple high-lift fins 

in an underwater vehicle using olivo-cerebellar 

dynamics. IEEE J Oceanic Eng. 33(4): 

563-78. doi: 10.1109/JOE.2008.2005356. 

Barrett, D., Grosenbaugh, M., & Triantafyllou, 

M. 1996. The optimal control of a flexible 

hull robotic undersea vehicle propelled by an 

oscillating foil. In: Proc. 1996 IEEE AUV 

Symp. 1-9. 


& MacIver, M.A. 2011. Mechanical properties 

of a bio-inspired robotic knifefish with an 

undulatory propulsor. Bioinspir Biomim. 

6(2). doi: 10.1088/1748-3182/6/2/026004. 

Dickinson, M.H., Lehman, F., & Sane, S.P. 

1999. Wing Rotation and the aerodynamic 

basis of insect flight. Science. 284:1954-60. 

doi: 10.1126/science.284.5422.1954. 

Ellington, C.P. 1984. The aerodynamics 

of hovering insect flight. IV. Aerodynamic 

mechanisms. Philos T R Soc Lond B. 

305:79-113. 

Forsythe, S.E., Leinhos, H.A., & 

Bandyopadhyay, P.R. 2008. Dolphininspired 

combined maneuvering and pinging 

for short-distance echolocation. J Acoust 

Soc Am. 124(4):255-61. 

Gottlieb, J.R., Tangorra, J.L., Esposito, 

C.J., & Lauder, G.V. 2010. A biologically 

derived pectoral fin for yaw turn maneuvers. 

Appl Bionics Biomech. 7:41-55. 

doi: 10.1080/11762320903537782. 

Green, M.H., Ho, R.K., & Hale, M.E. 2011. 

Movement and function of the pectoral fins 

of the zebrafish (Danio rerio) during slow 

swimming. J Exp Biol. in review. 

Green, M.H., & Hale, M.E. 2011. Motoneurons, 

muscles and movement: Neural 

control of the pectoral fins during slow and 

fast swimming. Integr Comp Biol. E51. 

Kazantsev, V.B., Nekorkin, V.I., Makarenko, 

V.I., & Llinas, R. 2004. Self-referential phase 

reset based on inferior olive oscillator dynamics. 

P Natl Acad Sci. 101(52):18183-8. 

doi: 10.1073/pnas.0407900101. 

Krieg, M., & Mohseni, K. 2010. Dynamic 

modeling and control of biologically inspired 

vortex ring thrusters for underwater robot 

locomotion. IEEE T Robot. 26(3):542-54. 

doi: 10.1109/TRO.2010.2046069. 

Llinas, R., Leznick, E., & Makarenko, V.I. 

2004. The olivo-cerebellar circuit as a universal 

motor control system. IEEE J Ocean Eng. 

29(3):631-9. doi: 10.1109/JOE.2004.833212. 

Mitchell, T., Fortune, E.S., & Cowan, N.J. 

2011. Reweighting of vision and electrosensation 

during locomotion in the Glass Knifefish. 

PLoS, to be submitted. 

Pennisi, E. 2011. Manta machines. Science. 

332:28-9. doi: 10.1126/science.332.6033.1028. 

Rosenberger, L.J. 2001. Pectoral fin locomotion 

in batoid fishes: Undulation versus 

oscillation. J Exp Biol. 204:379-94. 

Roth, E., Zhuang, K., Stamper, S., Fortune, 

E.S., & Cowan, N.J. 2011. Stimulus predictability 

mediates a switch in locomotor 

smooth pursuit performance for Eigenmannia 

virescens. J Exp Biol. 214:1170-80. 

doi: 10.1242/jeb.048124. 

Sfakiotakis, M., Lane, D.M., & Davies, 

J.B.C. 1999. Review of fish swimming modes 

for aquatic locomotion. IEEE J Oceanic Eng. 

24(2):237-52. doi: 10.1109/48.757275. 

Tangorra, J.L., & Lauder, G.V. 2011. Biorobotic 

models of highly deformable fins for 

Studies of the mechanics and control of 

fin forces. Integr Comp Biol. in press. 

doi: 10.1093/icb/icr036. 


COMMENTARY 

GhostSwimmer AUV: Applying Biomimetics 

to Underwater Robotics for Achievement 

of Tactical Relevance 

AUTHORS 

Michael Rufo 

Mark Smithers 

Boston Engineering Corporation 

Introduction 

I 

t is clear that unmanned underwater 

vehicles (UUVs), autonomous 

underwater vehicles (AUVs), and 

other waterborne robots are successfully 

addressing critical capability 

requirements for many customers, including 

those in defense and oil and 

gas. There remains, however, despite 

the best efforts of many entities, capability 

gaps for these systems. 

While UUVs share many of the 

same navigation, power, and logistical 

challenges as their unmanned ground 

vehicle (UGV) and unmanned aerial 

vehicle (UAV) counterparts, they are 

subject to additional challenges including 

limited communications and harsh 

environments. 

Developing biologically-inspired 

(or biomimetic) technologies can provide 

some guidance for next generation 

systems. The Advanced Systems 

Group (ASG) at Boston Engineering 

(Waltham, MA) is developing the 

GhostSwimmer 

 

AUV, for example, 

which endeavors to attack many of the 

problems facing current UUVs. The 

increasing interest in the use of longrange/long-duration 

UUVs for littoral 

observation, military surveillance, and 

other missions demands alternative or 

new technology development. As per 

the U.S. Navy UUV Master Plan 

(2004), the areas of autonomy, sensors, 

and communications are 

among the leading areas of interest. 

Energy and propulsion are also main 

areas of interest as per this document, 

where it states that “advanced energy 

and propulsion, in combination with 

other UUV technologies, will enable 

the use of smaller vehicles (reducing 

cost) in the long term, and will provide 

greater performance.” (The 

Navy UUV Master Plan, 2004). 

GhostSwimmer is a tactical, biomimetic 

autonomous “artificial fish” 

UUV that employs the mechanics 

and dynamics of biological systems to 

create efficient swimming and high 

maneuverability while remaining responsive 

to the needs of current riverine 

and littoral missions (among other 

possibilities). Its importance lies in its 

ability to provide advanced mobility 

in a system that employs payloads. 

Considerable work has been completed 

on understanding the propulsive 

characteristics of individual fish 

fins (extensive information on oscillating 

foil propulsion exists). The 

GhostSwimmer 

 

builds upon this 

and is modeled after a tuna. It is propelled 

by a composite construction lunate 

caudal tail fin (takingitsname 

from its crescent moon-like profile). 

In the biological tuna, almost all of 

theusablethrustisgeneratedbythe 

oscillating tail fin. Magnuson describes 

it as “a tapered hydrofoil with high 

aspect ratio, curved leading edge and 

moderate sweep back. In cross section, 

the fin is shaped like a thin symmetrical 

airfoil with a rounded anterior, or 

leading edge, and a sharp posterior or 

trailing edge” (Magnuson, 1978). 

This oscillating foil has the capability 

of producing thrust at high efficiencies. 

Triantafyllou and Triantafyllou 

(1993) stated “Oscillating foils are 

known under certain conditions to 

produce substantial thrust at high propulsive 

efficiency. Fish and cetaceans 

have developed through evolution a 

presumed optimal manner of propulsion 

primarily through flapping of 

the posterior part of their body. It is 

shown that thrust production depends 

significantly on the dynamics of the 

unstable wake formed behind the foil 

or fish tail: thrust develops through 

the formation of a reverse Karman 

street, whose preferred Strouhal numbers 

are between 0.25 and 0.35. Experimental 

data from flapping airfoils 

and data from fish observations confirm 

the theoretical predictions.” 

Triantafyllou and Barrett were able to 

construct a precision oscillating foil 

test apparatus with which propulsive 

efficiencies in the range of 80% 

were repeatedly measured (Barrett, 

1996). 

While based on this concept, 

GhostSwimmer 

 

strives to further 

the understanding of how fish use all 

their fins to enhance propulsive performance, 

maneuverability, and stealth 

characteristics (all the while providing 

a capability for the Warfighter). 


Overcoming the Challenges 

for the Future AUVs 

At this time, UUVs and AUVs 

struggle with complex underwater 

environments such as shallow and 

obstacle-filled waters. These vehicles 

are subject to an intrinsic paradox; 

they require stable dynamic behavior 

for maintaining intended path/heading 

without changing control surfaces or 

propulsion force. However, in complex 

environments, the vehicle must also be 

able to quickly change both path and 

speed safely. Dynamic stability and response 

aspects such as overshoot, rise 

time, and peak time become critical 

(Azarsinal et al., 2007). Based on their 

prowess in these areas, it makes sense to 

investigate biological sources of inspiration 

in the development of new 

man-made systems hoping to excel in 

these environments. 

The Boston Engineering Advanced 

Systems Group’s GhostSwimmer 

 

AUV (Figure 1) is intended to leverage 

this biological inspiration. Existing 

AUVs are generally propelled by rotary 

propellers driven by electric motors 

with energy stored in batteries. Small 

diameter propellers typically operate 

at low efficiencies and can suffer serious 

lag times in transient response 

(Barrett, 1996). Cost effective propeller 

improvements are often limited by 

a maximum practical diameter that can 

be mounted on a UUV. In complex 

underwater areas, rotating propellers 

FIGURE 1 

Boston Engineering’s GhostSwimmer PH I 

AUV (sponsor: ONR Code 341). 

also present a significant snagging 

risk. Energy technology, despite recent 

progress, is awaiting a breakthrough to 

provide significantly larger (and safer) 

power densities; therefore, producing 

a technology that can address both 

efficiency and control, independent 

of battery technology, has benefits. 

Additional challenges that UUVs 

are subject to range from being as simple 

as packaging electronics to be water 

tight at depth (versus splash-proof ), 

to using materials that are resistant to 

aggressive corrosion, to being able to 

localize underwater without the aid of 

conventional techniques (such as the 

GPS or radio frequency [RF] communications). 

As discussed later in this 

commentary, biomimetics can offer 

guidance in these areas as well. 

Biomimetics and 

Pragmatism 

Biomimetics (from the Greek bios 

[life] + mimesis [to imitate]) is not 

new. Mankind has been mimicking 

nature in products for centuries. Simple 

everyday items such as salad tongs 

(based on bird beaks) through complex 

sensors such as sonar (based on 

dolphins and bats) have each been 

generated by mimicking biology. 

Even human terminology mimics nature, 

consider saw “teeth”, computer 

“viruses” and “worms”, orlaptops 

that “hibernate.” Engineers and researchers 

have even mimicked plants 

in robotic systems (turning to face 

the sun for charging, for example) 

(Bar-Cohen, 2006). It should be 

noted that this discussion is not 

about synthetic life (involving using 

biological components to build biological 

systems). 

Nature evolves by responding to 

needs for surviving to the next generation. 

It benefits from millions of 

variations and trial and error experiments 

over the millennia. Engineers 

trying to develop tactically relevant 

technologies obviously must “evolve” 

much faster and in a pragmatic manner. 

As such, developing biomimetic 

technologies is as much about deciding 

what not to imitate as it is deciding 

what to imitate (Figure 2). This is because 

direct and absolute mimicry is 

often not appropriate or necessary. 

Consider that mankind flies without 

flapping wings as birds do. While understanding 

and mimicking the control 

surface design has direct benefits 

to engineer fast and high flying aircraft, 

engineers diverted from nature’s design 

with great success. However, the 

definition of the mission must always 

be at the core of any engineering effort. 

If the goal were to perch on a power 

line, even mankind’s best aircraft 

would fail. 

Evolutionary improvements in nature 

also do not always exactly coincide 

with the reasons for mimicking them. 

This should be considered when deciding 

whether the model is appropriate 

or ideal for mimicking in an engineered 

system designated for use in 

critical applications. Clearly, biologists 

and others trying to learn about “how 

biology works” are concerned with 

mimicking exactly how a biological 

FIGURE 2 

Early attempts at biomimetics often missed 

the mark (Fuller). 


model operates, its exact structure, its 

mechanical parameters, and more. 

The focus of this commentary is on 

the development of field-capable technologies 

for performing unmanned 

tasks in relevant mission spaces. In 

this sense, it is important to recognize 

that animals were evolved for survival 

in a particular environment with specific 

predators and other influences. 

In the GhostSwimmer 

 

case, extant 

tuna appeared roughly 60 million 

years ago (Dickson & Graham, 2004). 

These early tunas lived in a large circumtropical 

waterway that encircled 

theentireplanet(theTethysSea). 

This waterway existed for roughly 

50 million years during which the development 

of tunas was influenced by 

changes in the oceanography, induced 

primarily by tectonic activity. These 

changes increased productivity, expanded 

food webs, and opened potential niches 

(Dickson & Graham, 2004). 

Graham and Dickson suggest that 

the appearance of these more extensive 

ocean areas with high productivity and 

diversified food webs provided the 

catalyst for tuna evolution: enhanced 

locomotor performance and favored 

migratory behavior. Tuna’s specializations, 

thunniform swimming, capacity 

for regional endothermy, and an elevated 

aerobic capacity, are thought to 

be based on the need for extended 

swimming (Graham & Dickson, 

2004). Despite the fact that this is different 

than the pressing missions of a 

UUV (such as mine counter measures), 

this extended swimming implies 

endurance and propulsion that 

isofvaluetoapragmaticmission. 

This leads to the question the roboticist 

must ask: What aspects of a 

fish’s “design” or mechanics is applicable 

to the desired mission 

Another consideration is the different 

resources that are available. Evolution 

and robotics engineers have 

distinctly different tool sets, materials, 

actuators, and power and control systems 

at their disposal. In certain 

areas, engineers have an advantage; exotic 

materials such as titanium may 

allow advances that nature could not 

provide. However, nature still has the 

advantage (at least currently) in actuator 

power density (when considering 

bandwidth and other factors), regenerative 

capability, sensing, and control 

prowess. For example, many biomechanics 

sources have noted that muscle 

tendon series elasticity is critical for 

energetic and efficiency purposes 

(Paluska & Herr, 2006). Many advances 

are in the works in the areas of 

artificial muscles, neural networks, and 

similar technologies, but to date they 

lag behind their biological counterparts 

in key areas. Some areas in need 

of advancement for actuation include 

power to weight ratio, efficiency, fatigue 

life, and controllability. 

Where the Challenges 

of Underwater Operation 

and Biomimetics Intersect 

The challenges of achieving autonomy 

with unmanned systems are 

detailed in many articles and publications. 

From one perspective, autonomy 

intrinsically demands that 

engineered systems act like biological 

systems. For instance, these systems 

need obstacle detection and avoidance 

or situational awareness (“what 

is happening around me”), decisionmaking 

capacity (“should I respond 

aggressively or passively”), and health 

monitoring (“am I OK”). Interesting 

work in developing biologically 

inspired autonomy solutions based 

ontheseprinciplesiscurrentlyoccurring 

at Jet Propulsion Lab (JPL) 

(space mission systems) (Huntsberger, 

2001), Office of Naval Research 

(ONR) (sensory control, biomechanics) 

(ONR: Bio-Inspired Autonomous 

Systems), and Cornell (making increasingly 

complex machines) (Lipson). 

The modeling and understanding 

of unsteady hydrodynamics is also a 

challenge for underwater vehicle 

developers. While analogous to aerodynamic 

forces in some ways, hydrodynamics 

in unsteady, obstacle-laden 

environments is often more complex 

and, to date, difficult to model. In 

particular, controlled 6 degrees of freedom 

movement, particularly in unsteady 

conditions, is challenging for 

control, power, and propulsion systems. 

Underwater vehicles have a 

more challenging environment for 

controlled mobility than most UGVs 

due to this dynamic nature of their 

surroundings, and unless operating in 

open seas (blue water), they can face 

far more obstacles within tighter spaces 

than UAVs. 

Sensing and communication is another 

area where underwater systems 

have very different challenges. Biological 

systems have remarkable and innate 

abilities to detect obstacles or 

prey using a variety of sensory capabilities 

from echolocation (bottlenose 

dolphin biosonar is “probably the 

most sophisticated target location and 

analysis system in existence”) (Fulton, 

2010) to lateral lines (in fish such as 

the tuna, water flow parallel to the motion 

deflects cilium in the lateral line 

which in turn defines water velocity) 

(Martiny et al., 2009), to electric fields 

(weakly electric black ghost knife fish 

generate an electric field that causes 

voltage perturbations due to the differences 

in electrical conductivity between 

an object and the water, thereby 

sensing its surroundings) (Martiny 

et al., 2009). 


However, unmanned systems can 

leverage only those sensory capabilities 

developed by the technology community 

at large. Herein lies the challenge, 

UAVs may have similar collision 

avoidance sensory needs, but the technologies 

for in-air sensing are currently 

more advanced and, due to relatively 

low signal attenuation in air, can effectively 

sense objects hundreds or 

thousands of feet away (ignoring 

cloud cover and other aspects for sake 

of argument). The attenuation of 

various sensing signals in salt water is 

drastically more than in air, making 

technologies that many take for 

granted (WiFi and Bluetooth for example) 

impractical underwater. RF 

communications can work underwater 

but attenuation demands that low frequencies 

be used and the resulting 

trade-off is bandwidth (a reasonably 

sized acoustic modem can provide 

140 bps-15 kbps [www.benthos.com] 

where even limited WiFi [802.11 b for 

example] can provide 2 Mbps “in air” 

[Mitchell]). To achieve real-time 

or near real-time control, a UGV or 

UAV could use high-frequency, 

high-bandwidth RF communications. 

To date, this is a major challenge in underwater 

communications. Additionally, 

the change in medium 

(from air into and through water or 

vice versa) makes communication difficult 

due to refraction losses at the 

interface (losses are large for electromagnetic 

waves going from air into 

seawater and intrinsic wavelength differences 

make an underwater antenna 

for the same radio different than its 

in-air counterpart) (Butler, 2011). 

Standard acoustic and light-based 

sensors are greatly affected by the attenuation 

as well. Additionally, their 

operation often demands smooth, 

controlled motion at relatively slow 

speeds. When coupling the hydrodynamic 

maneuverability challenges 

for AUVs with these sensing challenges, 

one can only marvel at the ability of 

fast-moving fish and marine mammals 

to detect and maneuver. For these animals, 

this detection and maneuver is a 

system level process where sensing 

provides exterioception and proprioception 

and enables the vorticity and 

optimized flow control that is essential 

to maneuverability and fast swimming 

(Triantafyllou et al., 2002). 

The GPS is another technology 

taken for granted in the 21st century. 

However, current AUVs must surface 

to get GPS fixes because GPS will not 

work underwater; the signals from the 

satellites cannot sufficiently penetrate 

the water. This presents the AUV 

with a significant challenge in localization 

and navigation. Until these technologies 

advance in a cost effective 

manner for underwater applications, 

one of the most effective means for 

overcoming the challenge includes 

having the ability to surface quickly 

(and often covertly) where the manmade 

underwater system can then 

leverage in-air wireless and geopositioning 

technologies. This rapid 

surfacing can also benefit frombiologically 

inspired techniques, as the 

GhostSwimmer has implemented. 

Beyond Propulsion— 

Leveraging Biology 

to Advance the State 

of the Art 

Designing systems with bioinspired 

control systems can open avenues 

for efficiently performing, and 

switching between, necessary behavior 

states. Advances in computing technology 

(increased memory and computational 

power in smaller, less 

expensive packages) now enables incorporation 

of newer and more advanced 

biologically inspired control 

systems. This includes the development 

of intelligent control through 

distributed control in systems containing 

their own “nervous systems,” structured 

like their biological analog that 

can then “learn” through adaptive 

techniques (Thomopoulos & Braught, 

1995). 

The study of biology itself can provide 

advances in the understanding of 

unsteady hydrodynamics as mentioned 

above. Biological systems 

achieve high efficiencies and maneuverability 

through the sensing, manipulation, 

and creation of optimized flow 

around them. This includes the manipulation 

of tip vortices at control 

and propulsive surfaces as well as the 

optimization of output motion to generate 

smooth and effective jets in the 

flow behind the body. The understanding 

of these phenomena is directly 

applicable to the creation of 

higher performance AUVs and even 

manned systems (Barrett, 1996). 

Boston Engineering’s ASG is currently 

applying many of these principles to 

develop improved control surfaces for 

the U.S. Navy Sea Systems Command 

as well as advanced AUVs for ONR. 

Navigation is also an area that can 

benefit from biomimetics. Specific 

areas of interest in autonomy to the 

Navy include path planning, behavior 

development, localization, on-board 

mapping of environmental variability, 

and effective man-machine interfaces 

with a limited communication capability(Wernli,2001).Additionally,increasing 

uncertainty regarding Global 

Navigation Satellite Systems reliability 

has led unmanned systems developers 

to seek valuable alternatives. Integrating 

inertial systems with reference maps of 

Geophysical Fields of the Earth (GFE) 

is an area being explored by aerospace 


entities that offers promise through use 

of the recent advancements in embedded 

micro-processing, including 

memory devices’ capability and miniature 

size. GFEs, properties of the earth 

itself, are already well mapped in geographical 

system coordinates and can 

be considered a reliable navigation 

data source. Earth’s MagneticField 

(EMF) maps, models, and charts are 

currently in use for military and commercial 

entities (for directional information) 

and are available for at least 

98% of the earth’s surface (including 

water-covered areas) (Goldenberg, 

2006). 

Research and behavioral experiments 

have shown that various animals 

sense and use the earth’s magnetic field 

for navigation over both long and short 

distances (Johnsen & Lohmann, 

2005). Migratory animals, capable of 

sensing variations in geomagnetic 

fields, reference the earth’s mean field 

and its inclination at many points. 

Measurements of the field, including 

spatial variations and temporal evolutions, 

tabulated by the U.S. Geological 

Survey, allow the potential for geomagnetic 

field sensors to mimic animal 

behavior related to navigation (Zhai 

et al., 2007). 

Some animals, including certain 

birds, sea turtles, salamanders, and lobsters 

can discriminate small differences 

in some of the earth’s magnetic features. 

They use positional information 

in the earth’s field in several different 

ways and some actually learn the magnetic 

topography of the areas they call 

home. Some have postulated that animals 

have two separate magnetosensory 

systems (Goldenberg, 2006). A 

compass alone is rarely sufficient to 

guide animals to specific destinations 

or along a long and complex migratory 

route due to currents and other errorinducing 

phenomena. Navigation 

must be enhanced by the ability to 

determine position relative to a destination 

(human travelers use a GPS), 

i.e., positional information inherent 

in the earth’s magneticfield provides 

a similar, although less precise, assessment 

of location (Goldenberg, 2006). 

Animals also use other cues; research 

has shown that bees navigate 

relative to the sun by using the sun as 

a fixed point and orienting themselves 

by maintaining a fixed angle between 

its line of flight and the line to the 

sun (www.physics.ohio-state.edu). 

Ocean waves combine with the earth’s 

magnetic field to serve as orientation 

cues for newly hatched turtles; while 

older turtles are following a map they 

learned that enables them to establish 

their position relative to some distant 

target (Lohmann et al., 2004). Regardless 

of surface currents or other 

oceanographic features, sockeye salmon 

use an internal “map sense” to navigate 

home after several migrations and induced 

errors. An olfactory “imprint” 

is made on smelts as they leave their 

home stream, but approaching their 

stream from open sea demands one 

other imprint. Fish are perceptive of 

the azimuth and altitude of the sun, 

but during overcast days, ferromagnetic 

mineral magnetite in their brain 

may function as a biological compass. 

Another means for a fish to sense the 

magnetic field is by merely moving 

through the water (like a wire moved 

across a magnetic field, electrical 

current occurs in the wire) (Gedney, 

1984). It is possible that AUVs 

could mimic these approaches. 

The electric field of ocean currents 

indicate to sharks their drift relative to 

the bottom or to deeper water layers. 

Electric current of an ocean stream invading 

a quiet bay may provide a shark 

with directional cues in familiar territory. 

They may also explore the fields 

by occasionally diving deeper or to the 

bottom and their orientating in uniform 

DC electric fields has been proven 

behaviorally. The electric sense operates 

in a passive mode, whereas magnetic 

field detection is an active mode, allowing 

them to simultaneously sense drift 

with ocean streams and magnetic headings. 

Based on this research, measured 

potentials, sensitivity and noise effects, 

and amplification are relatively well 

known (Kalmijn, 2000). 

The sun’s movement across the sky 

gives orientation signals that vary with 

time of day. Migratory birds gain information 

from the sun better than 

humans, detecting changing patterns 

of polarization. Using magnetism and 

celestial rotation together can increase 

reliability (birds constantly cross check 

them and adjust) (www.teara.govt.nz). 

The sun’s position can be analyzed by 

looking at the polarization pattern of 

the sky arising from sunlight scattering 

(useful when the sun is obscured). 

Many insects use celestial polarization 

for compass orientation by using 

polarization-sensitive photoreceptors. 

Biological research has shown that 

some underwater animals use polarization 

of light for navigation, communication, 

and hunting. Studies have 

been performed using polarization of 

scattered light underwater to improve 

visibility. Polarization of light in water 

is caused by refraction through the surface, 

scattering light in water, refraction 

by polarizing objects, and emission 

from polarized light sources (Karpel & 

Schecher, 2011). 

The Present and 

Future of Biomimetic 

Underwater Robotics 

Rather than relying on the future 

development of a novel and 


FIGURE 3 

Boston Engineering’s GhostSwimmer PH I 

AUV in field testing. 

ground-breaking power source, Boston 

Engineering’sASG,alongwithitsteammate, 

Olin College Intelligent Vehicles 

Laboratory (Needham, MA), is taking 

inspiration from a comparably sized 

biological system to develop an AUV 

optimized for complex underwater 

environments. 

Thanks to funding from the ONR 

and a large internal R&D effort by 

both Boston Engineering and Olin 

College, this new generation of AUVs 

is emerging. By leveraging past research, 

the latest technologies, and innovative 

rapid prototyping techniques, 

the team developed an AUV that could 

autonomously swim like a fish in only 

6 months (Figure 3). The team is currently 

in Phase II with ONR and expects 

to showcase its next generation AUV in 

late 2011 to demonstrate advancements 

in efficiency and maneuverability while 

addressing and investigating the other 

challenges facing underwater vehicles 

as described above. 

While there have been many attempts 

to duplicate the swimming actions 

of a fish (Xianzhong DAI, 2003; 

Valdivia et al., 2006), a breakthrough 

achieved by the Boston Engineering 

research team has been in producing 

these movements efficiently within a 

tactically relevant vehicle. After proving 

that fish-like oscillating foil propulsion 

could indeed have higher 

propulsive efficiency and could be 

achieved by man-made methods, the 

next challenge is to properly control 

and coordinate the movements. The 

intendedsolutionforthiscontrolis 

itself bioinspired. The result is an 

AUV platform that could outperform 

conventional technologies deployed 

today in both endurance and mobility. 

This work holds the potential to provide 

a paradigm shift in underwater 

AUV capability and usefulness. 

The trend of mimicking biology 

in engineering certainly is not new, 

but through the efforts of the various 

parties working in the field of biologically 

inspired engineering and biological 

mechanics, a reduction in the 

number of limitations in enabling 

technologies and knowledge is being 

seen. Recent advances in the understanding 

of biological hydrodynamics, 

sensing, navigation techniques, and 

control coupled with low-power highthroughput 

computing (including 

Quad Core Processors, www.intel. 

com), advances in material science 

(ionomers and self healing materials, 

www.bimat.org), and improved prototyping 

techniques (such as direct metal 

laser sintering, www.morristech.com) 

has enabled robotics engineers as 

never before. As industry and academia 

continue to make progress in demonstrating 

how effective biomimetic designs 

can be, the more developers of 

next generation tactical systems such 

as Boston Engineering’s Advanced 

Systems Group will be able to rise to 

the technical challenges presented by 

current adopters of unmanned technology, 

especially in the underwater 

space. 

Authors: 

Michael Rufo and Mark Smithers 

Advanced Systems Group, 

Boston Engineering Corporation 

411 Waverley Oaks Road, 

Waltham, MA 02452 

Emails: mrufo@boston-engineering. 

com; msmithers@boston-engineering. 

com 

References 

Azarsinal, F., Bose, N., & Seif, M.S. 2007. 

An underwater vehicle maneuvering simulation: 

Focus on turning maneuvers. J Ocean 

Technol. 2(1):54-73. 

Bar-Cohen, Y. 2006. Biomimetics: Biologically 

Inspired Technologies. Boca Raton, 

FL: CRC Press. 469 pp. 

Barrett, D.S. 1996. Propulsive efficiency of 

a flexible hull underwater vehicle. Ph.D. 

thesis, Massachusetts Institute of Technology. 

Bee Navigation. http://www.physics.ohiostate.edu/~wilkins/writing/Samples/shortmed/ 

fiskemedium. Accessed June 20, 2011. 

Bird Migration. Navigating by the stars 

and sun. http://www.teara.govt.nz/ 

EarthSeaAndSky/BirdsOfSeaAndShore/ 

BirdMigration/7/en. Accessed June 20, 2011. 

Butler, L. 1987. Underwater Radio Communication. 

http://www.qsl.net/vk5br/ 

UwaterComms.htm. Accessed June 20, 2011. 

Dickson, K.A., & Graham, J.B. 2004. 

Evolution and consequences of endothermy 

in fishes. Physiol Biochem Zool. 77(6): 

998-1018. doi: 10.1086/423743. 

Fuller, J. Top 10 bungled attempts at oneperson 

flight. http://science.howstuffworks.com/ 

transport/flight/classic/ten-bungled-flightattempt4.htm. 

Accessed March 1, 2011. 

Fulton, J.T. 2010. Dolphin biosonar echolocation 

A case study. http://neuronresearch.net/ 


hearing/pdf/Dolphin_biosonar_echolocation. 

pdf. Accessed June 20, 2011. 

Gedney, L. 1984. Do Salmon navigate by 

the earth’s magnetic field Article #691. 

http://www.gi.alaska.edu/ScienceForum/ 

ASF6/691.html. Accessed June 20, 2011. 

Goldenberg, F. 2006. Geomagnetic navigation 

beyond the magnetic compass. Goodrich 

Corporation, Advanced Sensors Technical 

Center, 14300 Judicial Road, Burnsville, 

MN 55306. http://ieeexplore.ieee.org/xpl/ 

freeabs_all.jsparnumber=1650662. Accessed 

June 20, 2011. 

Graham, J.B., & Dickson, K.A. 2004. 

Tuna comparative physiology. J Exp Biol. 

207(iii):4015-24. doi: 10.1242/jeb.01267. 

Huntsberger, T. 2001. Biologically inspired 

autonomous rover control autonomous 

robots. Auton Robot. 11:341-6. 

doi: 10.1023/A:1012467829785. 

Intel. Intel core quad processors. http://www. 

intel.com/products/processor/core2quad/. 

Accessed June 20, 2011. 

Johnsen, S., & Lohmann, K. 2005. The 

physics and neurobiology of magnetoreception. 

Nature Reviews Neuroscience. 6:703-12. 

doi: 10.1038/nrn1745. 

Kalmijn, J. 2000. Detection and processing 

of electromagnetic and near-field acoustic 

signals in elasmobranch fishes. Philos T R 

Soc Lond B. 355(1401):1135-41. 

doi: 10.1098/rstb.2000.0654. 

Karpel, N., & Schechner, Y. 2011. Portable 

polarimetric underwater imaging system with 

a linear response. Dept. of Electrical Eng., 

Technion - Israel Institute of Technology, 

Haifa 32000, Israel. http://webee.technion. 

ac.il/~yoav/publications/polarimetric.pdf. 


Lipson, H. Associate Professor. Cornell 

University Mechanical and Aerospace Engineering. 

http://www.mae.cornell.edu/lipson/. 


Lohmann, K.J., Lohmann, C.M.F., Llewellyn, 

M., Ehrhart, L.M., Bagley, D.A., & Swing, T. 

2004. Animal behavior: Geomagnetic map 

used in sea-turtle navigation. Nature. 

428:909-10. doi: 10.1038/428909a. 

Magnuson, J.J. 1978. Locomotion by scombrid 

fishes: hydromechanics, morphology, 

and behavior. Fish Physiol. 7:239-313. 

doi: 10.1016/S1546-5098(08)60166-1. 

Martiny, N. 2009. Design of a lateral-line 

sensor for an autonomous underwater vehicle. 

In: 8th IFAC International Conference on 

Maneuvering and Control of Marine Craft, 

MCMC 2009. pp. 292-7. Sao Paulo, Brazil: 

IFAC. 

Mitchell, B. Wireless Standards-802.11b 

802.11a 802.11g and 802.11n: The 802.11 

family explained. http://compnetworking. 

about.com/cs/wireless80211/a/aa80211standard. 

htm. Accessed June 20, 2011. 

Morris Technologies. Direct metal laser sintering 

(DMLS). http://www.morristech.com/ 

dmls.asw. Accessed June 20, 2011. 

Office of Naval Research. Bio-inspired 

autonomous systems. http://www.onr. 

navy.mil/en/Media-Center/Fact-Sheets/ 

Bio-Inspired-Autonomous-Systems.aspx. 


Paluska, D., & Herr, H. 2006. Series 

elasticity and actuator power output. In: 

Proceedings of the 2006 IEEE International 

Conference on Robotics and Automation. 

Orlando, FL: IEEE. 

Teledyne benthos acoustic modems. 

http://www.benthos.com/acoustic-telesonarmodem-subsea-ATM885.asp. 

Accessed 

June 20, 2011. 

The Navy Unmanned Undersea Vehicle 

(UUV) Master Plan. November 9, 2004. 

Thomopoulos, S., & Braught, G. 1995. A 

biologically inspired mechanism of machine 

perception and intelligent control for multirobot 

coordination. Penn State University. 

Triantafyllou, M.S., Techet, A.H., Zhu, Q., 

Beal, D.N., Hover, F.S., & Yue, D.K.P. 

2002. Vorticity control in fish-like propulsion 

and maneuvering. Integr Comp Biol. 

42(5):1026-31. doi: 10.1093/icb/42.5.1026. 


1993. Propulsion through oscillating foils in 

nature and technology. J Fluids Struct. 

7:205-24. doi: 10.1006/jfls.1993.1012. 

Valdivia y Alvarado, P., & Youcef-Toumi, K. 

2006. Design of machines with compliant 

bodies for biomimetic locomotion in liquid 

environments. J Dyn Syst-T ASME. 

128(1,3):11. 

Wernli, R. 2001. Recent U.S. Underwater 

Vehicle Projects. San Diego, CA: Space and 

Naval Warfare Systems Center. 

Wernli, R. 2011. Recent US Navy Underwater 

Vehicle Projects. http://www.nmri.go. 

jp/main/cooperation/ujnr/24ujnr_paper_us/ 

Ocean_Engineering_and_Resource/OERD_ 

Wernli.pdf. Accessed June 20, 2011. 

Xianzhong, D. 2003. Some new progress of 

advanced robotics in China. In: Report to JCF 

2003 of IARP, Xianzhong DAI Robotics 

Expert Group of National High-Tech R&D 

Programme, China. Madrid, Spain. 

Zhai, J. 2007. Geomagnetic sensor based on 

giant magnetoelectric effect. Appl Phys Lett. 

91(12):1-3. doi: 10.1063/1.2789391. 


PAPER 

Autonomous Robotic Fish as Mobile Sensor 

Platforms: Challenges and Potential Solutions 

AUTHOR 

Xiaobo Tan 

Smart Microsystems Laboratory, 

Department of Electrical and 

Computer Engineering, 

Michigan State University 

Introduction 

ABSTRACT 

With advances in actuation and sensing materials and devices, there is a growing 

interest in developing underwater robots that propel and maneuver themselves 

as real fish do. Such robots, often known as robotic fish, could provide an engineering 

tool for understanding fish swimming. Equipped with communication 

capabilities and sensors, they could also serve as economical, dynamic samplers 

of aquatic environments. In this paper we discuss some of the major challenges in 

realizing adaptive, cost-effective, mobile sensor networks that are enabled by 

resource-constrained robotic fish. Such challenges include maneuvering in the 

presence of ambient disturbances, localization with adequate precision, sustained 

operation with minimal human interference, and cooperative control and sensing 

under communication constraints. We also present potential solutions and promising 

research directions for addressing these challenges, some of which are inspired 

by how fish solve similar problems. 

Keywords: robotic fish, adaptive sampling, mobile sensing platforms, aquatic 

sensor networks, water quality monitoring 

With 500 million years of evolution, 

fish and other aquatic animals 

areendowedwithavarietyofmorphological 

and structural features 

that enable them to move through 

water with speed, efficiency, and agility 

(Lauder & Drucker, 2004; Fish & 

Lauder, 2006). The remarkable feats 

in biological swimming have stimulated 

extensive theoretical, experimental, 

and computational research 

by biologists, mathematicians, and 

engineers, in an effort to understand 

and mimic locomotion, maneuvering, 

and sensing mechanisms adopted by 

aquatic animals. 

Over the past two decades, there 

has also been significant interest in 

developing underwater robots that 

propel and maneuver themselves 

like real fish do (Triantafyllou & 

Triantafyllou, 1995; Kato, 2000; 

Anderson & Chhabra, 2002; Alvarado 

& Youcef-Toumi, 2006; Hu et al., 

2006; Low, 2006; Epstein et al., 

2006; Morgansen et al., 2007; Lauder 

et al., 2007; Chen et al., 2010; Aureli 

et al., 2010; Smithers, 2011). Often 

termed robotic fish, these robots provide 

an experimental platform for 

studying fish swimming and hold 

strong promise for a number of underwater 

applications. Instead of using 

propellers, robotic fish accomplish 

swimming by deforming the body 

and/or fin-like appendages, mostly 

functioning as caudal fins and sometimes 

as pectoral fins. Body deformation 

and fin movements are typically 

achieved with motors. On the other 

hand, advances in smart materials 

have been explored to actuate robotic 

fish in a noiseless and compact way 

(Paquette & Kim, 2004; Tangorra 

et al., 2007; Chen et al., 2010; Aureli 

et al., 2010). Robotic fish produce 

wake signatures similar to those of 

real fish and are thus less detectable 

than propeller-driven underwater vehicles, 

which is an important advantage 

in applications requiring stealth. 

Recent advances in computing, 

communication, electronics, and 

materials have made it possible to 

create untethered robotic fish with 

onboard power, control, navigation, 

wireless communication, and sensing 

modules, which turns these robots 

into mobile sensing platforms in 

aquatic environments. Schools of robotic 

fish can form wireless sensor 

networks, which will have numerous 

promising applications, such as monitoring 

water quality, tracing oil spills, 

and patrolling harbors and coasts. 

Figure 1a shows a prototype of a robotic 

fish swimming in an inland 

lake. Figure 1b shows the close-up of 

another prototype, equipped with a 

dissolved oxygen (DO) sensor, global 

positioning system (GPS), and other 

electronic components, which has 

been developed for monitoring the 

DO level in aquafarms. Collected 

DO information will then be used to 

control the aerators to maintain a 

healthy environment for the aquatic 

animals on the farm. 

Autonomous robotic fish schools 

will provide a competitive alternative 


FIGURE 1 

Prototypes of autonomous robotic fish developed by the Smart Microsystems Laboratory at 

Michigan State University: (a) testing in a lake and (b) prototype for dynamically monitoring 

the DO level in aquafarms. 

to existing sensing technologies for 

aquatic and marine environments. 

Manual sampling, sometimes boat or 

ship-based, is still a common practice 

in environmental monitoring, which 

is labor-intensive with difficulty in 

capturing dynamic phenomena of 

interest. In-situ sensing with fixed or 

buoyedsensorsorverticalprofilers is 

another approach (Doherty et al., 

1999; Reynolds-Fleming et al., 

2002). However, these sensors have 

little freedom to move laterally, and it 

would require prohibitively many 

units for capturing distributed, spatially 

inhomogeneous information. The past 

decade has seen great progress in the use 

of robotic technology in aquatic sensing. 

Autonomous underwater vehicles 

(AUVs) (Bandyopadhyay, 2005), for 

example, are being used for hydrographic 

survey, fishery operations, and 

environmental monitoring (Hydroid, 

2009). Another highly successful technology 

is autonomous sea gliders, 

which has remarkable duration for 

continuous field operation because 

of highly energy-efficient design 

(Ericksen et al., 2001; Sherman et al., 

2001; Webb et al., 2001; Rudnick 

et al., 2004). The downside for both 

AUVs and gliders is their cost, starting 

at US $50,000 per unit (not including 

the cost of sensors), prohibiting the 

deployment of many of them for 

observing with high spatial resolution, 

and excluding them from many applications 

(such as aquafarm monitoring) 

where cost is critical. The size (meters 

long) and weight (at the order of 50 kg) 

of these vehicles also make them cumbersome 

to handle by a single person. 

Small autonomous robotic fish 

have the potential to address many of 

the aforementioned challenges. By a 

small robotic fish, we mean one that 

has length of 50 cm or less, displaces 

volume of up to 5 liters, and costs no 

more than US $5,000 (excluding that 

of aquatic sensors to be mounted). Its 

low cost, compact size, and light 

weight would make it affordable and 

convenient to deploy these robots in 

groups for versatile applications and 

various environments, such as ponds, 

lakes, rivers, and even oceans. Schools 

of robotic fish could form dynamic, 

adaptive sensor networks and provide 

distributed sensing coverage with desired 

spatiotemporal resolution. 

The realization of such a vision, 

however, is faced with a myriad of 

challenges. The size and cost considerations 

put stringent constraints on 

the robot’s locomotion, battery, computing, 

and communication capacities. 

The wide adoption of the 

robotic fish-based sensing technology 

will hinge on the robots’ ability to operate 

robustly in the unfriendly and 

often unpredictable environment, 

with their limited onboard resources 

and with minimal human intervention. 

This poses challenges across a 

wide spectrum, ranging from locomotion 

and maneuvering mechanisms, to 

energy-efficient designs, localization 

and communication schemes, and 

control and coordination strategies, 

to name a few. In this paper, we outline 

some of the most critical challenges 

and discuss potential approaches or 

opportunities in research and technology 

advancement for addressing the 

challenges. 

Maneuvering in 

Uncertain Environment 

As a sensor platform, the robotic 

fish often needs to survey a given 

path or hover over a particular region 

in the presence of ambient disturbances 

caused by wind, waves, currents, 

and turbulences. Regardless of 

its propulsion mechanism, however, a 

small robotic fish has limited actuation 

authority to counteract the disturbances. 

It is thus of great interest to 

be able to sense the flow and react in 

the most effective way under the actuation 

constraints. We can look to live 

fish for inspiration, because they deal 

with this very problem on a regular 

basis and have developed intricate 

sensing and actuation systems that 

offer us interesting insight. 

Artificial Lateral Line 

Most fish use the lateral line system 

as an important sensory organ to probe 

their environment (Coombs, 2001). 

A lateral line consists of arrays of so 


called neuromasts, each containing 

bundles of sensory hairs, encapsulated 

in a gelatinous structure called cupula. 

Under an impinging flow, the hairs are 

deflected, which elicits firing of the 

hair cell neurons and thus enables the 

animal to sense the flow field, perform 

hydrodynamic imaging, and identify 

new field objects of interest. The lateral 

line system plays an important role in 

various fish behaviors, including prey/ 

predator detection, schooling, rheotaxis, 

courtship and communication. 

A lateral line-like sensory module 

or an artificial lateral line will be very 

useful for a robotic fish to improve its 

maneuverability. For example, with 

feedback from the lateral line, the 

robot could manipulate vortices in 

the flow with its actuated fins and exploit 

the ambient flow energy for locomotion 

(Beal et al., 2006) or perform 

station-keeping by responding appropriately 

to the sensed ambient flow. 

Artificial lateral line systems, where arrays 

of beam or hair-like structures are 

used to measure flow velocities, have 

been proposed based on various physical 

transduction principles, including 

hot wire anemometry (Yang et al., 

2006), piezoresistivity (Yang et al., 

2010), capacitive sensing (Dagamseh 

et al., 2010), and encapsulated interface 

bilayers (Sarles et al., 2011). 

Recently, we have exploited the intrinsic 

mechanosensory property of 

ionic polymer-metal composites 

(IPMCs) to construct artificial lateral 

lines (Abdulsadda & Tan, 2011; 

Abdulsadda et al., 2011). As illustrated 

in Figure 2, an IPMC consists of three 

layers, with an ion-exchange polymer 

membrane (e.g., Nafion) sandwiched 

by metal electrodes. Inside the polymer, 

(negatively charged) anions covalently 

fixed to polymer chains are balanced 

by mobile (positively charged) cations. 

An applied mechanical stimulus, such 

FIGURE 2 

Illustration of the IPMC sensing principle. 

as a flow impinging on the IPMC, 

redistributes the cations inside and 

produces a detectable electrical signal 

(typically open-circuit voltage or 

short-circuit current) that is correlated 

with the mechanical or hydrodynamic 

stimulus (Chen et al., 2007). Conversely, 

an applied voltage across an 

IPMC leads to the transport of cations 

and accompanying solvent molecules, 

resulting in both differential swelling 

and electrostatic forces inside the 

FIGURE 3 

material, which cause the material to 

bend and hence the actuation effect 

(Shahinpoor & Kim, 2001). Figure 3a 

shows a prototype of an artificial lateral 

line consisting of four IPMC sensors. 

While the physical construction of 

robust and sensitive artificial lateral 

lines remains an active research area, 

it is of equal importance to make 

sense out of the data collected by the 

lateral line. Existing studies on biological 

and artificial lateral lines have 

Experimental results on localization of a dipole source with unknown location and vibration amplitude: 

(a) prototype of IPMC-based lateral line, consisting of four IPMC sensors, and (b) localization 

results along three different tracks, based on solving a model-based nonlinear estimation 

problem (Abdulsadda et al., 2011). 


mostly focused on the problem of 

localizing a vibrating sphere, known 

as a dipole, which is used to emulate periodic 

tail beating or other appendage 

movement of aquatic animals. Several 

approaches to signal processing have 

been reported, which include exploitation 

of the characteristic points (e.g., 

zero-crossings, maxima, etc.) in the 

measured velocity profile (Dagamseh 

et al., 2010), matching of the measured 

data with preobtained templates 

(Pandya et al., 2006), beamforming 

techniques (Yang et al., 2010), and 

artificial neural networks (Abdulsadda 

& Tan, 2011). We have further considered 

a source localization problem 

where both the source location and 

its vibrating amplitude are unknown. 

The posed problem is interesting, 

since a source far away but with large 

vibration could produce a signal that 

has similar amplitude as a signal produced 

by a source nearby but with 

small vibration. By formulating and 

solving a nonlinear estimation problem 

based on an analytical model for 

dipole-generated flow, we are able to 

resolve both the source location and 

the vibration amplitude simultaneously 

(Abdulsadda et al., 2011). As shown in 

Figure 3b, experimental results on an 

IPMC-based lateral line prototype 

(Figure 3a) have confirmed the effectiveness 

of the model-based estimation 

approach. 

Other than the dipole source localization 

problem, there are a few 

interesting directions for the signal 

processing of artificial lateral lines. 

The first is the detection and localization 

of multiple, more sophisticated 

moving sources (including vortices). 

With the sources moving, the resultingflow 

is no longer at a steady state, and 

the processing algorithm needs to 

localize the sources with minimal 

latency. Another major problem to 

consider is the information processing 

for a lateral line that is mounted on a 

robotic fish, where the motion of the 

robot itself and its fins adds significant 

“noise” to the lateral line signal. Biological 

fish deal with these problems 

effectively through biomechanical filtering 

for enhanced signal-to-noise 

ratio and through dynamic filtering 

in the central nervous system to remove 

the unwanted signal components 

(Coombs & Braun, 2003; Bodznick 

et al., 2003). For example, dynamic 

neural mechanisms have been identified 

for suppressing self-generated 

noise (Coombs & Braun, 2003). Such 

biological insight will prove valuable in 

devising the mechanical, electrical, and 

digital filtering mechanisms for solving 

complex processing problems faced by 

artificial lateral lines. 

Bioinspired Fin 

Achieving high-maneuverability 

hinges on the ability to manipulate 

the fluid in a delicate manner. Fish 

often use their pectoral fins to perform 

sophisticated maneuvers (Drucker & 

Lauder, 2001, 2003). These maneuvers 

involve complex conformational 

changes of the fins, involving cupping, 

twisting, and bending motions. 

Robotic fish fins, on the other hand, 

often use rigid foils (Kato, 2000; 

Morgansen et al., 2007). Recently, 

advances in soft actuation materials, 

e.g., IPMCs, have led to the exploration 

of these materials as flexible propulsors 

(Paquette & Kim, 2004). 

However, the resulting robotic fins 

typically have simple deformation 

modes, e.g., bending only (Chen et al., 

2010; Aureli et al., 2010), and fall 

short of emulating the complex deformation 

of biological fins. 

Understanding of the morphology 

and mechanics of fish fins has spurred 

effort on mimicking these features 

(typically at a higher level) in designing 

robotic fins (Lauder et al., 2007). In 

particular, the complex shape change 

of fish fins is enabled by multiple 

muscle-controlled, relatively rigid, bony 

fin rays that are connected via collagenous 

membrane (Lauder & Madden, 

2006). Coordinated movement of individual 

fin rays results in conformational 

changes of the fins desired in 

maneuvers. On the engineering side, 

by patterning electrodes of IPMC 

materials, one can expect to produce 

complex deformation by applying different 

voltage inputs to different electrode 

areas. The patterning can be 

achieved with masking during electroless 

plating or by selective removal of 

electrodes post-IPMC fabrication 

using laser or machining (Kim et al., 

2011). Inspired by the pectoral fins 

of bluegill sunfish, we have developed 

a lithography-based monolithic fabrication 

process for creating IPMC actuators 

capable of sophisticated shape 

changes (Chen & Tan, 2010). As 

shown in Figure 4, the fabricated sampleconsistsofmultipleactiveIPMC 

regions, coupled through much thinner 

passive regions. By phasing the voltage 

inputs to different active regions, we 

can realize various deformation modes 

including bending, twisting, and cupping 

(Figure 5). For example, a peak-topeak 

twisting angle of 16° is achieved 

with actuation voltages of 3 V (Chen 

& Tan, 2010). 

While the progress made in biomimetic 

fins is encouraging, significant 

further advances in both material 

fabrication and fin control are needed, 

before robotic fish are capable of manipulating 

the flow in a manner close 

to what their biological counterparts 

do. In particular, the materials need 

to be improved so that they can produce 

much larger deformation with 

reasonable bandwidth (a few Hz). On 


FIGURE 4 

Monolithically fabricated IPMC sample inspired by fish fins: (a) top view and (b) SEM picture of the 

cross section, showing that the passive area is much thinner than the active area (Chen & Tan, 2010). 

the control side, we need to model and 

understand the deformation and its 

hydrodynamic consequences of a given 

input by combining observation of 

kinetic patterns of fish fin movement, 

nonlinear elasticity modeling, computational 

fluid dynamics modeling, 

and experimental flow measurements 

using digital particle image velocimetry. 

Energy-Efficient 

Sustained Operation 

For the robotic fish-based sensing 

technology to gain widespread adoption, 

these robots will have to be able 

FIGURE 5 

to work continuously in the field 

with minimal human intervention. In 

particular, they need to operate for at 

least weeks, if not for months, before 

returning for battery recharge and 

other manual maintenance. Power is 

arguably the most crucial factor that 

limits the operational time. While 

fuel represents a potential energy 

source with high power and energy 

density, it is unclear when fuel-based 

propulsion will become feasible for 

small underwater robots. Therefore, 

battery is expected to be the primary 

power source for robotic fish, for at 

least the next 5–10 years. 

Examples of deformation modes demonstrated by the fabricated IPMC fin: (a) bending and 

(b) twisting. 

Thereareanumberofwaysone 

can potentially extend the run time 

of battery-powered robotic fish. For 

example, many onboard devices can 

be put to the sleep mode to save energy, 

when they are not active. Photovoltaic 

films can be mounted on the 

robot to harvest solar energy and replenish 

the battery, when the robot is 

on the water surface. Wave energy 

could be another source to tap into, 

but how to harvest it on an untethered 

and often goal-oriented robotic fish remains 

a challenge. 

Whilealltheaforementionedapproaches 

could stretch the mileage 

per battery charge to some extent, 

they are not game-changers. Design 

of energy-efficient locomotion mechanisms 

will be critical in realizing 

long-duration field operation, since 

locomotion is the biggest source of 

energy expenditure for autonomous 

robotic fish. To this end, we are currently 

developing a novel class of underwater 

robots, called gliding robotic 

fish (Figure 6a). Such a robot will represent 

a hybrid of underwater glider 

and robotic fish; for example, it will 

have wings for gliding and fins for maneuvering 

and assistive propulsion. 

Consequently, a gliding robotic fish 

is expected to possess both high energy 

efficiency and great maneuverability. 

Figure 6b further illustrates the 

gliding principle and why a gliding robotic 

fish will be energy-efficient. 

Under the combined influence of gravity 

and buoyancy, the body will experience 

vertical (up or down) motion. 

When the glider is properly pitched, 

the lift generated during buoyancyinduced 

vertical motion will enable 

horizontal travel. Through the control 

of pitch direction and buoyancy, one 

can switch between the descent/ascent 

gliding motion, resulting in a sawtoothshaped 

trajectory. Since buoyancy 


FIGURE 6 

Energy-efficient gliding robotic fish: (a) the concept of a gliding robotic fish with a hydrodynamic 

gliding body and a caudal fin and (b) illustration of the gliding principle. 

control and pitch control are the major 

sources of energy expenditure and take 

placeonlyduringascent/descent 

switching, the motion is very energyefficient, 

especially if the dive depth 

is relatively large. 

Communication and 

Localization 

Robotic fish need to communicate 

with a base station to receive commands 

and send back the collected environmental 

information. They also 

need to communicate with each 

other for information relay and motion 

coordination. Underwater communication, 

however, is particularly challenging 

for small robotic fish that 

have stringent power and size constraints. 

Radio frequency (RF) signals 

attenuate quickly in water, severely 

limiting the achievable communication 

range and data rate. Light communication 

is possible (Verzijlenberg 

& Jenkin, 2010), but again the range 

and data rate are very limited and it 

does not work in a turbid environment. 

Acoustic and sonar communication 

underwater has been studied for 

many decades and was recently explored 

for communication among robotic 

fish (Science Daily, 2008). 

However, the associated power and 

hardware required to achieve reasonably 

large communication distance 

and data rate are typically not affordable 

by small robotic fish. For these 

reasons, the most viable solution 

wouldbetocommunicatewhenthe 

robot surfaces, in which case lowpower, 

low-cost RF communication 

protocols such as ZigBee can be readily 

used. Unlike the Bluetooth protocol, 

which is intended for eliminating cables 

between electronic devices, the ZigBee 

protocol is built on top of the IEEE 

802.15.4 standard and it targets specifically 

wireless sensor network applications. 

For wider range communication, 

cellular networks could also be employed 

if such networks are available. 

Limiting the communication to the 

water surface entails additional challenges 

in robotic fish coordination, 

control, and networking. For effective 

FIGURE 7 

networking, we need to have a sufficient 

number of nodes on the surface. 

This can be achieved through joint 

motion planning and control. For example, 

we can hold robotic fish on 

the surface until a network with adequate 

density and coverage is formed 

and completes data transmission. 

Localization is another challenge in 

robotic fish-based sensor networks. 

For small robotic fish, having onboard 

localization capability is essential for 

successful navigation of the robot and 

for effective coordination of robotic 

fish networks. Accurate localization is 

also critical for tagging the sensed information 

so that the data collected 

by robotic fish are associated correctly 

to the physical location in water. While 

the GPS is readily available and does 

not take up much space, its typical precision 

of 5–10 m is inadequate for 

many applications of robotic fish due 

to their small size and relatively low 

speeds (50 cm/s or less). In addition, 

theGPSmaytakeafewminutesto 

lock satellites every time the robot 

emerges from underwater, which severely 

limits the networking and control 

performance. More agile and precise 

localization technology is needed. 

We have developed an efficient localization 

scheme for small robotic fish 

(Shatara & Tan, 2010). As illustrated 

in Figure 7a, the scheme is based on 

Underwater acoustic ranging-based localization: (a) schematic of the ranging protocol and 

(b) localization performance in a pool test (Shatara & Tan, 2010). 


acoustic ranging, which measures the 

time it takes an acoustic signal to travel 

from one node to the other. For example, 

node 1 simultaneously sends an 

RF packet and an acoustic pulse to 

node 2. When node 2 starts its onboard 

timer when it receives the RF 

packet and then stops its timer when 

it detects the acoustic pulse. Since the 

RF signal travels much faster than 

the acoustic signal, we can estimate 

thedistancebetweenthetwonodes 

based on the timer reading. The 

scheme involves simple hardware, a 

buzzer and a microphone, for each 

node. A sliding discrete Fourier transform 

algorithm, implemented on a 

digital signal controller, is employed 

for the detection of arrival of the acoustic 

signal. Figure 7b shows the results 

from experiments in a swimming 

pool, where a small robotic fish was 

towed across the deep side of the 

pool (about 13 m long) while its distances 

to the two beacon nodes 

mounted on the pool wall were measured 

through acoustic ranging. The 

resulting localization error was less 

than 1 m for the entire tested range, 

which was a significant improvement 

over the precision of a commercial 

GPS. 

Note that the above localization 

scheme works only when the robot 

surfaces, since it involves RF communication. 

The location of a robot when 

it is underwater can be inferred using 

dead reckoning. The scheme in 

Figure 7 does require beacon nodes 

(whose locations are known) to obtain 

the absolute location of a node. In the 

absence of such beacon nodes, the 

scheme can be used to get relative locations 

among nodes, which is of interest 

in coordinating schools of robotic fish. 

There are many other in-air localization 

schemes for wireless sensor networks, 

e.g., ranging based on received 

signal strength. While these schemes 

can be adapted for robotic fish-based 

aquatic networks, care must be taken 

to address the challenges associated 

with noises, disturbances, and signal 

attenuation at the air/water interface. 

Autonomous Control 

and Coordination 

With onboard communication, 

navigation, control, and sensing devices, 

robotic fish are desired to operate 

autonomously, as individuals and as 

schools, to carry out envisioned monitoring 

tasks. A few challenges arise in 

the control and coordination of these 

robots. A robotic fish needs to handle 

multiple functions subject to environmental 

uncertainties and resource constraints. 

In particular, the functions 

could include sampling the environment, 

processing and transmitting 

the measured data, maintaining network 

connectivity, and controlling 

its motion. There are various uncertainties 

that interfere with these functions, 

examples of which include 

motion perturbations due to waves 

and turbulences, imperfect sensor 

measurements, and localization error 

and communication packet drops. 

Furthermore, all of these functions 

compete for limited onboard computing 

and power resources. This is a classic 

multi-objective, multi-constraint 

optimization problem, and it demands 

asystematicapproachtothejoint 

consideration of control, networking, 

and sensor fusion. Evolutionary algorithms 

(Deb, 2001), which codify 

basic principles of genetic evolution, 

can offer a promising solution to this 

multi-objective optimization problem. 

Another challenge lies in coordinating 

a school of robotic fish. It is intriguing 

to deploy groups of robotic 

fish that cooperatively perform sensing 

tasks. In that case, it is often desirable 

not to use centralized control, because 

the centralized paradigm would entail 

prohibitive cost in communication, 

and it would paralyze the whole network 

if the command node fails. 

Therefore, individual robotic fish are 

expected to communicate only with 

their local neighbors and make decisions 

in a distributed manner. Animals 

including fish often exhibit coordinated 

collective movement facilitated 

by only local interactions, which has 

inspired great interest from the controls 

community in analyzing and synthesizing 

control laws for groups of 

unmanned vehicles. Significant progress 

has been made in this area, even 

with some demonstrated success in 

adaptive sampling using underwater 

gliders (Leonard et al., 2007). 

While these accomplishments can 

provide a sound starting point for the 

control and coordination of robotic 

fish schools, we need to recognize 

many new and subtle difficulties 

faced by the latter. For example, a robotic 

fish can only communicate with 

its peers when it surfaces, which renders 

communication and feedback 

intermittent and asynchronous. This 

again points to the need to jointly consider 

control, communication, and 

networking issues. 

Other Challenges 

As sensor platforms, the potential 

of robotic fish in environmental sensing 

will be ultimately limited by the 

availability of versatile sensors that are 

compact and easy to interface with. 

Most commercial sensors available 

today are not amenable to integration 

into small robotic fish, since sensor 

manufacturers have mostly been targeting 

handheld, fixed, or buoyed 


sensors where miniaturization is not 

critical. It is expected that, with the development 

of robotic fish and wireless 

networking technologies, manufacturers 

will see the growth opportunities 

in robotic fish-enabled aquatic 

sensing and start investing in the development 

of compact, economical, and 

robust aquatic sensors. 

There are other engineering challenge, 

one example of which is biofouling, 

where microorganisms and 

other organisms accumulate on the 

surface of robotic fish and their sensors, 

degrading movement and sensing 

performance. Periodic cleaning is 

an option; thanks to the mobility of 

robotic fish, access to these robots is 

relatively easy. Another possibility is 

to apply anti-fouling coatings. 

Conclusion 

In this paper, we have explored the 

potential of small robotic fish as 

mobile sensor platforms for aquatic 

and marine environments. Realization 

of this vision poses a rich set of challenges 

across a wide spectrum of areas, 

such as actuation/sensing materials, 

mechanism design, communication, 

control, and packaging. We have reviewed 

some of the major challenges 

and discussed possible routes to overcome 

them. The list of challenges outlined 

in this paper is by no means 

exhaustive, but even partial success 

in addressing them could have farreaching 

impact on aquatic environmental 

monitoring and other 

engineering applications. 


This work was supported in part by 

the Office of Naval Research under 

grant N000140810640 (program 

manager Dr. T. McKenna) and the 

National Science Foundation under 

grants ECCS 0547131, CCF 

0820220, EEC 0908810, IIS 

0916720, DBI 0939454, ECCS 

1050236, and ECCS 1029683. The 

author gratefully acknowledges the 

contributions of many former and 

current members of the Smart Microsystems 

Laboratory at Michigan State 

University, for the results and ideas 

presented in this paper. 

Author: 

Xiaobo Tan 

Smart Microsystems Laboratory 

Department of Electrical and 

Computer Engineering 

Michigan State University, 

East Lansing, MI 48824 

Email: xbtan@egr.msu.edu 

References 

Abdulsadda, A.T., & Tan, X. 2011. 

Underwater source localization using an 

IPMC-based artificial lateral line. In: 

Proceedings of the 2011 IEEE International 

Conference on Robotics and Automation, 

to appear. 2719-24. Shanghai, China: IEEE. 

Abdulsadda, A.T., Zhang, F., & Tan, X. 

2011. Localization of source with unknown 

amplitude using IPMC sensor arrays. In: 

Proceedings of Electroactive Polymer Actuators 

and Devices (EAPAD) XIII. Proc. SPIE, eds. 

Bar-Cohen, Y., & Carpi, F., 797627(1-11). 

San Diego, CA: SPIE or Society of Photo- 

Optical Instrumentation Engineers. 

Alvarado, P.V., & Youcef-Toumi, K. 2006. 

Design of machines with compliant bodies 

for biomimetic locomotion in liquid environments. 

Trans ASME J Dyn Syst Meas 

Control. 128:3-13. doi: 10.1115/1.2168476. 

Anderson, J.M., & Chhabra, N.K. 2002. 

Maneuvering and stability performance of a 

robotic tuna. Integr Comp Biol. 42:118-126. 

doi: 10.1093/icb/42.1.118. 

Aureli, M., Kopman, V., & Porfiri, M. 2010. 

Free-locomotion of underwater vehicles 

actuated by ionic polymer metal composites. 

IEEE/ASME Trans Mech. 15(4):603-14. 

doi: 10.1109/TMECH.2009.2030887. 

Bandyopadhyay, P.R. 2005. Trends in 

biorobotic autonomous undersea vehicles. 

IEEE J Oceanic Eng. 30:109-39. 

doi: 10.1109/JOE.2005.843748. 

Beal, D.N., Hover, F.S., Triantafyllou, 

M.S., Liao, J.C., & Lauder, G.V. 2006. 

Passive propulsion in vortex wakes. 


S0022112005007925. 

Bodznick, D., Montgomery, J., & Tricas, 

T.C. 2003. Electroreception: Extracting 

behaviorally important signals from noise. 

In: Sensory Processing in Aquatic Environments, 

eds. Collin, S.P., & Marshall, N.J., 

389-403. New York: Springer-Verlag. 

Chen, Z., Shatara, S., & Tan, X. 2010. 

Modeling of biomimetic robotic fish propelled 

by an ionic polymer-metal composite 

caudal fin. IEEE/ASME Trans Mech. 15(3): 

448-459. doi: 10.1109/TMECH.2009. 

2027812. 

Chen, Z., & Tan, X. 2010. Monolithic 

fabrication of ionic polymer-metal composite 

actuators capable of complex deformation. 

Sensor Actuat A: Phys. 157:246-57. 

doi: 10.1016/j.sna.2009.11.024. 

Chen, Z, Tan, X., Will, A., & Ziel, C. 2007. 

A dynamic model for ionic polymer-metal 

composite sensors. Smart Mater Struct. 16: 

1477-88. doi: 10.1088/0964-1726/16/4/063. 

Coombs, S. 2001. Smart skins: Information 

processing by lateral line flow sensors. 

Auton Robot. 11:255-61. doi: 10.1023/ 

A:1012491007495. 

Coombs, S., & Braun, C.B. 2003. Information 

processing by the lateral line system. 

In: Sensory Processing in Aquatic Environments, 

eds. Collin, S.P., & Marshall, N.J., 

122-38. New York: Springer-Verlag. 

Dagamseh, A., Lammerink, T., Kolster, M., 

Bruinink, C., Wiegerink, R., & Krijnen, G. 


2010. Dipole-source localization using biomimetic 

flowsensor arrays positioned as 

lateral-line system. Sensor Actuat A. 162(2): 

355-60. doi: 10.1016/j.sna.2010.02.016. 

Deb, K. 2001. Multi-objective Optimization 

using Evolutionary Algorithms. Chichester, 

UK: John Wiley & Sons. 

Doherty, K.W., Frye, D.E., Liberatore, 

S.P., & Toole, J.M. 1999. A moored profiling 

instrument. J Atmos Ocean Technol. 

16:1816-29. 2.0.CO;2"target=_blankdoi:- 

10.1175/1520-0426(1999)0162.0.CO;2. 

Drucker, E.G., & Lauder, G.V. 2001. 

Wake dynamics and fluid forces of turning 

maneuvers in sunfish. J Exp Biol. 204:431-42. 


Function of pectoral fins in rainbow trout: 

behavioral repertoire and hydrodynamic 

forces. J Exp Biol. 206:813-26. doi: 10.1242/ 

jeb.00139. 

Epstein, M., Colgate, J.E., & MacIver, 

M.A. 2006. Generating thrust with a 

biologically-inspired robotic ribbon fin. 

In: Proceedings of the 2006 IEEE/RSJ 

International Conference on Intelligent 

Robots and Systems. 2412-17. Beijing, China: 

IEEE. doi: 10.1109/IROS.2006.281681. 

Ericksen, C.C., Osse, T.J., Light, R.D., 

Wen, T., Lehman, T.W., Sabin, P.L., … 

Chiodi, A.M. 2001. Seaglider: A long-range 

autonomous underwater vehicle for oceanographic 

research. IEEE J Oceanic Eng. 26: 

424-36. doi: 10.1109/48.972073. 

Fish, F.E., & Lauder, G.V. 2006. Passive 

and active flow control by swimming fishes 

and mammals. Annu Rev Fluid Mech. 

38:193-224. doi: 10.1146/annurev. 

fluid.38.050304.092201. 

Hu, H., Liu, J., Dukes, I., & Francis, G. 

2006. Design of 3D swim patterns for 

autonomous robotic fish. In: Proc. 2006 

IEEE/RSJ Int. Conf. Intell. Robots Syst. 

2406-11. Beijing, China: IEEE. 

Hydroid, LLC. 2009. REMUS: Autonomous 

underwater vehicles - REMUS AUVs. 

Available at http://www.km.kongsberg. 

com/ks/web/nokb0240.nsf/AllWeb/ 

D5682F98CBFBC05AC1257497002976E4 

OpenDocument/ (accessed 5 July 2011). 

Kato, N. 2000. Control performance in 

the horizontal plane of a fish robot with 

mechanical pectoral fins. IEEE J Oceanic 

Eng. 25(1):121-9. doi: 10.1109/48.820744. 

Kim, K.J., Pugal, D., & Leang, K.K. 2011. 

A twistable ionic polymer-metal composite 

artificial muscle for marine applications. 

Mar Technol Soc J. This issue. 

Lauder, G.V., Anderson, E.J., Tangorra, J., 

& Madden, P.G.A. 2007. Fish biorobotics: 

Kinematics and hydrodynamics of selfpropulsion. 

J Exp Biol. 210:2767-80. 

doi: 10.1242/jeb.000265. 

Lauder, G.V., & Drucker, E.G. 2004. 

Morphology and experimental hydrodynamics 

of fish fin control surfaces. IEEE J Oceanic 

Eng. 29(3):556-71. doi: 10.1109/JOE. 

2004.833219. 

Lauder, G.V., & Madden, P.G.A. 2006. 

Learning from fish: Kinematics and experimental 

hydrodynamics for roboticists. Int 

J Automat Comput. 4:325-35. doi: 10.1007/ 

s11633-006-0325-0. 

Leonard, N.E., Paley, D.A., Lekien, F., 

Sepulchre, R., Fratantoni, D.M., & Davis, 

R.E. 2007. Collective motion, sensor 

networks, and ocean sampling. Proc IEEE. 

95(1):48-74. doi: 10.1109/JPROC.2006. 

887295. 

Low, K.H. 2006. Locomotion and depth control 

of robotic fish with modular undulating 

fins. Int J Automat Comput. 4:348-57. 

doi: 10.1007/s11633-006-0348-6. 

Morgansen, K.A., Triplett, B.I., & Klein, 

D.J. 2007. Geometric methods for modeling 

and control of free-swimming fin-actuated 

underwater vehicles. IEEE T Robot. 23(6): 

1184-99. doi: 10.1109/LED.2007.911625. 

Pandya, S., Yang, Y., Jones, D.L., Engel, J., 

& Liu, C. 2006. Multisensor processing 

algorithms for underwater dipole localization 

and tracking using MEMS artificial 

lateral-line sensors. EURASIP J Appl Signal 

Processing. 2006:76593. doi: 10.1155/ 

ASP/2006/76593. 

Paquette, J.W., & Kim, K.J. 2004. Ionomeric 

electroactive polymer artificial muscle for 

naval applications. IEEE J Oceanic Eng. 29(3): 

729-37. doi: 10.1109/JOE.2004.833132. 

Reynolds-Fleming, J.V., Fleming, J.G., & 

Luettich, R.A. 2002. Portable autonomous 

vertical profiler for estuarine applications. 

Estuaries. 25:142-7. doi: 10.1007/BF02696058. 

Rudnick, D.L., Eriksen, C.C., Fratantoni, 

D.M., & Perry, M.J. 2004. Underwater gliders 

for ocean research. Mar Technol Soc J. 38: 

48-59. doi: 10.4031/002533204787522703. 

Sarles, S.A., & Leo, D.J. 2011. Hair cell 

sensing with encapsulated interface bilayers. 

In: Proceedings of Bioinspiration, Biomimetics, 

and Bioreplication. Proc. SPIE, eds. Martin- 

Palma, R.J., & Lakhtakia, A., 797509(1-11). 

San Diego, CA: SPIE. 

Science Daily. 2008. School of robofish communicate 

with each other in underwater robot 

teams. Science Daily. http://www.sciencedaily. 

com/releases/2008/06/080606105454.htm 

(accessed 5 July 2011). 

Shahinpoor, M., & Kim, K. 2001. Ionic 

polymer-metal composites: I. Fundamentals. 

Smart Mater Struct. 10:819-33. doi: 10.1088/ 

0964-1726/10/4/327. 

Shatara, S., & Tan, X. 2010. An efficient, 

time-of-flight-based underwater acoustic 

ranging system for small robotic fish. IEEE 

J Oceanic Eng. 35(4):837-46. doi: 10.1109/ 

JOE.2010.2060810. 

Sherman, J., Davis, R.E., Owens, W.B., & 

Valdes, J. 2001. The autonomous underwater 

glider “spray”. IEEE J Oceanic Eng. 

26:437-46. doi: 10.1109/48.972076. 

Smithers, M. 2011. Designing future 

autonomous underwater vehicles. Defense 

Tech Briefs. Available at http://www. 

defensetechbriefs.com/component/content/ 

article/9075 (accessed 5 July 2011). 

Tangorra, J., Anquetil, P., Fofonoff, T., 

Chen, A., Del Zio, M., & Hunter, I. 2007. 


The application of conducting polymers to a 

biorobotic fin propulsor. Bioinspir Biomim. 

2:S6-17. doi: 10.1088/1748-3182/2/2/S02. 



Sci Amer. 272:64-71. doi: 10.1038/ 


Verzijlenberg, B., & Jenkin, M. 2010. 

Swimming with robots: Human robot 

communication at depth. In: Proceedings of 

the 2010 IEEE/RSJ International Conference 

on Intelligent Robots and Systems. 

4023-28. Taipei, Taiwan: IEEE. doi: 10.1109/ 

IROS.2010.5652751. 

Webb, D.C., Simonetti, P.J., & Jones, C.P. 

2001. “SLOCUM”: An underwater glider 

propelled by environmental energy. IEEE J 

Oceanic Eng. 26:447-52. doi: 10.1109/ 

48.972077. 

Yang, Y., Chen, J., Engel, J., Pandya, S., 

Chen, N., Tucker, C., … Liu, C. 2006. 

Distant touch hydrodynamic imaging with 

an artificial lateral line. Proc Nat Acad 

Sci. 103(50):18891-5. doi: 10.1073/ 

pnas.0609274103. 

Yang, Y., Nguyen, N., Chen, N., Lockwood, 

M., Tucker, C., Hu, H., … Jones, D.L. 2010. 

Artificial lateral line with biomimetic neuromasts 

to emulate fish sensing. Bioinspir 

Biomim. 5:016 001. 


PAPER 

Robotic Models for Studying Undulatory 

Locomotion in Fishes 

AUTHORS 


Jeanette Lim 

Ryan Shelton 

Museum of Comparative Zoology, 

Harvard University 

Chuck Witt 

Erik Anderson 

Department of Mechanical 

Engineering, Grove City College 



Engineering, Drexel University 

Introduction 

Fish moving through the water are 

capable of using a variety of locomotor 

modes. Some species swim nearly exclusively 

using their fins and generate 

propulsive and maneuvering forces 

using midline fins (dorsal and anal) 

or paired fins (pectoral and pelvic) 

(Drucker & Lauder, 1999, 2001; 

Hove et al., 2001; Standen, 2008; 

Standen & Lauder, 2007). Other 

species generate thrust primarily by 

activating body musculature to bend 

the body and generate waves passing 

from the head toward the tail (Gillis, 

1996; Jayne & Lauder, 1994, 1995c; 

Lauder & Tytell, 2006; Rome et al., 

1993). The caudal or tail fin is generally 

considered as an extension of the 

body in most analyses of fish locomotion, 

but the tail also possesses a 

substantial array of intrinsic muscles 

that appear to stiffen the tail during 

steady forward swimming and generate 

a variety of complex tail conformations 

during maneuvering (Flammang 

ABSTRACT 

Many fish swim using body undulations to generate thrust and maneuver in 

three dimensions. The pattern of body bending during steady rectilinear locomotion 

has similar general characteristics in many fishes and involves a wave of increasing 

amplitude passing from the head region toward the tail. While great progress has 

been made in understanding the mechanics of undulatory propulsion in fishes, the 

inability to control and precisely alter individual parameters such as oscillation frequency, 

body shape, and body stiffness, and the difficulty of measuring forces on 

freely swimming fishes have greatly hampered our ability to understand the fundamental 

mechanics of the undulatory mode of locomotion in aquatic systems. In this 

paper, we present the use of a robotic flapping foil apparatus that allows these parameters 

to be individually altered and forces measured on self-propelling flapping 

flexible foils that produce a wave-like motion very similar to that of freely swimming 

fishes. We use this robotic device to explore the effects of changing swimming 

speed, foil length, and foil-trailing edge shape on locomotor hydrodynamics, the 

cost of transport, and the shape of the undulating foil during locomotion. We 

also examine the passive swimming capabilities of a freshly dead fish body. Finally, 

we model fin-fin interactions in fishes using dual-flapping foils and show that thrust 

can be enhanced under correct conditions of foil phasing and spacing as a result of 

the downstream foil making use of vortical energy released by the upstream foil. 

Keywords: fish, robot, swimming, biomechanics 

& Lauder, 2008; Flammang & 

Lauder, 2009). 

Despite the large number of studies 

of fish undulatory locomotion 

over the last 20 years, there are still 

many unanswered questions about 

how the pattern of body deformation 

is generated, the effect of body stiffness 

on locomotor performance, what 

effect different tail shapes have on locomotor 

function, and how hydrodynamic 

interactions among different 

fins might influence the generation 

of swimming forces. Studies of freely 

swimming fishes have contributed 

enormously to our understanding of 

the mechanics of aquatic locomotion, 

but such an approach is necessarily 

limited to the behaviors voluntarily 

executed by living fishes. And measuring 

locomotor forces on freely 

swimming fishes is a challenging proposition. 

Furthermore, a wide array of 

interesting experimental manipulations, 

including changing the flexural 

stiffness of the body, altering the 

shape of the tail, and changing the 

spacing between fins, are clearly impossible 

to conduct in living animals. 

Robotics offers a complementary 

approach to studies of living fishes by 

allowing manipulation of variety of 

parameters such as flexural stiffness, 

aspect ratio, tail shape, and spacing between 

adjacent fins. A simple robotic 

flapping foil device can be used to generate 

undulatory locomotion in flexible 

fish-like materials, and forces can 


e measured and the effect of changes 

in body length, stiffness and tail shape 

can be quantified. 

The focus of this paper will be on 

undulatory locomotion in the water 

and the use of a robotic flapping apparatus 

to produce swimming in both 

flexible plastic foils and a passive 

freshly dead fish body. After a brief 

overview of undulatory locomotion 

in fishes, we discuss the design of a robotic 

controller for flapping flexible 

foils that allows measurement of selfpropelled 

speeds (SPS), forces, and 

torques and the cost of transport associated 

with foils of different shapes and 

stiffnesses. In addition, we address 

an important technical issue in studies 

of aquatic propulsion: the effect of 

swimming at non-SPS on body waveform, 

patterns of force production and 

wake flow patterns. Finally, we present 

data on the swimming performance 

of passive fish bodies and discuss the 

future for studies of robotically controlled 

undulatory locomotion. Our 

overall aim is to introduce a number 

of case studies with data that show 

the utility of this approach for studying 

underwater propulsion and to present 

an overview of how such studies can 

provide new ideas and tests for current 

views of how fish swim. 

Overview of Undulatory 

Propulsion in Fishes 

When fish swim using their bodies 

and tail fin as the primary thrust generators, 

they pass a wave of bending 

down the body that increases in amplitude 

from the head toward the tail 

(Donley & Dickson, 2000; Gillis, 

1996; Jayne & Lauder, 1995a, 1995b; 

Liao, 2002; Long et al., 1994). This 

bending wave is produced by a wave 

of muscular activity that also moves 

posteriorly and is created by spinal 

cord and hindbrain pattern generators 

(Bone et al., 1978; Fetcho & Svoboda, 

1993; Fetcho, 1986; Shadwick & 

Gemballa, 2006). Muscular power 

to generate thrust is produced primarily 

by the segmented myotomal 

musculature in the posterior region of 

fish, especially during slow swimming 

(Johnson et al., 1994; Rome et al., 

1993; Syme, 2006), and as speed increases 

more anterior body muscles 

are recruited to power locomotion. 

The division of the segmented body 

musculature into superficial “red” fibers 

and deeper and more complexly 

arranged “white” fibers also plays an 

important role in understanding the 

multiple “gaits” used by fishes; the relative 

roles of these two types of muscle 

fibers can change dramatically as fish 

change swimming speed and execute 

rapid locomotor behaviors such as 

fast-start escapes (Jayne & Lauder, 

1993; Tytell & Lauder, 2002). 

When fish swim slowly using body 

undulations, oscillation of the front 

third or so of the body is quite small 

even in species as diverse as eels, trout, 

and tuna (Donley & Dickson, 2000; 

Gillis, 1998; Lauder & Tytell, 2006), 

and a primary role of this reduced oscillation 

appears to be drag reduction 

by minimizing the frontal area of the 

fish that encounters oncoming flow. 

As swimming speed increases, the 

front region of fish shows increasingly 

large side-to-side oscillations, which is 

in part a reflection of the recruitment 

of body musculature in this more anterior 

region of the fish. 

Even when fish swim by undulatory 

propulsion using the production 

of traveling waves, other fins often 

are used too, and it is incorrect to suggest 

that fish locomotor modes represent 

completely distinct patterns of 

motion. For example, when trout or 

bluegill sunfish swim using body undulations, 

they are also actively using 

their dorsal and anal fins, which play 

a key role in balancing roll torques 

and generating thrust (Drucker & 

Lauder, 2001, 2005; Standen & 

Lauder, 2005, 2007). In addition, 

the pelvic fins play an important role 

in controlling body stability during 

locomotion in fishes (Harris, 1936, 

1938; Standen, 2008, 2010). Fishes 

also vary in tail shape, and the distinction 

between the externally symmetrical 

(homocercal) tail of teleost fishes 

and the asymmetrical tail in sharks 

and fish such as sturgeon (Lauder, 

1989, 2000) is well known. The heterocercal 

tail shape induces torques around 

the body center of mass that requires 

compensatory changes in body position 

to allow steady horizontal swimming 

(Liao & Lauder, 2000; Wilga 

& Lauder, 2002, 2004b). 

Although considerable progress has 

been made in studies of fish locomotion 

by investigating the kinematics, 

muscle activity, and hydrodynamics 

of live fishes swimming steadily and 

maneuvering, there are many limits 

to studies of this kind. Perhaps the 

two greatest limitations to research 

on live fishes are (1) the considerable 

difficulty in measuring locomotor 

forces and torques produced by the 

bending body as fish swim freely and 

(2) the inability to manipulate key 

variables such as body length, aspect 

ratio, tail shape, and stiffness. Without 

an ability to alter such key components 

that govern locomotor dynamics, we 

will be limited in our understanding 

of the factors that influence undulatory 

propulsion in the water. 

The purpose of this paper is to present 

a number of new case studies 

using a robotic flapping device for 

generating undulatory locomotion in 

engineered materials as a means of 


etter understanding how fish swim 

and the dynamics of undulatory propulsion. 

We discuss different examples 

to show the utility of this approach and 

how use of simple flexible foils informs 

studies of fish locomotion and points 

to new avenues of research. 

Simple Robotic Models of 

Fish Undulatory Locomotion 

We have designed a robotic apparatus 

that produces controlled heave and 

pitch motions of flapping foils. The 

most important features of this device 

are (1) that it can be set up to be selfpropelling 

(allowing locomotion by 

flapping foils at their natural swimming 

speed and not only at imposed 

speeds) and (2) that forces and torques 

can be measured on flapping foils 

during self-propelled swimming so 

that within-cycle patterns of force and 

torque oscillation can be compared 

among foils with different shapes 

and stiffnesses at different swimming 

speeds. This apparatus was designed 

with two sets of flapping foils in series 

in order to be able to model, with foils, 

the interactions that can occur between 

fins of fishes that are arranged 

in series such as the dorsal and anal 

fins and the tail fin (discussed further 

in the section on fin-fin interactions 

below). 

A general description of the first 

generation of this apparatus is presented 

in Lauder et al. (2007). Briefly, 

heave and pitch motors and rotary 

encoders that allow readouts of foil position 

are mounted on a carriage above 

a recirculating flow tank. This carriage 

is supported on low friction air bearings, 

which allow the carriage to move 

in response to foil thrust and drag forces 

generated during flapping motions. 

The second generation version of this 

apparatus has an ATI Nano-17 sixaxis 

force/torque sensor mounted on 

the shaft supporting the foils (Figure 

1A). This permits measurement 

of three forces and three torques during 

self-propulsion at sample rates that 

allow quantification of within-cycle 

patterns of force production even during 

self-propulsion. In addition, a linear 

encoder mounted on the carriage 

allows a readout of carriage position, 

FIGURE 1 

and this is used by a Labview program 

to calculate SPS from data generated 

by a series of swimming tests at a 

range of speeds. Synchronizing signals 

from the LabView program controlling 

the heave and pitch motors are 

used to trigger data acquisition from 

theATIsensorandalsototrigger 

image acquisition from three synchronized 

Photron high-speed video 

Images of a variety of flexible foils and freshly dead trout (Oncorhynchus mykiss) attached to a 

robotic flapping foil apparatus (see Lauder et al., 2007, for details on the basic design). (A) Flexible 

foil actuated at its leading edge in heave and pitch, suspended in a recirculating flow tank. The red 

arrow points to an ATI Nano-17 6-axis force/torque transducer on the foil shaft. (B, C) Flexible foils 

with different trailing edge shapes are used to study the effect of tail shape on swimming performance. 

(D, E, F) Images of a trout held behind the head to allow imposition of heave and pitch motions to 

study passive body properties. (E) An image from a trout self-propelling under an imposed heave 

motion; the body waveform produced is very similar to that generated during swimming. 


cameras. Images of the flapping foils to 

quantify both foil motion and hydrodynamic 

flow patterns are thus 

synchronized with force and torque 

measurements on the foil and with 

the imposed heave and pitch motions 

on the foil. 

The overall goals of using a flapping 

foil robotic device are to simplify 

and control as much as possible patterns 

of motion imposed on flexible 

and rigid foils that swim through the 

water and to allow direct testing of 

the effect on propulsion of a variety 

of key factors relevant to understanding 

fish locomotion: tail shape (Figures 

1B and 1C), foil flexibility, and 

interactions among fins. This robotic 

apparatus can also be used to examine 

the passive swimming capabilities of a 

freshly dead fish body (Figures 1D, 1E, 

and 1F ), and below we present data on 

the ability of fish bodies to passively 

propel and generate propulsive waveforms 

under imposed motions. Foils 

of various kinds are attached to a stainless 

steel sandwich bar (to hold the 

leading edge of flexible materials; Figures 

1A, 1B, and 1C) to a solid 8-mm 

shaft (for rigid NACA 0012 foils; see 

Lauder et al., 2007, and data shown 

in Figure 11), and flexible fish bodies 

aremountedinaholderthatisattached 

behind the head (Figures 1D 

and 1F). Holding systems are designed 

not to flex in response to imposed 

heave and pitch motions and to allow 

forces and torques generated by swimming 

foils to be transmitted to the 

force/torque sensor on the shaft. Holding 

systems such as the sandwich bar 

system do have their own drag, and 

this drag is time-dependent due to 

the heaving and pitching motion of 

the holding system as the foils selfpropel. 

It is thus not possible to give 

asinglevalueforthedragofthefoil 

holding system. Since all foil comparisons 

within a single experimental type 

used the same holding system and were 

treated identically, we do not present 

data on the performance of the holding 

apparatus alone. 

Sample data from a self-propelled 

flapping foil actuated in heave only 

are shown in Figure 2. Monitoring 

foil shaft position and measuring forces 

in the X (upstream-downstream) and 

Y (side to side) directions allows calculation 

of the heave velocity and other 

derived quantities such as the instantaneous 

power required by the foil to 

swim and the coefficients of thrust 

and power. The cost of transport is calculated 

by measuring the foil mass and 

dividing the cost/meter by this value 

for each foil (see Table 1). Typical 

peak thrust coefficients for highly flexible 

foils (flexural stiffness in the range 

of 10 −4 to 10 −6 Nm 2 ) are in the range 

of±0.2,lowerthantypicalforrigid 

foils, but these flexible foils nonetheless 

are capable of self-propulsion 

at speeds of 10-30 cm s −1 . 

FIGURE 2 

A key technical issue arises in studies 

of flapping foil propulsion that 

hope to imitate the self-propelled condition 

achieved by swimming fishes: 

foils that are not self-propelling may 

exhibit patterns of thrust oscillation 

that are not centered around zero. 

Any foil that is truly self-propelling 

(and not being dragged through the 

wateratspeedsslowerorfasterthan 

it would naturally move) should generate 

thrust in an oscillatory pattern, and 

thrust integrated over a single flapping 

cycle should equal zero. Graphs in the 

literature of thrust coefficients during 

foil-based locomotion that are not 

centered around zero indicate that 

the foil was being towed above or 

below the SPS and do not reflect the 

self-propelled condition. Data from 

recordings of foil forces generated 

during self-propulsion and at speeds 

below and above the SPS are shown 

in Figure 3. During self-propelled 

swimming, the thrust coefficient has 

a mean of zero over each flapping 

Sample data from a self-propelling flexible plastic foil (flexural stiffness = 9.2 × 10 −5 Nm 2 ) actuated 

in heave at the leading edge at 2-Hz frequency. Heave position is monitored by rotary encoders, 

and X andY forces are measured by a force transducer mounted on the foil shaft. From the data on 

foil position, force, and velocity, we make calculations of the instantaneous power, and dimensionless 

thrust and power coefficients. The dashed lines indicate zero for each trace. 


TABLE 1 

Locomotor properties of flexible plastic foils of three lengths while self-propelling. 

Foil length SPS (m/s) Reynolds number Strouhal number Work/cycle (mJ) Cost/meter (mJ/m) Cost of transport (mJ/g/m) 

20 0.105 21,114 0.831 2.248 42.59 106.47 

25 0.101 25,250 0.775 2.257 44.69 89.39 

35 0.091 31,850 0.349 2.233 49.08 70.11 

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 

shorter lengths. 

Reynolds and Strouhal numbers are dimensionless. 

Foils were actuated at their leading edge with ±1 cm heave, no pitch, at 2 Hz. 

Foil span was 6.8 cm for all three foils. 

Standard errors for all parameters ranged from 0.3% to 1.5% of the mean values in each column. 

cycle. When foils are forced to swim 

above their SPS, the thrust coefficient 

curves shift above the zero baseline, 

and when forced swimming occurs at 

speeds below the SPS, data are shifted 

below the baseline. 

Change in foil swimming speed 

above and below the SPS can also 

have dramatic effects on the kinematics 

of the foil (Lauder et al., 2011) and 

on the hydrodynamic wakes displayed 

by swimming foils. Figure 4 shows 

FIGURE 3 

the shape observed during swimming 

of a flexible foil (flexural stiffness, 

3.1 × 10 −6 Nm 2 )andthehydrodynamic 

wakes that result from swimming 

at, below, and above the average 

SPS. During swimming at an imposed 

speed below the SPS, the trailing edge 

ofthefoilhasalargeamplitudeand 

irregular motion that produces a wide 

bifurcating wake with separate momentum 

jets to each side (Figure 4A). 

At the SPS where the foil is allowed to 

Graph showing the dimensionless thrust coefficient versus time for a flexible plastic foil (flexural 

stiffness = 9.2 × 10 −5 Nm 2 ), 20 cm long, 6.8 cm high, actuated in heave ±1 cm at the leading 

edge at 2-Hz frequency. The red curve shows data for the self-propelled condition (18.8 cm/s), 

during which the thrust coefficient integrated over a single flapping cycle equals zero. Note that 

when experiments are done under non-self-propelling conditions (green and blue curves) the 

plots shift up or down so that the integrated coefficient over a flapping cycle is no longer zero. 

Data shown have been digitally filtered with a bandpass filter. Strouhal number for this experiment 

= 0.3 at the SPS (red curve). (Color versions of figures available online at: http://www. 

ingentaconnect.com/content/mts/mtsj/2011/00000045/00000004.) 

swim freely with no imposed constraints 

on speed, the foil bends into 

a regular ribbon-like pattern with 

a fish-likebodywake(seeNauen& 

Lauder, 2002a, 2002b) and alternating 

centers of vorticity with a fluid jet that 

meanders in between these vortical 

centers (Figure 4B). Above the SPS 

where the external free-stream flow 

is increased above SPS and the foil is 

forced to swim against the increased 

flow, the foil shape exhibits large 

amplitude wave-like motion and a 

substantial drag-like wake with fluid 

velocities below that of the free stream 

in the wake (Figure 4C). These changing 

patterns of foil kinematics and 

hydrodynamics during swimming 

that are not under conditions of selfpropulsion 

are most easily seen in highly 

flexible materials (flexural stiffnesses in 

the range of 10 −3 to 10 −6 Nm 2 ) where 

the fluid-structure interaction is most 

evident visually. These flexible materials 

which have flexural stiffnesses similar 

to those of fish (McHenry et al., 1995) 

show fish-like propulsion and deform 

into wave-like patterns with a fish-like 

wake (Figure 4B) when allowed to selfpropel 

in a low-friction system. 

The structure of the wake behind 

flapping flexible foils has a significant 

three-dimensional component due 

to the finite chord and span, and we 


FIGURE 4 

Hydrodynamics of propulsion in a flexible foil (flexural stiffness = 3.1 × 10 −6 Nm 2 ) swimming 

below, at, and above its SPS (8.55 cm/s). The foil was actuated at the leading edge with amplitude 

±0.5 cm, no pitch, at 3Hz, and is 20 cm long; Strouhal number at the SPS is 0.83 (see Table 1). The 

bottom margin of the foil is marked in white. Yellow arrows indicate water velocity; red color indicates 

counterclockwise vorticity; blue color indicates clockwise vorticity. The panels on the left 

show the whole foil swimming, while the matched panels on the right show a close-in view of the 

wake structure. The effect of swimming at a non SPS is dramatic, both on foil shape and on wake 

structure. In (A) the foil swam at an imposed speed of 4.75 cm/s, and in (C) the foil swam at an 

imposed speed of 25.7 cm/s. 

quantified this aspect of foil locomotor 

dynamics using the volumetric flow 

visualization system described by 

Flammang et al. (2011a, 2011b) and 

Troolin and Longmire (2010). Figure 

5 shows how each of the centers 

of vorticity in the ribbon-like pattern 

shown in Figure 4B actually represent 

vortical columns on each side of the 

foil, which connect to each other above 

and below the foil. These inter-column 

connections occur both to upstream 

and downstream columns on the 

same side and also across the foil to 

vortical columns on the opposite side 

(Figure 5B). To date no other studies 

have provided three-dimensional 

wake snapshots during self-propulsion 

in highly flexible foils, but such studies 

in the future may reveal interesting 

patterns of wake interaction among 

components of the foil and may contribute 

to explaining the dynamics of 

propulsion in the flexing bodies. 

These results also suggest that at 

least some of the diversity of fish 

wakes reported in the literature results 

from situations in which the fishes 

were not swimming steadily, either 

in still water or against imposed 

flows, as wakes that look like those 

in Figures 4A and 4C are frequently 

presented. Fish commonly accelerate 

and execute small maneuvers during 

locomotion and considerable effort is 

needed to ensure that kinematics and 

wake flow patterns are taken at moments 

when the fish is swimming 

steadily. Even small accelerations can 

substantially change fish wake flows 

(Tytell, 2004), and data from the 

flapping foil robot here illustrate that 

these kinematic and hydrodynamic alterations 

can be reproduced with flexible 

foils. 

One of the most straightforward 

questions that could be asked about 

undulatory propulsion and one that 

is difficult to study with live fishes is 

the effect of changing length alone on 

locomotor performance. We clearly cannot 

alter fish length experimentally 

and expect reasonable swimming performance, 

and comparing fish of 

different lengths (while useful for studies 

of scaling) does not account for the 

many other changes in the musculature 

and skeleton that occur as fish 

grow. Table 1 shows data obtained 

from three flexible foils of different 


FIGURE 5 

Volumetric flow visualization using the V3V technique (see Flammang et al., 2011a, 2011b) to 

image the 3D wake structure behind the flexible foil shown in Figure 4B. The trailing edge of 

the foil is located at the −60 mm position on the x-axis, and the fluid structures shown are all 

in the wake of the foil. This foil was self-propelling and was actuated using the same parameters 

shown for Figure 4B. (A) The flexible foil achieved a ribbon-like shape and columns of vorticity 

extend vertically on either side of the foil. Vorticity is isosurfaced at a value of 3.1 (blue surface), 

and a horizontal slice through the wake is shown to correspond to that shown using 2D piv in 

Figure 4B (green plane with velocity vectors). (B) The vortical columns on each side of the 

ribbon-like foil motion connect to each other across the top and bottom of the same side (yellow 

arrows) and opposite sides (red arrows). 

The work per cycle stays nearly constant, 

as does the cost per meter 

(Table 1), but the cost of transport 

decreases substantially as the increased 

mass of the longer foils does not result 

in increased energy requirements for 

propulsion. Foils such as these that 

are composed of a very flexible material 

self-propel at shorter lengths with a 

Strouhal number that is quite large 

relative to most self-propelling bodies. 

At longer lengths, the Strouhal number 

approaches that of many swimming 

fishes or rigid foils (Table 1). 

Figure 6 shows changes in shape 

of self-propelling foils made of the 

same material (flexural stiffness, 3.1 × 

10 −6 Nm 2 ) that occur due to change 

FIGURE 6 

Graphs to show the effect of length on the 

shape (amplitude envelope) of a flexible foil 

(flexural stiffness = 3.1 × 10 −6 Nm 2 )swimming 

at its SPS. This foil was 6.85 cm in 

chord, and of varying length, given in the 

color-coded legend. The leading edge was actuated 

at 2 Hz and with amplitude of ±1 cm. 

Each plot shows the shape of the foil during 

self-propulsion as indicated by the peak-topeak 

amplitude of the sideways flapping motion. 

(A, B) Foil shapes as the absolute distance 

along the foil and as percentage of the total foil 

length, respectively. 

lengths made of the same material and 

actuated in heave only at the leading 

edge. Altering the length of this flexible 

swimming foil from 20 to 35 cm produces 

only minor changes in the SPS 

(from10toabout9cms −1 )and 

hence in Reynolds number but dramatically 

lowers the Strouhal number 

(from0.83to0.34;Table1)dueto 

changes in foil trailing edge amplitude. 


in length. Longer foils show peak excursions 

at the same locations as shorter 

foils (Figure 6A) but amplitudes that 

are lower. Each of the foils of this material 

generates a ribbon-like shape 

when self-propelling with a consistent 

wave-like pattern down its length (see 

Figure 4B). When the amplitude of 

each foil as a percentage of foil length 

is plotted (Figure 6B), changes in 

amplitude are more evident with 

side-to-side excursion amplitudes decreasing 

as length increases while the 

wave-like pattern is retained. The 

shortest foils have high amplitudes 

near the trailing edge and an amplitude 

envelope that grows along the foil 

while the longer foils display a tapering 

amplitude envelope (compare blue and 

black curves in Figure 6B). These data 

show that length alone can have significant 

effects on locomotor efficiency 

and foil kinematics and that, all other 

things held constant, foil length increases 

on the order of 100% can provide 

reduced costs of transport. 

The Effect of Trailing 

Edge Shape on 

Swimming Performance 

The shape of the trailing tail edge 

of swimming fishes has been the subject 

of considerable discussion in the 

literature, with various authors considering 

the advantages or disadvantages 

of fish tails with symmetrical, 

asymmetrical, or forked shapes (Affleck, 

1950; Aleev, 1969; Lauder, 1989; 

Plaut, 2000; Thomson, 1971, 1976; 

Wilga & Lauder, 2004a). One advantage 

of a robotic flapping foil approach 

is that the trailing edge of a flexible 

flapping foil can be altered and a variety 

of configurations constructed that 

allow the effect of trailing edge shape 

alone on locomotor performance to 

FIGURE 7 

Propulsion by flexible foils of different shapes (material flexural stiffness = 3.1 × 10 −4 Nm 2 ) swimming 

at their SPS. Each foil was actuated at ±1 cm heave at 2 Hz. Error bars are ±2 SE. Strouhal 

numbers for these experiments range from 0.2 to 0.3. Foils were constructed to be of differing 

shapes and areas as follows: foil 1= square trailing edge, area = 131.2 cm 2 , dimensions = 6.85 cm × 

19.15 cm; foil 2 = angled trailing edge, same length as foil 1, area = 107.71 cm 2 ; foil 3 = angled 

trailing edge, same area as foil 1; foil 4 = angled trailing edge: same length and area as foil 1; foil 5 = 

forked trailing edge: same area as foil 1. 

be investigated. We designed five different 

flexible foils (material flexural 

stiffness = 3.1 × 10 −4 Nm 2 ) that control 

for total length and foil area and 

allow different tail shapes to be compared 

for the effect of these changes 

on swimming speed (Figure 7). Foil 1 

corresponds to a highly abstracted 

“trout-like” body shape with a mostly 

vertical trailing tail edge, while foils 3 

and 4 present a shark-like tail trailing 

edge shape. Many fish have forked 

tails, and this shape is represented by 

foil 5. 

The fastest swimming foil (Figure 

7, foil 4) is the one with the 

most area near the axis of actuation, 

even though it has the same area as 

several of the other foils. This result 

corresponds with our previous results 

showing that foils with higher aspect 

ratios and hence more material near 

the actuator swim significantly faster 

(Lauderetal.,2011).Interestingly, 

the foil with the angled trailing edge 

and the shark tail shape (Figure 7, 

foil 3) swims significantly faster (P < 

0.003) than a foil with the same 

area but a straight trailing edge (Figure 

7, foil 1). The foil shape with 

the significantly lowest swimming 

speed was the notched shape (Figure 

7, foil 5) even though it possesses 

the same area as foils 1 and 3. The reduced 

performance of the notched 

shape may be due to bending of the 

upper and lower “lobes” of the tail 

during the flapping motion, and kinematic 

data obtained for these foils 

does show that each lobe of this shape 

twists during the flapping cycle. Twisting 

of the tail could be reduced by introducing 

stiffening elements, and this 

may be one reason that fishes with 

high-performance tail shapes such as 

tuna possess substantial stiffening of 

both the upper and lower tail lobes 

(Fierstine & Walters, 1968; Westneat 

& Wainwright, 2001). 

Although the angled foil shape 

swims significantly faster than a foil 

of the same area with a straight trailing 

edge (Figure 7: compare foils 1 and 3), 

more power is required for this foil to 

swim at this (self-propelled) speed 

(Figure 8A). The foils were made of 


FIGURE 8 

Graphs of power consumed (A) and the cost of 

transport (B) of foils 1 and 3 (see Figure 7) selfpropelling. 

Both foils have the same surface 

area (131.2 cm 2 ) and were actuated with 

±1 cm heave at 2 Hz. Power values are significantly 

different at P =0.003.Costoftransport 

values are not significantly different at P =0.07. 

thesamematerialandhavethesame 

area, so the cost of transport can be 

calculated in mJ/m. Figure 8B shows 

that there is no significant difference 

(P = 0.07) between the foils in cost of 

transport, although there is a trend in 

the data toward the angled foil edge 

costing less per meter resulting in the 

marginally non-significant difference. 

Further experiments on both foils 

will be needed to extend this result 

and to determine if in fact there is a 

small difference in cost of transport between 

these two foil types. 

fishes that minimizes the complexity 

of the foil and maintains constant material 

properties along the foil length, 

fish bodies are clearly different in 

showing changing material properties 

from head to tail (McHenry et al., 

1995). To better understand the locomotor 

properties of the passive fish 

body alone, we used freshly dead rainbow 

trout (Oncorhynchus mykiss) and 

attached the body to the robotic flapping 

foil apparatus (Figures 1D, 1E, 

and 1F). We took care to ensure 

that rigor mortis had not set in during 

the experiments and to ensure that 

the body had thus not stiffened during 

the time the flapping trials were 

conducted. By actuating the passive 

trout body in heave and pitch in various 

combinations just behind the 

head,wewereabletoconstructa 

swimming performance surface for 

the passive fish body (Figure 9). 

Body waveforms that are remarkably 

like those occurring in live trout 

were generated by the passive fish 

bodies. These data show that heave 

amplitude overall has the greater effect 

on swimming performance than 

FIGURE 9 

changes in pitch alone. At any given 

heave value increases in pitch further 

increase swimming speed, but at 

higher heave amplitudes near ±2 cm, 

changes in pitch only produce modest 

increases in SPS (Figure 9). These 

data provide an interesting comparison 

to our previous results showing 

the effects of heave and pitch actuation 

on flexible foil propulsion (Lauder 

et al., 2011). In that study adding 

pitch motions to baseline heave actuation 

for foils of varying flexural stiffness 

did not produce significant 

increases in foil SPS but did allow 

stiffer foils to maintain a relatively 

high swimming speed that would 

have declined with heave actuation 

only. 

As driving frequency increases, the 

tail beat amplitude of the passive flapping 

trout body remains relatively 

constant from 0.1 to 2.0 Hz before 

increasing steadily to a peak at 

3.5 Hz (Figure 10). These tail beat 

amplitude values are very similar to 

those observed in live trout swimming 

under a variety of locomotor conditions 

(Liao et al., 2003a; Webb, 

Performance surface of a freshly dead trout (Oncorhynchus mykiss) attached to a robotic controller 

(see Figures 1D, 1E, 1F) driven at 2 Hz at a variety of heave and pitch amplitudes. This trout was 

25.3 cm in total length. The graph shows how SPS of the passive trout body varies with different 

pitch and heave actuation parameters. 

Undulatory Locomotion 

of a Passive Fish Body 

Although flapping flexible foils 

provide a reasonable and simple 

model for undulatory propulsion in 


FIGURE 10 

Graph of tail-tip amplitude versus heave actuation frequency for a freshly dead trout (Oncorhynchus 

mykiss) attached to a robotic controller driving the passive body at a variety of frequencies. This trout 

was 25.3 cm in total length and was actuated with a constant heave (±2 cm) and pitch (±5°). Error 

bars are ±1 SE of the mean. 

1971; Webb et al., 1984) and increases 

in tail beat amplitude of the 

magnitude shown here for the passive 

trout body are similar to those observed 

previously as fish increase 

swimming speed and alter both frequency 

(primarily) and amplitude 

(to a lesser extent) of the tail beat. 

Although during routine undulatory 

swimming fish bodies are not 

passive and are certainly stiffened by 

body musculature as fish swim, for 

all but the fastest swimming speeds 

locomotion is powered by red muscle 

fibers, which can make up a rather 

small percentage of body mass 

(around 1.5% in largemouth bass; 

see Johnson et al., 1994). The bulk 

of the locomotor musculature is composed 

of white fibers that are not activated 

until the fastest swimming 

speeds are needed (Jayne & Lauder, 

1994, 1995a). Thus, the red muscle 

fibers can be considered as acting to 

bend a mostly passive fish body composed 

of white muscle fibers and associated 

skeletal tissues that may not be 

much different in flexural stiffness 

from the passive bodies studied here. 

Also, data on the propulsion of passive 

fish bodies are relevant to fishes 

swimming in turbulent flows. Trout 

swimming in a vortex street have been 

shown to greatly alter body kinematics 

and to utilize vortical energy shed from 

objects in the flow (Liao, 2004; Liao 

et al., 2003b). The amplitude of the 

center of mass oscillation by trout 

swimming in the Karman gait is up to 

±2 cm for a 10-cm-long trout, which 

corresponds on a length-specific basis 

to the middle region of the performance 

surface in Figure 9 (Liao et al., 2003a). 

By greatly reducing muscle activity 

when trout enter a vortex street, the 

body becomes largely passive and deforms 

in response to the oncoming 

flows. Trouts are able to maintain 

their position in such flows entirely passively 

and allow their bodies to extract 

energy from the oncoming flow and 

generate thrust. This phenomenon 

was further demonstrated in a study 

using passive foils and freshly dead 

fish bodies to show that the passive 

trout body alone in a vortex wake can 

generate sufficient thrust to maintain 

position in flow (Beal et al., 2006). 

Fin-Fin Hydrodynamic 

Interactions 

One of the most intriguing aspects 

of fish functional design is the arrangement 

of two fins in series that could 

allow enhanced locomotor efficiency 

through hydrodynamic interactions 

between the fins. For example, the 

dorsal finandtheanalfininmostfishes 

are located upstream of the caudal fin, 

and the wakes shed from these fins 

could interact with the tail fin during 

locomotion (Drucker & Lauder, 2001, 

2005; Tytell, 2006). Undulatory locomotion 

using the body also causes the 

attached median fins to oscillate from 

side to side, and in the fish species studied 

so far these fins have been shown also 

to be actively oscillated by intrinsic fin 

musculature (Jayne et al., 1996) and 

thus can generate thrust on their own. 

The possibility of fin-fin hydrodynamic 

interactions during locomotion 

has been explored in a number of 

experimental papers on living fishes 

(Drucker & Lauder, 2001, 2005; 

Standen, 2008; Standen & Lauder, 

2007; Tytell, 2006; Webb & Keyes, 

1981) as well as using computational 

approaches (Akhtar et al., 2007; Weihs 

et al., 2006). Akhtar et al. (2007) used 

data on the kinematics of the bluegill 

dorsal fin and tail from Drucker and 

Lauder (2001) and performed a twodimensional 

computational analysis of 

the effect of having the fins in series 

and found that vortex shedding from 

the dorsal fin can increase thrust of the 

tail and that the amount of this thrust 

increase depends on the phasing of 

dorsal fin andtailmotion. 

In order to experimentally evaluate 

hydrodynamic fin-fin interactions 

using an apparatus in which phasing 

and the distance between fins can be 

experimentally manipulated, we used 

our robotic flapping foil apparatus 

in the dual-foil configuration with 


upstream and downstream foils. Foils 

were rigid aluminum plates in a 

NACA 0012 airfoil shape, and each 

foil could be moved in heave and 

pitch, and both foils were driven from 

a common carriage mounted on air 

bearings as described above. Foils were 

moved according to the parameters 

measured for bluegill sunfish dorsal 

and caudal fins (Drucker & Lauder, 

2001) and used by Akhtar et al. 

(2007) for their computational study 

of this problem: the upstream foil was 

moved with heave amplitude of 

2.5 cm, the downstream foil at 

3.5 cm heave amplitude. Pitch amplitude 

for both foils was ±20°, and the 

frequency for both foils was 1.7 Hz, 

corresponding to the frequency of fin 

flapping in the bluegill sunfish model 

case (Drucker & Lauder, 2001). 

Increases in SPS as a result of 

changing foil phase and distance reflect 

thrust enhancement beyond the 

thrust achieved by the two foils operating 

separately (assessed by offsetting 

the foils from each other so that the 

downstream foil was no longer in 

the wake of the upstream foil). A general 

description of the dual-foil configuration 

and images of flows 

around and between the foils are presented 

in Lauder et al. (2007). Data 

for SPS were plotted over a range of 

phase differences between the upstream 

and downstream foils and 

polynomial fits to these data points 

were used to determine the change 

in SPS with phase (e.g., Figure 11). 

Here we present the results of experiments 

measuring the SPS of dual 

foil flapping. Figure 11a shows the effect 

of changing the phase of sinusoidal 

motion of the downstream 

foil relative to the upstream foil on 

SPS for three different spacings between 

the foils. In each case, there is 

a clear peak in swimming speed, and 

FIGURE 11 

Robotic model of fish fin hydrodynamic interactions. Dual NACA 0012 flapping foils in series are 

driven by a robotic flapping foil apparatus. Details of this device and of the experimental setup are 

given in Lauder et al. (2007). (A) SPS plotted against the phase difference between the upstream 

and downstream foils. Foils were arranged with 0 side-to-side offset (in cm) and are thus moving 

in line with each other with the downstream foil at varying chord length separations from the 

upstream foil: 0.5, 1, and 2 chord lengths separation; foil chord length was 6.85 cm. Further 

experimental details are provided in the text. (B) Effect of changing the foil offsets (in cm) so 

that the downstream foil is not moving in the direct wake of the upstream foil. The one chord length 

spacing with zero offset curve (dark blue) is the same as in panel (A) and is shown for reference. 

The other three plots show the effect of a 3.5 cm offset of the midline motion of the downstream 

foil relative to the upstream foil, removing it from the upstream foil wake. 

as the distance between the foils is increased 

the peak swimming speed 

shifts toward a larger phase lag of 

the downstream foil. Interestingly, 

the maximal swimming speed of the 

two foils together does not change significantly 

as the distance between foils 

changes, and alterations in phasing 

between the foils can thus be used 

to compensate for changes in spacing. 

For each interfoil spacing, however, 

there are phase relationships that significantly 

reduce swimming performance, 

indicating that thrust of the 

entire two-foil system is sensitive to 

phase relationships between two foils 

flapping. These data correspond very 

well to the computational results of 

Akhtar et al. (2007), which showed 

peak thrust enhancement at a phase 


of about 40°, very similar to our data 

(Figure11A,bluecurve)wherethe 

plateau around the SPS peak includes 

the 40° value. 

Figure 11B illustrates the changes 

in propulsion that result from moving 

the downstream foil to the side by an 

offset of 3.5 cm to move it out of the 

wake of the upstream foil. This effectively 

creates a tandem foil configuration 

in which fluid dynamic 

interactions between the foils are 

minimized, although not completely 

eliminated. Comparison of the three 

offset curves to the zero offset curve 

shows that offsetting the foils reduces 

both the SPS and the effect of foil 

phasing. Offset plots show reduced 

effects of foil phasing (with lower 

maxima and higher minima) and less 

distinct overall peaks in SPS, showing 

that the downstream foil was not able 

to improve swimming speed of the 

two foils together when removed 

from the upstream foil wake. 

Computational work and preliminary 

flow visualization (Lauder et al., 

2007) data indicate that the behavior 

of tandem foils can be explained in 

terms of how the downstream foil interacts 

with fluid structures generated by 

the upstream foil. The shifting of the 

peak SPS with increased spacing of 

the tandem foils can be explained by 

the fact that, for larger spacings, the convection 

of those structures to the downstream 

foil takes longer. That time, tc,is 

simply the spacing between the foils, s, 

divided by the convection velocity, Uc. 

The increase in phase lag, Δϕ, needed 

so that the downstream foil meets the 

fluid structures at the right time is simply 

ðS 2 S 1 Þ 

Δϕ ¼ 2πf 

U c 

where f is the frequency of flapping 

and s 2 and s 1 are two different spacings. 

If we assume that Uc is close to the 

SPS, this equation estimates the Δϕ 

between the peak SPS well for the 

three spacings we used. The equation 

predicts Δϕ = 0.648 radians, or 37°, 

for a spacing difference of 0.5 chord 

lengths, and 74° for a difference of 

one chord length. The Δϕ from our 

data are approximately 30°and 75° for 

the corresponding spacing differences. 

The nearness of the foils and/or higher 

convection velocities at the smaller 

spacings may explain the lower than 

predicted Δϕ for the peaks in SPS in 

these cases. Overall, the agreement is 

very good considering that the peaks 

are somewhat broad and the equation 

for Δϕ is a simplification of the fluid 

dynamics. 

The Future of 

Undulatory Biorobotics 

In this paper, we use a robotic tool 

for investigating a variety of phenomena 

relating to undulatory propulsion 

in fishes and present experimental 

data that would be difficult if not impossible 

to obtain from studying live 

animals. The promise of robotic models 

for studying the biomechanics of 

locomotion in fishes has just begun 

to be realized (Curet et al., 2011; 

Long et al., 2006, 2010; Tangorra 

et al., 2010, 2011), and fundamental 

questions relating to the mechanics of 

undulatory propulsion remain to be 

addressed. In particular, key unresolved 

issues are the extent to which 

changes in body stiffness during propulsion 

affect locomotor performance 

(see Long & Nipper, 1996) and how 

active modulation of stiffness during 

an undulatory cycle and across 

changes in swimming speed are 

achieved and affect propulsive speed 

and efficiency. 

To address these questions, a new 

generation of robotic undulatory devices 

will be needed that allow for 

controlled modulation of body stiffness 

and the phasing of stiffness 

changes with undulatory cycles of 

compression and tension on the 

bending fish or foil body. An additional 

arena that is key to making 

progress in understanding undulatory 

mechanics is the ability to perturb the 

locomotor system to assess how stiffness 

of the body relates to the ability 

to recover from perturbations. There 

have been very few studies of perturbations 

of undulatory locomotor systems 

(see Webb, 2004), and yet fishes 

oftenswiminchallenginghydrodynamic 

environments in which they 

are forced to recover from impulsive 

challenges to their undulatory 

pattern. 

Acknowledgments 

This work was supported the Office 

of Naval Research grant N00014- 

09-1-0352 on fin neuromechanics 

monitored by Dr. Thomas McKenna 

and by the National Science Foundation 

grant EFRI-0938043. We 

thank the members of the Lauder 

and Tangorra labs for many helpful 

discussions on fish fins and flexible 

flapping foil propulsion and Nate 

Jackson for his assistance with the 

dual-flapping foil experiments. Many 

thanks to Brooke Flammang, Tyson 

Strand, and Dan Troolin for assistance 

with the V3V volumetric flow 

imaging experiments on the undulating 

plastic foil. 

Lead Author: 


The Museum of 

Comparative Zoology 


26 Oxford Street, Harvard University 

Cambridge, MA 02138 

Email: glauder@oeb.harvard.edu 

References 

Affleck, R.J. 1950. Some points in the 

function, development, and evolution of the 

tail in fishes. Proc Zool Soc Lond. 120:349-68. 

doi: 10.1111/j.1096-3642.1950.tb00954.x. 

Akhtar, I., Mittal, R., Lauder, G.V., & 

Drucker, E. 2007. Hydrodynamics of a 

biologically inspired tandem flapping foil 

configuration. Theor Comp Fluid Dyn. 

21:155-70. doi: 10.1007/s00162-007-0045-2. 

Aleev, Y.G. 1969. Function and Gross 

Morphology in Fish. Jerusalem: Keter Press. 

Beal, D.N., Hover, F.S., Triantafyllou, 

M.S., Liao, J., & Lauder, G.V. 2006. 

Passive propulsion in vortex wakes. 


S0022112005007925. 

Bone, Q., Kiceniuk, J., & Jones, D.R. 1978. 

On the role of the different fibre types in fish 

myotomes at intermediate swimming speeds. 

Fish Bull. 76:691-9. 




locomotive strategy. J R Soc Interface. 

8:1041-50. doi: 10.1098/rsif.2010.0493. 

Donley, J., & Dickson, K.A. 2000. Swimming 

kinematics of juvenile Kawakawa tuna 

(Euthynnus affinis) and chub mackerel (Scomber 

japonicus). J Exp Biol. 203:3103-16. 


Locomotor forces on a swimming fish: Threedimensional 

vortex wake dynamics quantified 

using digital particle image velocimetry. 

J Exp Biol. 202:2393-412. 


Locomotor function of the dorsal fin in teleost 

fishes: Experimental analysis of wake forces 

in sunfish. J Exp Biol. 204:2943-58. 


Locomotor function of the dorsal fin in 

rainbow trout: Kinematic patterns and hydrodynamic 

forces. J Exp Biol. 208:4479-94. 

doi: 10.1242/jeb.01922. 

Fetcho, J., & Svoboda, K.R. 1993. Fictive 

swimming elicited by electrical stimulation 

of the midbrain in goldfish. J Neurophysiol. 

70:764-80. 

Fetcho, J.R. 1986. The organization of the 

motoneurons innervating the axial musculature 

of vertebrates: I. Goldfish (Carassius 

auratus) and mudpuppies (Necturus maculosus). 

J Comp Neurol. 249:521-50. doi: 10.1002/ 

cne.902490408. 

Fierstine, H.L., & Walters, V. 1968. Studies 

in locomotion and anatomy of scombroid 

fishes. Mem Southern Calif Acad Sci. 6:1-31. 

Flammang, B.E., & Lauder, G.V. 2008. 

Speed-dependent intrinsic caudal fin muscle 

recruitment during steady swimming in 

bluegill sunfish, Lepomis macrochirus. J Exp 

Biol. 211:587-98. doi: 10.1242/jeb.012096. 

Flammang, B.E., & Lauder, G.V. 2009. 

Caudal fin shape modulation and control 

during acceleration, braking and backing 

maneuvers in bluegill sunfish, Lepomis 

macrochirus. J Exp Biol. 212:277-86. 

doi: 10.1242/jeb.021360. 

Flammang, B.E., Lauder, G.V., Troolin, 

D.R., & Strand, T. 2011a. Volumetric 

imaging of shark tail hydrodynamics reveals 

a three-dimensional dual-ring vortex wake 

structure. P Roy Soc B. 278. doi: 10.1098/ 

rspb.2011.0489. 

Flammang, B.E., Lauder, G.V., Troolin, 

D.R., & Strand, T.E. 2011b. Volumetric 

imaging of fish locomotion. Biol Lett. 

doi: 10.1098/rsbl.2011.0282. 

Gillis, G.B. 1996. Undulatory locomotion 

in elongate aquatic vertebrates: Anguilliform 

swimming since Sir. James Gray. Am Zool. 

36:656-65. 

Gillis, G.B. 1998. Environmental effects on 

undulatory locomotion in the American eel 

Anguilla rostrata: Kinematics in water and 

on land. J Exp Biol. 201:949-61. 

Harris, J.E. 1936. The role of the fins in the 

equilibrium of the swimming fish: I. Wind 

tunnel tests on a model of Mustelus canis 

(Mitchell). J Exp Biol. 13:476-93. 

Harris, J.E. 1938. The role of the fins in the 

equilibrium of the swimming fish: II. The role 

of the pelvic fins. J Exp Biol. 16:32-47. 

Hove, J.R., O’Bryan, L.M., Gordon, M.S., 

Webb, P.W., & Weihs, D. 2001. Boxfishes 

(Teleostei: Ostraciidae) as a model system for 

fishes swimming with many fins: Kinematics. 

J Exp Biol. 204:1459-71. 

Jayne, B.C., & Lauder, G.V. 1993. Red and 

white muscle activity and kinematics of the 

escape response of the bluegill sunfish during 

swimming. J Comp Physiol A. 173:495-508. 

doi: 10.1007/BF00193522. 

Jayne, B.C., & Lauder, G.V. 1994. How 

swimming fish use slow and fast muscle fibers: 

Implications for models of vertebrate muscle 

recruitment. J Comp Physiol A. 175:123-31. 

doi: 10.1007/BF00217443. 

Jayne, B.C., & Lauder, G.V. 1995a. Are 

muscle fibers within fish myotomes activated 

synchronously Patterns of recruitment within 

deep myomeric musculature during swimming 

in largemouth bass. J Exp Biol. 

198:805-15. 

Jayne, B.C., & Lauder, G.V. 1995b. Red 

muscle motor patterns during steady swimming 

in largemouth bass: Effects of speed 

and correlations with axial kinematics. 

J Exp Biol. 198:1575-87. 

Jayne, B.C., & Lauder, G.V. 1995c. 

Speed effects on midline kinematics during 

steady undulatory swimming of largemouth 

bass, Micropterus salmoides. JExpBiol. 

198:585-602. 

Jayne, B.C., Lozada, A., & Lauder, G.V. 

1996. Function of the dorsal fin in bluegill 

sunfish: Motor patterns during four locomotor 

behaviors. J Morphol. 228:307-26. 

doi:10.1002/(SICI)1097-4687(199606) 

228:33.0.CO;2-Z. 

Johnson, T.P., Syme, D.A., Jayne, B.C., 

Lauder, G.V., & Bennett, A.F. 1994. Modeling 

red muscle power output during steady and unsteady 

swimming in largemouth bass (Micropterus 

salmoides). Am J Physiol. 267:R481-8. 


Lauder, G., Madden, P.G.A., Tangorra, J., 

Anderson, E., & Baker, T.V. 2011. 

Bioinspiration from fish for smart material 

design and function. Smart Mater Struct. 

in press. 

Lauder, G.V. 1989. Caudal fin locomotion 

in ray-finned fishes: Historical and functional 

analyses. Am Zool. 29:85-102. 

Lauder, G.V. 2000. Function of the caudal 

fin during locomotion in fishes: Kinematics, 

flow visualization, and evolutionary patterns. 

Am Zool. 40:101-22. doi: 10.1668/0003- 

1569(2000)040[0101:FOTCFD]2.0.CO;2. 

Lauder, G.V., Anderson, E.J., Tangorra, J., 

& Madden, P.G.A. 2007. Fish biorobotics: 

Kinematics and hydrodynamics of selfpropulsion. 

J Exp Biol. 210:2767-80. 

doi: 10.1242/jeb.000265. 

Lauder, G.V., & Tytell, E.D. 2006. Hydrodynamics 

of undulatory propulsion. In: 

Fish Biomechanics. Volume 23 in Fish 

Physiology, eds. Shadwick, R.E., & Lauder, 

G.V., 425-68. San Diego: Academic Press. 

Liao, J. 2002. Swimming in needlefish 

(Belonidae): Anguilliform locomotion with 

fins. J Exp Biol. 205:2875-84. 

Liao, J. 2004. Neuromuscular control of trout 

swimming in a vortex street: Implications for 

energy economy during the Karman gait. 

J Exp Biol. 207:3495-506. doi: 10.1242/ 

jeb.01125. 

Liao, J., Beal, D.N., Lauder, G.V., & 

Triantafyllou, M.S. 2003a. The Kármán gait: 

Novel body kinematics of rainbow trout 

swimming in a vortex street. J Exp Biol. 

206:1059-73. doi: 10.1242/jeb.00209. 

Liao, J., & Lauder, G.V. 2000. Function 

of the heterocercal tail in white sturgeon: Flow 

visualization during steady swimming and 

vertical maneuvering. J Exp Biol. 203:3585-94. 

Liao, J.C., Beal, D.N., Lauder, G.V., & 

Triantafyllou, M.S. 2003b. Fish exploiting 

vortices decrease muscle activity. Science. 

302:1566-9. doi: 10.1126/science.1088295. 

Long, J.H., Jr., Koob, T.J., Irving, K., 

Combie, K., Engel, V., Livingston, N., … 

Schumacher, J. 2006. Biomimetic evolutionary 

analysis: Testing the adaptive value of 

vertebrate tail stiffness in autonomous swimming 

robots. J Exp Biol. 209:4732-46. 

doi: 10.1242/jeb.02559. 

Long, J.H., McHenry, M.J., & Boetticher, 

N.C. 1994. Undulatory swimming: How 

traveling waves are produced and modulated 

in sunfish (Lepomis gibbosus). J Exp Biol. 

192:129-45. 

Long, J.H., & Nipper, K.S. 1996. The 

importance of body stiffness in undulatory 

propulsion. Am Zool. 36:678-694. 

Long, J.H., Porter, M.E., Root, R.G., & 

Liew, C.W. 2010. Go reconfigure: How 

fish change shape as they swim and evolve. 

Integr Comp Biol. 50:1120-39. doi: 10.1093/ 

icb/icq066. 

McHenry, M.J., Pell, C.A., & Long, J.A. 

1995. Mechanical control of swimming speed: 

Stiffness and axial wave form in undulating 

fish models. J Exp Biol. 198:2293-305. 

Nauen, J.C., & Lauder, G.V. 2002a. Hydrodynamics 

of caudal fin locomotion by chub 

mackerel, Scomber japonicus (Scombridae). 

J Exp Biol. 205:1709-24. 

Nauen, J.C., & Lauder, G.V. 2002b. 

Quantification of the wake of rainbow trout 

(Oncorhynchus mykiss) using three-dimensional 

stereoscopic digital particle image velocimetry. 

J Exp Biol. 205:3271-9. 

Plaut, I. 2000. Effects of fin size on swimming 

performance, swimming behavior and 

routine activity of zebrafish Danio rerio. 

J Exp Biol. 203:813-20. 

Rome, L.C., Swank, D., & Corda, D. 1993. 

How fish power swimming. Science. 261: 

340-3. doi: 10.1126/science.8332898. 

Shadwick, R., & Gemballa, S. 2006. 

Structure, kinematics, and muscle dynamics in 

undulatory swimming. In: Fish Biomechanics. 

Volume 23 in Fish Physiology, eds. Shadwick, 

R.E., & Lauder, G.V., 241-280. San Diego: 


Standen, E.M. 2008. Pelvic fin locomotor 

function in fishes: Three-dimensional kinematics 

in rainbow trout (Oncorhynchus mykiss). 

J Exp Biol. 211:2931-42. doi: 10.1242/ 

jeb.018572. 

Standen, E.M. 2010. Muscle activity and 

hydrodynamic function of pelvic fins in 

trout (Oncorhynchus mykiss). J Exp Biol. 

213:831-41. doi: 10.1242/jeb.033084. 

Standen, E.M., & Lauder, G.V. 2005. Dorsal 

and anal fin function in bluegill sunfish 

(Lepomis macrochirus): Three-dimensional 

kinematics during propulsion and maneuvering. 

J Exp Biol. 205:2753-63. doi: 10.1242/ 

jeb.01706. 

Standen, E.M., & Lauder, G.V. 2007. 

Hydrodynamic function of dorsal and anal 

fins in brook trout (Salvelinus fontinalis). 

J Exp Biol. 210:325-39. doi: 10.1242/ 

jeb.02661. 

Syme, D.A. 2006. Functional properties of 

skeletal muscle. In: Fish Biomechanics. 

Volume 23 in Fish Physiology, eds. Shadwick, 

R.E., & Lauder, G.V., 179-240. San Diego: 


Tangorra, J., Phelan, C., Esposito, C., & 

Lauder, G. 2011. Biorobotic models of highly 

deformable fins for studies of the mechanics 

and control of fin forces. Integr Comp Biol. 

in press. doi: 10.1093/icb/icr036. 

Tangorra, J.L., Lauder, G.V., Hunter, I., 

Mittal, R., Madden, P.G., & Bozkurttas, M. 

2010. The effect of fin ray flexural rigidity 

on the propulsive forces generated 

by a biorobotic fish pectoral fin. J Exp 

Biol. 213:4043-54. doi: 10.1242/ 

jeb.048017. 

Thomson, K.S. 1971. The adaptation and 

evolution of early fishes. Q Rev Biol. 46: 

139-66. doi: 10.1086/406831. 

Thomson, K.S. 1976. On the heterocercal 

tail in sharks. Paleobiology. 2:19-38. 

Troolin, D., & Longmire, E. 2010. Volumetric 

velocity measurements of vortex rings from 

inclined exits. Exp Fluids. 48:409-20. 

doi: 10.1007/s00348-009-0745-z. 

Tytell, E.D. 2004. Kinematics and hydrodynamics 

of linear acceleration in eels, 


Anguilla rostrata. P Roy Soc Lond B. 

271:2535-40. doi: 10.1098/rspb.2004.2901. 

Tytell, E.D. 2006. Median fin function 

in bluegill sunfish, Lepomis macrochirus: 

Streamwise vortex structure during steady 

swimming. J Exp Biol. 209:1516-34. 

doi: 10.1242/jeb.02154. 

Wilga, C.D., & Lauder, G.V. 2004b. 

Hydrodynamic function of the shark’s tail. 

Nature. 430:850. doi: 10.1038/430850a. 

Tytell, E.D., & Lauder, G.V. 2002. The 

c-start escape response of Polypterus senegalus: 

Bilateral muscle activity and variation during 

stage 1 and 2. J Exp Biol. 205:2591-603. 

Webb, P.W. 1971. The swimming energetics 

of trout: I. Thrust and power output at 

cruising speeds. J Exp Biol. 55:489-520. 

Webb, P.W. 2004. Response latencies to 

postural disturbances in three species of 

teleostean fishes. J Exp Biol. 207:955-61. 

doi: 10.1242/jeb.00854. 

Webb, P.W., & Keyes, R.S. 1981. Division 

of labor between median fins in swimming 

dolphin (Pisces: Coryphaenidae). Copeia. 

1981:901-04. doi: 10.2307/1444198. 

Webb, P.W., Kostecki, P.T., & Stevens, 

E.D. 1984. The effect of size and swimming 

speed on the locomotor kinematics of rainbow 

trout. J Exp Biol. 109:77-95. 

Weihs, D., Ringel, M., & Victor, M. 2006. 

Aerodynamic interactions between adjacent 

slender bodies. AIAA J. 44:481-4. 

doi: 10.2514/1.18902. 

Westneat, M., & Wainwright, S.A. 2001. 

Mechanical design for swimming: Muscle, 

tendon, and bone. In: Tuna: Physiology, 

Ecology, and Evolution, eds. Block, B., & 

Stevens, E.D., 271-311. San Diego: Academic 

Press. doi: 10.1016/S1546-5098(01)19008-4. 

Wilga, C.D., & Lauder, G.V. 2002. Function 

of the heterocercal tail in sharks: Quantitative 

wake dynamics during steady horizontal 

swimming and vertical maneuvering. 

J Exp Biol. 205:2365-74. 

Wilga, C.D., & Lauder, G.V. 2004a. 

Biomechanics of locomotion in sharks, rays 

and chimeras. In: Biology of Sharks and Their 

Relatives, eds. Carrier, J.C., Musick, J.A., 

& Heithaus, M.R., 139-64. Boca Raton: 

CRC Press. 


PAPER 

Thrust Production in Highly Flexible Pectoral 

Fins: A Computational Dissection 

AUTHORS 

Srinivas Ramakrishnan 1 

ANSYS, Inc. 

Meliha Bozkurttas 

Franklin W. Olin 

College of Engineering 

Rajat Mittal 

Department of Mechanical Engineering, 

Johns Hopkins University 


Department of Organismic 

and Evolutionary Biology, 


Introduction 

Robust design based on natural 

systems is a significant engineering 

challenge. Evolution-based design is 

inherently a multi-objective optimization 

problem. Natural selection puts 

pressure on organisms to produce 

locomotion abilities that balance competing 

requirements of speed, efficiency, 

and effectiveness. The goals 

of ongoing research efforts are to elucidate 

the competing requirements that 

have enabled the evolution of highly 

maneuverable propulsion/locomotion 

at low speeds. Prominent natural systems 

of interest are flapping flight in 

air and aquatic locomotion and a common 

feature among these systems is 

the presence of highly compliant control 

surfaces. Organisms that employ 

these models of locomotion appear to 

exploit the flexibility of their wings/ 

1 Work presented here was done while the author 

was a postdoctoral scientist at The George Washington 

University prior to joining Ansys Inc. 

ABSTRACT 

Bluegill sunfish pectoral fins represent a remarkable success in evolutionary 

terms as a means of propulsion in challenging environments. Attempts to mimic 

their design in the context of autonomous underwater vehicles have overwhelmingly 

relied on the analysis of steady swimming. Experimental observations of 

maneuvers reveal that the kinematics of fin and wake dynamics exhibit characteristics 

that are distinctly different from steady swimming. We present a computational 

analysis that compares, qualitatively and quantitatively, the wake hydrodynamics 

and performance of the bluegill sunfish pectoral fin for two modes of swimming: 

steady swimming and a yaw turn maneuver. It is in this context that we comment on 

the role that flexibility plays in the success of the pectoral fin as a versatile propulsor. 

Specifically, we assess the performance of the fin by conducting a “virtual dissection” 

where only a portion of fin is retained. Approximately 90% of peak thrust for 

steady swimming is recovered using only the dorsal half. This figure drops to 70% 

for the yaw turn maneuver. Our findings suggest that designs based on fin analysis 

that account for various locomotion modes can lead to more robust performance 

than those based solely on steady swimming. 

Keywords: computational fluid dynamics (CFD), immersed boundary methods 

(IBM), bluegill sunfish, biological locomotion 

fins to achieve high maneuverability 

at low speeds. This paper presents the 

analysis of one such control surface: 

the bluegill sunfish pectoral fin. 

Atypicalsunfish pectoral fin 

consists of 14 fin raysasshownin 

Figure 1. We see the fin raysnumbered 

sequentially starting from the 

dorsal edge (ray 1) to the ventral edge 

(ray 14). These rays support an asymmetric 

planform shape for the pectoral 

fin. Figure 3 shows different frames of 

the sunfish executing a maneuver from 

a ventral view. The motion of the pectoral 

fin and body are captured using 

multiple high-speed video cameras 

simultaneously operating at 250 or 

more frames per second with a 1024 × 

1024 resolution (Lauder et al., 2006). 

The wing surface is digitized at about 

300 spatial locations at several points 

during the fin cycle.Thus,thekinematics 

of the fin motionisacquired 

for the simulation. The collaboration 

with experimentalists (biologists and 

FIGURE 1 

Bluegill sunfish pectoral fin consists of 

14 rays, which form the full planform. The 

dissected planform is interpolated from rays 

1–8 to investigate the flow and performance. 


FIGURE 2 

Bioinspired design paradigm. 

engineers), through a multi-disciplinary 

effort (Lauder et al., 2006; Mittal et al., 

2006), has enabled high-fidelity data 

to be used in the computational analysis 

(see Figure 2). 

It is clear from looking at the fin 

motion during the maneuver (see Figure 

3) that the kinematics of the fin 

involves both deformation and translation. 

This poses severe challenges for 

traditional body-fitted computational 

methods. Here, the immersed boundary 

method, with its ability to handle 

complex deforming structures, enables 

us to undertake high-fidelity computational 

fluid dynamics (CFD) analysis of 

the pectoral fin hydrodynamics(see 

Computational Methodology). It has 

been used to gain valuable insight into 

pectoral fin hydrodynamics in steady 

swimming (Bozkurttas et al., 2009; 

Dong et al., 2010). The experimentally 

obtained steady swimming kinematics 

was analyzed, and an efficient reconstruction 

of the kinematics using proper 

orthogonal decomposition (POD) 

was obtained. The POD modes using 

a combination of the first three modes 

(hereafterreferredtoasMode1+2+3) 

were successful in reproducing two 

thirds of the full fin kinematics. More 

significantly, this combination of 

modes was found to retain 92% 

of the thrust produced using the actual 

kinematics (Bozkurttas, 2007; 

Bozkurttas et al., 2009). Further, detailed 

analysis of the pressure distribution 

over the full fin surface (rays 

1-14) during steady swimming also revealed 

that most of the thrust was produced 

by the dorsal part mainly around 

the spanwise tip region (Bozkurttas 

et al., 2009; Dong et al., 2010) (see Figure 

5). Since different sections of the 

pectoral fin trace different trajectories 

during a fin stroke, the contribution 

of each region of the fin to its overall 

performance may not be uniform. Naturally, 

this leads us to the central theme 

of this paper, the idea of examining the 

thrust production of different sections 

of the fin. The goal is to enable a virtual 

“dissection” or “ablation” of the pectoral 

fin dynamics and the effect of this 

ablation on the fin performance. It is 

expected that this will yield useful insight 

into the hydrodynamic function 

of the fin in various swimming modes. 

FIGURE 3 

A bluegill sunfish during a maneuver: ventral (bottom) view. Images are frames from a highspeed 

video. Note the differential motion of the left and right side fins. Top row: t/T =0,t/T = 

0.23, t/T = 0.30. Bottom row: t/T = 0.46, t/T = 0.70, t/T = 0.84. 

Computational 

Methodology 

We present a brief description of 

the Cartesian grid-based immersed 

boundary method for moving boundaries 

starting with the governing 

equations. The three-dimensional 

unsteady, viscous incompressible 

Navier-Stokes equations are given as 

∂u i 

¼ 0 

∂x i 

∂u i 

∂t þ ∂ u 

iu j 

¼ 1 ∂p 

þ ν ∂ 

∂u i 

∂x j ρ ∂x i ∂x j ∂x j 

ð1Þ 

where i; j =1,2,3,u i are the velocity 

component, p is the pressure, and ρ 


and ν are the fluid density and kinematicviscosity.Wehaveemployeda 

conventional notation where repeated 

indices imply summation. 

1. Numerical Method 

The Navier-Stokes equations 

(Eq. 1) are discretized using a cellcentered, 

collocated (non-staggered) 

arrangement of the primitive 

variables (u i , p). In addition to the 

cell-centered velocities (u i ), the 

face-centered velocities, U i ,are 

computed. A second-order Adams- 

Bashforth scheme is employed for 

the convective terms while the diffusion 

terms are discretized using 

an implicit Crank-Nicolson scheme 

whicheliminatestheviscousstability 

constraint. The spatial derivatives 

are computed using a 

second-order accurate central difference 

scheme. The equations 

are integrated in time using the 

fractional step method (Chorin, 

1967). In the first sub-step of this 

method, a modified momentum 

equation is solved and an intermediate 

velocity u* obtained. The second 

sub-step requires the solution 

of the pressure correction equation 

which is solved with the constraint 

that the final velocity u i 

n+1 

be divergence-free. This gives a 

Poisson equation for the pressure 

correction and a Neumann boundary 

condition imposed on this pressure 

correction at all boundaries. 

This Poisson equation is solved 

with a highly efficient geometric 

multigridmethodwhichemploys 

a Gauss-Siedel line-SOR smoother. 

Once the pressure correction is obtained, 

the pressure and velocity are 

updated (see Dong et al., 2006 and 

Mittal et al., 2008, for additional 

details). These separately updated 

face velocities satisfy discrete mass 

conservation to machine accuracy 

and use of these velocities in estimating 

the non-linear convective flux 

leads to a more accurate and robust 

solution procedure. The advantage 

of separately computing the facecentered 

velocities was initially proposed 

by Zang et al. (1994) and 

discussed in the context of the 

Cartesian grid methods in Ye et al. 

(1999) and Mittal et al. (2008). 

2. Immersed Boundary Treatment 

The immersed boundary method 

used here employs a multidimensional 

ghost cell methodology 

to impose the boundary conditions 

on the immersed boundary. The 

current solver is designed from the 

start for fast, efficient, and accurate 

solution of flows with complex 

three-dimensional, moving boundaries. 

Also, the current method is 

a “sharp“ interface method in that 

the boundary conditions on the 

immersed boundary are imposed 

at the precise location of the immersed 

body, and there is no spurious 

spreading of boundary forcing 

into the fluid as what usually occurs 

with diffuse interface methods 

(Mittal & Iaccarino, 2005). 

3. Geometric Representation 

of Immersed Boundary 

The current method is designed to 

simulate flows over arbitrarily complex 

2D and 3D immersed stationary 

and moving boundaries and the 

approach chosen to represent the 

boundary surface should be flexible 

enough so as not to limit the type 

of geometries that can be handled. 

A number of different approaches 

are available for representing the 

surface of the immersed boundary, 

including level sets (Osher & 

Sethian, 1988; Tran & Udaykumar, 

2004), and unstructured surface 

grids. In the current solver, we 

choose to represent the surface of 

the immersed boundary by an unstructured 

mesh with triangular elements. 

This approach is very well 

suited for the wide variety of engineering 

and biological configurations 

that are of interest to us and 

is compatible with the immersed 

boundary methodology used in 

the current solver. 

4. Boundary Motion 

Boundary motion can be included 

into immersed boundary formulation 

with relative ease. In advancing 

the field equations from time level 

n to n + 1 in the case of a moving 

boundary, the first step is to move 

from its current location to the 

new location. This is accomplished 

by moving the nodes of the surface 

triangles with a known velocity. 

Thus, we employ the following 

equation to update the coordinates 

(X i ) of the surface element vertices, 

X nþ1 

i 

Δt 

X n 

i 

¼ V nþ1 

i 

ð2Þ 

where V i is the vertex velocity. The 

vertex velocity can either be prescribed 

or it can be computed 

from a dynamical equation if the 

body motion is coupled to the 

fluid. The next step is to determine 

the ghost cells for this new immersed 

boundary location and 

recompute interpolation weights 

associated with the ghost point 

methodology. Subsequently, the 

flow equations, which are written 

in Eulerian form, are advanced 

in time. The general framework 

described above can, therefore, be 

considered as Eulerian-Lagrangian, 

wherein the immersed boundaries 

are explicitly tracked as surfaces in 

a Lagrangian mode, while the flow 

computations are performed on a 

fixed Eulerian mesh. Additional 


details regarding the current immersed 

boundary methodology 

may be found in Mittal et al. (2008). 

Computational Setup 

All simulations are conducted in 

a rectangular computational domain. 

The boundary conditions on the 

bounding box of the domain are freestream 

on the left (x direction), outflow 

on the right while the remaining 

boundaries (top and bottom ( y direction) 

and front and back (z direction)) 

employ slip boundary conditions (see 

Figure 4). The fin surface and fish 

body are considered as no-slip boundaries. 

The fins are treated as deforming 

membranes while the body, where applicable, 

is treated as rigid body undergoing 

general motion. The Reynolds 

number in the present work is defined 

as Re = UL s /ν where U, L s ,andν are 

the swimming velocity, spanwise fin 

length, and the kinematic viscosity of 

FIGURE 4 

water (ν =1.007×10 −6 m 2 s −1 at room 

temperature), respectively. 

Based on a swimming speed of 

1.1 body length per second, the 

Reynolds number for the steady swimming 

is 6300. However, a comparison 

of the force coefficients obtained 

at Re = 1440 with those at the experimental 

Reynolds number appear to be 

in good agreement both quantitatively 

and qualitatively (Bozkurttas, 2007). 

So, for computational expediency, we 

use the lower Reynolds number in the 

steady swimming analysis (Dong et al., 

2010). As mentioned earlier, low dimensional 

model performance analyses 

have shown that Mode 1 + 2 + 3 

gait that accounts for 67% of the fin 

motion still produces 92% of the 

thrust (Bozkurttas et al., 2009). Therefore, 

in lieu of the experimentally 

extracted fin kinematics, this simplified 

model has been used here. The 

grid size in these simulations is 153 × 

161 × 97, which is about 2.35 million 

Cartesian grid (4.8 million grid points) and unstructured mesh employed for yaw maneuver: 

(a) x-y plane section, (b) x-z plane section, (c) y-z plane section (strongside fin ontheleft 

and weakside on right of the body), and (d) unstructured surface mesh (pectoral fin only, number 

of nodes = 10,000, number of elements = 19,602). 

grid points. A domain size of 3.8L s × 

4.5L s ×1.8L s is selected where L s is 

the span wise size of the fin. Comprehensive 

studies have been carried out 

to assess the effect of the grid resolution 

and domain size on the salient 

features of the flow and also to demonstrate 

the accuracy of the selected grid 

(Bozkurttas, 2007). 

The Reynolds number for the turning 

maneuver based on a freestream 

velocity of 0.5 body lengths per second 

is approximately 3500. The domain 

size employed for the maneuver is 

7.5L s ×5L s ×5L s . The pectoral fins 

and an idealized body, immersed in 

the computational grid, are shown in 

Figure 4. The nominal grid size used 

in the current simulation is 241 × 

145 × 145 (see Figure 4). Finally, the 

domain size for the maneuver with 

just the strongside (outside) fin is 

4L s ×4L s ×4L s with a non-uniform 

grid using 128 points in all three 

dimensions. 

We note in passing that all the steady 

swimming cases and ablated fin simulations 

(for the maneuver) do not include 

the fish body. This is reasonable 

sincewehaveobservedthatthedifference 

in the thrust coefficients with 

and without the body is minimal. As 

we shall see shortly, the wake dynamics 

for both steady swimming and maneuver 

are dominated by vortex structures 

generated far from the fish body (see 

Figures 5 and 8). Thus, the interaction 

between the body and the fin hydrodynamics 

is minimal. 

The performance of the fin is evaluated 

using the computed force coefficients 

which are defined as, 

C T ¼ 

2F x 

ρU∞ 2A fin 

; C L ¼ 2F y 

ρU∞ 2A ; 

fin 

C Z ¼ 

2F z 

ρU 2 ∞ A fin 

ð3Þ 


FIGURE 5 

The anatomy of the principal vortex dynamics 

involved in steady swimming. 

where F x , F y and F z are the forces respectively 

in the streamwise (drag/ 

thrust), vertical (lift), and spanwise 

(lateral) directions, A fin is the nominal 

fin area,andρ is the density of the 

fluid. U ∞ is the forward swimming 

velocity. The force components are 

calculated by directly integrating the 

computed pressure and shear stress 

on the fin surface. 

Results 

Steady Swimming 

A snapshot of the vortex dynamics 

at the end of a steady swimming fin 

beatisshowninFigure5.Notethe 

FIGURE 6 

complex interaction of among vortices 

generated by the path traversed by the 

fin tip during a fin beat. Clearly, both 

adduction and abduction appear to 

produce distinct vortex structures. 

This is in stark contrast with a simple 

ring vortex created during the maneuver 

(see Figure 8). The time variations 

of the force coefficients (C T , C L and 

C Z ) for three fin planforms are plotted 

in Figure 6. Note the presence of two 

distinct and comparable peaks corresponding 

to the adduction and abduction 

phases. This force signature 

bears the trademark of efficiency where 

the fin sustains net forward thrust 

throughout its fin beat. Clearly, the 

chordwise and spanwise compliance 

of the fin allows the simultaneous formation 

and persistence of two distinct 

vortex structures within a single fin 

beat. A rigid planform would lead to 

a more restrictive envelope for the fin 

tip path resulting in vortex dynamics 

that have stronger interactions detrimental 

to sustained net thrust production 

(see Akhtar et al., 2007). 

We now construct two different 

ablated fin models: one that contains 

only the rays 1-4 and one that contains 

rays 1-8 (see Figure 1). The motion of 

these dissected fins is precisely the 

same as that for the full fin andwe 

carry out flow simulations for both of 

these cases. Examining the results from 

our virtual dissection, we notice that 

the dorsal half of the fin (rays 1-8) captures 

the two main peaks of the thrust 

and preserves 90% of the thrust production 

of the full fin planform. Consequently, 

the ventral contribution of 

the fin, represented by rays 9-14 in 

Figure 1, to the thrust production is 

found to be insignificant. Also, the 

planform interpolated from rays 1-4 

has a similar trend in thrust variation 

during the entire fin-beat cycle albeit 

with smaller amplitudes. Interestingly, 

it has two main peaks and even the two 

local peaks in the abduction phase as in 

the full fin case. This further reinforces 

the notion that the dorsal leading edge 

of the bluegill’s pectoral fin dominates 

the overall performance during steady 

swimming propulsion. This planform 

produces almost 40% of the thrust 

produced by the fish fin while undergoing 

Mode 1 + 2 + 3 gait. Finally, 

we observe similar tendencies for lift 

and spanwise force coefficients for the 

three planforms, except the case with 

just rays 1-4 where the values show attenuation. 

The key observation here is 

that the dorsal half of the pectoral fin 

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) 

streamwise force, (b) vertical force, and (c) lateral force. 


(rays 1-8) is responsible for producing 

a majority of the thrust. These results 

bring into question the need for the 

ventral portion of the fin. We explore 

this in detail as we consider the case of 

the yaw turn maneuver. 

FIGURE 7 

Comparison of time variation of force coefficients: (a) strongside and (b) weakside. 

Yaw Turn Maneuver 

The evolution of wake structure 

from the strongside fin, that drives 

the maneuver, is shown from two vantage 

points: lateral (Figure 8 (a,c,e)) 

anddorsal(Figure8(b,d,f)).The 

well-defined vortex ring formed during 

the outstroke (abduction) produces a 

lateral jet oriented normal to the fish 

body (see Figure 9). This type of vortex 

ring and associated lateral jet shown in 

Figures 8 and 9 have also been observed 

in experimental visualization 

(Drucker & Lauder, 2001). The peak 

lateral velocity is found to be greater 

than three times the freestream velocity. 

Consequently, the lateral forces 

developed are several times that observed 

in forward thrust for the steady 

swimming case (Bozkurttas, 2007). 

Preliminary estimates for stroke-averaged 

force coefficients ratio between 

lateral force in maneuvering ( ― C Z = 

6.1) to steady swimming thrust ( ― C T = 

1.29) is approximately 4 (( ‐ ) denotes 

averageoverstroke).Thisfactorisin 

reasonable agreement with the forces 

measured experimentally (Drucker & 

Lauder, 2001). 

Returning to Figure 7(a), we note 

that the C Z peak is reached between 

t/T = 0.15 and t/T = 0.3. Shortly thereafter, 

the C T peak occurs between 

t/T = 0.3 and t/T = 0.4. As expected, 

the first priority in the maneuver is 

to evade the stimulus (an obstacle or 

predator in the wild) by quickly generating 

a strong lateral force (maximum 

occurs at t/T = 0.2). Thereafter, the 

drag force developed in the streamwise 

direction is likely used to modulate the 

direction of the resultant force as the 

sunfish turns away from the stimulus. 

The evolving vortex ring, clearly seen 

in Figures 8(d) and 8(f ), continues to 

be oriented nearly parallel to the fish 

body. Consequently, the lateral jet orientation 

ensures that the maximum 

lateral force continues to act normal 

to the fish body for the duration of 

the maneuver. Here, the inherent flexibility 

of the pectoral fin structure and 

the ability to continuously alter planform 

area is likely to be very useful. 

Finally, an examination of the force 

histories for the dissected fin reveals 

that the peak lateral thrust developed 

by the dorsal part (rays 1-8) is approximately 

70% of the total as opposed to 

30% for the ventral (rays 8-14) portion 

(see Figure 10 and Figure 11c). The 

streamwise drag is slightly more comparable, 

although the dorsal part peak is 

higher (see Figure 11a). Overall, while 

the dorsal portion contributes to the majority 

of lateral force production, the ratio 

of dorsal to ventral contribution appears 

to be more equitable than the steady 

swimming case. 

Conclusions 

A comparative analysis of the pectoral 

fin performance in steady swimming 

and yaw turn maneuver reveals 

that the dorsal part of the pectoral fin 

is responsible for the majority of force 

production. The chordwise and spanwise 

flexibility of the pectoral fin and 

its ability to have them function either 

in concert or independently seems to 

enable the bluegill sunfish to achieve 

a variety of maneuvers. The virtual 

dissection reveals a significant loss of 

performance with maneuvering with 

respect to peak lateral thrust when 

the ventral portion is removed. Thus, 

a fin design using just the dorsal portion 

of the pectoral fin might perform 

as well as the full fin insteadyswimming 

but will not retain the same 

maneuverability. Hence, any effective 

design based on the pectoral fin that 

aims to preserve its performance over 

all locomotion mode needs to retain 

a greater portion of the fin thanthat 

suggested by steady swimming alone. 

The pectoral fins of fishes display a 

diversity of shapes (e.g., Drucker & 

Lauder, 2002; Thorsen & Westneat, 

2005), and although some general 

conclusions about correlations of fin 

shape with fish ecology have been possible 

(see Wainwright et al., 2002), 

there are very few data on functional 

regionalization of pectoral fins and on 

therolethatdifferentfin rays within 


FIGURE 8 

Formation of the vortex ring due to the strongside pectoral fin motion: (a), (c), and (e) are lateral views at t/T = 0.22, t/T = 0.49, and t/T = 0.66, 

respectively; (b), (d), and (f) are the corresponding dorsal views at t/T = 0.22, t/T = 0.49, and t/T = 0.66, respectively. 


FIGURE 9 

The strongside lateral jet associated with the vortex structures in Figure 8 (c) at t/T = 0.49. 

FIGURE 10 

Formation of the vortex ring due to the strongside pectoral fin motion: (a) full, (b) dorsal, and 

(c) ventral portion of the fin sections. 

the pectoral fin might play in controlling 

locomotor performance. Taft et al. 

(2008) discussed functional regionalization 

during steady swimming in 

sculpin, but the role that different fin 

rays play during maneuvering behaviors 

has not previously been analyzed. 

The results presented here suggest that 

the ventral region of the fin playsan 

important role in modulating maneuvering 

forces, and future studies on the 

diversity of fish pectoral fin shapes 

could focus on the surface area and 

mechanical properties of this region 

of the fin incorrelationwithmaneuvering 

performance. No data are currently 

available that would permit 

even general conclusions about the diversification 

of pectoral fin structure in 

relation to maneuvering capability, 

and this represents a new and very interesting 

direction for future work that 

integrates approaches from biomechanics 

and fluid dynamics with behavioral 

and ecological studies of fish 

locomotion. 


This work was done while the first 

three authors were at The George 

FIGURE 11 

Comparison of forces produced on the dorsal and ventral halves of the strongside fin with respect to the full fin: (a) streamwise force, (b) vertical 

force, and (c) lateral force. 


Washington University, and the work 

wassupportedunderONR-MURI 

grant N00014-03-1-0897 monitored 

by Dr. Thomas McKenna. 

Lead Author: 

Srinivas Ramakrishnan 

ANSYS, Inc. 

10 Cavendish Court, 

Lebanon, NH 03766 

Email: srinivas.ramakrishnan@ 

ansys.com 

References 

Akhtar, I., Mittal, R., Lauder, G.V., & 

Drucker, E.G. 2007. Hydrodynamics of a 

biologically inspired tandem flapping foil 

configuration. Theor Comp Fluid Dyn. 

21:155-70. doi: 10.1007/s00162-007-0045-2. 

Bozkurttas, M. 2007. Hydrodynamic performance 

of fish pectoral fins with application to 

autonomous underwater vehicles. Ph.D. Thesis, 

The George Washington University. p. 9. 

Bozkurttas, M., Mittal, R., Dong, H., 

Lauder, G.V., & Madden, P. 2009. A lowdimensional 

models and performance scaling 

of a highly deformable fish pectoral fin. 


S0022112009007046. 

Chorin, A.J. 1967. A numerical method 

for solving incompressible viscous flow 

problems. J Comput Phys. 2:1-9. 

doi: 10.1016/0021-9991(67)90037-X. 

Dong, H., Bozkurttas, M., Mittal, R., Lauder, 

G.V., & Madden, P. 2010. A Computational 

modelling and analysis of the hydrodynamics 

of a highly deformable fish pectoral fin. 


S0022112009992941. 

Dong, H., Mittal, R., & Najjar, F.M. 2006. 

Wake topology and hydrodynamic performance 

of low-aspect-ratio flapping foils. 


S002211200600190X. 

Drucker, E.G., & Lauder, G.V. 2001. Wake 

dynamics and fluid forces of turning maneuvers 

in sunfish. J Exp Biol. 204:431-42. 

Drucker, E.G., & Lauder, G.V. 2002. Experimental 

hydrodynamics of fish locomotion: 

Functional insights from wake visualization. 

Integr Comp Biol. 42:243-57. doi: 10.1093/ 

icb/42.2.243. 

Lauder, G.V., Madden, P.G.A., Mittal, R., 

Dong, H., & Bozkurttas, M. 2006. Locomotion 

with flexible propulsors: I. Experimental 

analysis of pectoral fin swimming in sunfish. 

Bioinspir Biomim. 1:35-41. doi: 10.1088/ 

1748-3182/1/4/S04. 

Mittal, R., Dong, H., Bozkurttas, M., & 

Lauder, G.V. 2006. Locomotion with flexible 

propulsors: II. Computational modeling of 

pectoral fin swimming in sunfish. Bioinspir 

Biomim. 1:25-34. doi: 10.1088/1748-3182/ 

1/4/S05. 

Mittal, R., Dong, H., Bozkurttas, M., Najjar, 

F.M., Vargas, A., & Von Loebbecke, A. 2008. 

A versatile sharp interface boundary method 

for incompressible flows with complex boundaries. 

J Comput Phys. 227:1-9. doi: 10.1016/ 

j.jcp.2008.01.028. 

Mittal, R., & Iaccarino, G. 2005. Immersed 

boundary methods. Annu Rev Fluid Mech. 

37:239-61. doi: 10.1146/annurev.fluid.37. 

061903.175743. 

Osher, S., & Sethian, J.A. 1988. Fronts 

propagating with curvature-dependent. 


0021-9991(88)90002-2. 

Taft, N., Lauder, G.V., & Madden, P.G. 

2008. Functional regionalization of the 

pectoral fin of the benthic longhorn sculpin 

during station holding and swimming. 

J Zool Lond. 276:159-67. doi: 10.1111/ 

j.1469-7998.2008.00472.x. 

Thorsen, D.H., & Westneat, M. 2005. 

Diversity of pectoral fin structure and function 

in fishes with labriform propulsion. J Morphol. 

263:133-50. doi: 10.1002/jmor.10173. 

Tran, L.B., & Udaykumar, H.S. 2004. A 

particle-level set-based sharp interface cartesian 

grid method for impact, penetration, and 

void collapse. J Comput Phys. 193:469-510. 

doi: 10.1016/j.jcp.2003.07.023. 

Wainwright, P., Bellwood, D.R., & Westneat, 

M. 2002. Ecomorphology of locomotion in 

labrid fishes. Environ Biol Fish. 65:47-62. 

doi: 10.1023/A:1019671131001. 

Ye, T., Mittal, R., Udaykumar, H.S., & 

Shyy, W. 1999. An accurate Cartesian grid 

method for simulation of viscous incompressible 

flows with complex immersed boundaries. 


jcph.1999.6356. 

Zang, Y., Streett, R.L., & Koseff, J.R. 1994. 

A non-staggered fractional step method for 

time-dependent incompressible Navier- 

Stokes equations in curvilinear coordinates. 


jcph.1994.1146. 


PAPER 

Learning From the Fins of Ray-Finned 

Fish for the Propulsors of Unmanned 

Undersea Vehicles 

AUTHORS 



Engineering, Drexel University 

Timo Gericke 


Museum of Comparative Zoology, 


Introduction 

Military and civilian studies have 

identified that two of the most significant 

technological obstacles to deploying 

unmanned undersea vehicles 

(UUVs) are energy and autonomy 

(Nicholson & Healey, 2008; Office 

of the Secretary of Defense, 2009). 

The energy and the rate it is used 

(power) limit the duration and distance 

of operations and bound the 

type of activity that can occur even 

for short periods. Autonomy defines 

the degree to which humans must supervise 

UUV operations and provides 

UUVs with the ability to react to 

external stimuli without human intervention. 

Among the enabling technologies 

that are critical for solving the 

challenges associated with energy and 

autonomyaremoreeffectivepropulsors 

(Office of the Secretary Defense, 

2009). Propulsors are required to provide 

increased maneuverability, stealth, 

and endurance for the widespread range 

of missions envisioned for UUVs, from 

long duration sensing in the open ocean 

to mine countermeasures in very shallow, 

high-energy water. 

ABSTRACT 

Advanced propulsors are required to help unmanned undersea vehicles (UUVs) 

overcome major challenges associated with energy and autonomy. The fins of rayfinned 

fish provide an excellent model from which to develop propulsors that can 

create forces efficiently and drive a wide range of behaviors, from hover to lowspeed 

maneuvers to high-speed travel. Although much is known about the mechanics 

of fins, little is known about the fin’s sensorimotor systems or how fins 

are regulated in response to external disturbances. This information is crucial for 

implementing propulsive and control systems that exploit the same phenomena as 

the biological fins for efficiency, effectiveness, and autonomous regulation. Experiments 

were conducted to evaluate the in vivo response of the sunfish and its pectoral 

fins to vortex perturbations applied directly to the fish and to the fins. The fish 

and the fins responded actively to perturbations that disturbed the motion of the fish 

body. Surprisingly, perturbations that deformed the fins extensively did not cause a 

reaction from either the fins or the body. These results indicate that the response of the 

pectoral fins to large deformations is not reflexive and that fin motions are regulated 

when it is necessary to correct for disturbances to the motion of the fish. The results 

also demonstrate a benefit of compliance in propulsors, in that external perturbations 

can disturb the fins without having its impact be transferred to the fish body. 

Keywords: biorobotics, flapping fins, vortex pertubations, sensory-based control 

Fish are important biological models 

from which to learn methods of 

propulsion that are effective and efficient 

over a wide range of operating 

conditions. Bony fish, such as the 

bluegill sunfish (Lepomis macrochirus) 

and the swordfish (Xiphias gladius), 

are able to hover, swim and maneuver 

at low speeds, manipulate the orientation 

of their bodies, conduct acrobatics 

to escape or to attack prey, and, especially 

for the swordfish, sustain high 

swimming speeds. These behaviors 

can be accomplished in smooth water 

and in high-energy flows and relate 

directly to the behaviors desired for 

UUVs. The remarkable swimming 

abilities of these fish are due, in large 

part, to the fish having multiple, highly 

actuated, flexible fins that are able to 

create and to modulate large-magnitude 

forces. 

A great deal is known about the 

mechanisms that contribute to the 

production of hydrodynamic forces 

by flapping the fins and the fish 

body. Forces are created through the 

dynamic interaction of the fins, the 

body, and the fluid, which results in 

energy being added to, or taken from, 

the fluid. A review of seminal work 

that explains the way in which marine 

animals control vorticity is presented 

in Triantafyllou et al. (2002) and 


Zhu et al. (2002). Numerical and 

experimental studies of flexible fins 

with two-dimensional kinematics 

(heaving and pitching) include, but 

are in no way limited to, studies 

of McHenry (1995), Liu and Bose 

(1997), Prempraneerach et al. (2003), 

Triantafyllou et al. (2005), Fish et al. 

(2006), Lauder et al. (2006), Mittal 

et al. (2006), Lauder and Madden 

(2007), and Zhu and Shoele (2008). 

Recent studies that considered deformable 

fins with complex kinematics 

are presented in, for example, Shoele 

and Zhu (2009), Dong et al. (2010), 

and Tangorra et al. (2010). 

In contrast to our understanding 

of the mechanics of fins and of hydrodynamic 

forces, little is known about 

how fishes sense their interaction 

with the water and use sensory information 

to regulate the fins. Knowledge 

of fin sensorimotor control is critical 

if engineered systems are to take full 

advantage of the mechanisms used by 

fins to create forces efficiently and to 

react to changes in the environment. 

The focus of this paper will be on 

the pectoral fins of sunfish and, in particular, 

on how and when sunfish alter 

the use of the pectoral fins in response 

to external perturbations. We begin 

with an overview of pectoral fin swimming 

in sunfish and briefly present robotic 

fins that produce and modulate 

forces like the biological fins. A series 

of experiments where the biological 

fin is perturbed during steady swimming 

is then presented. These experiments 

address the response of the fins 

in the context of using the fins to control 

and stabilize the fish body. 

Ray-Finned Fish and Robots 

Sunfish Swimming 

The ability of the sunfish to control 

the magnitude and direction of its propulsive 

forces is due to its ability to 

modulate the kinematics, coordination, 

and mechanical properties of 

its fins and muscular tail (Tangorra 

et al., 2010, 2011). Hydrodynamic 

forces are created through an exchange 

of energy between the propulsive surfaces 

and the surrounding fluid. As the 

fish moves through the water, vortices 

develop along the body and fins, the 

propulsive structures bend and store 

energy, and the vortices are shed into 

the flow along with directed jets 

(Triantafyllou et al., 2002; Dong 

et al., 2010). The complex motions 

that cause this exchange of energy are 

the result of driven motions of the 

fin rays and a dynamic interaction of 

the deformable fin surfaces with the 

water. The forces created by fins are, 

therefore, modulated through changes 

to the kinematics of the fin and active 

adjustments of fin’s mechanical properties 

(Lauder et al., 2006; Mittal et al., 

2006; Akhtar et al., 2007; Tangorra 

et al., 2010). The changes may be subtle, 

as in steady swimming where the 

stiffness of the fin rays is gradually increased 

with speed, but where the motions 

of the fins are approximately the 

same. Or the changes may be obvious, 

as when the fish interrupts a cyclic 

swimming pattern and uses a stiff, impulsive 

fin motion to slow the fish and 

turn it away from an obstacle (Gottlieb 

et al., 2010). 

Ray-Finned Robotic Systems 

Robotic fins(Figure1)havebeen 

developed that produce motions, 

forces, and flows like the biological 

fins (Tangorra, Davidson et al., 2007; 

Phelan, Tangorra et al., 2010; Tangorra, 

Lauder et al., 2010). These fins were 

designed originally as physical models 

with which to conduct experimental 

studies that would have been difficult 

to conduct with the living fish 

FIGURE 1 

Biorobotic models of the sunfish pectoral fin 

(A) and caudal fin (B). The pectoral fin is instrumented 

with strain gages along the fin 

rays and pressure sensors along the body 

plate in order to model distributed sensing in 

the sunfish. Modified versions of the robotic 

pectoral and caudal fins,aswellasdorsaland 

anal fins, are implemented on a fish robot (C). 

The fish robot can swim freely or be attached 

to a rigid mast (shown) so that forces can be 

measured. The grooves in the side of the fish 

body are used for pressure lines and ports. 

(Tangorra, Phelan et al., 2011). The 

fins are comprised of fin rays, each 

with multiple actuated degrees of freedom 

(DOF), within a thin, flexible 

webbing. The geometries of the fin 

rays were defined so that the stiffness 

of the robotic fin was proportional to 

that of the biological fin across the 

fin’s chord and span. The architecture 

of the robotic fin provides a great degree 

of control over the fin’s motions 

and mechanical properties, which enables 

the magnitude and direction of 

the force produced by the fin tobe 

easily modulated (Figure 2). Gross 

changes to the profile of the fin’s force 

can be made by changing the fin’s gait 


FIGURE 2 

Thrust (horizontal) and lift (vertical) forces for 

pectoral fins executing normal and modified 

steady swimming gaits. By making relatively 

small changes to fin stiffness (A) and fin motions 

(B), the forces can be moved throughout 

the thrust-lift plane. Normal, full-fin steady 

swimming for three levels of stiffness (A). The 

gait was modified slightly by altering the phase 

angle between fin rays by 30° (B, red) or by 

using just the upper or lower half of the fin 

(B, green). The magnitude of the force can 

also be altered simply by changing the frequency 

of the fin beat. 

pattern, for example, by switching 

from a steady swimming gait to the 

pattern used by the fish for a turn maneuver. 

Smaller changes to the force 

profile can be made by changing the 

frequency of the fin beat and/or by 

changing phase relationships between 

fin rays (Figure 2A). Considerable 

changes to the magnitude and direction 

of the force can also be made by 

adjusting the mechanical properties 

of some, or all, of the fin rays. When 

the mechanical properties of the fin 

rays are under active control, as in 

the fish, changes to the force profile 

can happen very quickly since the 

driven motions of the fin do not have 

to be changed. 

The designs of the robotic fins were 

modified and the fins implemented 

on a freely swimming biorobotic fish 

(Figure 1). Modifications included 

placing actuators within the fish body 

adjacent to each fin, using a network of 

microcontrollers to drive fin motions, 

and minimizing the number of actuated 

DOF for each fin ray.Themotions 

and orientation of the robotic 

fish are controlled by adjusting the 

propulsive forces created by five rayfinned 

fins. In this first implementation, 

the forces are modulated by 

switching between several fin gaits 

and by making predetermined changes 

to fin beat frequencies and to the phase 

relationships between fin rays. 

Sensory-Based Control of Fins 

What is clearly missing in this robotic 

system is the ability to automatically 

modulate the kinematics and 

mechanical properties of the fins 

based on sensory information about 

the fins and their interaction with the 

water. The motions of the fins are adjusted 

based on the forces required to 

control the robot’s body, but sensory 

information is not being used to exploit 

the phenomena that are critical 

to the efficient production of force 

(e.g., vorticity) nor to adjust behaviors 

in response to changes in the flow (e.g., 

speed and turbulence). This is due 

to the fact that very little is known 

about the sensory-based control of 

ray-finned fins (Phelan et al., 2010). 

The fine level of control that the sunfish 

has over fin motions and mechanical 

properties suggests strongly that 

there is closed-loop control of the fins. 

However, fundamental questions 

about the existence of sensory systems 

intrinsic to fins, about the types of 

stimuli that elicit responses from fins, 

about information in the flow that 

is relevant to propulsive forces, and 

about the behavior of the fins in response 

to external perturbations have 

not yet been answered. This knowledge 

is vital for the development of 

fin-based propulsors that take advantage 

of the phenomena used by fish 

to produce forces efficiently and that 

automatically adjust their behavior in 

response to disturbances and changes 

in operating requirements. 

Experimental Methods 

and Equipment 

Experimentation 

Experiments were conducted to 

evaluate the response of the sunfish’s 

pectoral fins to external perturbations 

applied to the fin andtothefish’s body 

during steady swimming. Perturbations 

were created using a vortex generator 

(Figure 3), which produces a 

vortex ring that moves through the 

waterandimpartsashortduration 

impulse to the fish (Figure 4). The 

strength of the vortex was sufficient 

to deform the pectoral fin ortodisplace 

the fish laterally by several millimeters. 

The vortex is not visible, so 

it does not elicit a visually mediated response 

from the fish. The vortex does, 

however, produce a pressure wave that 

may be sensed by the fish. 

Two bluegill sunfish, with body 

lengthsof160±10mmandintact 

pectoral fins, were used for the experiments. 

For the experimental trials, 

FIGURE 3 

The vortex ring generator and vortex (left). 

Blue dye was added to the vortex generator’s 

cavity to make the vortex visible to the naked 

eye. The vortex generator comprises an orifice 

plate (1), two cavity plates (2), a latex membrane 

(3), and a connector plate (4), which 

enables the air-line to be connected to the vortex 

generator. 


FIGURE 4 

Sunfish in flow tank with vortex generator (A). The sunfish kindly positioned itself in the center 

of the test area and laser sheet (B). The laser sheet is used with PIV to characterize the vortex as 

it travels toward the fish. 

asunfish was placed in the working 

area (280 × 280 × 800 mm) of a 

600-l flow tank and was allowed to acclimate 

for 2 h. The flow rate was set to 

100 mm s −1 , which equates to a steady 

swimming speed of approximately 

0.6 body lengths s −1 .Atthisspeed, 

sunfish generate swimming forces 

using primarily their pectoral fins. The 

tail and the caudal, anal, dorsal, and 

paired pelvic fins are moved very little 

but are important for stability. The 

vortex generator was positioned approximately 

150 mm above the tank 

floor and placed either perpendicular 

to the fish in order to perturb the fish’s 

body or at a 45° angle to the fish in 

order to perturb the pectoral fin during 

its outstroke. A horizontal light sheet 

(Figure 4) used for particle image 

velocimetry (PIV) was positioned so 

that it had the same height as the middle 

of the vortex generator. The fish 

was directed into the middle of the 

test area and light sheet by coaxing it 

with a wooden dowel. Once the fish 

was positioned properly, the vortex 

was launched to strike the fish. Vortices 

impacted the fish (1) on the body 

near the tip of the left pectoral fin 

while the fin rested against the body 

during the pause between fin beats 

and (2) at the tip of the left pectoral 

fin as the fin completed its outstroke. 

High-speed (500 fps), highdefinition 

video (1024 × 1024 pixels) 

was used to capture the motions of 

the fish and of the fish’s fins. Two 

cameras (Photron 1024 PCI, Photron 

USA, Inc., San Diego, CA) were synchronized 

and positioned so that the 

ventral and posterior views of the fish 

were captured. 

Analysis 

The linear and rotational velocities 

of the vortices were analyzed using 

DaVis (LaVision GmbH, Göttingen, 

Germany). 

The motions of the fish and of the 

fins were analyzed for two fin beats before 

and two beats after the impact of 

the vortex. The coordinates of eight 

points along the fish body and pectoral 

were digitized using Matlab (The 

Mathworks Inc., Natick, MA) and 

tracked through time. Deformations 

and curvatures were calculated for 

the pectoral fin during the impact of 

thevortexring.Threepointsalongthe 

fin were selected to characterize the 

shape of the fin andtodefine the radius 

of curvature. 

Design of the Vortex 

Ring Generator 

Vortex rings are commonly generated 

using a piston that moves within 

a cylindrical cavity and pushes a volume 

of fluid (the slug) out of the cavity 

and past an orifice with sharp 

edges. The movement of the piston 

causes the boundary layer that develops 

in the cavity to separate at the orifice’s 

edge and to roll up into a vortex 

ring that has a toroidal shape. The 

speed of the piston, the diameter of 

the orifice pate, and the ratio of cavity 

length to cavity diameter influence the 

formation of the vortex and the speed 

at which the vortex travels. Excellent 

discussions of vortex generation are 

presented in Gharib et al. (1998), 

Allen and Auvity (2002), Shusser 

et al. (2002), and Mohseni (2006). 

The vortex generator that was developed 

for our experiments is similar 

to a piston based vortex generator, 

but the design was modified so that it 

would be more appropriate for the 

testing of swimming fish. Two requirements 

that influenced the design were 

(1) the vortex generator had to be silent, 

so that the fish did not hear a 

mechanism and anticipate the arrival 

of the vortex, and (2) the system had 

to be small, so that it could be placed 

at the side of the flow tank without interfering 

with the swimming fish. The 

vortex generator consists of two acrylic 

plates(45×55×12mm)inwhich 

a cylindrical cavity is cut (Figure 3). 

The plates are covered by a 0.3-mm 

thick aluminum plate with either 

a 4.0- or 7.5-mm diameter orifice. A 

latex membrane is sandwiched between 

the cavity plates and another 

acrylic plate in which a cylindrical 

well is cut. This plate is connected via 

a 6-mm diameter air line (Polyurethane 

Tubing, NewWay Air Bearings, Aston, 

PA) to a 60-ml syringe (Becton 

Dickinson and Company, Franklin 

Lakes, NJ). A fast push on the syringe 

plunger causes the latex membrane to 

expand into the cavity and to exhaust 

the fluid and create the vortex. The 


effective length of the cavity can be increased 

by drawing the syringe plunger 

back. This draws the latex membrane 

back into the cylindrical well. Dye 

was introduced into the chamber via 

a1.6-mmdiameterholedrilledinto 

the acrylic plate, radial to the cavity. A 

steel tube was inserted into the hole, 

and was connected via medical tubing 

(Scientific Commodities, Inc., Lake 

Havasu City, AZ) to a syringe filled 

with food-grade dye. The vortex generator 

was mounted to an aluminum 

arm (80/20 Inc., Columbia City, IN) 

so that it could be positioned within 

the flow tank. 

The force, impulse, and linear velocity 

of 12 vortices were characterized 

to better understand the properties 

of the vortex and how best to actuate 

the plunger. The force generated by 

the impact of the vortex was measured 

(Figure 5b) by shooting the vortex 

against a plate that was connected to 

a2.5g force transducer (LSB200, JR 

S-Beam Load Cell, Irvine, CA). The 

plate was located 100 mm from the orifice 

of the vortex generator. The vortex 

was imaged using the high-speed 

camera as it travelled within the 2-mm 

thick light sheet. Mean values for vortices 

created using a 5-mm diameter 

cavity were: 13 mN force (0.6 mN SE), 

0.13 mNs impulse (0.002 mNs SE), 

and 0.99 m/s velocity (0.01 m/s SE). 

A 13-mm diameter cavity produced 

a more powerful but slower vortex: 

67 mN force (2.3 mN SE), 1.0 mNs 

impulse (0.02 mNs SE), and 0.85 m/s 

velocity (0.01 m/s SE). These values 

compare well with estimates we have 

made for the peak force and impulse 

created by a sunfish pectoral fins. At 

a swimming speed of 0.5 body length 

per second, average fin forces are less 

than approximately 10 mN and the 

impulse over the fin beat is less than 

2.5 mNs. 

FIGURE 5 

Evaluation of vortex ring’s velocity using PIV (A). Force from vortex ring during impact with rigid 

plate attached to force transducer (B). 

Response to 

Vortex Perturbations 

Perturbation experiments that 

involved hitting the swimming fish 

with a vortex ring showed that the 

fish did not alter the pectoral fin beat 

during the time course of a single fin 

stroke but did change the amplitude 

and timing of the pectoral fin beats 

subsequent to a vortex impact that 

perturbed the fish’s position. 

Response to Vortex Perturbations 

Applied to the Body 

Vortex perturbations that impacted 

the side of the fish displaced the fish 

FIGURE 6 

laterally by several millimeters (Figure 

6), which is significant relative to 

the thickness of the fish’s body (maximum 

of approximately 25 mm). The 

lateral displacement occurred whether 

the fish had been drifting toward or 

away from the vortex generator prior 

to the disturbance and was not accompanied 

by any obvious change to 

the roll or yaw of the fish. An active 

response of the fish to the vortex perturbation 

was evident in the fishes’ 

motion after a short delay. The soonest 

the active response occurred was 

0.05 s, while the longest delay before 

a response was evident was 0.20 s. In 

the majority of trials, the fishes actively 

Distance from orifice plate of the left pectoral fin (purple), the right pectoral fin (blue), and the 

fish at a point between the pelvic fins (green). The distance between the fish and the orifice plate 

is amplified relative to the fins and is measured at the scale on the right. In this trial, the fish was 

moving towards the vortex generator and was hit by the vortex at about t = 1.49 (red). The active 

response of the fish occurred by t = 1.50 (gray). (Color versions of figures available online at: 

http://www.ingentaconnect.com/content/mts/mtsj/2011/00000045/00000004.) 


moved away from the vortex generator 

after being hit by the vortex (Figure 7). 

The movement was not particularly 

quick, but was always faster than the 

fish’s lateral velocity before the perturbation 

had occurred. In some cases (e.g., 

Figure 6), the fish actively moved toward 

the vortex generator after being 

pushed away from the vortex generator 

by the impulse. This occurred only 

when the fish had been drifting towards 

the vortex generator before the perturbation. 

In some trials, the fish was startled 

by the vortex and swam out of the test 

area. The startled motions were not 

analyzed quantitatively. 

The motions of the pectoral fins 

during the fin beat subsequent to the 

perturbation were significantly differentfromthemotionsofthepectoral 

fins prior to the perturbation. However, 

the pectoral fins did not seem to 

react quickly to the stimulus. In fact, 

the initial movement of the fish’s body 

in response to the vortex generally 

occurred between pectoral fin beats, 

while the pectoral fins were against 

the fish body (Figures 6 and 7). 

Thus, the active motion of the fish 

was initiated by other fins, which reacted 

within as little as 0.05 s. Active 

FIGURE 7 

Distance from orifice plate of the left pectoral fin (purple), the right pectoral fin (blue), and the 

fish at a point between the pelvic fins (green). The distance between the fish and the orifice plate 

is amplified relative to the fins and is measured at the scale on the right. In this trial, the fish was 

moving away from the vortex generator and was hit by the vortex at about t = 1.25 (red). The 

active response of the fish occurred by t = 1.35 (gray). The fish continued to move away from 

the vortex generator until approximate t = 1.8 s. 

movement of the pectoral fins did 

not usually resume until 0.10-0.20 s 

after the vortex. The frequency of the 

pectoral fin beats did not change 

consistently after the perturbation. In 

three of the eight trials, the frequency 

of the pectoral fin beat increased from, 

on average, 1.37 Hz (SD = 0.15) to 

1.83 Hz (SD = 0.28). In the other 

five trials, the frequency of the fin beat 

decreased from, on average, 1.69 Hz 

(SD = 0.26) to 1.34 Hz (SD = 0.21). 

The amplitude of the pectoral fin motions 

also changed. This altered the 

force balance between the two pectoral 

fins and contributed to the movement 

of the fish body. In the beat after the 

vortex stimulus, the amplitude of the 

right pectoral fin (opposite the side of 

the impact) was consistently smaller 

than before the vortex. Its motion decreased 

in all eight trials, on average by 

18.9% (SD = 10.9%). The amplitude 

of the left pectoral fin alsochanged, 

but the changes were less consistent. 

In four trials, the amplitude decreased 

by, on average, 37.4% (SD = 22.8), 

while in the other four trials, the 

amplitude increased by, on average, 

8.9% (SD 7.0%). By the second fin 

beat after the perturbation, the motions 

of the left and right pectoral 

fins were much more similar to the 

motions before the fin beat, and were 

similar to each other. 

Response to Vortex Perturbations 

Applied to the Fin 

The pectoral fins were deformed 

significantly when struck by the vortex 

during the fin beat (Figures 8 and 9). 

The vortex made contact with the left 

pectoral fin near the end of the fin’s 

outstroke. The vortex bent the tips 

of the fin rays and progressively bent 

larger portions of the fin as the vortex 

travelled towards the fish body. The fin 

seemed to bend and fold as if it were 

made from thin paper and exhibited 

deformations from the tip to the 

base. The maximum measured curvature 

of the fin (alongfin ray6)increased 

from 0.054 mm −1 near the tip 

and 0.024 mm −1 near the base during 

unperturbed swimming to 0.113 mm −1 

near the tip and 0.029 mm −1 near 

the base when in contact with the vortex. 

The vortex remained in contact 

with the fin while it travelled towards 

the fish body. This resulted in the 

pectoral fin being pushed back to the 

body faster than during an unperturbed 

instroke. Times ranged from one third 

to one half of the duration of a normal 

instroke and were dependent on many 

FIGURE 8 

Pectoral fin perturbed by vortex during swimming. 

The mean fin ray curvature after impact 

was 14.4 mm −1 (0.05 mm SE). Reflective 

particles are used so that the fluid movement 

is visible. 


FIGURE 9 

Ventral view of the fish as the left pectoral fin is hit by a vortex (no dye). The left pectoral fin 

is hit by the vortex (1). The fin is deformed (2, 3, and 4) and is pushed to the body by the vortex. 

(5 and 6) The right pectoral fin continues to beat normally. The body is not deflected by the vortex. 

variables, including the speed of the 

vortex, how well contact was made 

with the fin, and the time of impact 

within the fin beat. 

Despite the severity with which the 

vortex changed the shape and trajectory 

of the perturbed fin, the fish did 

not appear to react to the perturbation 

or to change its behavior subsequent to 

the perturbation. During the perturbation, 

the observed motions of the 

unperturbed fin andofthefish body 

were not visibly different from motions 

prior to the perturbation. Subsequent 

to the perturbation, the perturbed fin 

remained against the fish body until 

the unperturbed fin completed its 

instroke. Both fins then resumed 

what appeared to be a normal fin 

beat. Small differences in the pectoral 

fin beat and the use of other fins likely 

occurred to accommodate for differences 

in propulsive forces produced 

during the perturbation, but these 

changes were not visible. Nor were 

there changes in the motion of the 

fish body, which was not observed to 

move laterally or to rotate in yaw. 

Discussion 

Theobjectiveoftheexperiments 

was to determine how sunfish respond 

to perturbations applied to the body 

and fins during steady swimming. 

These experiments provided a contextual 

understanding of sensory based 

modulation of pectoral fin function. 

The experiments produced a mix of 

expected and surprising results. 

As expected, the fishes did alter the 

amplitude and timing of the pectoral 

fin beats subsequent to a perturbation 

that disturbed the lateral position of 

the fish body. However, the pectoral 

fins did not respond quickly to the 

disturbance, but remained against the 

fish body for durations that were only 

slightly different from the pauses 

between fin beats prior to the disturbance. 

Active movement of the fish’s 

body after the disturbance occurred 

with a latency of as little as 0.05 s, 

which is similar to the 0.08 s latency 

measured by Webb (2004) in response 

to roll disturbances. The movement of 

the fish body is believed to have been 

caused by fins other than the pectoral 

fins, since the pectoral fins remained 

against the body for 0.10–0.20 s after 

the perturbation. When the pectoral 

fins were moved, the amplitudes of 

the fins seemed to have been adjusted 

to help equilibrate the movement of 

the fish. By the second fin beatafter 

the disturbance, the motions of the 

two pectoral fins were synchronized 

and had amplitudes similar to those 

before the disturbance. 

The delay in the response of the 

pectoral fins to the vortex and lateral 

disturbance is different from the response 

of the fins during experiments 

where an obstacle was placed in front 

of the swimming fish (Gottlieb et al., 

2010). In those experiments, sunfish 

altered the motions of the left 

and right pectoral fins during the outstroke 

of a steady swimming beat. 

The changes were not subtle, and the 

fish did not seem to wait for the next 

cycle as in the present studies. The pectoral 

fin onthesideoftheobstacle 

stiffened and the fin rays were moved 

through trajectories that were very different 

from steady swimming. The fin 

on the side opposite to the obstacle 

nearly stopped and served to stabilize 

the motion of the fish. The difference 

in the pectoral fins’ response to the 

obstacle and to the vortex and lateral 

displacement may be related to the 

fish’s perception of the stimuli. The 

obstacle may have been more threatening 

than the vortex, which the fish may 

have interpreted as a common fluidic 

event. The fish therefore disrupted 

the steady swimming gait in order to 

produce large lateral forces that turned 

the fish away from an unknown obstacle 

that may have posed a threat. 

In contrast, the disturbance in motions 

caused by a fluidic stimulus could be 

accommodated simply by adjusting 

motions of the fins within their normal 

gaits. This would allow the central 

pattern generator that drives the motions 

of pectoral fins (Westneat et al., 

2004) to continue to produce similar 

output characteristic rather than having 

to switch between gaits. 

Most surprising was the lack of reaction 

to the vortex when the vortex 

deformed the pectoral fin attheend 


of the fin’s outstroke and throughout 

the instroke. Nerves and free nerve 

endings exist throughout the fin rays 

and the fin webbing (experimental 

findings, M. Hale, University of 

Chicago), and so it was expected that 

at least one of the phenomena that 

the vortex created—pressure, impact, 

bending—would have elicited a sensory 

mediated response. The vortex 

was in contact with the fin for over 

100ms,andsothedurationofthe 

stimulus was certainly sufficient for a 

sensory-mediated response to occur. 

It was also surprising that neither the 

motions of the body, nor subsequent 

beats of the pectoral fins, were clearly 

different from those before the vortex 

perturbation. It is highly likely that the 

left pectoral fin, while being deformed, 

produced forces that were different 

from normal. During a normal steady 

swimming gait, each pectoral fin will 

produce lateral forces that are similar 

in magnitude to thrust and lift. Since 

the fins typically beat synchronously, 

the lateral forces from the left and 

right fins balance and cancel. This 

would not have been the case when 

the left pectoral fin was deformed, 

and the unbalanced forces should 

have accelerated the fish body laterally 

and/or in roll and yaw. The lack of obvious 

lateral motion and adjustment to 

the pectoral fin beat may be due simply 

to the fish being insensitive to lateral 

forces. To move the fish laterally, 

forces must accelerate the mass of the 

fish and also overcome drag forces 

and the load from the mass of water 

against which the side of the fish 

pushes. Thus, the loss of lateral force 

during a single fin beat can be easily 

tolerated because it is difficult for the 

fish to move sideways. So although 

studies of biorobotic models of the 

pectoral fins have shown that the 

fin’s kinematics and mechanical properties 

must be controlled very carefully 

to produce forces like the fish 

(Tangorra et al., 2007, 2010), the 

mechanics of the fish body do not 

necessarily require the careful control 

of forces at all times in all directions. 

Conclusions 

Perturbation experiments which 

involved hitting the swimming sunfish 

with a vortex ring showed that the 

fish did not alter the pectoral fin beat 

during the time course of a single fin 

stroke but did change the amplitude 

and timing of its motions in beats subsequent 

to an impact that disturbed 

the fish’s position. Vortices that struck 

the pectoral fin during the fin’s outstroke 

deformed the fin extensively, 

but the perturbations did not cause 

the stroke of the unaffected pectoral 

fin to change, nor did the perturbation 

cause changes in the motions of the 

fish body or in the subsequent strokes 

of either pectoral fin. 

These outcomes suggest that the 

kinematics of the pectoral fins is modulated 

by sensory information only 

when a perturbation results in a disturbance 

to the fish body, which is the system 

that the fins are working to control. 

The pectoral fins did not react quickly 

when the vortex displaced the fish’s 

body but modulated their motions to 

help stabilize the displaced fish after 

other fins had already been engaged. 

The pectoral fins also did not react reflexively 

to vortex perturbations that 

deformed the fins’ webbing and fin 

rays. The fin did not appear to move 

away from the vortex or to resist the deformation 

by stiffening. The compliant 

fin allowed itself to bend and perhaps to 

shed the load from the vortex, and then 

altered its motions during the course of 

the subsequent fin beat. 

Theresultsillustrateabenefit of 

compliant mechanisms within a highly 

controllable system. The fins of rayfinned 

fish have the ability to control 

forces precisely by altering the kinematics 

and mechanical properties of 

individual fin rays. Small changes in 

either the trajectories or stiffness of 

fin rays can significantly alter the 

force that is transferred to the fish 

(Tangorra et al., 2010). However, it 

is not always necessary to regulate the 

fins precisely. By maintaining its flexibility 

and allowing itself to be deformed, 

the fin was able to be hit by 

the vortex—which had sufficient 

forces to displace the fish—without 

transferring the full impact of the perturbation 

to the fish body. Therefore, 

the fin’s passive mechanics made it unnecessary 

for the fins to be modulated 

in order to restore the fish to equilibrium. 

However, when the fish does want to 

move the quickly—as in a maneuver 

away from the obstacle—the pectoral 

fin can be stiffened, the gait changed, 

and large lateral forces from the fin 

can be transferred to the fish body. 


This work was supported by the Office 

of Naval Research grant N00014- 

0910352 on fin neuromechanics monitored 

by Dr. Thomas McKenna and 

by the National Science Foundation 

EFRI 0938043. We are very grateful 

to the members of the Lauder and the 

Tangorra laboratories for many helpful 

discussions about fish fins and robotic 

fins and for assistance in executing 

experiments and analyzing results. 

Lead Author: 

James Tangorra 

3141 Chestnut St., 

Randell 115 

Drexel University, 

Philadelphia, PA 19104 

Email: tangorra@coe.drexel.edu 


References 

Akhtar, I., Mittal, R., Lauder, G., & Drucker, 

E. 2007. Hydrodynamics of a biologically 

inspired tandem flapping foil configuration. 

Theoret Comput Fluid Dynamics. 21(3): 

155-70. doi: 10.1007/s00162-007-0045-2. 

Allen, J., & Auvity, B. 2002. Interaction of a 

vortex ring with a piston vortex. J Fluid Mech. 

465(1):353-78. 

Dong, H., Bozkurttas, M., Mittal, R., 

Madden, P., & Lauder, G. 2010. Computational 

modelling and analysis of the hydrodynamics 

of a highly deformable fish pectoral 

fin. J Fluid Mech. 645:345-73. doi: 10.1017/ 

S0022112009992941. 

Fish, F.E., Nusbaum, M.K., Beneski, J., & 

Ketten, D. 2006. Passive cambering and 

flexible propulsors: cetacean flukes. Bioinspir 

Biomim. 1(4):S42-48. doi: 10.1088/ 

1748-3182/1/4/S06. 

Gharib, M., Rambod, E., & Sharriff, K. 

1998. A universal time scale for vortex ring 

formation. J Fluid Mech. 360(1):121-40. 

doi: 10.1017/S0022112097008410. 

Gottlieb, J., Tangorra, J., Esposito, C., & 

Lauder, G. 2010. A biologically derived 

pectoral fin for yaw turn manoeuvres. 

Appl Bionics Biomech. 7(1):41-55. 

doi: 10.1080/11762320903537782. 

Lauder, G.V., & Madden, P.G. 2007. Fish 

locomotion: kinematics and hydrodynamics 

of flexible foil-like fins. Exp Fluids. 43(5): 

641-53. doi: 10.1007/s00348-007-0357-4. 

Lauder, G.V., Madden, P.G., Mittal, R., 

Dong, H., & Bozkurttas, M. 2006. Locomotion 

with flexible propulsors: I. Experimental 

analysis of pectoral fin swimming in sunfish. 

Bioinspir Biomim. 1:S25. doi: 10.1088/ 

1748-3182/1/4/S04. 

Liu, P., & Bose, N. 1997. Propulsive performance 

from oscillating propulsors with 

spanwise flexibility. P Roy Soc Lond A Mat. 

453(1963):1763. 

McHenry, M. 1995. Mechanical control of 

swimming speed: stiffness and axial wave 

form in undulating fish models. J Exp Biol. 

198(11):2293-305. 

Mittal, R., Dong, H., Bozkurttas, M., Lauder, 

G., & Madden, P.G. 2006. Locomotion 

with flexible propulsors: II. Computational 

modeling of pectoral fin swimming in 

sunfish. Bioinspir Biomim. 1(4):S35-41. 

doi: 10.1088/1748-3182/1/4/S05. 

Mohseni, K. 2006. Pulsatile vortex generators 

for low-speed maneuvering of small underwater 

vehicles. Ocean Eng. 33(16):2209-23. 

doi: 10.1016/j.oceaneng.2005.10.022. 

Nicholson, J. W., & Healey, A.J. 2008. The 

present state of autonomous underwater 

vehicle (AUV) applications and technologies. 

Mar Technol Soc. 42(1):8. 

Office of the Secretary of Defense, USA.2009. 

Unmanned Systems Roadmap (2009-2034). 

Phelan, C.T., Tangorra, J.L., Lauder, G.V., 

& Hale, M.E. 2010. A biorobotic model of 

the sunfish pectoral fin for investigations of fin 

sensorimotor control. Bioinspir Biomim. 5. 

doi: 10.1088/1748-3182/5/3/035003. 

Prempraneerach, P., Hover, F.S., & 

Triantafyllou, M.S. 2003. The effect of 

chordwise flexibility on the thrust and 

efficiency of a flapping foil. Autonomous 

Undersea Systems Institute. In: 13th Int. 

Symp. Unmanned Untethered Submersible 

Techn. Durham, NH. 

Shoele, K., & Zhu, Q. 2009. Fluid–structure 

interactions of skeleton-reinforced fins: performance 

analysis of a paired fin in lift-based 

propulsion. J Exp Biol. 212(16):2679. 

doi: 10.1242/jeb.030023. 

Shusser, M., Gharib, M., Rosenfeld, M., & 

Mohseni, K. 2002. On the effect of pipe 

boundary layer growth on the formation of 

a laminar vortex ring generated by a piston/ 

cylinder arrangement. Theoret Comput Fluid 

Dynamics. 15(5):303-16. doi: 10.1007/ 

s001620100051. 

Tangorra, J., Lauder, G.V., Hunter, I.W., 

Mittal, R., Madden, P.G., & Bozkurttas, M. 

2010. The effect of fin ray flexural rigidity on 

the propulsive forces generated by a biorbotic 

fish pectoral fin. J Exp Biol. 213:4043-54. 

doi: 10.1242/jeb.048017. 

Tangorra, J., Phelan, C.T., Esposito, C.J., & 

Lauder, G.V. 2011. Use of biorobotic models 

of highly deformable fins for studying the 

mechanics and control of fin forces in fishes. 

Integr Comp Biol. in press. doi: 10.1093/ 

icb/icr036. 

Tangorra, J.L., Davidson, S.N., Hunter, I.W., 

Madden, P.G., Lauder, G.V., Dong, H., … 

Mittal, R. 2007. The development of a biologically 

inspired propulsor for unmanned 

underwater vehicles. IEEE J Oceanic Eng. 

32(3):533-50. doi: 10.1109/JOE.2007.903362. 

Triantafyllou, M., Techet, A., Zhu, Q., 

Beal, D., Hover, F., & Yue, D. 2002. 

Vorticity control in fish-like propulsion and 

maneuvering. Integr Comp Biol. 42(5):1026. 

doi: 10.1093/icb/42.5.1026. 

Triantafyllou, M.S., Hover, F.S., Techet, A., 

& Yue, D. 2005. Review of hydrodynamic 

scaling laws in aquatic locomotion and fishlike 

swimming. Appl Mech Rev. 58(4):226-37. 

doi: 10.1115/1.1943433. 

Webb, P. 2004. Response latencies to postural 

disturbances in three species of teleostean 

fishes. J Exp Biol. 207(6):955-61. doi: 10.1242/ 

jeb.00854. 

Westneat, M., Thorsen, D., Walker, J., & 

Hale, M. 2004. Structure, function, and 

neural control of pectoral fins in fishes. IEEE J 

Oceanic Eng. 29(3):674-83. doi: 10.1109/ 

JOE.2004.833207. 

Zhu, Q., & Shoele, K. 2008. Propulsion 

performance of a skeleton-strengthened fin. 

J Exp Biol. 211(13):2087. doi: 10.1242/ 

jeb.016279. 

Zhu, Q., Wolfgang, M.J., Yue, D.K.P., & 

Triantafyllou, M.S. 2002. Three-dimensional 

flow structures and vorticity control in fishlike 

swimming. J Fluid Mech. 468:1-28. 

doi: 10.1017/S002211200200143X. 


PAPER 

Bioinspired Design Process for an Underwater 

Flying and Hovering Vehicle 

AUTHORS 

Jason D. Geder 

John S. Palmisano 

Ravi Ramamurti 

Marius Pruessner 

Banahalli Ratna 

Naval Research Laboratory 

William C. Sandberg 

Science Applications 

International Corporation 

Background 

Biologists and zoologists have 

been studying fish swimming for 

many decades, and several comprehensive 

texts and papers exist (Breder, 

1926; Lindsey, 1978; Alexander, 

1983; Azuma, 1992; Blake, 1983; 

Webb, 1975, 1984; Videler, 1993; 

Sfakiotakis et al., 1999). The experimental 

studies of fish swimming biodynamics 

have become increasingly 

quantitative as measurement technology 

has improved. The use of highspeed 

photography shed light upon 

the details of fin deformation (Gibb 

et al., 1994; Walker & Westneat, 

1997). Laser light scattering techniques 

enabled observations of not 

only the fish body and fin dynamics 

but also the velocity field about 

the fish and in the wake (Drucker & 

Lauder, 1999, 2002; Gharib et al., 

2002; Bartol et al., 2003). Vehicle designers 

are able to draw upon such rich 

data sets as they embark upon designs. 

But where does one begin 

A bioinspired vehicle design should 

begin, according to Webb (2004), by 

specifying the performance goals of 

ABSTRACT 

We review here the results obtained during the past several years in a series of 

computational and experimental investigations aimed at understanding the origin of 

high-force production in the flapping wings of insects and the flapping and deforming 

fins of fish and the incorporation of that information into bioinspired vehicle 

designs. We summarize the results obtained on pectoral fin force production, flapping 

and deforming fin design, and the emulation of fish pectoral fin swimming in 

unmanned vehicles. In particular, we discuss the main results from the computational 

investigations of pectoral fin force production for a particular coral reef fish, 

the bird wrasse (Gomphosus varius), whose impressive underwater flight and hovering 

performance matches our vehicle mission requirements. We describe the 

tradeoffs made between performance and produceability during the bio-inspired 

design of an actively controlled curvature pectoral fin and the incorporation of it 

into two underwater flight vehicles: a two-fin swimming version and four-fin swimming 

version. We describe the unique computational approach taken throughout 

the fin and vehicle design process for relating fin deformation time-histories to 

specified desired vehicle dynamic behaviors. We describe the development of the 

vehicle controller, including hardware implementation, using actuation of the multiple 

deforming flapping fins as the only means of propulsion and control. Finally, 

we review the comparisons made to date between four-fin vehicle experimental 

trajectory measurements and controller simulation predictions and discuss the 

incorporation of those comparisons into the controller design. 

Keywords: bio-inspired robotics, pectoral fin, unmanned systems, computational 

fluid dynamics 

the desired mission first and then examining 

those living creatures whose 

performance is relevant. These creatures 

have evolved to meet all their 

needs, and the maneuvering enabled 

by these needs may intersect with 

the performance requirements driving 

a vehicle design. However, since the 

living creatures selected for study are 

most likely not optimized for the mobility 

characteristics that are driving 

the design, one should not copy nature 

but instead be guided by it. 

This point of caution to designers 

has been made many times by biologists 

(Combes & Daniel, 2001; 

Wainwright et al., 2002; Collar et al., 

2008). 

The mission selected for the Naval 

Research Laboratory (NRL) swimming 

vehicle, which we describe 

below, requires precise low-speed maneuvering 

and excellent hovering in a 

complex near-shore environment in 

addition to excellent position-keeping 

in tidal currents. Cost and mechanical 

simplicity constraints demanded a 

rigid hull. This eliminated undulatory 

swimming fish as a primary means of 

design inspiration, and instead we 

were lead to consider fin-based swimming 

creatures. Kato and Furushima 


(1996) had already proceeded down 

the path of paired fin swimming, drawing 

inspiration from the black bass to 

design rigid pectoral fins and incorporate 

them into a test-bed vehicle. 

He subsequently pursued low-speed 

maneuvering using rigid pectoral fins 

(Kato, 2000) and then went on to develop 

a passive flexible fin and an active 

pneumatic actuator pectoral fin (Kato 

et al., 2008). Barrett and Triantafyllo 

(1995) on the other hand had taken 

their inspiration from the undulating 

body and oscillating caudal fin ofthe 

tuna to design and build their flexible 

“Robotuna.” We looked for a fish 

which possessed the dynamic performance 

characteristics we needed and 

for which experimental measurements 

of swimming dynamics, including fin 

kinematics, already existed. Experimental 

observations of pectoral fin 

muscle activity, kinematics, and dynamics 

in coral reef wrasses (Westneat 

& Walker, 1997; Walker & Westneat, 

1997) have shown that the body is 

essentially held rigid during straightline 

motion, thus satisfying our rigid 

hull constraint. Very rapid (∼10 body 

lengths per second) translational 

motions were observed (Walker & 

Westneat, 2000) to give comparable 

swimming performance to that seen 

in body-caudal fin swimmers of comparable 

size even though there was no 

contribution from body undulation 

or caudal fin oscillation in the wrasses. 

This solution offered the potential of 

fast forward swimming (or positionkeeping 

in strong currents) while 

avoiding the complexity of a flexible 

hull. In addition, Walker and 

Westneat (1997) had observed high 

maneuverability by these swimmers 

in their complex reef habitats. These 

habitats are reasonable representations 

of our vehicle’s projected operating 

environment, hence we saw the possibility 

of obtaining both high forward 

speed and excellent low-speed maneuvering 

and hovering performance as 

well. 

The fish we selected to inspire our 

design was one of the coral reef pectoral 

fin swimmers, the bird wrasse 

(Gomphosus varius), shown in Figure 1. 

FIGURE 1 

Bird wrasse (Gomphosus varius). 

Controlled Deformation 

Fin Technology 

Development 

Living creatures, such as insects, 

birds, and pectoral fin swimmers, 

generate lift and thrust by executing 

large-amplitude wing/fin flapping, 

often with substantial shape deformation 

from root to tip and leading edge 

to trailing edge. The flow for these 

motions is three-dimensional and unsteady, 

and conventional steady-state 

aerodynamics is unable to correctly 

compute the corresponding time history 

of flapping-force generation. 

Comprehensive reports on the research 

carried out to study the fluid dynamics 

of flapping fins and wings 

(Rozhdestvensky & Ryzhov, 2003) 

and of biomimetic fins for underwater 

vehicles (Triantafyllo et al., 2004) 

exist, in addition to an extensive number 

of studies, too great to list here, on 

unsteady lift production by flapping 

insect wings. Three-dimensional unsteady 

computations are necessary to 

correctly predict the lift and thrust variation 

throughout the flapping stroke 

cycle. Such computations, for creaturesorvehicleswithmovingand 

deforming surfaces, provide the timevarying 

pressure distribution on all 

surfaces, which in turn can provide insights 

into how the flapping forces and 

maneuvering moments are being generated. 

This information can be coupled 

with computational visualization 

of the time-varying flow about the 

fish to analyze the origin of body and 

fin vorticity, its growth, and eventual 

shedding into the wake. This is the 

approach we developed for tuna caudal 

fin force production analysis 

(Ramamurti et al., 1996, 1999), subsequently 

validated against wrasse experimental 

data (Ramamurti et al., 

2002), and which we also followed 

throughout our fin and vehicle development 

efforts described below. 

There are 13 multiply bifurcating 

fin rays in the bird wrasse pectoral fin 

shown in Figure 2, each contributing 

to the fin curvature time-variation 

throughout the stroke cycle. For ease 

of design, manufacture, actuation, 

and control, it is ideal to have the fewest 

possible number of rays (which we 

refer to as ribs) and have each of them 

FIGURE 2 

Bird wrasse fin structure (from Walker & 

Westneat, 1997). 


e as simple in shape and structure 

as possible. But for more effective fin 

propulsion, it is ideal to maximize the 

number of ribs since more control 

points result in a smoother fit to 

desired fin curvature time-histories. 

Our computational investigations 

(Ramamurti et al., 2004) have shown 

that the loss of force magnitude over 

the stroke cycle is very small if we 

considerably reduce the number of 

ribs, as long as we maintain the ability 

to properly modify the surface curvature. 

The computations showed that 

when the fin was made rigid by specifying 

the motion with just the leading 

edge of the fin tip, the thrust produced 

during the upstroke was less than half 

of the peak thrust produced by the 

flexible fin computations. During the 

downstroke, the computations for 

the rigid and nearly rigid fin produced 

no positive thrust, while the 

partially and fully flexible cases produced 

substantial thrust. In the case 

of the rigid fin, there was also a substantial 

penalty in lift during the 

FIGURE 3 

Effect of fin flexibility on the time variation of thrust forces (from Ramamurti et al., 2004). 

(Color versions of figures available online at: http://www.ingentaconnect.com/content/mts/ 

mtsj/2011/00000045/00000004.) 

FIGURE 4 

Representative rib cross-sectional geometry 

and bending analysis showing a 20° rib 

deflection. 

upstroke. An example from these 

computational investigations is shown 

below in Figure 3. 

Assessment of these findings led us 

to reduce the number of ribs from 13 

to 5. Five was selected since it enabled 

reduced fin complexity, thus substantially 

reducing fin size and weight, 

while maintaining the critically important 

flexural capability. Each of the five 

fin rays were individually designed 

and constructed from compliant ABS 

plastic material using a 3-D printer to 

achieve the desired tip deflection with 

an achievable linear actuation force applied 

to each individual rib at the root 

(Trease et al., 2003), as illustrated in 

Figure 4. The pushing and pulling of 

the ribs at the root of the fin is similar 

to how fish bend their ribs using muscle 

actuation. 

The root section of the fin was selected 

to be of rectangular cross section 

with rounded leading edges and a tapered 

trailing edge in order to accommodate 

the actuators at the base of the 

ribs as shown in Figure 5a. A translucent 

silicone rubber membrane skin, 

optimized in thickness to approximately 

0.5 mm via finite-element analysis 

(Palmisano et al., 2007), provided 

the continuous surface covering for 

the rays as shown in Figure 5b. 

After several parametric 3-D unsteady 

computational fluid dynamics 

(CFD) studies had been carried out 

(Ramamurti & Sandberg, 2006), various 

fin parameters were chosen that 

optimized thrust performance, given 

the mechanical constraints of the fin. 

An example of these computations 

showing the variation of lift and 

thrust as a function of fin flexibility, 

stroke amplitude, and fin strokebias 

is shown in Figure 6. 

The angle of attack of the root section 

of the fin was chosen to be 20°, the 

amplitude of the oscillation to be 114°, 

the flapping frequency to be 1 Hz, and 

the rib spacing to be the minimum 

possible value of 1.2 cm dictated by 

the size of the actuators. These specific 

values, selected for maximizing fin 

forceproduction,areforasolitary 

flapping and deforming fin. Incorporation 

of the fin into a vehicle design 

presents challenges to retain fin performance 

while maintaining vehicle 

simplicity. 

Two-Fin Test-Bed Vehicle 

A simple vehicle was designed and 

built to serve as a test-bed for evaluating 

the performance of all aspects of 


FIGURE 5 

Mechanical fin (a) CAD image without skin and (b) actual with skin. 

the controlled curvature fin technology, 

including its capability for vehicle 

propulsion, low-speed maneuvering, 

and hovering. It was, therefore, intentionally, 

a minimalist design that 

served to house the fins, actuators, 

and a battery. We described above 

how the parameters that govern the 

force production by the fin were chosen. 

However, due to mechanical 

constraints and a desire to easily 

manufacture a vehicle prototype, additional 

3-D unsteady CFD studies of 

the flapping fins incorporated in the 

FIGURE 6 

test-bed vehicle led us to modify the 

fin-alone values. The angle of attack 

of the root section of the fin was chosentobe0°,andtheribspacingwas 

reduced to 0.8 cm. It is during construction 

tradeoffs of this type that 

one relies upon what has been learned 

from the studies of nature to meet operational 

performance goals while balancing 

that with the desire to create a 

vehicle that is producible at a reasonable 

cost. Fixing the fin root at a specific 

angle deviates from nature, but 

the construction simplicity and cost 

Mean fin generated (a) thrust and (b) lift as functions of non-dimensionalized stroke amplitude 

and fin flex, or curvature. The surfaces indicate a bias in the fin stroke angle of 0° (red), 20° 

(green), and 40° (blue). 

savings are so substantial that some 

performance penalty was accepted. 

The flapping stroke amplitude and 

frequency, as well as the phasing of 

the individual fin tipdeflection time 

histories, were retained as controllable 

parameters, and we have performed 

CFD analyses of force production 

with this finconfiguration (Ramamurti 

et al., 2010). 

In keeping with the review nature 

of this paper, we are emphasizing the 

process carried out for our specific 

bio-inspired vehicle design. The details 

of the controlled curvature fin design, 

the linear actuator design and construction, 

the isolated fin construction 

and testing, the vehicle design, the vehicle 

construction, the experimental 

testing, and the validation of computations 

were previously reported 

(Palmisano et al., 2007, 2008; Sandberg 

& Ramamurti, 2008). The fin technology 

test-bed demonstration vehicle 

incorporated two actively controlled deformation 

fins. The two-fin vehicle is 

shown in Figure 7. 

Four-Fin Test-Bed Vehicle 

The test results for the two-fin vehicle 

demonstrated that the controlled 

deformation fin force production was 

capable of meeting our vehicle propulsion 

(position-keeping in a current) 

requirements (Geder et al., 2008). 

However, by design, this test-bed fin 

technology demonstration vehicle 

had restricted options for sensor payload 

and did not have the fore-aft 

force production capability needed 

for heave-pitch control. Hence, the 

design of a larger 41-cm long four-fin 

vehicle was initiated (Figure 8). The 

fore-aft symmetry of the four-fin design 

enables hover and higher precision 

positioning capabilities by decoupling 

vehicle pitch and heave control. 


FIGURE 7 

Two-fin technology test-bed demonstration vehicle: (a) exterior view and (b) interior view. 

The current four-fin vehicle design shown here employs a water-tight cylinder 

for housing the power source and electronics with a flooded space in the nose and 

tail for buoyancy trimming and supplemental sensors. Hardware control and all 

computations are performed by a 16-MHz ATmega2560 microcontroller. Computations 

(Ramamurti et al., 2010) and experimental tests carried out to date 

(Geder et al., 2011) characterized how changes in fin stroke amplitude, frequency, 

bias angle, and curvature affect the thrust and lift forces for zero free stream flow 

speed. Further testing and computations are ongoing to fully characterize the fin 

forces and vehicle dynamics. However, current models have the necessary fidelity 

to accurately predict vehicle performance as outlined in the following sections. 

Four-Fin Vehicle Control 

With the vehicle state variables defined as in Figure 8, the vehicle dynamics can 

be written as, 

Mv →̇ þ Cv ð 

→ Þv → þ Dv ð → 

Þv → þ g → 

ðη → 

Þ ¼ τ → ; 

where M is a matrix of rigid body mass and inertial terms, C is a matrix of centripetal 

and Coriolis terms, D is a matrix of hydrodynamic lift and drag terms, g is a 

vector of hydrostatic terms, v =[uvwpqr] T , η =[xyzϕθψ] T is the position and 

orientation vector in the earth-fixed frame where ϕ, θ, and ψ are roll, yaw, and 

pitch angles, and τ is a vector of all forces and moments external to the rigid body. 

The portion of the vector, τ, that is effected by the fins is represented as, 

→ 

τfins ¼ 

2 

6 

4 

3 

f T ;LF þ f T ;LB þ f T ;RF þ f T ;RB 

0 

f L;LF f L;LB f L;RF f L;RB 

 

 

y L f L;LF þ f L;LB y R f L;RF þ f L;RB ; ð2Þ 

7 

x F f L;LF þ f L;RF þ xB f L;LB þ f L;RB 

5 

 

f T ;LF þ f T ;LB f T ;RF þ f T ;RB 

y L 

y R 

ð1Þ 

where f T is fin thrust and f L is fin lift. 

Subscripts ‘LF’, ‘LB’, ‘RF’, and‘RB’ 

identify the left front, left back, right 

front, and right back fins, respectively. 

The x-position of the center of pressure 

on the fins is denoted by x F for the 

front fins and x B for the back fins. 

The y-position of the center of pressure 

on the fins is denoted by y L for the left 

fins and y R for the right fins. 

Mathematical models representing 

thedynamicperformanceofthefins 

have been developed to include the effects 

on force production of inflow velocities 

to the leading edge (or trailing 

edge for reverse motion) and to include 

the effects of fin interactions with each 

other, namely the effects of the trailing 

vortices off the front fins on the inflow 

to the back fins (Geder et al., 2011). 

Other fin dynamic representations 

modeled controllable parameters including 

fin curvature and stroke amplitude(Ramamurtietal.,2010).In 

the earlier 3-D unsteady CFD studies, 

these two key parameters were found 

to have a direct relationship with thrust 

generation—increasing stroke amplitude 

or fin curvature increased thrust 

(Ramamurti & Sandberg, 2006). 

However, since both stroke amplitude 

and flapping frequency are limited mechanically 

in the vehicle, optimal combinations 

of amplitude and frequency 

were experimentally found for high 

thrust and lift fin gaits. The best mix 

of these parameters for our vehicle 

was determined to be 100° for stroke 

amplitude and 1.8 Hz flapping frequency, 

which yielded not only high 

force output but also relatively low 

power consumption (Palmisano et al., 

2007). These findings allowed us to fix 

stroke amplitude and frequency as 

constants and to focus on fin curvature 

as the primary thrust control parameter. 

Further, biasing the fin stroke up 

or down, as in Figure 9, affects fin lift 


FIGURE 8 

Four-fin technology test-bed vehicle, (a) exterior view, (b) interior view. 

generation while maintaining constant 

thrust (Geder et al., 2008). 

In previous work we evaluated the 

benefits of two vehicle control methods 

(Geder et al., 2008). The first 

method, called weighted gait combination, 

used combinations of thrustgenerating 

and lift-generating fin 

gaits to produce vectored propulsive 

forces. The second method, called 

mean bulk angle bias (MBAB), used 

weighted forward-reverse gait control 

with stroke bias angle control. Between 

these two control methods, 

our results showed that MBAB better 

decoupled control over body-fixed 

thrust and lift forces and yielded better 

vehicle response characteristics in simulation. 

As such, the MBAB method 

is used to control the four-fin vehicle. 

The vehicle controller commands 

changes to the fins to effect changes 

in the forces and moments imparted 

on the vehicle, as shown in equation 2. 

These fin commandsarebasedon 

errors in the vehicle dynamic states, 

computed as the commanded values 

minus the computed values. States 

are computed onboard the vehicle 

using a suite of sensors (three axes 

of accelerometers, three axes of rate 

gyros, magnetic compass, and pressure 

sensor) and sensor fusion and filtering 

schemes (Geder et al., 2009). Errors in 

surge motion (x-axis translation) dictate 

commands for fin thrust changes 

to all fins. Errors in heave motion 

(z-axis translation) dictate commands 

for fin lift changes to all fins. Errors 

in roll motion (x-axis rotation) dictate 

commands for differential lift changes 

between left and right fins. Errors in 

FIGURE 10 

Four-fin vehicle operating in a Naval Research Laboratory test facility. 

pitch motion (y-axis rotation) dictate 

commands for differential lift changes 

in forward and back fins. Errors in yaw 

motion (z-axis rotation) dictate commands 

for differential thrust changes 

in left and right fins. The vehicle has 

no direct control over sway motion 

( y-axis translation), and instead commands 

yaw motion changes to move 

in this direction. The direction of 

yaw motion depends on the sway 

error. The output of a proportionalintegral-derivative 

(PID) controller 

for each vehicle state is used to determine 

forward and reverse fin gait percentage 

and bulk angle commands. 

Four-Fin Vehicle 

Performance 

Initial measurements of the dynamic 

performance of the four-fin vehicle 

have been conducted in two test 

facilities, one a 6 × 2.5 × 2 foot water 

tank and the other a 50-foot diameter 

by 50-foot deep water tank (Figure 10). 

The experimental measurements have 

served to validate the vehicle dynamic 

model and to begin the assessment of 

vehicle performance. 

FIGURE 9 

Vehicle images showing all four fins with 

strokes (a) biased down to produce positive 

lift and (b) biased up to produce negative lift. 


FIGURE 11 

FIGURE 12 

Comparison of experimental and simulated open-loop heading angle responses (from Geder et al., 

2011). 

Comparison of experimental and simulated closed-loop heading angle responses with (a) proportional 

control and (b) PID control (from Geder et al., 2011). The solid curve represents the actual 

simulated heading response of the vehicle based on the modeled dynamics. The dashed curve 

represents the actual experimental heading response observed in vehicle testing. The square 

data represent the measured heading response of the vehicle based on the output of sensor models. 

The circle data represent the measured experimental vehicle heading from the output of onboard 

sensors. 

An open loop test was conducted to 

characterize vehicle heading angle response 

and to compare model simulation 

performance with experimental 

performance (Geder et al., 2011). 

The right finsweresetatfullreversekinematics, 

and the left fins were set to 

closely match the opposite thrust of 

the right fins. At t =11s,thegaitweighting 

inputs were reversed. The simulated 

and experimental responses show very 

good agreement (Figure 11) with a maximum 

difference between the two responses 

at any given time of 5°. Both 

simulated and experimental results exhibit 

a 30°/s maximum turning rate, 

4 s time from zero to maximum speed, 

and braking angle of 35°—the amount 

of residual turning distance after fin 

kinematics are reversed. 

After validating the four-fin vehicle 

dynamic model in yaw motion and implementing 

state feedback control, 

initial closed-loop experiments were 

done to test heading angle control 

(Geder et al., 2011). In Figure 12, a 

comparison of experimental and simulated 

results is given. For a simple proportional 

control algorithm (Figure 12a), 

we see the results match well with a 

30-40° amplitude and 6.5-s period. Differences 

between measured experimental 

heading from onboard sensors 

and simulated heading can be attributed 

to the noise and sensitivity characteristics 

of the sensors (Geder et al., 

2009). External magnetic disturbances 

in our test facility cause errors up to 

10° in heading measurements, also factoring 

into the variation between experimental 

and simulated heading 

responses, as our calibration of the onboard 

compass was not perfect. Adding 

in a derivative gain to damp the 

heading response and a small integral 

gain to eliminate any steady-state error, 

we see the response to a 180° step 

command in heading in Figure 12b. 

With PID control over heading angle, 

the response is nearly critically damped 

with a rise time of 7 s. We also see close 

agreement between measured and actual 

angles, and again differences in 

the responses can be attributed to sensor 

noise and sensitivity, as well as compass 

calibration errors. 


Conclusions 

We have reviewed the history 

of our bioinspired vehicle design process. 

The process began by collaborating 

with biologists in order to 

understand the dynamics of flapping 

and deforming fins in pectoral fin 

swimmers. The detailed kinematics 

they measured in fish swimming experiments 

provided the time-varying 

fin surface curvature data necessary 

for computing the 3-D unsteady flow 

about the swimming fish. Examination 

of the computed flow variations 

about the flapping and deforming 

fins and the fish body provided insights 

into the relationship between the fin 

flows and the fin force time-histories 

throughout the stroke cycle. Parametric 

variations of key stroke parameters in 

3-D unsteady flow computations 

yielded the sensitivity of the timevarying 

fin forces, which in turn provided 

the information needed to modify 

our fin designfromthatofthebird 

wrasse. It is during such computations 

that insights from nature can be 

blended with design constraints to 

yield a range of possible bio-inspired 

designs. A controlled curvature fin utilizing 

individually designed fin ribs, 

each actuated at the rib root by a linear 

actuator was built and tested. The successful 

fin tests were followed by the 

design and construction of a two-fin 

test-bed vehicle to demonstrate the 

mobility potential of the concept. Incorporation 

of the fins into a vehicle 

necessitated further compromises 

where we balanced biomimetic performance 

with produceability and cost. 

These test-bed vehicle tests were followed 

by the design and construction 

of a more capable four-fin vehicle, incorporating 

the same fins. A series of 

controllers were developed and built 

to enable assessment of vehicle propulsion 

and maneuvering performance as 

a function of the varying kinematics of 

each fin. Deforming fin force time histories, 

including fin-fin unsteady interactions, 

have been incorporated into 

the vehicle dynamic model used for 

controller development. The preliminary 

four-fin vehicle dynamic performance 

measurements indicate very 

good agreement with computed performance. 

These initial results are 

very encouraging and indicate that 

our efforts to emulate the positionkeeping, 

low-speed maneuvering, and 

hovering performance of the bird 

wrasse into a producible and low cost 

vehicle are bearing fruit. 

Lead Author: 

Jason D. Geder 

Laboratory for Computational 

Physics and Fluid Dynamics 

Naval Research Laboratory 

Overlook Avenue, SW, 

Washington, DC 

Email: jgeder@lcp.nrl.navy.mil 

References 

Alexander, R.M. 1983. The history of fish 

biomechanics. In: Fish Biomechanics, eds. 

Webb, P.W., & Weihs, D., 1-35. 

New York: Praeger. 

Azuma, A. 1992. The Biokinetics of Flying 

and Swimming. Tokyo: Springer-Verlag. 

265 pp. 

Barrett, D.S., & Triantafyllo, M.S. 1995. 

The design of a flexible hull undersea vehicle 

propelled by an oscillating foil. In: Proc. 

9th Int. Symp. on Unmanned Untethered 

Submersible Technology. Durham, NH: 

Autonomous Undersea Systems Institute 

(AUSI). 

Bartol, I.K., Gharib, M., Weihs, D., Webb, 

P.W., Hove, J.R., & Gordon, M.S. 2003. 

Hydrodynamic stability of swimming in 

ostraciid fishes: role of the carapace in the 

smooth trunkfish Lactophrys triqueter 

(Teleostei: Ostraciidae). J Exp Biol. 206: 

725-44. doi: 10.1242/jeb.00137. 

Blake, R.W. 1983. Fish Locomotion. 

Cambridge, UK: Cambridge University 

Press. 228 pp. 

Breder, C.M. 1926. The locomotion of fishes. 

Zoologica. 4:159-297. 

Collar, D., Wainwright, P., & Alfaro, M. 

2008. Integrated diversification of locomotion 

and feeding in labrid fishes. Biol Lett. 4:84-6. 

doi: 10.1098/rsbl.2007.0509. 

Combes, S., & Daniel, T. 2001. Shape, 

flapping, and flexion: wing and fin design 

for forward flight. J Exp Biol. 204:2073-85. 


Locomotor forces on a swimming fish: 

Three-dimensional vortex wake dynamics 

quantified using digital particle image 

velocimetry. J Exp Biol. 203:2393-412. 


Experimental hydrodynamics of fish 

locomotion: Functional insights from wake 

visualization. Integr Comp Biol. 42:243-57. 

doi: 10.1093/icb/42.2.243. 

Geder, J.D., Palmisano, J., Ramamurti, R., 

Sandberg, W.C., & Ratna, B. 2008. Fuzzy 

logic PID-based control design and performance 

for pectoral fin-propelled unmanned 

underwater vehicle. In: Proc. Int. Conf on 

Control, Automation and Systems. Seoul, 

Korea: IEEE. doi: 10.1109/ICCAS.2008. 

4694526. 

Geder, J.D., Ramamurti, R., Palmisano, J., 

Pruessner, M., Ratna, B., & Sandberg, W.C. 

2009. Sensor data fusion and submerged 

test results of a pectoral fin propelled UUV. 

In: Proc. 16th Intl Symp on Unmanned 

Untethered Submersible Technology. Durham, 

NH: Autonomous Undersea Systems Institute 

(AUSI). 

Geder, J.D., Ramamurti, R., Palmisano, J., 

Pruessner, M., Sandberg, W.C., & Ratna, B. 

2011. Four-fin bio-inspired UUV: modeling 


and control solutions. ASME Intl Mech Eng 

Congress Exposition. IMECE. 2011-64005. 

Gharib, M., Pereira, F., Dabiri, D., Hove, 

J.R., & Modarress, D. 2002. Quantitative 

flow visualization: Toward a comprehensive 

flow diagnostic tool. Integr Comp Biol. 

42:964-70. doi: 10.1093/icb/42.5.964. 

Gibb, A.C., Jayne, B.C., & Lauder, G.V. 

1994. Kinematics of pectoral fin locomotion 

in the bluegill sunfish Lepomis Macrochirus. 

J Exp Biol. 189:133-61. 

Kato, N. 2000. Control performance in 

the horizontal plane of a fish robot with 

mechanical pectoral fins. IEEE J Oceanic Eng. 

25:125-9. doi: 10.1109/48.820744. 

Kato, N., Ando, Y., Tomokazu, A., Suzuki, H., 

Suzumori, K., Kanda, T., & Endo, S. 2008. 

Elastic pectoral fin actuators for biomimetic 

underwater vehicles. In: Bio-mechanisms 

of Swimming and Flying, eds. Kato, N., 

Kamimura, S., 271-82. Tokyo: Springer. 

Kato, N., & Furushima, M. 1996. Pectoral 

fin model for maneuver of underwater vehicles. 

In: Proc. Symp. on Autonomous 

Underwater Vehicle Technology. Monterrey, 

CA: IEEE. doi: 10.1109/AUV.1996.532400. 

Lindsey, C.C. 1978. Form, function, and 

locomotory habits in fish. In: Fish Physiology 

Vol. II, eds. Hoar, W.S., & Randall, D.J., 

1-100. New York: Academic Press. 

Palmisano, J., Geder, J.G., Ramamurti, R., 

Liu, K.J., Cohen, J.J., Mengesha, T., … 

Ratna, B. 2008. Design, development and 

testing of flapping fin with actively-controlled 

curvature for an unmanned underwater vehicle. 

In: Bio-mechanisms of Swimming and 

Flying, eds. Kato, N., & Kamimura, S., 

283-94. Tokyo: Springer. doi: 10.1007/ 

978-4-431-73380-5_23. 

Palmisano, J., Ramamurti, R., Liu, K.J., 

Cohen, J., Sandberg, W.C., & Ratna, B. 

2007. Design of a bio-mimetic controlledcurvature 

robotic pectoral fin. In: Proc. 

IEEE Intl. Conf. on Robotics and Automation. 

Rome, Italy: IEEE. doi: 10.1109/ROBOT. 

2007.363110. 

Ramamurti, R., Geder, J.D., Palmisano, J., 

Ratna, B., & Sandberg, W.C. 2010. Computations 

of flapping flow propulsion for unmanned 

underwater vehicle design. AIAA J. 

48(1):188-201. doi: 10.2514/1.43389. 

Ramamurti, R., Lohner, R., & Sandberg, 

W.C. 1996. Computation of the unsteady 

flow past a tuna with caudal fin oscillation. 

In: Advances in Fluid Mechanics, eds. Rahman, 

M.,&Brebbia,C.,vol.9,169-78.Southampton, 

UK: Computational Mechanics Publishing. 

Ramamurti, R., Lohner, R., & Sandberg, 

W.C. 1999. Computation of the 3-D 

unsteady flow past deforming geometries. 

Int J Comput Fluid D. 13:83-99. 

doi: 10.1080/10618569908940891. 

Ramamurti, R., & Sandberg, W.C. 2006. 

Computational fluid dynamics study for 

optimization of a fin design. In: Proc. 24th 

AIAA Applied Aerodynamics Conference. 

AIAA. AIAA-2006-3658. 

Ramamurti, R., Sandberg, W.C., & Lohner, R. 

2004. The influence of fin rigidity and 

gusts on the force production in fish and 

insects: a computational study. In: Proc. 42nd 

AIAA Aerospace Sciences Meeting and Exhibit. 

Am Inst Aeronautics and Astronautics. 

AIAA-2004-0404. 

Ramamurti, R., Sandberg, W.C., Lohner, R., 

Walker, J.A., & Westneat, M.W. 2002. Fluid 

dynamics of flapping aquatic flight in the 

bird wrasse: 3-D unsteady computations with 

fin deformation. J Exp Biol. 205:2997-3008. 

Rozhdestvensky, K.V., & Ryzhov, V.A. 

2003. Aerohydrodynamics of flapping-wing 

propulsors. Prog Aerosp Sci. 39:585-633. 

doi: 10.1016/S0376-0421(03)00077-0. 

Sandberg, W.C., & Ramamurti, R. 2008. 

3-D Unsteady computations of flapping flight 

in insects, fish, and unmanned vehicles. 

In: Bio-mechanisms of Swimming and 

Flying, eds. Kato, N., & Kamimura, S., 

205-217. Tokyo: Springer. doi: 10.1007/ 

978-4-431-73380-5_17. 

Sfakiotakis, M., Lane, D.M., & Davies, J.B.C. 

1999. Review of fish swimming modes for 

aquatic locomotion. J Exp Biol. 24:241-52. 

Trease, B.P., Liu, K.J., & Kota, S. 2003. 

Biomimetic compliant system for actuatordriven 

aquatic propulsor: preliminary results. 

In: ASME Intl. Mech. Eng. Congress and 

Exposition. IMECE 2003-41446. 

Triantafyllo, M.S., Techet, A.H., & Hover, 

F.S. 2004. Review of experimental work in 

biomimetic foils. IEEE J Oceanic Eng. 

29:585-94. doi: 10.1109/JOE.2004.833216. 

Videler, J.J. 1993. Fish Swimming. London, 

UK: Chapman and Hall. 260 pp. 

Wainwright, P., Bellwood, D., & Westneat, 

M.W. 2002. Ecomorphology of locomotion 

in labrid fishes. Environ Biol Fish. 65:47-62. 

doi: 10.1023/A:1019671131001. 

Walker, J.A., & Westneat, M.W. 1997. 

Labriform propulsion in fishes: Kinematics 

of flapping aquatic flight in the bird wrasse 

Gomphosus varius (Labridae). J Exp Biol. 

200:1549-69. 


Mechanical performance of aquatic rowing 

and flying. P Roy Soc Lond B Bio. 267: 

1875-81. doi: 10.1098/rspb.2000.1224. 

Webb, P.W. 1975. Hydrodynamics and 

energetics of fish propulsion. Bull Fish Res Bd 

Can. 190:1-158. 

Webb, P.W. 1984. Form and function in fish 

swimming. Sci Amer. 251:58-68. 

Webb, P.W. 2004. Maneuverability— 

General issues. IEEE J Oceanic Eng. 

29:547-53. doi: 10.1109/JOE.2004.833220. 

Westneat, M.W., & Walker, J.A. 1997. 

Motor patterns of underwater flight: an 

electromyographic study of the pectoral 

muscles in the bird wrasse, Gomphosus varius. 

J Exp Biol. 200:1881-93. 


PAPER 

A Twistable Ionic Polymer-Metal Composite 

Artificial Muscle for Marine Applications 

AUTHORS 

Kwang J. Kim 

David Pugal 

Active Materials and Processing 

Laboratory, Department of 

Mechanical Engineering, 

University of Nevada-Reno 

Kam K. Leang 

Electroactive Systems and Controls 

Laboratory, Department of 


University of Nevada-Reno 

Introduction 

I 

onic polymer-metal composite 

(IPMC) material is one of the most 

promising active (smart) materials 

for developing novel soft biomimetic 

actuators and sensors, preferably for 

underwater applications (Shahinpoor 

et al., 1998; Shahinpoor & Kim, 2001; 

Kim & Shahinpoor, 2003; Shahinpoor 

& Kim, 2005). The advantages of 

the IPMC include low driving voltage 

(

FIGURE 1 

(a) Scanning electron microscope image of the cross-section of a Nafion-based IPMC. (b) Illustrative 

movement of cations and water molecules inside of an IPMC. 

saturated in a polar solvent (such as 

water) and then an electric field is applied 

across the electrodes, the composite 

bends. The bending is caused 

by induced swelling on the cathode 

side of the composite and shrinking 

on the anode side [see Figure 1(b)] 

due to a sudden flux of cations and 

polar solvent (such as water). An oppositely 

applied voltage causes bending in 

the opposite direction. Conversely, 

when an IPMC is mechanically deformed, 

charges develop on the electrodes 

and thus IPMCs can function 

as current or voltage sensor (Alici 

et al., 2008; Pugal et al., 2010a). 

The electromechanical behavior of 

IPMCs have many noteworthy applications. 

Due to their biocompatibility, 

IPMC actuators show great promise in 

biomedical devices such as active endoscopes 

(Yoon et al., 2007) and smart 

catheters (Fang et al., 2007). Strips of 

IPMCs can be used as sensors in 

hand prostheses (Biddiss & Chau, 

2006). But one of the most promising 

applications is innovative propulsion 

systems for underwater autonomous 

systems (Fish et al., 2008; Kamamichi 

et al., 2006; Kim et al., 2005, 

2007; Lauder, 2007). Specifically, 

IPMCs can replace or enhance the 

design of propulsors for underwater 

walking and swimming machines that 

are currently based on traditional actuators 

such as DC motors (Ayers & 

Witting, 2007; Bozkurttas et al., 2008; 

Buchholz et al., 2008; Kato, 1998; 

Krieg & Mohseni, 2008; Tangorra 

et al., 2007), pneumatic actuators 

(Cai et al., 2010), and magnetic actuators 

(Tortora et al., 2010). In fact, 

strips of IPMCs have been used to construct 

artificial tentacles for a jellyfishlike 

robot (Guo et al., 2007). The 

walking speed of the jellyfish robot 

was controlled through the frequency 

of the input voltage applied to the 

IPMC-based legs. Likewise, an artificial 

caudal fin to propel a robotic fish 

was created from an IPMC actuator 

(Chen et al., 2010; Guo et al., 2006; 

Aureli et al., 2010a). The achievable 

peak swimming speed of the robotic 

fish in (Chen et al., 2010) was reported 

at 22 mm/s. In terms of performance, 

the maximum (stall) torque to weight 

ratio between a prototype twistable 

IPMC AM with dimensions 50 mm × 

25 mm × 1 mm is comparable to a 

small ungeared DC motor as illustrated 

in the comparison shown in Figure 2. 

The comparable performance of 

IPMCs and traditional actuators suggests 

that IPMCs can play a critical 

role in the development of highly 

maneuverable and efficient marine 

systems, e.g., the system described in 

Menozzi et al. (2008). 

Due to the nature of the material 

deformation caused by cation and 

solvent flux, bending motion is the 

most commonly studied and applied 

for IPMCs (Kim et al., 2007). Particularly, 

when an IPMC strip is mounted 

in the cantilever configuration, with 

one end fixed and the other free, an 

applied electric field to the IPMC as 

illustrated causes the actuator to bend 

as illustrated in Figure 3(a). This single 

degree-of-freedom bending motion 

has wide applications, such as a singlelink 

(Aureli et al., 2010a) and multilink 

(Yim et al., 2007) oscillatory 

propulsor. However, IPMCs with 

multiple degrees of freedom are highly 

desirable to create control surfaces, 

which can undergo complex motion 

and deformation, for both station 

keeping (Bandyopadhyay et al., 

2008) as well as propulsion and maneuvering. 

It has been observed that 

the propulsion and maneuvering 

capabilities of the Bluegill (Lepomis 

marcrochirus) sunfish is primarily due 

to its highly deformable pectoral fin 

(Bozkurttas et al., 2008). Thus, multiple 

degrees-of-freedom IPMC actuator 

technology offers many possibilities for 

mimicking such behavior to create 

more efficient and maneuverable 

underwater systems (Chen & Tan, 

2010). As depicted in Figure 3(b), 

twisting motion can be achieved by 

patterning sectored electrodes combined 

with independent control each 

isolated region. By carefully creating 

electrodes on the surface of the 


FIGURE 2 

Comparison of torque-to-weight ratios for traditional motors to twistable IPMC AM fin. IPMC AM fin 

was fixed on one end, then 5-V DC input was applied to each electrode, with opposing polarity, and 

the blocking torque of the free end was measured with an ATI Nano17 force/torque sensor. 

polymer with proper electrical isolation 

between adjacent units, sections 

of the AM fin can be independently 

controlled to achieve complex shapes 

FIGURE 3 

IPMC AM fin motion: (a) bending and (b) twisting. 

and deformations. Additionally, isolated 

electrodes can be patterned for 

sensing motion and deformation of 

the AM fin (Kruusamae et al., 2009). 

IPMC Manufacturing 

Basic Structure of IPMC and 

Nafion Membrane Fabrication 

A basic IPMC consists of an ion exchange 

polymer, such perfluorinated 

alkenes or styrene/divinylbenzenebased 

polymers, sandwiched between 

two noble metallic electrodes as 

shown in Figure 1(a). The conducting 

media can be palladium, silver, gold, 

carbon, graphite, and even nanotubes; 

however, platinum is the most commonly 

used. The metal electrodes are 

often chemically deposited on the 

polymer’s surface through a reduction 

process (Kim & Shahinpoor, 2003). 

Conductive paint can also be applied 

to the surface of the membrane to 

serve as an electrode; however, this approach 

is not as robust and effective as 

electrochemical plating. 

The commonly used ion exchange 

membrane Nafion (Dupont) for manufacturing 

IPMCs is easily available 

from distributors. In the case of 

Nafion, the typical chemical structure 

is shown in Figure 4, and it consists of 

fluorocarbons, oxygen, sulfonate 

groups, and a mobile cation, which 

are typically either hydrogen, sodium, 

or lithium. Commercially available 

Nafion membrane such as Nafion 

115, Nafion 117, and Nafion 1110 

for fabricating IPMCs have nominal 

dry thicknesses of 127, 178, and 

254 μm, respectively (Aureli et al., 

2010a). Methods to enhance the performance 

of IPMCs include boosting 

the capacitance of the composite 

(Akle et al., 2005; Aureli et al., 

2009), where this is motivated by studies 

that correlate actuation and sensing 

performancewithcapacitance(Akle 

et al., 2005). Enhancements in performanceandblockingforcehavealso 

been made by incorporating nanoparticulates 

into the polymer matrix 


FIGURE 4 

The chemical structure of Nafion, which includes fluorocarbons, oxygen, sulfonate groups, and 

a mobile cation X + that can be hydrogen, sodium, or lithium. The K is usually 5–11, and the L is 

usually 1 (Shahinpoor & Kim, 2001). 

(Nam et al., 2003; Nguyen et al., 2007). 

In these studies, the nanocompositebased 

IPMCs were observed to have 

higher water uptake and slower water 

loss, thus leading to larger bending 

displacement and blocking force. It 

was also found that by using a dispersing 

agent in the reduction process 

to form fine platinum polycrystals 

whichsubsequentlyleadtodeeper 

penetration of the platinum layer, the 

blocking force was increased significantly 

(Shahinpoor & Kim, 2001). 

More recent work to improve output 

force is by increasing the thickness of 

the Nafion membrane (Kim & 

Shahinpoor, 2002). Since relatively 

thick Nafion membrane is not readily 

available, researchers have explored the 

solution casting process (Kim & 

Shahinpoor, 2002; Kim et al., 2003; 

Pak et al., 2004; Shan & Leang, 

2009) and the hot pressing technique 

(Lee et al., 2006). The output force 

enhancement for thicker IPMCs is evident 

by considering two IPMC strip 

actuators, both having the same length 

and width, but each have a different 

thicknesses, such as t and 2t (twice as 

thick). Assuming that for both actuators 

the same tip displacement is required, 

then the required strain for 

the thick actuator is 2ɛ. As the stress 

tensor for linear beam with thickness 

of t, width b, and length L can be expressed 

as (Kim & Shahinpoor, 2002) 

σ t ¼ 6F tL 

bt 2 ; ð1Þ 

the ratio of the stresses is 

σ 2t 

σ t 

≈ 2 ¼ F 2t 

2 2 F t 

; ð2Þ 

FIGURE 5 

hence F 2t =8F t .Therefore,athicker 

IPMC will produce a larger blocking 

force, motivating the need to create 

thicker membranes for fabricating 

IPMCs. At the same time, the measurements 

show that actuation speed 

reduces with the increase of the thickness. 

Therefore, further study is necessary 

to find an optimal thickness of the 

membrane for marine applications. 

Pretreatment and Platinum 

Plating Process 

The manufacturing of IPMCs begins 

with the pretreatment of the ion 

exchange membrane (Nafion) and 

the platinum plating process as 

outlined in the flowchart shown in 

Figure 5. First, the surface of the membrane 

is either mechanically roughened 

or chemically etched (Yoon et al., 

2007) to either enhance the capacitance 

or to improve adhesion of the 

metal electrode to the surface. Then, 

organic and metallic impurities on 

thebareNafion membrane are removed 

through a pretreatment process 

by initially chemically cleaning the 

Nafion membrane in 3% hydrogen 

peroxide (H 2 O 2 ). Next, the cleaned 

membrane is rinsed in 0.5 M sulfuric 

acid (H 2 SO 4 )at80°C.Afterwards, 

the pretreated and cleaned Nafion 

membrane is immersed in an appropriate 

metal salt solution such as 

tetraamineplatinum (II) chloridemonohydrate 

[(Pt(NH 3 ) 4 )Cl 2 H 2 O] for 

2 h, followed by several washings in 

de-ionized water. Platinum particles 

are metalized on the surface of the 

Nafion membrane by reducing the 

Left: The process flow for pretreating the Nafion membrane and applying platinum electrodes. 

Right: fabricated IPMCs from commercially available Nafion membrane. 


membrane in a sodium borohydride 

(NaBH 4 ) or lithium borohydride 

(LiBH 4 ) solution for 3 h. Finally, the 

platinum-plating process is repeated to 

achieve at least three layers of platinum 

on the surface of the Nafion membrane 

to enhance surface conductivity 

and overall performance. The platinum 

particulate layer is often buried 

1-20 μm withintheIPMCsurface 

and is highly dispersed. 

Electrode Patterning Process 

The sectored electrodes for the 

twistable IPMC AM fin can be created 

by masking, surface machining, or ablating 

the electrode using a high-power 

laser. The first approach involves the 

use of a mask to cover areas of the 

membrane that should not be exposed 

to the platinum-plating process, 

subsequently creating an isolation 

region between adjacent electrodes. 

The second approach uses a precision 

computer-controlled milling machine 

to mechanically remove the platinum 

electrodes from the surface of an IPMC 

membrane to create the isolation region. 

Masking Technique: The masking 

technique to create IPMCs with 

sectored electrodes involves the use 

of UHMW (Ultra-high-molecularweight) 

polyethylene tape (3M). The 

tape is used to cover specific regions 

of the bare Nafion membrane to inhibit 

the plating of platinum on the surface 

of the membrane. For example, 

the tape is applied to the bare Nafion 

membrane, then the taped Nafion 

membrane is processed using the pretreatment 

and plating process described 

above and outlined in Figure 5. Sample 

IPMCs with sectored electrodes are 

shown in Figure 6(a), and an outline 

of the process for the masking technique 

is illustrated in Figure 7(a). 

Machining Technique: The 

surface machining method utilizes a 

computer-controlled milling machine, 

such as an automated circuit board 

router (e.g., ProtoMat S42, LPKF) to 

mechanically remove the plated platinum 

metal on the surface of the Nafion 

membrane. The surface machining 

process is outlined in Figure 7(b) and 

described as follows: 

1. Create the machining path to create 

the electrode pattern using CAD 

software (e.g., Solidworks) or a circuit 

board layout program (e.g., 

Eagle). The result of this step is a 

CAD/CAM file, which is ran by 

the milling machine. 

2. Attach an IPMC sample (with platinum 

electrodes) to the working 

surface of the milling machine 

using an adhesive layer, for example 

double-sided tape (see Figure 7(b)) 

or a vacuum system. Air bubbles 

trapped underneath the IPMC 

sample should be removed. Additionally, 

the locations of the corners 

of the IPMC on the working surface 

are marked for aligning the 

sample. 

3. Load the CAD/CAM file onto the 

milling machine and start the milling 

process. 

4. Remove the machined IPMC sample, 

then flip it over and attach the 

sample to the working surface making 

sure the corners are aligned with 

the markings created in Step 2. 

Repeat Steps 2 and 3. 

5. Remove the machined IPMC sample 

and trim away excess material. 

Sample IPMCs with sectored 

electrodes created by the surface 

machining process are shown in Figures 

6(b1)-6(b4). During the machining 

process, care is taken to 

avoid removing too much material 

for thin IPMC membranes. In general, 

the machine should be set to remove 

only the platinum material (approximately 

25-50 μm deep). Comparing 

the masking and surface-machining results, 

the machining process allows 

better control of the shape and pattern 

of the electrode. One of the major 

challenges of the masking process is 

ensuring that the tape adheres to the 

IPMC membrane’s surface all through 

the plating process. During machining, 

if the depth of cut is not well-controlled, 

removal of excess material affects the 

mechanical properties, for example, 

stiffness, of the IPMC actuator. 

Modeling 

Electromechanical 

Transduction 

Two approaches are discussed to 

model the electromechanical bending 

and twisting response of the sectoredelectrode 

IPMCs: (1) a simplified 

finite-element analysis (FEA) model 

derived from the piezoelectric effect 

and (2) a more comprehensive physicsbased 

model. The former model has 

the advantage of being easy to implement 

due to fact that software packages 

are available and specifically tailored to 

solve the problem. Additionally, the 

established algorithms are computationally 

efficient. However, the first 

method lacks consideration of the 

physical processes that governs the behavior 

of the IPMC. The latter model, 

on the other hand, considers the underlying 

physics that includes electrostatic 

forces, osmotic pressure, charge 

imbalance and the effects of local strains 

(Kim & Shahinpoor, 2003; Nemat- 

Nasser & Jiang, 2000; Tadokoro 

et al., 2000; Chen & Tan, 2008; Leo 

et al., 2005). This type of model offers 

valuable insight on the physical behavior 

of the composite material and 

the model can be used to help guide 

the development of the material on 

many levels. Despite being more 

realistic—and sometimes more 

accurate—the physics-based model 


FIGURE 6 

Fabricated IPMCs with sectored electrodes: samples created using (a) the masking technique and (b1)-(b4) the surface-machining approach. 

Dimensions are in millimeters (mm). 

is more computationally demanding 

to solve. Simulation results comparison 

with experimental data are 

presented in Performance Characterization 

and Discussion 

Simplified FEA Model for 

Electromechanical Response 

Asimplified finite-element model 

is created to predict and gain insight 

on the bending and twisting capabilities 

of the sectored-electrode IPMC 

AM fin. Such a model can also be 

applied to optimize the design of an 

AM fin for specific applications. The 

key feature of this simple model is 

the deformation is estimated using an 

equivalent bimorph beam model analogous 

to a piezoelectric actuator (Kim 

& Tadokoro, 2007). As a result the 

model only captures the basic electromechanical 

behavior and is thus easy to 

implement and computationally efficient. 

Figure 8 shows the boundary conditions 

for the finite-element model, 

where the ‘clamped area’ (25 mm × 

5 mm) is considered fixed. The bimorph 

finite-element model consists of two 

electro-mechanical actuation layers, 

layers (i) and (iii), as shown in Figure 8. 


FIGURE 7 

Electrode patterning process: (a) the masking and (b) the surface-machining technique. 

The material separating patterned areas 

of the material are modeled by layer (ii). 

The IPMC material is treated as a 

homogenous material (uniform stiffness 

and density). The standard stiffness 

matrix k, piezoelectric strain matrix d, 

and permittivity matrix ɛ are used in 

the modeling (Moheimani & Fleming, 

2006). The material properties are as 

follows: density is 2930 kg/m 3 ; 

Poisson’s ratio is 0.49; E=1.16GPa; 

d 31 = d 32 =4.11×10 -6 m/V (bending); 

d 31 = d 32 =2.67×10 −7 m/V (twisting). 

The finite-element model is created 

using ANSYS software (Canonsburg, 

PA). A SOLID98 tetrahedral coupled 

field solid element is chosen for the 

electro-mechanical material, while a 

SOLID187 tetrahedral structural solid 

element is chosen for the electrode separation 

material. Both material properties 

are assigned a tetrahedral solid to compensate 

for the irregular mess interface 

caused by large aspect ratio between the 

IPMCs thickness and length. By selecting 

these elements, a uniform tetrahedral 

mesh throughout the entire model is 

achieved. 

Physics-Based Model for 

Electromechanical Transduction 

When a voltage is applied to the 

electrodes of an IPMC, the freely movable 

cations inside the polymer start 

migrating due to the imposed electric 

FIGURE 8 

Finite-element model structure. Layers (i) and (iii) are the bimorph (electro-mechanical) layers, and layer (ii) is assumed to be a nonconducting 

electrode separation layer (absent of electro-mechanical properties). 


field. However, the attached anions do not migrate nor diffuse. In case of waterbased 

IPMCs, migrating cations drag the water molecules along, causing 

osmotic pressure changes and therefore swelling near the cathode and contraction 

of the polymer near the anode. This in turn results in bending of the material 

towards the anode side. Furthermore, the migrated cations cause charge imbalance 

near the electrodes and this possibly results in the electrostatic force contribution 

to the local strain. In the following, the basic underlying equations 

are presented and how to apply the calculated charge imbalance near the 

electrodes in modeling the IPMC actuation in a 3-D domain is discussed 

(see Figure 9). 

First, the ionic migration and diffusion are described in the polymer domain by 

Nernst-Planck and Poisson equations, 

∂C 

∂t 

þ ∇· ð D∇C zμFC∇ϕÞ ¼0; 

∇ 2 ϕ ¼ F ð C C 0Þ 

; ð4Þ 

ɛ 

where C is the cation concentration, μ is the mobility of counter ions, D is the diffusion 

constant, F is the Faraday constant, z is the charge number, ϕ is the electric 

potential, and C 0 is the anion concentration with initial value of C 0 =1200mol/m 3 . 

For the electrode, Ohm’s law for the current density is 

σ∇V ¼ → j; ð5Þ 

where σ is electric conductivity, V is the voltage, and → j is the current density in the 

electrode. It must be noted that the electric potential ϕ inside the polymer and the 

electric potential V in the electrode are different. To relate the variables of 

the electrodes V and → j to the ionic flux inside the polymer, the Ramo-Shockley 

FIGURE 9 

Conceptual physics-based model. Calculations are done in three domains—the polymer domain 

and two electrode domains. 

ð3Þ 

theorem is used (Ramo, 1939; Shockley, 

1938), i.e., 

I ¼ 1 φ ∑ → 

n q nW ð 

→ 

r n Þ· →v n ; 

ð6Þ 

where → r n ; → v n ,and → q n are the position vector, 

instantaneous velocity, and charge 

of cation n, respectively. The electric 

field that would be produced by 1 V applied 

potential without any charges, 

neither mobile nor fixed, is denoted by 

→ 

W . The current in the external circuit is 

represented by I, andϕ is a constant 

with value of 1 V. Equation (6) can be 

simplified for a 2-D domain with parallel 

electrodes 

j ¼ 1 h ∫ f dl; 

ð7Þ 

where f is an ionic flux in the electrode 

direction and has a unit of Cm 2 s and j is 

local current density at the inner boundary 

of an electrode. The term dl is the 

integration element along the path 

where the particle moves, which in this 

case is assumed to be from one electrode 

to other. 

The Equations (3)-(7) are used to 

calculate the time dependent cation 

concentration C and corresponding 

charge density ρ = C − C 0 in the polymer 

of IPMC in response to an applied 

voltage. To relate the charge density 

ρ to the physical bending, the force 

coupling similar to the one shown in 

Pugal et al. (2008) is used. A set of 

continuum mechanics equations were 

implemented for the polymer domain. 

Normal and shear strain are by 

definition 

ɛ i ¼ ∂u i 

; ɛ ij ¼ 1 

∂u i 

þ ∂u 

j 

; ð8Þ 

∂x i 2 ∂x j ∂x i 

where u is the displacement vector, 

x denotes a coordinate and indices i 


̂ 

and j are in the range of 1-3 and denote 

components correspondingly to x, y,or 

z direction. The stress-strain relationship 

is 

σ f ¼ Dɛ; 

ð9Þ 

where D is a 6 × 6 elasticity matrix, 

consisting of components of Young’s 

modulus and Poisson’s ratio. The system 

is in equilibrium, if the relation 

∇· σ f ¼ → F ð10Þ 

is satisfied. This is the Navier’s equation 

for displacement (Heinbockel, 

2001). The body force and charge coupling 

is defined as 

→ 

F ¼ Aρx; ̂ 

ð11Þ 

the range of hundreds of thousands 

of degrees of freedom. The challenges 

are described in more detail in Pugal 

et al. (2010b). To reduce the problem 

size without significant loss in the 

calculation precision, the following 

modeling approach for 3-D actuation 

of IPMC is developed. 

Firstly, the cation concentration 

C (from which the charge density ρ 

can be directly calculated) and voltage 

ϕ are calculated in a 2-D domain. 

As the 2-D domain is scaled in the longitudinal 

(x) direction of an IPMC, the 

electrode conductivity value is also 

linearly scaled. The electrode currents 

and corresponding voltage gradient in 

the electrodes are taken into account in 

the 2-D model to obtain more precise 

cation concentration. See, for instance, 

Figure 10—there the molar flux f 

and electric current j are depicted. As 

a result of the calculations, spatial 

and temporal C and ϕ in 2-D are 

found and stored. This is done for 

each patterned electrode that is subjected 

to a different voltage. Due to 

thesmallDebyescreeninglengthof 

the charges near the electrodes (Porfiri, 

2009), the 2-D model calculations are 

done on a mesh that is extremely fine 

near the boundaries. 

Secondly, the calculated C and ϕ are 

extruded onto slightly coarser mesh in 

the 3-D domain as illustrated in 

Figure 11. For instance, Figure 12 

shows the extruded voltage ϕ values 

on the 3-D domain boundary. It 

should be noted that in the current 

model, it is not necessary to extrude 

the values of ϕ, but it was done for 

illustration purposes. Finally, the 

where A is a parametrically determined 

constant and x is IPMC’s longitudinal 

direction. This approach allows calculating 

deformation that is in the typical 

IPMC actuation range. When a very 

large deformation is expected, for instance, 

in case of a very thin membrane, 

geometric nonlinearity and 

more precise force coupling must be 

used in the model. The conceptual 

model with the variables is illustrated 

in Figure 9. 

The finite element method is used 

to solve Equations (3), (4), and (5) 

with Equation (7) as the boundary 

condition between the electrode and 

polymer domain and Equation (10). 

The equations are implemented in 

Comsol Multiphysics software package 

(Multiphysics, 2011). The high aspect 

ratio of the domain representing 

an IPMC and the nonlinear nature 

of the problem make it difficult to 

directly solve the equations in a 

full-scale 3-D domain. Namely, the 

problem size would be very large, in 

FIGURE 10 

Electric current streamlines in the electrodes and total ionic flux streamlines in the polymer 

A 

domain. The color depicts the total current density m2 

in the electrodes. (Color versions of 

figures available online at: http://www.ingentaconnect.com/content/mts/mtsj/2011/00000045/ 

00000004.) 


FIGURE 11 

Extruding the cation concentration C that has been calculated for each 

time step in 2-D into a 3-D domain. 

FIGURE 12 

Extruded data from 2-D domain. The surface depicts the extruded voltage 

values on the electrodes. Notice the voltage gradient in the longitudinal 

x direction. 

coupling Equations (11) and (10) are 

carried out to calculate the timedependent 

bending of an IPMC in 3-D. 

Performance 

Characterization 

and Discussion 

The bending and twisting performance 

of the fabricated IPMC AM 

fins are characterized and then compared 

to the electromechanical models. 

The experimental setup consists of 

control electronics connected to the 

IPMC AM fin and a measurement system 

(laser displacement sensors and 

data acquisition system) for collecting 

the output response. Each set of electrodes 

are independently controlled 

by a separate custom-designed voltage 

amplifier as depicted in Figure 13. A 

complete description of the experimental 

setup is described in Riddle 

et al. (2010). All measurements are 

taken while the subject IPMC is actuated 

in de-ionized water. 

Measured Bending and 

Twisting Performance 

Themeasuredresponseforaselected 

masked and machined electrode 

IPMC, Figure 6(b2), are shown in Figure 

14. The results in Figure 14 show 

that both types of actuators provided 

approximately the same degree of 

FIGURE 13 

twist. Furthermore, the actuators also 

exhibited non-smooth twisting 

motion, which may have been caused 

by the effects of the fabrication process. 

A peak twist angle of 7.3° is measured 

for the machined IPMC [see 

Figures 14(c) and 14(d)]. The frequency 

of the actuation of 1 Hz is 

Experimental setup for measuring bending and twisting response using two non-contact laser 

sensors (Micro-Epsilon, optoNCDT 1402). Voltage amplifier gain is A. 


sufficient for underwater propulsion 

(Chen et al., 2010). The bending 

angle could be enhanced in various 

ways, depending on the application 

of interest. For instance, rigid extensions 

with different shapes can be 

used (Anton, 2008). 

FIGURE 14 

Measured IPMC twisting response for 5 V sinusoidal input at 1 Hz: (a) input voltage applied to 

electrodes; (b) twisting response for masked electrode IPMC; (c) twisting response for machined 

electrode IPMC; (d) sequence images of machined electrode IPMC showing twisting behavior. 

Simulated Responses 

The bending and twisting response 

of the IPMC, Figure 6(b2), produced 

by the simple FEA model for an 

input voltage of 2 V are shown in 

Figures 15(a) and 15(b). For bending, 

the tip displacement is predicted at 

approximately 2.5 mm. For twisting, 

the angle of twist is estimated using 

θ ¼ tan 1 a b 

, where a and b are 

shown in Figure 15(c). The predicted 

maximum angle of twist is approximately 

0.7° for a 2-V input. In Figures 

15(d) and 15(e), the FEA results 

and the response for the bending and 

twisting motion measured along the 

length of the IPMC actuator are 

compared. For bending, the predicted 

and measured results agreed well, 

where the maximum root-meansquared 

(RMS) error is approximately 

0.2 mm. For bending, however, the 

RMS error increased significantly for 

3 V and beyond. It is noted that the 

finite-element approach assumes the 

material is isotropic and operating 

within the linear-elastic range. Due to 

the complex nature of the IPMC 

material, the simplified FEA may be 

limited in its ability to accurately 

predict the performance for large 

electric fields. For example, as shown 

in Figure 14, the measured twist 

angle is approximately 7° for an input 

voltage with magnitude of 5 V. This 

result indicates that the IPMC’s response 

can be highly nonlinear for 

large input voltages. Therefore, one 

of the challenges is developing detailed 

enough models to aid in designing the electrode patterns to meet a specific 

application. 

To adequately model the twisting deformation with the physics-based electromechanical 

model presented in Physics-Based Model for Electromechanical 

Transduction, the sinusoidal voltages 

u 1 ðÞ¼5 t ½V 

Šsinð2πtÞ; u 2 ðÞ¼ t 

5½V 

Šsinð2πtÞ; ð12Þ 

are set as time-dependent boundary conditions to the electrodes 1, 2 and 3, 4, 

respectively (see Figure 13). The calculated displacement fields at two different 

times are shown in Figure 16. The model is validated against measured time response 

[Figure 14(c)]. The model estimation versus measurements are shown in 

Figure 17. It can be seen that the model predicts the twisting deformation well. 

Slight discrepancies in the peak values can be attributed to the fact that the model 


FIGURE 15 

Finite-element modeling results for (a) bending and (b) twisting for 2-V input. (c) Definition of twist angle θ. Comparison of finite-element model 

results and measured response: (d) bending and (e) twisting response for different applied voltages. 

does not take into account electrochemical 

currents and therefore does 

not provide correct voltage gradient on 

the electrodes for higher applied voltages. 

It must be noted that the hydrodynamic 

effects are not considered in 

the calculations due to the low frequency 

and rather small twisting amplitude. 

Conclusions 

IPMC AMs are suited for creating 

artificial fish-like propulsors that can 

mimic the undulation, flapping, and 

complex motions of fish fins. A newly 

developed IPMC AM fin withpatterned 

electrodes was introduced for 

realizing multiple degrees-of-freedom 

motion, such as bending and twisting. 

These twistable AM fins are suited for 

creating artificial fish-like propulsors 

that can mimic the undulatory, flapping, 

and complex motions of fish 

fins as well as novel control surfaces 

for applications in a wide spectrum of 

micro-autonomous robots and marine 

systems. The masking and surface machining 

fabrication process are viable 

approaches to create a twistable AM 

fin. Experimental characterization 

showed that peak twisting for the 

sectored-electrode IPMC exceeded 7°. 

A finite-element bimorph beam 

model was used to predict the bending 

and twisting behavior of a selected 

IPMC actuator, where good agreement 

between the measured response and 

model output was found for low 

electric fields (2 V). A full-scale 3-D 


FIGURE 16 

Calculated twisting of IPMC at t = 0.25 s (left) and at t = 0.75 s (right). The color shows y-directional displacement. 

physics-based model to simulate electromechanical 

twisting transduction 

of IPMC was developed. Comparison 

between the experimental results and 

model output for the electromechanical 

transduction indicated that the 

FIGURE 17 

Comparison of experimental (solid line) and simulated twisting angle (dash line). 

model predicts the twisting deformation 

well. The future work will explore 

complex electrode patterns, integrated 

sensing electrodes, an alternative 

approach to create a twistable fin structure 

using a soft boot, and the development 

of a prototype autonomous 

marine system powered by the twistable 

IPMC AM fin. 


The authors gratefully thank 

the financial support from the U.S. 

Office of Naval Research (grant 

N0001409102183) and Dr. Tom 

McKenna. The authors also thank 

S.M. Kim, Y.S. Jung, S. Song, 

R. Riddle, and Y. Shan for their help 

with the experiments. KJK expresses 

his special thanks to Dr. Promode 

Bandyopadhyay of the Naval Undersea 

Warfare Center (NUWC) in 

Newport, RI, for his thoughtful 

encouragement. 

Lead Authors: 

Kwang J. Kim 

Active Materials and 

Processing Laboratory 


Engineering, University 

of Nevada-Reno 

Email: kwangkim@unr.edu 


Kam K. Leang 

Electroactive Systems and 

Controls Laboratory 


Engineering, University 

of Nevada-Reno 

Email: kam@unr.edu 

References 

Anton, M. 2008. Mechanical modeling 

of IPMC actuators at large deformations. 

Ph.D. Thesis, University of Tartu. 

Akle, B.J., Leo, D.J., Hickner, M. A., & 

McGrath, J.E. 2005. Correlation of capacitance 

and actuation in ionomeric polymer 

transducers. J Mater Sci. 40:3715-24. 

doi: 10.1007/s10853-005-3312-x. 

Alici, G., Spinks, G.M., Madden, J.D., 

Wu, Y., & Wallace, G.G. 2008. Response 

characterization of electroactive polymers as 

mechanical sensors. IEEE/ASME T Mech. 

13(2):187-96. 

Aureli, M., Kopman, V., & Porfiri, M. 

2010a. Free-locomotion of underwater 

vehicles actuated by ionic polymer metal 

composites. IEEE/ASME T Mech. 

15(4):603-14. 

Aureli, M., Lin, W., & Porfiri, M. 2009. 

On the capacitance-boost of ionic polymer 

metal composites due to electroless plating: 

Theory and experiments. J Appl Phys. 

105:104911. doi: 10.1063/1.3129503. 

Aureli, M., Prince, C., Porfiri, M., & 

Peterson, S.D. 2010b. Energy harvesting from 

base excitation of ionic polymer metal 

composites in fluid environments. Smart 

Mater Struct. 19:015003. doi: 10.1088/ 

0964-1726/19/1/015003. 

Ayers, J., & Witting, J. 2007. Biomimetic 

approaches to the control of underwater 

walking machines. Philos T R Soc A. 

365:273-95. doi: 10.1098/rsta.2006.1910. 

Bandyopadhyay, P.R., Beal, D.N., & 

Menozzi, A. 2008. Biorobotic insights into 

how animals swim. J Exp Biol. 211:206-14. 

doi: 10.1242/jeb.012161. 

Bhat, N.D., & Kim, W.-J. 2004. Precision 

force and position control of ionic polymer-metal 

composite. J Syst Contr Eng. 218(6):421-32. 

Biddiss, E., & Chau, T. 2006. Electroactive 

polymeric sensors in hand prostheses: Bending 

response of an ionic polymer metal composite. 

Med Eng Phys. 28(6):568-78. doi: 10.1016/ 

j.medengphy.2005.09.009. 

Bozkurttas, M., Tangorra, J., Lauder, G., & 

Mittal, R. 2008. Understanding the hydrodynamics 

of swimming: From fish fins to 

flexible propulsors for autonomous underwater 

vehicles. Adv Sci Tech. 58:193-202. 

doi: 10.4028/www.scientific.net/AST.58.193. 

Brufau-Penella, J., Puig-Vidal, M., Giannone, 

P., Graziani, S., & Strazzeri, S. 2008. Characterization 

of the harvesting capabilities of 

an ionic polymer metal composite device. 

Smart Mater Struct. 17(1):015009. 

doi: 10.1088/0964-1726/17/01/015009. 

Buchholz, J.H.J., Clark, R.P., & Smits, A.J. 

2008. Thrust performance of unsteady propulsors 

using a novel measurement system, 

and corresponding wake patterns. Exp 

Fluids. 45:461-72. doi: 10.1007/s00348- 

008-0489-1. 

Cai, Y., Bi, S., & Zheng, L. 2010. Design 

and experiments of a robotic fish imitating 

wow-nosed ray. J Bionic Eng. 7:120-126. 

doi: 10.1016/S1672-6529(09)60204-3. 

Chen, Z., Shatara, S., & Tan, X. 2010. 

Modeling of biomimetic robotic fish propelled 

by an ionic polymer-metal composite caudal 

fin. IEEE/ASME T Mech. 15:448-59. 

Chen, Z., & Tan, X. 2008. A controloriented 

and physics-based model for ionic 

polymer-metal composite actuators. IEEE/ 

ASME T Mech. 13(5):519-29. 

Chen, Z., & Tan, X. 2010. Monolithic 

fabrication of ionic polymer-metal composite 

actuators capable of complex deformation. 

Sensor Actuat A-Phys. 157:246-57. 

doi: 10.1016/j.sna.2009.11.024. 

Chen, Z., Um, T., & Bart-Smith, H. 2011. 

A novel fabrication of ionic polymer-metal 

composite actuator capable of 3-dimensional 

kinematic motions. Sensor Actuat A-Phys. 

168(1):131-9. doi: 10.1016/j. 

sna.2011.02.034. 

Fang, B.-K., Ju, M.-S., & Lin, C.-C. K. 2007. 

A new approach to develop ionic polymermetal 

composites (IPMC) actuator: Fabrication 

and control for active catheter systems. Sensor 

Actuat A-Phys. 137(2):321-9. doi: 10.1016/ 

j.sna.2007.03.024. 

Fish, F.E., Howle, L.E., & Murray, M.M. 

2008. Hydrodynamic flow control in marine 

mammals. Integr Comp Biol. 48:1-13. 

doi: 10.1093/icb/icn029. 

Guo, S., Ge, Y., Li, L., & Liu, S. 2006. 

Underwater swimming micro robot using 

IPMC actuator. In: Proceedings of the 

2006 IEEE International Conference on 

Mechatronics and Automation, June 25-28. 

pp. 249-54. Piscataway, NJ: IEEE. 

doi: 10.1109/ICMA.2006.257505. 

Guo, S., Shi, L., Ye, X., & Li, L. 2007. A 

new jellyfish type of underwater microrobot. 

In: Proc. IEEE International Conference on 

Mechatronics and Automation. pp. 509-14. 

Piscataway, NJ: IEEE. 

Heinbockel, J. 2001. Introduction to 

Tensor Calculus and Continuum Mechanics. 

Bloomington, IN: Trafford Publishing. 

Jeon, J.H., & Oh, I.K. 2009. Selective growth 

of platinum electrodes for MDOF IPMC 

actuators. Thin Solid Films. 517:5288-92. 

doi: 10.1016/j.tsf.2009.03.111. 

Jeon, J.H., Oh, I.K., Han, J.H., & Lee, 

S.W. 2007. Development of bio-mimetic 

patterned IPMC actuators with multiple 

electrodes. Key Eng Mat. 334:1005-8. 

doi: 10.4028/www.scientific.net/KEM. 

334-335.1005. 

Kamamichi, N., Yamakita, M., Asaka, K., 

& Luo, Z.-W. 2006. A snake-like swimming 

robot using IPMC actuator/sensor. 

In: IEEE International Conference on 

Robotics and Automation. pp. 1812-7. 


Kang, S., Shin, J., Kim, S.J., Kim, H.J., & 

Kim, Y.H. 2007. Robust control of ionic 

polymer-metal composites. Smart Mater 


Struct. 16(6):2457-63. doi: 10.1088/ 

0964-1726/16/6/049. 

Kato, N. 1998. Locomotion by mechanical 

pectoral fins. J Mar Sci Technol. 3:113-21. 

doi: 10.1007/BF02492918. 

Kim, B., Kim, B.M., Ryu, J., Oh, I.-H., 

Lee, S.-K., Cha, S.-E., & Pak, J. 2003. 

Analysis of mechanical characteristics of the 

ionic polymer metal composite (IPMC) 

actuator using cast ion-exchange film. In: 

Proceeding of the SPIE Smart Structures 

and Materials: Electroactive Polymer Actuators 

and Devices (EAPAD). pp. 486-95. 


Kim, B., Kim, D.-H., Jung, J.J., & Park, 

J.-O. 2005. A biomimetic undulatory tadpole 

robot using ionic polymer-metal composite 

actuators. Smart Mater Struct. 14:1579-85. 

Kim, K.J., & Shahinpoor, M. 2002. A novel 

method of manufacturing three-dimensional 

ionic polymer-metal composites (ipmcs) 

biomimetic sensors, actuators and artificial 

muscles. Polymer. 43(3):797-802. 

doi: 10.1016/S0032-3861(01)00648-6. 

Kim, K.J., & Shahinpoor, M. 2003. Ionic 

polymer-metal composites: II. Manufacturing 

techniques. Smart Mater Struct. 12(1):65-79. 

doi: 10.1088/0964-1726/12/1/308. 

Kim, K.J., & Tadokoro, S. 2007. Electroactive 

Polymers robotics Applications: Artificial 

Muscles and Sensors. London: Springer- 

Verlag. 

Kim, K.J., Yim, W., Paquette, J.W., & Kim, D. 

2007. Ionic polymer-metal composites 

for underwater operation. J Intel Mat Syst 

Str. 18(2):123-31. doi: 10.1177/ 

1045389X06063468. 

Krieg, M., & Mohseni, K. 2008. Thrust 

characterization of a bioinspired vortex ring 

thruster for locomotion of underwater 

robots. IEEE J Oceanic Eng. 33(2):123-32. 

doi: 10.1109/JOE.2008.920171. 

Krishen, K. 2009. Space applications for ionic 

polymer-metal composite sensors, actuators, 

and artificial muscles. Acta Astronaut. 

64(11-12):1160-6. 

Kruusamae, K., Brunetto, P., Graziani, S., 

Punning, A., Di Pasquale, G., & Aabloo, A. 

2009. Self-sensing ionic polymer-metal 

composite actuating device with patterned 

surface electrodes. Polym Int. 59(3):300-4. 

doi: 10.1002/pi.2752. 

Lauder, G.V. 2007. How fish swim: Flexible 

fin thrusters as an EAP platform. In: Proceedings 

of the SPIE Smart Structures and 

Materials: Electroactive Polymer Actuators 

and Devices (EAPAD). pp. 652402. San 

Diego, CA: SPIE. 

Lavu, B.C., Schoen, M.P., & Mahajan, A. 

2005. Adaptive intelligent control of ionic 

polymer metal composites. Smart Mater 

Struct. 14(4):466-74. doi: 10.1088/ 

0964-1726/14/4/002. 

Lee, S.J., Han, M.J., Kim, S.J., Jho, J.Y., Lee, 

H.Y., & Kim, Y.H. 2006. A new fabrication 

method for ipmc actuators and application to 

artificial fingers. Smart Mater Struct. 15: 

1217-24. doi: 10.1088/0964-1726/15/5/008. 

Leo, D.J., Farinholt, K., & Wallmersperger, T. 

2005. Computational models of ionic 

transport and electromechanical transduction 

in ionomeric polymer transducers. In: Proceedings 

of the SPIE Smart Structures and 

Materials: Electroactive Polymer Actuators 

and Devices (EAPAD). pp. 170-81. 

Menozzi, A., Leinhos, H.A., Beal, D.N., & 

Bandyopadhyay, P.R. 2008. Open-loop control 

of a multifin biorobotic rigid underwater 

vehicle. IEEE J Oceanic Eng. 33(2):59-68. 

doi: 10.1109/JOE.2008.918687. 

Moheimani, S.O.R., & Fleming, A.J. 2006. 

Piezoelectric Transducers for Vibration Control 

and Damping (Advances in Industrial 

Control). London: Springer. 

Multiphysics, C. 2011. User’s Guide, 

version 4.1 a. COMSOL AB. Los Angeles, CA. 

Nam, J.D., Choi, H.R., Tak, Y.S., & Kim, 

K.J. 2003. Novel electroactive, silicate 

nanocomposites prepared to be used as 

actuators and artificial muscles. Sensor 

Actuat A-Phys. 105:83-90. 

Nemat-Nasser, S., & Jiang, Y.L. 2000. 

Electromechanical response of ionic 

polymer-metal composites. J Appl Phys. 

87(7):3321-31. doi: 10.1063/1.372343. 

Nguyen, V.K., Lee, J.W., & Yoo, Y. 2007. 

Characteristics and performance of ionic 

polymer-metal composite actuators based 

on nafion/layered silicate and nafion/silica 

nanocomposites. Sensor Actuat B. 120: 

529-37. doi: 10.1016/j.snb.2006.03.015. 

Pak, J.J., Kim, J., Oh, S.W., Son, J.H., 

Cho, S.H., Lee, S.-K., … Kim, B. 2004. 

Fabrication of ionic-polymer-metal-composite 

(IPMC) micropump using a commercial 

nafion. In: Proceedings of the SPIE Smart 

Structures and Materials: Electroactive Polymer 

Actuators and Devices (EAPAD). pp. 272-80. 


Porfiri, M. 2009. Influence of electrode 

surface roughness and steric effects on the 

nonlinear electromechanical behavior of 

ionic polymer metal composites. Phys Rev E. 

79(4):041503. doi: 10.1103/ 

PhysRevE.79.041503. 

Pugal, D., Jung, K., Aabloo, A., & Kim, 

K.J. 2010a. Ionic polymer metal composite 

mechanoelectrical transduction: Review and 

perspectives. Polym Int. 59(3):279-89. 

doi: 10.1002/pi.2759. 

Pugal, D., Kim, K.J., & Aabloo, A. 2010b. 

Modeling the transduction of IPMC in 

3D configurations. In: Proceedings of the 

SPIE Smart Structures and Materials and 

Nondestructive Evaluation and Health 

Monitoring. pp. 76441T. San Diego, CA: 

SPIE. 

Pugal, D., Kim, K.J., Punning, A., Kasemagi, 

H., Kruusmaa, M., & Aabloo, A. 2008. 

A self-oscillating ionic polymer-metal composite 

bending actuator. J Appl Phys. 

103(8):084908-6. doi: 10.1063/1.2903478. 

Ramo, S. 1939. Currents induced by 

electron motion. Proc IRE. 27(9):584-5. 

doi: 10.1109/JRPROC.1939.228757. 

Richardson, R.C., Levesley, M.C., Brown, 

M.D., Hawkes, J.A., Watterson, K., & 

Walker, P.G. 2003. Control of ionic polymer 

metal composites. IEEE/ASME T Mech. 

8(2):245-3. 


Riddle, R.O., Jung, Y., Kim, S.-M., Song, S., 

Stolpman, B., Kim, K.J., & Leang, K.K. 

2010. Sectored-electrode IPMC actuator for 

bending and twisting motion. In: Proceedings 

of the SPIE Smart Structures and Materials 

and Nondestructive Evaluation and Health 

Monitoring. San Diego, CA: SPIE. 

Shahinpoor, M., Bar-Cohen, Y., Simpson, 

J.O., & Smith, J. 1998. Ionic polymer-metal 

composites (IPMCs) as biomimetic sensors, 

actuators and artificial muscles—a review. 

Smart Mater Struct. 7(6):R15-30. 

doi: 10.1088/0964-1726/7/6/001. 

Shahinpoor, M., & Kim, K.J. 2001. Ionic 

polymer-metal composites: I. Fundamentals. 

Smart Mater Struct. 10:819-33. 

doi: 10.1088/0964-1726/10/4/327. 

Shahinpoor, M., & Kim, K.J. 2005. Ionic 

polymer-metal composites: IV. Industrial and 

medical applications. Smart Mater Struct. 

14(1):197-214. doi: 10.1088/0964- 

1726/14/1/020. 

Shan, Y., & Leang, K.K. 2009. Frequencyweighted 

feedforward control for dynamic 

compensation in ionic polymer-metal 

composite actuators. Smart Mater Struct. 

18(12):125016. doi: 10.1088/0964-1726/18/ 

12/125016. 

unmanned underwater vehicles. IEEE 

J Oceanic Eng. 32(3):533-50. 

doi: 10.1109/JOE.2007.903362. 

Tiwari, R., Kim, K.J., & Kim, S. 2008. Ionic 

polymer-metal composite as energy harvesters. 

Smart Struct Syst. 4(5):549-63. 

Tortora, G., Caccavaro, S., Valdastri, P., 

Menciassi, A., & Dario, P. 2010. Design of an 

autonomous swimming miniature robot based 

on a novel concept of magnetic actuation. 

In: IEEE International Conference on 

Robotics and Automation. pp. 1592-7. 

Piscataway, NJ: IEEE. doi: 10.1109/ 

ROBOT.2010.5509599. 

Yim, W., Lee, J., & Kim, K.J. 2007. An 

artificial muscle actuator for biomimetic 

underwater propulsors. Bioinspir Biomim. 

2:S31-41. doi: 10.1088/1748-3182/2/2/S04. 

Yoon, W.J., Reinhall, P.G., & Seibel, E.J. 

2007. Analysis of electro-active polymer 

bending: A component in a low cost 

ultrathin scanning endoscope. Sensor Actuat 

A-Phys. 133(2):506-17. doi: 10.1016/ 

j.sna.2006.04.037. 

Shockley, W. 1938. Currents to conductors 

induced by a moving point charge. J Appl 

Phys. 9(10):635-6. doi: 10.1063/1.1710367. 

Smela, E., Kallenbach, M., & Holdenried, J. 

1999. Electrochemically driven polypyrrole 

bilayers for moving and positioning bulk 

micromachined silicon plates. IEEE/ASME 

T Microelectromech Syst. 8(4):373-83. doi: 

10.1109/84.809051. 

Tadokoro, S., Yamagami, S., & Takamori, T. 

2000. An actuator model of ICPF for robotic 

applications on the basis of physicochemical 

hypotheses. In: IEEE International Conference 

on Robotics and Automation. pp. 1340-6. 


Tangorra, J.L., Davidson, S.N., Hunter, 

I.W., Madden, P.G.A., Lauder, G.V., Dong, 

H., … Mittal, R. 2007. The development 

of a biologically inspired propulsor for 


PAPER 

Batoid Fishes: Inspiration for the Next 

Generation of Underwater Robots 

AUTHORS 

Keith W. Moored 

Mechanical and Aerospace 

Engineering Department, 

Princeton University 

Frank E. Fish 

Department of Biology, 

University of West Chester 

Trevor H. Kemp 

Hilary Bart-Smith 


Engineering Department, 

University of Virginia 

Introduction 

ABSTRACT 

For millions of years, aquatic species have utilized the principles of unsteady 

hydrodynamics for propulsion and maneuvering. They have evolved high-endurance 

swimming that can outperform current underwater vehicle technology in the areas 

of stealth, maneuverability and control authority. Batoid fishes, including the manta 

ray, Manta birostris, the cownose ray, Rhinoptera bonasus, and the Atlantic stingray, 

Dasyatis sabina, have been identified as a high-performing species due to their 

ability to migrate long distances, maneuver in spaces the size of their tip-to-tip 

wing span, produce enough thrust to leap out of the water, populate many underwater 

regions, and attain sustained swimming speeds of 2.8 m/s with low flapping/ 

undulating frequencies. These characteristics make batoid fishes an ideal platform 

to emulate in the design of a bio-inspired autonomous underwater vehicle. The 

enlarged pectoral fins of each ray undergoes complex motions that couple spanwise 

curvature with a chordwise traveling wave to produce thrust and to maneuver. Researchers 

are investigating these amazing species to understand the biological principles 

for locomotion. The continuum of swimming motions—from undulatory to 

oscillatory—demonstrates the range of capabilities, environments, and behaviors 

exhibited by these fishes. Direct comparisons between observed swimming motions 

and the underlying cartilage structure of the pectoral fin have been made. A 

simple yet powerful analytical model to describe the swimming motions of batoid 

fishes has been developed and is being used to quantify their hydrodynamic performance. 

This model is also being used as the design target for artificial pectoral 

fin design. Various strategies have been employed to replicate pectoral fin motion. 

Active tensegrity structures, electro-active polymers, and fluid muscles are three 

structure/actuator approaches that have successfully demonstrated pectoral-finlike 

motions. This paper explores these recent studies to understand the relationship 

between form and swimming function of batoid fishes and describes attempts 

to emulate their abilities in the next generation of bio-inspired underwater vehicles. 

Keywords: biomimicry, bioinspired, autonomous underwater vehicle, manta ray, 

tensegrity structures 

There has been an explosion of activity 

in the area of biomimicry and 

bioinspired engineering research. Biomimicry 

directly emulates the form 

and function of species to illuminate 

the physical principles behind nature’s 

designs. Bioinspired engineering takes 

advantage of the knowledge gained 

through biomimetic studies to judiciously 

apply novel physical principles 

to develop solutions with added functionality 

over conventional engineering 

approaches. Biology, biomimetics 

and bioinspired engineering are intimately 

linked. Thus it comes as no surprise 

that biologists and engineers are 

collaborating by developing biorobotic 

devices to: (1) elucidate key insights 

into biological form and function and 

(2) develop bioinspired autonomous 

underwater vehicles (BAUVs) to improve 

functionality of autonomous 

underwater vehicles (AUVs). 

Aquatic species outperform conventional 

AUVs in the areas of maneuverability 

and control authority 

(Bandyopadhyay, 2005), while having 

a low-noise signature that blends into 

the background and high swimming 

efficiencies. Batoid rays excel in all 

of these areas, giving them an abundance 

of recent attention. The focus 

of this paper is to present the growing 

body of work being done to understand 

and quantify the swimming performance 

of batoid fishes (i.e., skates, 

sting rays, manta rays) and the stateof-the-art 

in robotic mimicry. Of particular 

interest are the mechanisms 

associated with the swimming of 

these fishes, which employ flattened 

pectoral fins to propel and maneuver 

in the ocean and in rivers. 

Understanding Biological Foundation 

will discuss our current biological 


understanding of batoid rays. Rationale 

for Mimicking Batoid Rays will 

present compelling reasons for scientists 

and engineers to study batoids 

rays, highlighting their swimming 

characteristics that would be desired 

in an underwater vehicle. Bioinspired 

Robotics delves into the expanding 

worldofbio-inspiredunderwatervehicles, 

with particular emphasis on 

ray-like platforms. Concluding Comments 

and Future Directions concludes 

with a discussion on critical areas that 

need to be addressed in order for the 

next generation of underwater vehicles 

to be truly bioinspired. 

Understanding Biological 

Foundation 

Fish swim by imparting momentum 

to water from the movements 

of a variety of propulsors, which can 

include the body, median fins, and 

paired fins (Sfakiotakis et al., 1999). 

Although primitive batoid fishes use 

the body and caudal fin toswim, 

more advanced batoids have become 

specialized to swim with enlarged pectoral 

fins. It is emphasized that pectoral 

fin locomotion can have significant 

advantages in maneuvering and stationkeeping 

(Sfakiotakis et al., 1999). In 

recent years, more attention is being 

paid to pectoral fin hydrodynamics as 

their importance is being realized in not 

only steady-state locomotion but also in 

transient maneuvers (Bandyopadhyay, 

FIGURE 1 

2005; Lauder et al., 2002; Combes & 

Daniel, 2001; Palmisano et al., 2007). 

Batoid rays take pectoral fin locomotion 

to an evolutionary extreme 

(Figure 1). Rays have a dorso-ventrally 

flattened body with enlarged pectoral 

fins that are seamlessly merged with 

their body to form a biological blended 

wing-body configuration. Propulsive 

waves are passed through the fins by 

serial contraction of the appendicular 

musculature. The waves have their 

greatest amplitude toward the periphery 

of the fin. 

Even though among rays there is 

similar morphology, their locomotor 

strategies can be very different. Undulatory 

motion, defined as having 

greater than one or more waves present 

on a fin (Rosenberger, 2001), is one 

extreme of kinematic motion and was 

termed ‘rajiform’ by Breder (1926). 

These fishes swim just over the ocean 

floor. The other extreme is oscillatory 

motion, defined as having less than 

half of a wave present (flapping) on a 

fin, and was coined ‘mobuliform’ by 

Webb (1994). The mobuliform swimming 

mode appears as a wing-like flapping 

motion and is associated with rays 

that have a more pelagic existence. The 

various species of batoids exhibit a 

continuum of kinematic motions between 

the two extremes of undulation 

and oscillation (Rosenberger, 2001). 

Myliobatoids, i.e., the mid-water 

rays, including manta, eagle, bat, and 

(a) Image of a manta ray. (b) Image of a cownose ray. Both rays are part of the batoid family and 

swim via an oscillatory motion. 

cownose rays, nearly exclusively utilize 

oscillatory motion (Klausewitz, 1964; 

Sasko et al., 2006; Heine, 1992). 

Some research has been done to characterize 

the biology and behavior of 

myliobatoids (Schaefer & Summers, 

2005; Summers, 2000). Heine (1992) 

studied the kinematics of the cownose 

ray by videotaping live rays swimming 

in a flow tank. Rosenberger (2001) 

compared the kinematics of many 

batoid rays spanning the undulationoscillation 

continuum and suggested 

that oscillatory rays have evolved to 

have efficient locomotion. Klausewitz 

(1964) describes the kinematic motions 

of the manta ray while Moored 

et al. (Moored, 2010; Moored et al., 

2011b) developed a simple yet powerful 

analytical model to quantify the kinematics 

of different species of batoid 

rays (such as the manta ray, Atlantic 

stingray, and the cownose ray). This 

model is used as a target deformation 

field for a bio-inspired fin (Moored 

et al., 2011b) and to calculate the 

swimming performance of different 

batoid ray species (Moored et al., 

2011b; Pederzani et al., 2011). 

Morphometrics 

The greatly enlarged pectoral fins 

form wide lateral extensions of the 

body that range in morphology from 

a circular disc to triangular, wing-like 

planforms. Species of batoids show 

overa90-foldrangeinsizewiththe 

largest being the manta (Manta birostris). 

Rays that swim by undulations 

of the fin intherajiformmodehave 

fin shapes with relatively low aspect 

ratios (the ratio of span to chord). Oscillatory 

swimmers, using the mobuliform 

mode, possess higher aspect ratio 

fins with longer spans. 

The cross-sectional geometry of 

batoid rays has a streamlined appearance. 

Rajiform swimmers have a 


ody and pectoral fins with a flattened 

ventral side and low vaulted dorsum, 

giving a design similar to a cambered 

wing. Although the central portion of 

the body shows a slight asymmetry 

with a flattened ventral surface and 

convex dorsal surface, the pectoral 

fins of mobuliform swimmers display 

symmetrical cross-sectional profiles 

reminiscent of engineered foils 

(Abbott & von Doenhoff, 1949). 

The internal skeleton of the pectoral 

fins of batoids are composed of numerous 

short, cylindrical cartilaginous 

elements (Heine, 1992; Schaefer & 

Summers, 2005). These cartilaginous 

elements are the supportive radials of 

the fin. The radials are stacked end to 

end. The radial cartilages are mineralized 

to varying degree depending on 

the species of batoid, where the mineralization 

is found on the exterior 

of the cartilaginous element with the 

core being unmineralized (Schaefer & 

Summers, 2005). Rajiform swimming 

rays display joint staggering with little 

calcification of the joints, whereas 

the skeleton of oscillatory swimmers 

shows cross-bracing and calcification. 

The skeleton is moved by long thin 

muscles that run from the expanded 

pectoral girdle along each fin rayto 

every radial. The range of motion of 

the articulated radials is small (∼15°), 

but the large number of components 

in the pectoral fins permits sufficient 

spanwise and chordwise flexibility 

for propulsion and maneuvering 

(Rosenberger, 2001; Klausewitz, 

1964; Heine, 1992; Schaefer & 

Summers, 2005). 

Rationale for Mimicking 

Batoid Rays 

With respect to pursuing bioinspired 

engineering, a key question 

to be answered is to explain/justify 

why a particular species is a good candidate 

to emulate. These reasons can 

be very diverse and are motivated by 

the particular application envisioned. 

In the case of AUVs, compelling reasons 

to consider biology as a starting 

point for the development of the next 

generation vehicle are (1) a stealthy 

signature, (2) high efficiency and economy, 

(3) expanded working environment, 

and (4) scalability/payload 

capacity. Additionally, a key justification 

for this approach is that there are 

tangible improvements that can be 

made over current AUV technologies. 

The pool of species to emulate is vast. 

However, recent studies of batoid rays 

have demonstrated significant swimming 

abilities that would be desirable 

in an underwater vehicle. 

Stealth means either quiet operation 

or the ability to blend into the 

background noise. A biomimetic approach 

naturally fulfills these requirements 

by creating a vehicle whose 

minimal noise signature blends in 

with the environment. The noise signature 

of a fish is very different to 

that of a propeller (Bandyopadhyay, 

2005). Even biomimetic sensor arrays 

such as an artificial lateral line (Yang 

et al., 2006) or artificial seal whiskers 

(Stocking et al., 2010) that are sensitive 

to hydrodynamic wake signatures 

would presumably delineate a flapping 

fin wake as an animal and a propeller 

wake as a man-made device. Given recent 

advances in underwater imaging 

technology including LIDAR systems 

(Jaffe et al., 2001), synthetic aperture 

sonar (Kocak & Caimi, 2005) and 

biomimetic sonar systems (Dobbins, 

2007), the shape and movements of 

an AUV are becoming increasingly important. 

By mimicking the body form 

of an aquatic species, the identification 

of such a stealthy vehicle as a manmade 

device becomes difficult. 

Batoid rays offer an intriguing design 

solution for a high-endurance vehicle. 

Pelagic rays, such as the manta 

ray or cownose ray, migrate thousands 

of miles a year. This suggests that these 

species have evolved to become highendurance 

swimmers. As discussed 

previously, myliobatoid rays have a 

dorso-ventrally flattened body with 

enlarged pectoral fins, forming a natural 

gliding morphology. In terms of 

avehicle,abatoid-inspiredUVisan 

advance on current underwater glider 

technology. This bioinspired platform 

wouldenableavehicletohavehighendurance 

capabilities like current underwater 

gliders, such as the Slocum 

AUV (Webb & Simonetti, 1999). Additionally, 

this system has the potential 

to transition to a faster, more maneuverable 

vehicle that can operate in dynamic 

environments such as the 

littoral zone or areas with large currents 

and high wave action. Observations 

of various rays show them to be 

highly maneuverable and adaptable 

to local conditions—for example, to 

station keep and even swim backwards. 

Their ability to control their 

stability via the pectoral fins, especially 

when compensating for challenging 

environments, must also be considered 

as a desirable characteristic to emulate 

in an underwater vehicle. 

Scalability is an attractive feature 

in any artificial system. With respect 

to bioinspired underwater vehicles, batoids 

display an extraordinary range of 

dimensions, growing in excess of 

9 m tip-to-tip in the case of manta 

rays.Thus,thesizeandspeedthat 

batoids perform at are equivalent to 

the operation range of marine vehicles. 

The size of the vehicle will very much 

depend on the mission requirements, 

but by using the batoid as the foundation, 

it is feasible to produce a variety 

of sized vehicles that can explore and 


traverse a wide range of ocean space 

while performing a wide range of tasks. 

Moreover, a batoid-inspired vehicle 

would have a large planform surface 

area, making this platform an excellent 

candidate for flexible solar cells to extend 

its range (Dennler et al., 2008), 

similar to the solar powered SAUV II 

vehicle (Jalbert et al., 2003). In addition, 

the rigid body of batoids permits 

space for control systems, sensory 

devices and increased payload. 

Bioinspired Robotics 

There has been growing use of 

AUVs in recent years with over 

240 different AUV platforms developed 

and used in the field (Bandyopadhyay, 

2005). These AUVs typically are built 

for reconnaissance/surveying and were 

originally designed for endurance 

(Blidberg, 2001). This gave rise to 

the design of underwater gliders 

using conventional design principles 

(i.e., steady-state hydrodynamics) 

that have high endurance but little maneuverability 

(Webb & Simonetti, 

1999). From another perspective, 

biology has created thousands of swimming 

platforms that can outmaneuver 

the best AUVs while still 

having highly efficient, high-speed 

and high-endurance performance. 

Moreover, many of these biological 

systems can also hover in place with no 

forward locomotion, generate large 

enough forces to hold station under 

adverse environmental conditions, 

burst with incredible acceleration and 

have a significantly reduced noise signature 

compared to man-made AUVs 

(Fish & Lauder, 2006). In an attempt 

to bridge the performance gap between 

conventional AUVs and biological 

systems, engineers have been 

shifting focus to BAUVs, which is a 

highly multi-disciplinary research area 

(Bandyopadhyay, 2005; Colgate & 

Lynch, 2004). To reach some of these 

goals, there is a spectrum of first generation 

BAUVs that have been developed. 

Some form an exotic collection 

mimicking lamprey (Ayers et al., 

2000; Crespi et al., 2004), tuna 

(Barrett et al., 1996; Yu et al., 2004; 

Anderson & Chhabra, 2002), and dolphins 

(Yu et al., 2007), while others 

aremoreconventionalstyleAUV 

designs outfitted with bioinspired flapping 

propulsors (Fish et al., 2003; Low 

&Willy,2006;Listaketal.,2005; 

Borgen et al., 2003; Mojarrad, 2000; 

Licht et al., 2004). These different 

BAUV designs were made possible 

partly from advances in our understanding 

of unsteady hydrodynamics 

and the biology of nektonic (swimming) 

organisms. However, this BAUV 

technology still has a long way to go 

before the performance gap is bridged. 

Recently, researchers have turned 

to pectoral fin locomotion for inspiration. 

Pectoral fin motions utilized by 

sunfish, perch, bass and bird wrasse 

for low-speed swimming and maneuvering 

have been studied (Gibb et al., 

1994; Drucker & Jensen, 1997; 

Lauder & Jayne, 1996; Walker & 

Westneat, 1997). To understand the 

forces and moments produced by 

these pectoral fins, biorobotic solutions 

began with paddle-like fins that 

mimic the bulk kinematics of labriform 

swimming (Kato, 1998; Kato & 

FIGURE 2 

A biomimetic sunfish pectoral fin (Tangorra et al., 2008a, 2008b). 

Furushima, 2002). In recent years, devices 

have been constructed to more 

accurately replicate the kinematics 

utilized by the fish with the advent of 

actively flexible fins that can produce 

chordwise undulations as well as 

spanwise curvature (Yan et al., 2010; 

Palmisano et al., 2008; Kato et al., 

2008; Tangorra et al., 2007). Actively 

flexible fins deform due to the presence 

of actuators instead of undergoing 

rigid body motions, like the heave 

and pitch of oscillating airfoils. The 

motion of actively flexible fins is fully 

prescribed. In contrast, passively flexible 

fins deform under fluid loading 

such that their motion is not fully prescribed, 

but a function of the forces applied 

to the fin. Tangorra et al. (2008a, 

2008b) advanced their artificial pectoral 

fin (Figure 2) by not only matching 

the kinematics of the sunfish but also 

replicating the internal fin structure 

and material properties, which allowed 

the fin to have a greater degree of passive 

flexibility. This fin wasusedto 

fully characterize how sunfish produce 

and manipulate fluid forces to propel 

themselves and maneuver. With 

equivalent passive flexibility as the 

fins of the sunfish, the artificial fin 

was able to produce thrust on both 

the outstroke and instroke of its fin 

beat, as observed of the animal. 

An excellent example demonstrating 

the link between biology, biomimetics, 

and bioinspired engineering 


is in the development of an artificial 

ghost knifefish (Curet et al., 2010). 

Through observation of biology, a 

biorobotic device was developed and 

used to understand the locomotion 

strategies of this fish. Particle image 

velocimetry, in conjunction with computational 

fluid dynamics, were employed 

to explore the propulsive and 

station-keeping characteristics of this 

fascinating fish. 

Batoid-Inspired Devices 

There have also been attempts by 

researchers to develop batoid-inspired 

fins and AUVs. These devices mimic 

both the undulatory swimming seen 

inbenthicrays(similartothelocomotion 

of the ghost knifefish) as well 

as the oscillatory swimming seen in pelagic 

rays, such as the manta. Many of 

these batoid-inspired devices are used 

as a platform for exploring actuation 

technologies. 

Motors and servomotors are used 

in ray-like devices due to their simple 

controllability, high-speed operation 

and repeatability. Some motor-driven 

devices mimic undulatory rays (Low 

& Willy, 2006; V. y Alvarado et al., 

2010), while others mimic oscillatory 

rays (Yang et al., 2009; Zhou & 

Low, 2010; Gao et al., 2007). Researchers 

have developed oscillatory 

ray-like vehicles based on pneumatic 

pectoral fins (Brower, 2006; Cai 

et al., 2010; Suzumori et al., 2007). 

Sfakiotakis et al. (2005) also used 

pneumatically driven “fin rays” to produce 

an undulatory ray-like device. 

Festo (2008) has built a BAUV called 

AquaRay, utilizing fluidic muscles. 

This robot uses an oscillatory flapping 

motion to swim, but no quantitative 

data on the performance is given. 

Takagi et al. (2007) utilized ionic 

polymer-metal composites (IPMCs) 

actuators to develop a stingray-like 

device that could achieve a swimming 

speed of 0.24 BL/s. Chen et al. (2011) 

developed a novel fabrication method 

to produce IPMCs that can deform 

with complex three-dimensional kinematics. 

This fabrication technology 

was used to produce a manta ray-like 

device. Shape memory alloys have 

also been employed in the design of artificial 

pectoral fins (Yong-hua et al., 

2007; Wang et al., 2008). Wang 

et al. (2008) presented a robotic squid 

utilizing a rajiform mode of swimming 

to achieve 0.24 BL/s swimming 

speed. The best swimming speed performance 

of these actuator platforms 

was 1.4 BL/s achieved by the servomotor 

driven devices; however, the 

associated power cost is not given. 

These studies showcase the plethora 

of actuator technologies that can 

be utilized to produce deformations 

similar to that of rays. One concern 

with this approach is that the actuator 

choice is directly coupled with the fin 

technology. An alternative approach is 

to start with a fin design that is actuator 

independent and so the choice of 

actuator is dependent on the application 

of the vehicle. For instance, if actuator 

efficiency is not a concern but 

noiseless operation is of prime importance, 

an SMA actuator could be chosen. 

Also, this approach opens up the 

possibility of replacing current actuators 

with new technologies that may 

be superior. Solutions like this are discussed 

next. 

In a study to increase our understanding 

of the hydrodynamics of 

batoid locomotion, Clark and Smits 

(2006) designed and built an active 

artificial oscillating fin that was independent 

of the actuator. They quantified 

the performance of the fin by 

measuring the efficiency and thrust 

production, as well as determining an 

optimal traveling wave wavelength. 

Furthermore, by using dye flow visualization, 

they characterized the wake 

structure as a series of interacting trailing 

edge vortices forming a threedimensional 

reverse von Kármán 

vortex street. In free swimming tests 

(Moored et al., 2011a), a swimming 

speed of 2 BL/s and a swimming economy, 

ζ (ζ = U = P f ,whereU is the 

swimming speed and P f is the power 

consumption), of 0:132 BL/J was 

reached for an actively flexible single 

fin. When some passive flexibility 

was introduced, the swimming speed 

dropped to 1:7 BL/s while the economy 

rose to 0:18 BL/J at the same flapping 

frequency of 2 Hz. This work 

has also highlighted the prime importance 

of the traveling wave in ray-like 

propulsion. 

Another actuator-independent approach 

has been developed using active 

tensegrity structures. Tensegrity structures 

are truss-like structures where 

some of the rigid elements have been 

replaced by cable elements (Figure 3). 

The cable elements must be in a 

state of tension for the structure to 

have integrity, giving rise to the contraction 

of “tensional-integrity” to 

tensegrity. Tensegrity structures act 

as a “skeleton-tendon” foundation 

that can use any actuator type to support 

the generation of large loads, 

match the kinematics of batoid rays, 

and perform with minimal actuation 

energy (Moored et al., 2011b; Moored 

& Bart-Smith, 2009). 

Various tensegrity actuation strategies 

are explored that are capable of 

matching the key kinematic features 

of batoid-propulsion: a chordwise 

traveling wave coupled with a large 

amplitude curved spanwise deformation. 

The strategies involve either 

embedding the actuators into the tensegrity 

structure (embedded actuation) 

or migrating the actuators outside of 


FIGURE 3 

(a) Three-dimensional tensegrity structures (three, four, and six strut prismatic structures). (b) Tensegrity-based fin concept. The fin can deform 

with coupled curved spanwise motion and chordwise undulation to mimic the kinematics of the manta ray and the batoid family in general. The 

tensegrity deforms when active elements contract or expand. 

the structure (remote actuation). With 

respect to embedded actuation, optimal 

solutions have been calculated 

that give the location and actuation 

strain necessary to match a target displacement 

field (Moored & Bart- 

Smith, 2007). However, embedded 

actuation is problematic, as it requires 

many actuators to match the complex 

ray kinematics, adds mass to the 

active structure (thereby requiring 

more power to flap) and limits the 

scalability of solutions to the size of 

the actuator. Remote actuation overcomes 

these limitations by placing 

the actuators outside of the active region 

and connecting to the structure 

via a routed cable. A general numerical 

model––applicable to any topology 

and any actuation strategy––has 

been derived (Moored & Bart-Smith, 

2009). 

Moored et al. (2011) derive analytical 

solutions for active planar tensegrity 

beam structures. These solutions coupled 

with the numerical solution are 

utilized to identify optimal stiffnessto-mass 

and strength-to-mass strategies. 

Structural performance metrics 

were calculated showing that the fin 

structure can closely match the kinematics 

of the manta ray, under external 

loading, using open-loop actuation of 

four actuators remotely located outside 

of the active structure (Moored et al., 

2011b). In an attempt to simplify the 

experimental design of an artificial 

fin, actuated via remote actuation, a 

single tensegrity beam was built 

andplacedwithinanelastomerfin. 

Figure 4a shows images of a single 

tensegrity beam as it is actuated. The 

beam enables leading-edge actuation 

of the artificial fin (Figure4b).This 

fin was then tested in a flow tank to 

observe the influence of frequency on 

the wake topology. Figures 4c and 4d 

show the actuating fin inwaterfrom 

the side and below. The black lines 

superimposed on these images represent 

the kinematical model for the kinematics 

of a cownose ray—note the 

excellent agreement between experiment 

and theory, especially with the 

passive response of the elastomer fin. 

This approach costs minimal power 

consumption and shows the simple design 

of a high-performance tensegritybased 

artificial pectoral fin. 

Concluding Comments 

and Future Directions 

The idea to look to nature for inspiration 

is not new, and this rich arena 


FIGURE 4 

Example of a tensegrity-based actuating fin. (a) Photos of a tensegrity beams employing remote 

actuation. (b) Dorsal view of elastomer fin with leading edge actuation via tensegrity beam. 

(c) Posterior view of actuating fin shown in (b). (d) Lateral view of actuating fin. Note the lines 

in (c) and (d) represent the mathematical model derived to describe kinematic motions of a manta 

ray pectoral fin. 

continues to be a source for engineers 

to aid in solving challenging problems. 

From Leonardo Da Vinci’s flying machine, 

Helical Air Screw, to leading 

edge whale-like tubercles on wind turbine 

blades to improve efficiencies 

(Fish et al., 2011). The opportunities 

to learn from nature and emulate its 

unique approach to overcoming challenges 

seem endless. In this paper, we 

have touched upon the challenge of 

creating efficient, economic, and maneuverable 

underwater swimming 

platforms. We have focused on batoid 

fishes for inspiration in the design of 

the next generation of bioinspired underwater 

vehicles, as emulation of its 

swimming characteristics—efficiency, 

maneuverability, stealth, working 

environments, scalability—has the 

potential to significantly improve 

upon current state-of-the-art in AUV 

technologies. 

The development of a batoidinspired 

underwater vehicle can be 

classified in terms of the approach 

taken to achieve ray-like swimming. 

The first is developing a batoidinspired 

AUV that performs as a platform 

to test the capabilities of a variety 

of traditional and novel actuating technologies. 

The motivation here is to 

demonstrate the capabilities of such 

devices—usually in terms of force 

and stroke—and is not necessarily a 

desire to truly replicate the biological 

system. These actuators include 

electroactive polymers, fluidic muscles, 

shape memory alloys, motors 

and servomotors. Of particular interest 

in testing these actuators has been the 

challenge of quantifying the swimming 

performance of the particular 

vehicle, many of which do not necessarily 

mimic the kinematics of batoid 

fishes. As biology has limitations due 

to the materials available to construct 

abodyandtheevolutionaryprocess 

that produces an organism, possible 

improvements to the basic body plan 

can perhaps be engineered to enhance 

performance beyond the capabilities of 

nature. 

The second approach considers fundamental 

questions associated with 

biology’s solutions to propulsion, 

maneuverability, stability, and stealth. 

Technology is used in this case to replicate 

the biology to help answer these 

questions. Using the underlying biology 

as the basis for inquiry, the 

mechanisms that dictate batoid swimming 

performance are explored. This is 

being done through the design and development 

of artificial systems—either 

real or virtual—that can achieve near 

identical kinematic motions of the rays 

being studied. Specifically, researchers 

are working to elucidate the dominant 

mechanisms in batoid swimming that 

dictate efficiency and maneuverability. 

A key outcome of this work is to fully 

explore nature’s design space and 

beyond. By mimicking biology, we 

attempt to elucidate the key features 

that control and optimize function. 

Nature evolves solutions that satisfy 

multiple constraints; engineers and 

scientists can design for a single desired 

outcome. By identifying and quantifying 

the key features/characteristics that 

dictate optimality, we can judiciously 

choose to build these into an artificial 

system, depending on the required 

functionality of the device. For example, 

one may desire a vehicle that 

can swim for as long as possible or 

as fast as possible—two different solutions 

may be necessary for these two 

requirements. 

As mentioned in Understanding 

Biological Foundation, there has 

been an extensive study of the underlying 

cartilage structure of various 


FIGURE 5 

batoid rays (Schaefer & Summers, 

2005). Biomechanical studies of the 

cartilage arrangement have been carried 

out to examine the relationship 

between the form of the underlying 

structure and its impact on the function 

(Russo et al., 2011). In this 

work, Russo et al. have taken the 

cartilage architecture and developed 

a numerical model to study the kinematic 

function of this form. This 

initial study is beginning to explore 

the relationship between form and 

function. 

One of the most exciting developments 

in the creation of these bioinspired 

devices and vehicles is in the 

development of rapid prototyping fabrication 

(Figure 5). This has opened 

the possibility to build a cartilage 

structure that uses the same design 

principles observed in nature, as described 

by Schaefer & Summers 

(2005). In this physical model, cartilage 

elements are connected in the 

spanwise direction with cross-bracing 

in the chordwise direction to mimic 

the architecture of the Atlantic stingray 

(www.bartsmithlabs.com). This 

technology enables the design to be 

quickly and easily varied so as to answer 

questions regarding the influence 

of the architecture on kinematic 

performance. The images in Figure 5 

are compelling, as they demonstrate 

kinematic capabilities and possibilities 

of such a structure. By mimicking 

the underlying structure of biology, 

we can explore the capabilities of 

these species and potentially expand 

upon them. 

Significant progress has been made, 

but there is still much to be done. Information 

on the kinematics of swimming 

is being generated, but there is 

still much to be learned, especially 

with respect to some of the more fine 

motor skills observed. Also, not much 

Example of artificial cartilage structure design using rapid-prototyping technology. The individual 

elements represent the cartilage elements found in the pectoral fins of batoid rays. The arrangement 

and connectivity are similar to portions found in the Atlantic stingray. 

is known about the hydroacoustic 

properties of the biology. Material 

properties of the constituent parts of 

the pectoral fin are needed to improve 

the fidelity biomechanical models that 

describe form and function. Lastly, 

more investigation of the sensing and 

control strategies of batoids is needed. 

This improved understanding will 

providevaluableinsightwhenmore 

sophisticated vehicles are developed. 

With regard to engineering a batoidinspired 

AUV, there are huge opportunities 

in actuator design and development. 

Structural and material design and selection 

also are areas that need to be 

addressed. For example, how do we design 

a skin that can accommodate both 

the out-of-place hydrodynamic forces 

and the potential in-plane stretching 

experienced during actuation. How 

do the properties of the artificial system 

scale with the biological properties 

Actuation technology is an area 

that has the potential to revolutionize 

the field of biomimicry and bioinspired 

engineering. 

In this paper, we have focused on 

a small subset of bio-inspired underwater 

vehicles. We have presented a 

review of work related to the development 

of a bio-inspired underwater 

robot—actuation technology integrated 

into a batoid-like vehicle and 

understanding the biological foundation 

to explore the full design space. 

It is clear though that these two categories 

are very closely related. Without 

actuator development, it may not be 

possible to achieve anything close to 

what biology achieves. But a clear picture 

of biology function is needed 

so that actuator requirements can be 

quantified. Synergy between biology, 

biomimicry, and bio-inspired engineering 

is essential if we want to 

develop the next generation of underwater 

vehicles. 



The authors would like to acknowledge 

funding from the Office of Naval 

Research through the MURI program 

on Biologically Inspired Autonomous 

Sea Vehicles (Program Manager: Dr. 

R. Brizzolara, Contract No. N00014- 

08-1-0642) and the David and Lucille 

Packard Foundation. 

References 

Abbott, I.H., & von Doenhoff, A.E. 1949. 

Theory of Wing Sections: Including a 

Summary of Airfoil Data. New York: 

McGraw-Hill. 

Ayers, J., Wilbur, C., & Olcott, C. 2000. 

Lamprey robots. In: Proceedings of 1st International 

Symposium on Aqua Bio-mechanisms, 

eds. Wu, T., & Kato, N. Honolulu, Hawaii: 

Tokai University Pacific Center. 

Anderson, J.M., & Chhabra, N.K. 2002. 

Maneuvering and stability performance of 

a robotic tuna. Integr Comp Biol. 42(1): 

118-26. doi: 10.1093/42.1.118. 

Bandyopadhyay, P.R. 2005. Trends in biorobotic 

autonomous undersea vehicles. 


doi: 10.1109/JOE.2005.843748. 

Barrett, D., Grosenbaugh, M., & Triantafyllou, 

M. 1996. The optimal control of a flexible 

hull robotic undersea vehicle propelled by 

an oscillating foil. In: Proceedings of the Symposium 

on Autonomous Underwater Vehicle 

Technology, AUV’96. 1-9. Monterey, CA: 

IEEE. doi: 10.1109/AUV.1996.532833. 

Blidberg, D.R. 2001. The development of 

autonomous underwater vehicles (AUVs): 

A brief summary. In: IEEE ICRA Conference, 

Seoul Korea, May 21-25, 2001. 

Borgen, M.G., Washington, G.N., & Kinzel, 

G.L. 2003. Design and evolution of a piezoelectrically 

actuated miniature swimming 

vehicle. IEEE-ASME T Mech. 8(1):66-76. 

doi: 10.1109/TMECH.2003.809131. 

Breder, C.M., Jr. 1926. The locomotion of 

fishes. Zoologica. 4:159-297. 

Brower, T.P.L. 2006. Design of a manta 

ray inspired underwater propulsive mechanism 

for long range, low power operation. Master’s 

thesis, Tufts University. 

Cai, Y., Bi, S., & Zheng, L. 2010. Design 

and experiments of a robotic fish imitating 

cow-nosed ray. J Bionic Eng. 7(2):120-6. 

doi: 10.1016/S1672-6529(09)60204-3. 

Chen, Z., Um, T.I., & Bart-Smith, H. 2011. 

A novel fabrication of ionic polymer-metal 

composite membrane actuator capable of 

3-dimensional kinematic motions. Sensor 

Actuat A-Phys. 168(1):131-9. doi: 10.1016/ 

j.sna.2011.02.034. 

Clark, R.P., & Smits, A.J. 2006. Thrust 

production and wake structure of a batoidinspired 

oscillating fin. J Fluid Mech. 562: 

415-29. doi: 10.1017/S0022112006001297. 

Colgate, J.E., & Lynch, K.M. 2004. 

Mechanics and control of swimming: A 

review. IEEE J Oceanic Eng. 29(3):660-73. 

Combes, S.A., & Daniel, T.L. 2001. Shape, 

apping and exion: wing and fin design for 

forward flight. J Exp Biol. 204(12):2073-85. 

Crespi, A., Badertscher, A., Guignard, A., & 

Ijspeert, A.J. 2004. An amphibious robot 

capable of snake and lamprey-like locomotion. 

In: Proceedings of the 35th International 

Symposium on Robotics ISR 2004. Paris, 

France: ISR publisher. 



with counter-propagating waves: a novel 


8(60):1041-50. 

Dennler, G., Scharber, M., Ameri, T., 

Denk, P., Forberich, K., Waldauf, C., & 

Brabec, C. 2008. Design rules for donors 

in bulk-heterojunction tandem solar cells 

Towards 15% Energy-Conversion Eficiency. 

Adv Mater. 20(3):579-83. doi: 10.1002/ 

adma.200702337. 

Dobbins, P. 2007. Dolphin sonar: Modelling 

a new receiver concept. Bioinspir Biomim. 2:19. 

doi: 10.1088/1748-3182/2/1/003. 

Drucker, E.G., & Jensen, J.S. 1997. Kinematic 

and electromyographic analysis of steady 

pectoral fin swimming in the surfperches. 

J Exp Biol. 200:1709-23. 

Festo, M. 2008. Fluidic Muscle Brochure. 

http://www/festo.com/cms/en-us_us/5030.htm. 

Fish, F.E., & Lauder, G.V. 2006. Passive and 

active ow control by swimming fishes and 

mammals. Annu Rev Fluid Mech. 38:193-224. 

doi: 10.1146/annurev.fluid.38.050304.092201. 

Fish, F.E., Lauder, G.V., Mittal, R., Techet, 

A.H., Triantafyllou, M.S., Walker, J.A., & 

Webb, P.W. 2003. Conceptual design for the 

construction of a biorobotic AUV based on 

biological hydrodynamics. In: Proceedings of 

the 13th International Symposium on Unmanned 

Untethered Submersible Technology. 

Durham, NH: AUSI (Autonomous Undersea 

Systems Institute). 


Howle, L.E. 2011. The humpback whale’s 

ipper: Application of bio-inspired tubercle 

technology. Integr Comp Biol. 211:1857-67. 

Gao, J., Bi, S., Xu, Y., & Liu, C. 2007. 

Development and design of a robotic manta 

ray featuring flexible pectoral fins. In: IEEE 

International Conference on Robotics and 

Biomimetics, December 15-18, 2007. ROBIO 

2007, 519-23. Yalong Bay of Sanya, China: 

IEEE. 

Gibb, A., Jayne, B., & Lauder, G. 1994. 

Kinematics of pectoral fin locomotion in the 

bluegill sunfish Lepomis machrochirus. 

J Exp Biol. 189:133-61. 

Heine, C. 1992. Mechanics of apping fin 

locomotion in the cownose ray, Rhinoptera 

bonasus (Elasmobranchii, Myliobatidae). 

Ph.D. thesis, Duke University. 

Jaffe, J., Moore, K., McLean, J., & Strand, 

M. 2001. Underwater optical imaging: status 

and prospects. Oceanography. 14(3):66-76. 

Jalbert, J., Baker, J., Duchesney, J., Pietryka, 

P., Dalton, W., Blidberg, D., … Holappa, K. 

2003. Solarpowered autonomous underwater 

vehicle development. In: Proceedings of the 

Thirteenth International Symposium on Unmanned 

Untethered Submersible Technology. 

Durham, NH: Autonomous Undersea Systems 

Institute. 


Kato, N. 1998. Locomotion by mechanical 

pectoral fins. J Mar Sci Technol. 3(3):113-21. 

doi: 10.1007/BF02492918. 

Kato, N., Ando, Y., Tomokazu, A., Suzuki, 

H., Suzumori, K., Kanda, T., & Endo, S. 

2008. Elastic pectoral fin actuators for biomimetic 

underwater vehicles. In: Bio-mechanisms 

of Swimming and Flying, eds. Kato, N., & 

Kamimura, S., Part 3, 271-82. Springer. 

doi: 10.1007/978-4-431-73380-5_22. 

Kato, N., & Furushima, M. 2002. Pectoral 

fin model for maneuver of underwater 

vehicles. In: Autonomous Underwater 

Vehicle Technology, 1996. AUV’96, 

Proceedings of the 1996 Symposium on 

IEEE. pp. 49-56. Durham, NH: Autonomous 

Undersea Systems Institute. 

Klausewitz, W. 1964. Der lokomotionsmodus 

der Flugelrochen (Myliobatoidei). 

Zool Anz. 173:111-20. 

Kocak, D., & Caimi, F. 2005. The current 

art of underwater imaging with a glimpse 

of the past and vision of the future. Mar 

Technol Soc J. 39(3):5-26. doi: 10.4031/ 

002533205787442576. 

Lauder, G.V., & Jayne, B.C. 1996. Pectoral 

fin locomotion in fishes: Testing drag-based 

models using three-dimensional kinematics. 

Am Zool. 36(6):567-81. 

Lauder, G.V., Nauen, J.C., & Drucker, E.G. 

2002. Experimental hydrodynamics and evolution: 

function of median fins in ray-finned 

fishes 1. Integr Comp Biol. 42(5):1009-17. 

doi: 10.1093/icb/42.5.1009. 

Licht, S., Polidoro, V., Flores, M., Hover, F.S., 

& Triantafyllou, M.S. 2004. Design and 

projected performance of a apping foil AUV. 


doi: 10.1109/JOE.2004.833126. 

Listak, M., Martin, G., Pugal, D., Aabloo, A., 

& Kruusmaa, M. 2005. Design of a semiautonomous 

biomimetic underwater vehicle for 

environmental monitoring. In: Computational 

Intelligence in Robotics and Automation. 

Espoo, Finland: IEEE. 

Low, K.H., & Willy, A. 2006. Biomimetic 

motion planning of an undulating robotic 

fish fin. J Vib Control. 12:1337-59. 

doi: 10.1177/1077546306070597. 

Mojarrad, M. 2000. AUV biomimetic 

propulsion. In: OCEANS 2000 MTS/IEEE 

Conference and Exhibition. pp. 2141-6. 

Providence, RI: IEEE. 

Moored, K.W. August 2010. The design of a 

novel tensegrity-based synthetic pectoral fin 

for bio-inspired propulsion. Ph.D. thesis, 

University of Virginia. 

Moored, K.W., & Bart-Smith, H. 2007. The 

analysis of tensegrity structures for the design 

of a morphing wing. J Appl Mech. 74:668-76. 

doi: 10.1115/1.2424718. 

Moored, K.W., & Bart-Smith, H. 2009. Investigation 

of clustered actuation in tensegrity 

structures. Int J Solids Struct. 46:3272-81. 

doi: 10.1016/j.ijsolstr.2009.04.026. 

Moored, K.W., Dewey, P.A., Leftwich, M.C., 

Bart-Smith, H., & Smits, A.J. 2011a. 

Bio-inspired propulsion mechanisms based 

on manta ray locomotion. Mar Technol Soc J. 

45(4):110-8. 

Moored, K.W., Kemp, T.H., Houle, N.E., & 

Bart-Smith, H. 2011b. Analytical predictions, 

optimization and design of a tensegrity-based 

artificial pectoral fin. Int J Solids Struct. 

(Accepted for Publication). 

Palmisano, J., Geder, J., Ramamurti, R., 

Liu, K.J., Cohen, J., Mengesha, T., … Ratna, 

B. 2008. Design, development, and testing 

of flapping fins with actively controlled curvature 

for an unmanned underwater vehicle. 

In: Bio-mechanisms of Swimming and Flying: 

Fluid Dynamics, Biomimetic Robots, and 

Sports Science, eds. Kato, N., & Kamimura, 

S., 283-94. Berlin: Springer. 

Palmisano, J., Ramamurti, R., Lu, K.J., 

Cohen, J., Sandberg, W., & Ratna, B. 2007. 

Design of a biomimetic controlled-curvature 

robotic pectoral fin. In: IEEE International 

Conference on Robotics and Automation 

(ICRA). 966-73. IEEE. 

Pederzani, J.-N., Moored, K.W., Fish, F.E., 

& Haj-Hariri, H. 2011. A numerical investigation 

of the hydrodynamic signature of 

batoid swimming. In: Society for Integrative 

and Comparative Biology, Annual meeting. 

Salt Lake City, UT: Oxford Univ Press Inc. 



oscillation. J Exp Biol. 204(2):379-94. 

Russo, R.S., Blemker, S., Fish, F.E., Moored, 

K.W., & Bart-Smith, H. 2011. Form function 

relationship between ray skeletal architecture 

and ray locomotion. In: Annual meeting of 

the Society for Integrative and Comparative 

Biology. Salt Lake City, UT: Oxford Univ 

Press Inc. 

Sasko, D.E., Dean, M.N., Motta, P.J., & 

Hueter, R.E. 2006. Prey capture behavior 

and kinematics of the Atlantic cownose ray, 

Rhinoptera bonasus. Zoology. 109(3):171-81. 

doi: 10.1016/j.zool.2005.12.005. 

Schaefer, J.T., & Summers, A.P. 2005. 

Batoid wing skeletal structure: Novel 

morphologies, mechanical implications, and 

phylogenetic patterns. J Morphol. 264(3): 

298-313. doi: 10.1002/jmor.10331. 

Sfakiotakis, M., Laue, D., & Davies, B. 

2005. An experimental undulating-fin device 

using the parallel bellows actuator. In: 

Robotics and Automation, Proceedings 2001 

ICRA. IEEE International Conference on. 

pp. 2356-62. Seoul, Korea: IEEE. 

Sfakiotakis, M., Lane, D.M., & Davies, 

J.B.C. 1999. Review of fish swimming modes 

for aquatic locomotion. IEEE J Oceanic Eng. 

24(2):237-52. doi: 10.1109/48.757275. 

Stocking, J.B., Eberhardt, W.C., Shakhsheer, 

J.R., Paulus, M.A., & Calhoun, B.H. 2010. 

A capacitance-based whisker-like artificial 

sensor for fluid motion sensing. In: IEEE 

Sensors. Waikoloa, Big Island, Hawaii. 

Summers, A. 2000. Stifiening the stingray 

skeleton: An investigation of durophagy in 

myliobatid stingrays (Chondrichthyes, Batoidea, 

Myliobatidae). J Morphol. 243(2):113-26. 

doi: 10.1002/(SICI)1097-4687(200002) 

243:23.0.CO;2-A. 

Suzumori, K., Endo, S., Kanda, T., Kato, N., 

& Suzuki, H. 2007. A bending pneumatic 

rubber actuator realizing soft-bodied manta 

swimming robot. In: 2007 IEEE International 


Conference on Robotics and Automation. 

pp. 4975-980. IEEE. 

Takagi, K., Yamamura, M., Luo, Z., Onishi, 

M., Hirano, S., Asaka, K., & Hayakawa, Y. 

2007. Development of a rajiform swimming 

robot using ionic polymer artificial muscles. 

In: Intelligent Robots and Systems, 2006 

IEEE/RSJ International Conference on. 

pp. 1861-6. Beijing, China: IEEE. 

Tangorra, J., Anquetil, P., Fofonoff, T., Chen, 

A., Zio, M., & Hunter, I. 2007. The application 

of conducting polymers to a biorobotic 

fin propulsor. Bioinspir Biomim. 2:S6. 

doi: 10.1088/1748-3182/2/2/S02. 

Tangorra, J., Davidson, S., Hunter, I., 

Madden, P., Lauder, G., Haibo, D., … 

Mittal, R. 2008a. The development of a 

biologically inspired propulsor for unmanned 

underwater vehicles. IEEE J Oceanic Eng. 

32(3):533-50. 

Tangorra, J., Lauder, G., Madden, P., Mittal, R., 

Bozkurttas, M., & Hunter, I. 2008b. A 

biorobotic apping fin for propulsion and 

maneuvering. In: Robotics and Automation, 

2008. ICRA 2008. IEEE International 

Conference on. pp. 700-5. Pasadena, CA: 

IEEE. 

V. y Alvarado, P., Chin, S., Larson, W., 

Mazumdar, A., & Youcef-Toumi, K. 2010. A 

soft body under-actuated approach to multi 

degree of freedom biomimetic robots: A 

stingray example. In: Biomedical Robotics and 

Biomechatronics (BioRob), 2010 3rd IEEE 

RAS and EMBS International Conference 

on. pp. 473-8. Tokyo, Japan: IEEE. 


Labriform propulsion in fishes: kinematics 

of flapping aquatic flight in the bird wrasse 

Gomphosus varius (Labridae). J Exp Biol. 

200:1549-69. 

Wang, Z., Hang, G., Wang, Y., Li, J., & 

Du, W. 2008. Embedded SMA wire actuated 

biomimetic fin: a module for biomimetic 

underwater propulsion. Smart Mater 

Struct. 17:025039. doi: 10.1088/0964- 

1726/17/2/025039. 

underwater glider. In: International Symposium 

on Unmanned Untethered Submersible 

Technology. 75-85. Durham, NH: Autonomous 

Undersea Systems Institute. 

Webb, P.W. 1994. Mechanics and Physiology 

of Animal Swimming: The Biology of Fish 

Swimming. Cambridge, UK: Cambridge 

University Press. 45-62 pp. 

Yan, Q., Zhang, S., & Yang, J. 2010. Initial 

implementation of basic actuated unit of a 

flexible pectoral fin driven by SMA. In: 

Mechatronics and Automation (ICMA), 

2010 International Conference on. 

pp. 899-904. Xi’an, China: IEEE. 

Yang, S., Qiu, J., & Han, X. 2009. Kinematics 

modeling and experiments of pectoral 

oscillation propulsion robotic fish. J Bionic 

Eng. 6(2):174-9. doi: 10.1016/S1672-6529 

(08)60114-6. 

Yang, Y., Chen, J., Engel, J., Pandya, S., 

Chen, N., Tucker, C., … Liu, C. 2006. 

Distant touch hydrodynamic imaging with 

an artificial lateral line. P Natl Acad Sci. 

103(50):18891. 

Yong-hua, Z., Shi-wu, Z., Jie, Y., & Low, K. 

2007. Morphologic optimal design of bionic 

undulating fin based on computational fluid 

dynamics. In: Mechatronics and Automation, 

2007. ICMA 2007. International Conference 

on. pp. 491-6. Harbin, Heilongjian, China: 

IEEE. 

Yu, J., Hu, Y., Fan, R., Wang, L., & Huo, J. 

2007. Mechanical design and motion 

control of a biomimetic robotic dolphin. 

Adv Robotics. 21(3):499-513. 

doi: 10.1163/156855307780131974. 

Yu, J., Tan, M., Wang, S., & Chen, E. 2004. 

Development of a biomimetic robotic fish 

and its control algorithm. IEEE Sys Man 

Cybern B. 34(4), 1798-810. 

Zhou, C., & Low, K. 2010. Better endurance 

and load capacity: An improved design of 

manta ray robot. J Bionic Eng. 7:S137-44. 

doi: 10.1016/S1672-6529(09)60227-4. 

Webb, D., & Simonetti, P. 1999. The 

Slocum AUV: An environmentally propelled 


PAPER 

Bioinspired Propulsion Mechanisms 

Based on Manta Ray Locomotion 

AUTHORS 

Keith W. Moored 

Peter A. Dewey 


Engineering, Princeton University 

Megan C. Leftwich 

Physics Division, Los Alamos 

National Laboratory 

Hilary Bart-Smith 


Engineering, University of Virginia 




Introduction 

I 

n recent years, there has been 

considerable interest in developing 

novel underwater vehicles that use 

propulsion systems inspired by biology 

(Colgate & Lynch, 2004; 

Bandyopadhyay, 2005). Such vehicles 

have the potential to open up new mission 

capabilities and improve maneuverability, 

efficiency, and speed (Fish 

et al., 2003, 2011). Here we explore 

how various aspects of biological locomotion 

relate to performance in the 

particular case of ray-like swimming, 

with the aim of informing the design 

of new vehicles. 

The kinematic motion of batoid fish 

(rays) is based on the chordwise traveling 

wave that is a hallmark of their motion 

(Rosenberger, 2001). Species are classified 

as being oscillatory if the traveling 

wave wavelength is longer than the 

chord of their fin and undulatory if the 

wavelength is less than the chord of 

ABSTRACT 

Mobuliform swimmers are inspiring novel approaches to the design of underwater 

vehicles. These swimmers, exemplified by manta rays, present a model for new classes 

of efficient, highly maneuverable, autonomous undersea vehicles. To improve our 

understanding of the unsteady propulsion mechanisms used by these swimmers, 

we report detailed studies of the performance of robotic swimmers that mimic aspects 

of the animal propulsive mechanisms. We highlight the importance of the undulatory 

aspect of producing efficient manta ray propulsion and show that there is a strong interaction 

between the propulsive performance and the flexibility of the actuating surfaces. 

Keywords: mobuliform, manta ray, unsteady, swimming, flexible actuators 

their fin. The manta ray is an example 

of an oscillatory swimmer. Previously, 

Clark and Smits (2006) explored the 

thrust production and efficiencies of an 

artificial pectoral fin thatcapturedthe 

traveling wave motion and observed 

efficiencies upwards of 50% for an oscillatory 

motion at a fixed flow velocity. 

To study the swimming of mantas, 

we use artificial or robotic devices that 

generate a simple baseline motion that 

approximates biological kinematics. 

The complexity of the motion is then 

progressively increased by adding more 

kinematic features until the motion 

resembles the biology very closely. At 

each level of complexity, various performance 

metrics are measured. We explore 

theroleofspanwisecurvature,theeffects 

ofaspanwisetravelingwaveandtip 

speed modulation, which have not been 

previously investigated, and, the role 

of a chordwise traveling wave motion 

is investigated in the performance of 

an actively and passively flexible fin. 

Experiments 

Two biorobotic devices were developed 

and tested. First, an artificial pectoral 

fin able to produce root-fixed 

pure heaving motions was developed 

and will be referred to as the heaving 

fin. Thefin wascastfromaflexible 

plastic and actuated with variable degrees 

of spanwise curvature. The measurements 

on the heaving fin were 

conducted in a tow tank (Figure 1). 

This facility tows a fin throughstill 

water at a fixed velocity, U, inatank 

measuring 5-m long, 1-m deep, and 

1.5-m wide, and we directly measure 

thenetforceproduced,T, andthe 

power imparted into the fluid, P f , 

by a flapping fin structure. The 

propulsive efficiency of the motion, 

η p ¼ TU= ‐ P ‐ f , can then be calculated 

from the thrust and power measurements 

averaged over a cycle, 

‐ 

T and P ‐ f , respectively. If the fin were 

unconstrained and free-swimming, 

then the net force would cause the 

fin to accelerate or decelerate to a 

new velocity where there is no average 

net force. Constraining the fin allows 

for the measurement of force production 

and is a commonly used experimental 

approach (Anderson et al., 

1998). 


FIGURE 1 

Tow tank facility consisting of an artificial pectoral fin, tow tank, motor, carriage, and a control 

and measurement system. 

Second, an artificial pectoral fin capable 

of generating a chordwise traveling 

wave motion was developed and 

will be referred to as the traveling 

wave fin. This fin was used to measure 

the free-swimming performance in a 

tank measuring 6.7-m long, 1-m 

deep, and 1-m wide. This fin was 

also cast from a flexible plastic, but it 

was activated using a number of rigid 

spars in the spanwise direction. By reducing 

the number of actuating spars, 

the degree of passive flexibility of the 

fin could be varied. The traveling 

wave wavelength was controlled by 

changing the phase differences between 

adjacent spars. The steady 

swimming speed U and mean power 

input over a cycle P ‐ f were measured, 

and hence, the energy economy ζ 

could be found, where ζ ¼ U = P ‐ f . 

Energy economy is the inverse of 

cost of transport, both of which are 

used extensively in the biological literature 

(Schmidt-Nielsen, 1972; Fish 

et al., 1991; Liao et al., 2003; Liao, 

2004). Energy economy, however, is 

a more appropriate engineering metric 

as the dimensions are distance/per unit 

energy (the units could be miles per 

gallon for instance). Efficiency is also 

an appropriate performance measure, 

but from an experimental point-ofview 

it can only be measured when 

there is net thrust production or the 

fin is not in a free-swimming mode. 

Thus, for the experiments in the tow 

tank, efficiency was a measurable 

performance metric, but in the freeswimming 

cases, economy is used instead 

because the efficiency could not 

be directly measured. 

The performance is measured as a 

function of the Strouhal number, 

St = fA/U, where f is the frequency of 

motion, A is the peak-to-peak trailingedge 

amplitude of motion at the midspan, 

and U is the free-stream velocity. 

The Strouhal number is a measure 

of the lateral to streamwise spacing of 

the shed vortices in the wake and to a 

large extent governs the structure of 

the wake. It has been shown to be a 

critical parameter in describing the efficient 

propulsion of oscillating foils 

and plates (Anderson et al., 1998; 

Buchholz & Smits, 2008) and for 

swimming (Clark & Smits, 2006; 

Borazjani & Sotiropoulos, 2008, 

2009) and flying animals (Taylor 

et al., 2003). 

Fixed Velocity Experiments: 

Heaving Motion 

The skeletal structure of the heaving 

fin is composed of three connected 

hinged plates. The angular position of 

each hinge is individually controlled by 

a linear actuator. The fin allows for 

out-of-plane motion with no pitching 

or undulation in the chordwise direction. 

The skeletal structure is embedded 

into a compliant PVC polymer. 

The PVC is molded around the structure 

into a fin with a trapezoidal planform 

shape. This shape was chosen to 

be a simple representative shape of the 

manta ray as well as to maximize spacing 

for the internal structure. The fin 

has a span length of b =28cmand 

an average chord length of ― c =19 

cm, with an aspect ratio, AR = b 2 /S, 

of 1.47, wherein S is the planform 

area. The cross-sectional shape is a 

NACA 0020 airfoil. The trailing edge 

is stiffened by a thin metal sheet attached 

to the main structure. 

Three flapping mode shapes were 

explored: flat root-fixed heave (Figure 

2(a)), curved root-fixed heave (Figure 

2(b)), and curved root-fixed heave 

with a span-wise traveling wave (Figure 

2(c)). For each of these mode 

shapes the tip speed of the fin can be 

modulated. Previous kinematics studies 

of ray locomotion (Heine, 1992; 

Rosenberger, 2001) found that in 

order to swim faster oscillatory rays 

do not vary their beat frequency, but 

instead they vary the tip speed of 

their fin while holding frequency and 

amplitude constant. This can be 

viewed as modifying the actuation 

waveform to suit a particular mode of 

swimming. To implement this mode 

in our experiments, the time-varying 

waveform was varied from a pure sinusoid 

towards an almost square wave 

form (Figure 3). This allows the frequency 

and amplitude to be held 

fixed while the maximum fin tip 

speed is increased. The thrust and propulsive 

efficiency were measured for 

each prescribed motion. 


FIGURE 2 

Swimming modes ranging from an artificial/simple motion that approximates manta ray locomotion 

to a more complex biologically inspired motion: (a) flat root-fixed heaving motion, (b) curved 

root-fixed heaving motion, and (c) curved root-fixed heaving motion with a span-wise traveling 

wave (tip lag effect). 

Free-Swimming Experiments: 

Traveling Wave Motion 

The traveling wave fin was tested 

under free-swimming conditions in a 

stationary tank to explore the role 

of flexibility in ray-like propulsion. 

Low-friction carts were attached to 

the fin actuation mechanism to form 

a carriage that was mounted on tracks 

above the tank (see Figure 4). The 

chord of the fin was aligned parallel 

to the tracks, which allowed for a single 

degree of freedom and made the 

swimming direction of the carriage 

unidimensional, and no transverse 

motion of the carriage was permitted. 

The root of the fin abutted against an 

acrylic sheet that was in contact with 

thefreesurfaceofthewatertominimize 

surface waves. Upon actuating 

the fin, the cart propelled itself down 

the length of the tank. 

FIGURE 3 

An elliptical planform fin withan 

aspect ratio of 1.6 and a NACA 0020 

cross-section was cast using a flexible 

PVC plastic. Four aluminum spars 

were embedded into the fin to provide 

actuation. A push-rod connected each 

spar to a gear in a gear-train, driven by 

a DC motor, that produced a sinusoidal 

rotation of the spar about a pivot 

point located at the root chord of the 

fin. This results in a linearly increasing 

amplitude of motion along the span of 

the fin (Figure 5). The wavelength of 

the traveling wave λ could be varied 

by changing the phase difference between 

the actuating spars. 

In addition, the number of actuating 

spars could be varied to allow for a 

certain degree of passive flexibility in 

the fin. When four actuating spars 

are used, the locomotion of the fin 

was prescribed for the entire chord of 

Holding frequency and amplitude constant while modulating tip speed can be achieved by varying 

the actuation waveform from a sine wave to a square wave. 

the fin, with a traveling wave wavelength 

λ a , and this fin will be referred 

to as the active fin. When the two trailing 

edge spars are removed, the trailing 

half of the chord of the fin willpassively 

respond to the leading edge actuation 

and external fluid forces. This fin 

will be referred to as the passive fin 

(Figure 5). For the passive fin, the 

traveling wave along the chord is generated 

by the first two gears in the gear 

train and the passive response of the 

trailing edge. Due to these compounding 

factors, the precise wavelength 

along the chord for the passive fin is 

unknown, and we define an analogous 

wavelength, λ p such that λ a = 

λ p when the offset between the first 

two gears is identical for both the active 

and passive fin. We define the dimensionless 

wavelengths λ a ¼ λ a=C 

and λ p ¼ λ p=C, whereC is the root 

chord of the fin. This study focuses on 

cases with λ a ; λ p > 1, representative 

of oscillatory swimmers (Rosenberger, 

2001). 

Results and Discussion 

Flat and Curved Modes 

The first two modes of swimming, 

the flat mode (Figure 2(a)) and 

thecurvedmode(Figure2(b)),were 

studied using the heaving fin. 

The thrust coefficient is defined by 

C T ¼ T = 1 2 ρU 2 S,whereinT is the 

thrust, ρ is the water density, and S 

is the planform area of the fin. Similarly, 

the power coefficient is defined 

by C p ¼ P= 1 2 ρU 3 S, wherein P is the 

power input to the water (as defined 

by Clark & Smits, 2006). Figure 6(a) 

shows that the thrust production and 

power input increases as the Strouhal 

number increases, as found in previous 

studies by Anderson et al. (1998) and 

Dong et al. (2006). The flat mode of 

swimming produces more thrust and 


FIGURE 4 

Tank facility for the chordwise traveling wave experiments: (a) perspective view and (b) front 

view. Drawings not to scale. 

FIGURE 5 

Traveling wave actuation system for (left) active fin and (right) passive fin. 

FIGURE 6 

Flat mode compared to curved mode: (a) thrust performance as a function of St and (b) power 

coefficient as a function of St. 

uses more power than the curved mode 

throughout the Strouhal range. This 

results is not unexpected as the flat 

mode of swimming sweeps out a larger 

volume of fluid than the curved 

mode of swimming, causing the overall 

increase in the thrust and power 

coefficients. 

Because the two modes of swimming 

are dissimilar, comparing the 

thrust or efficiency as a function of 

Strouhal number may be misleading. 

For a fixed Strouhal number, different 

amounts of thrust and power are produced 

by each swimming mode. A 

better comparison is the thrust or efficiency 

as a function of the power 

coefficient because each swimming 

mode can be compared at a fixed 

power input. 

Figure 7a shows this comparison. 

For both modes of swimming the 

power input increases as the thrust increases, 

although at higher power 

input the thrust increases at a slower 

rate. For all power inputs, more thrust 

is produced for the flat mode of swimming 

than for the curved mode of 

swimming, while the thrust tends to 

the drag of the motionless fin asthe 

power input tends to zero. Interestingly, 

observations of swimming 

manta rays do not indicate that they 

use this higher-performance flat 

mode and instead use the curved 

mode for swimming. This suggests 

that the curved mode, coupled with 

another mechanism (perhaps a chordwise 

traveling wave), more fully characterizes 

the swimming mechanics of 

manta rays. 

Figure 7(b) shows the propulsive 

efficiency as a function of Strouhal 

number. The efficiency of the curved 

mode first rises quickly rise with 

increasing Strouhal number, and then 

apeakinefficiency is attained, followed 

by a slow decline in efficiency 


FIGURE 7 

Flat mode compared to curved mode: (a) thrust coefficient as a function of the power coefficient 

and (b) efficiency as a function of Strouhal number. 

at the higher Strouhal numbers. This 

trend is characteristic of efficiency 

curves for oscillating foils and plates 

(Clark & Smits, 2006; Anderson 

et al., 1998; Heathcote et al., 2006a, 

2006b; Heathcote & Gursul, 2007; 

Buchholz & Smits, 2008). In general, 

the efficiency crosses from negative 

(net drag) to positive (net thrust) and 

continues to increase while at high 

values of St there is a decline in efficiency 

that follows potential flow theory 

(Jones et al., 1998). Thus, a peak 

in efficiency is expected at an intermediate 

Strouhal number. The peak efficiency 

for the curved mode is about 

20% occurring at a St = 0.2, which is 

in the range of 0.2 < St

FIGURE 9 

Tip lag mode: (a) efficiency as a function of St and (b) peak efficiency dependence on tip lag. 

function was used to modulate the tip speed while fixing the frequency and 

amplitude, as given by 

xt ðÞ¼at ð ϕÞ 4 þbt ð ϕÞ 2 þA; 0 ≤ t ≤ T =2 

a ¼ 

U sqr 

max 

2ϕ 3 þ A ϕ 4 ; b ¼ 2A 

ϕ 2 

ϕ ¼ 1=4f ; U sqr 

max ¼ 2παU ∞ St 

þ U sqr 

max 

2ϕ 

The amplitude, A, the frequency, f, and the maximum tip speed, U sqr 

max, all determine 

the shape of the quartic function. The period is T. By holding the frequency 

and amplitude constant the tip speed can be modulated by varying α 

between 1 and 2.667. When α is 1, the quartic function matches a sine wave, 

but when α is 2.667, the maximum tip speed is 2.667 times faster than the maximum 

tip speed of a sine wave without varying the frequency or amplitude. It 

should be noted, however, that many animals increase their swimming speed 

by modulating their frequency of motion while fixing the amplitude. Frequency 

modulation will increase the Strouhal number by increasing frequency. 

Tests were conducted at two Strouhal numbers, 0.2 and 0.25, and the amplitude 

was fixed at A/b = 0.44 for α =1-2.4. Figure 10a shows the thrust coefficient 

dependence on the tip speed, α, for a Strouhal number of 0.2. As the tip speed is 

increased the thrust coefficient increases linearly, indicating that tip speed modulation 

can be used to increase thrust production, as exhibited by rays (Heine, 

1992; Rosenberger, 2001). 

Figure 10(b) shows the thrust coefficient plotted against the power coefficient 

for both frequency modulation and tip speed modulation. Tip speed modulation 

produces less thrust than frequency modulation for the same input power, suggesting 

that tip speed modulation is not an efficiency strategy. However, incorporating 

a chordwise traveling wave may change this outcome. Additionally, as the 

value of α is increased, the fin has an increasing period of effectively no motion. In 

a free-swimming test, the increasing resting time for the fin would result in an 

unpowered gliding period over part of the flapping cycle. This would result in a 

burst-and-coast behavior that could improve the economy since forward motion 

would occur without any input power for part of the flapping cycle. Alternatively, 

(1) 

tip speed modulation could an effective 

flight/sprinting mode, where efficiency 

is not important. 

Chordwise Traveling Wave 

We now investigate the effects of a 

chordwise traveling wave, using the 

traveling wave fin (the experimental 

arrangement was described in Free- 

Swimming Experiments: Traveling 

Wave Motion). 

Clark and Smits (2006) investigated 

the thrust and efficiency of an artificial 

pectoral fin using a chordwise 

traveling wave motion and found efficiencies 

peaking near 50% for optimal 

conditions (St ≈ 0.25 and λ a ≈ 4‐6). 

The efficiencies were measured at predetermined 

Strouhal numbers. For the 

current work, this constraint is not imposed, 

and the fin was instead actuated 

at a given frequency and wavelength 

and allowed to freely swim down the 

length of a tow tank. In doing so, the 

fin attains its self-propelled swimming 

speed and Strouhal number for that 

frequency and wavelength. 

Figure 11 shows the steady velocity 

achieved as a function of input flapping 

frequency for different wavelengths 

of actuation. The velocity is 

giveninroot-chordlengths(CL)per 

second, where C root = 0.254 m. For 

the active fin, an almost linear increase 

in velocity is observed with increasing 

frequency, a result in agreement with 

previous studies that found thrust 

coefficients increasing with frequency 

(Anderson et al., 1998). Peak velocities 

were found to occur for λ a ¼ 6 at the 

highest flapping frequencies. In these 

instances the velocity was upwards of 

2 CL/s, corresponding to a dimensional 

velocity of 0.51 m/s, highlighting 

that this form of propulsion may 

prove fruitful for future underwater vehicle 

designs that demand relatively 

high speeds. 


FIGURE 10 

(a) Thrust coefficient increasing with tip speed increase and (b) tip speed modulation compared 

to frequency modulation. The parameter α is the ratio of maximum tip speed of the square wave 

actuation compared to the maximum tip speed of a sine wave of the same frequency and amplitude, 

α ¼ Umax=U sqr 

max sin . 

Figure 11 also reveals the role of 

passive flexibility. Here, the fin was actuated 

using only the two anterior 

spars, leaving the posterior half of the 

fin to respond passively to the forcing 

by the actuators and the fluid forces. 

The velocity of this fin still increases 

with increasing frequency, but the 

trend is no longer linear. Tests in still 

water indicate that the passive fin has a 

resonant frequency of about 2.4 Hz, 

defined as the frequency at which the 

trailing edge amplitude is maximized. 

FIGURE 11 

Fin velocity in chord lengths per second as a 

function of flapping frequency. The wavelength 

λ a refers to the active fin whileλ p refers to the 

passive fin. (Color versions of figures available 

online at: http://www.ingentaconnect.com/ 

content/mts/mtsj/2011/00000045/00000004.) 

FIGURE 12 

Strouhal number as a function of steady-state 

swimming velocity for the active fin compared 

with biological data of the manta ray 

(Fish, 2010). 

As resonance is approached, the amplitude 

of the trailing edge motion increases, 

resulting in the fin velocity 

increasing at an enhanced rate (in comparison 

to a linear trend). It should be 

noted that the swimming velocities for 

the passive fin were approximately 

80% of those of the active fin. 

The free-swimming Strouhal number 

for the active fin, along with manta 

ray field data (Fish, 2010), are displayed 

in Figure 12 (the Strouhal 

number for the passive fin isnot 

shown since the trailing edge excursion 

of the passive fin is unknown). The experimental 

data collapse onto a single 

curve, indicating that the Strouhal 

number for the freely swimming active 

fin does not depend on the wavelength. 

As the swimming velocity increases, 

the Strouhal number begins 

to enter the regime presumed to be efficient 

(St = 0.2-0.4; Taylor et al., 

2003). Hence, the optimal swimming 

speed for the traveling wave fin, from 

the perspective of efficiency, is likely 

to be ≥2 CL/s. The biological and 

experimental data display the same 

trend, whereby an increase in swimming 

velocity yields a decrease in 

Strouhal number. Borazjani and 

Sotiropoulos (2008, 2009) were able 

to show, for carangiform and anguiliform 

swimming, that the Strouhal 

number for self-propulsion approaches 

the efficient regime observed in nature 

only with increasing swimming velocity. 

The current study supports this 

conclusion for mobuliform swimming 

as well. 

The energy economy, ζ =V c /P f , for 

the active and passive fins is shown in 

Figure 13. In the case of the shortest 

wavelength(λ* = 3), the active fin has 

a higher energy economy than the passive 

fin, but for the longer wavelengths 

the energy economy for the passive fin 

FIGURE 13 

Energy economy as a function of flapping frequency. 

The wavelength λ a refers to the active 

fin while λ p refers to the passive fin. 


exceeds that of the active fin despite a 

decrease in the overall swimming speed 

(Figure 11). Clearly, by removing the 

two trailing edge spars in creating the 

passive fin, the power consumption 

decreases significantly compared to 

the active fin. The highest energy 

economy recorded was 0.18 CL/J 

for the passive fin withλ p ¼ 12 and 

f = 2 Hz, but its economy is still trending 

upward with increasing frequency, 

which may reflect the fact that the 

maximum test frequency was still 

below the resonant frequency of the 

fin (2.4Hz).LeftwichandSmits 

(2010) found thrust production of a 

passively flexible artificial lamprey tail 

to increase as resonance is approached. 

The current work suggests that the energy 

economy also benefits by a system 

exploiting the resonant modes of a fin. 

The increase in energy economy 

with frequency may also be a Strouhal 

number effect. Figure 12 indicates that 

the fin begins to enter the “efficient” 

regime (St = 0.2-0.4) with increasing 

swimming velocity (which occurs at 

higher flapping frequencies; see Figure 

11). While the energy economy is 

not a direct measure of efficiency, the 

two parameters are inherently linked, 

so it is not altogether surprising that 

the energy economy increases as the 

Strouhal number approaches the supposedly 

optimal range. The active fin 

with λ a ¼ 12 displays a maximum at 

f = 1.6 Hz. It is believed that this 

peak is related to changes in the wake 

structure that result from the threedimensionality 

of the wake associated 

with this fin. Dewey et al. (2011) 

found that an increasing wavelength 

causes the wake to bifurcate into a double 

wake structure that is less efficient, 

and it is believed to be responsible for 

the maximum observed in Figure 13. 

It may be that the other cases will 

eventually reach a maximum due to a 

similar mechanism, but further testing 

is required to support this suggestion. 

Conclusions 

This study has explored various kinematic 

modes associated with biological 

propulsion based on the manta ray. 

The results highlight the interdependence 

of the kinematic motions and 

fluid–structure interactions on the performance 

characteristics of the animal. 

A purely heaving fin was used to examine 

three kinematic modes of swimming 

(flat, curved, and tip lag) as well 

as variation in the actuation waveform 

(sine waveform to square waveform). It 

was found that the flat mode of swimming 

produces higher efficiency and 

thrust compared to the curved mode 

of swimming. Furthermore, there was 

no performance benefit foundbyincorporating 

tip lag into the motion or 

by modulating tip speed instead of frequency. 

These results are counter to 

the hypothesis that as the motion becomes 

more biologically similar the 

thrust or efficiency performance will 

increase. The maximum thrust coefficient 

was found to be about 0.7 at 

St =0.45,andthemaximumpropulsive 

efficiency was about 22% at 

St = 0.15. These performance metrics 

were low, as expected, due to the absenceofachordwisetravelingwave 

motion.Whatwasnotexpectedwas 

that the performance would be insensitive 

to curvature, tip lag, and tip 

speed variations. Without flexibility, 

the motion of the fin may be too constrained 

and not “natural” enough to 

achieve the performance benefits of 

the kinematic variations explored. 

The presence of a chordwise traveling 

wave to generate an undulatory 

motion appears to be of prime importance 

for efficient propulsion. For example, 

Clark and Smits (2006) found 

efficiencies upwards of 50%. In freeswimming 

experiments of artificial 

fins similar to that studied by Clark 

and Smits, we found that they were 

able to generate speeds similar to 

those observed in nature, of the order 

2 CL/s. When the fin was actively actuated, 

that is to say that the traveling 

wave motion was defined for all points 

on the chord of the fin, the steady-state 

Strouhal number obtained by the oscillating 

fin was found to be independent 

of the wavelength and exhibited 

thesametrendasthemantarayin 

nature. That is, at low swimming 

velocities, both the manta ray and the 

artificial fin display high steady-state 

Strouhal numbers, but the Strouhal 

numbers decrease with increasing 

swimming speed and they approach 

the regime where efficient propulsion 

is hypothesized to exist (St =0.2- 

0.4). Introducing passive flexibility 

into the fin, by restricting the actuationtotheleadingedgeandletting 

the rest of the fin respond passively to 

the actuation and the external fluid 

forces, improved the energy economy. 

It was found that the passively actuated 

fin achieved steady state swimming 

speeds that were approximately 80% 

of that of the actively actuated fin, 

but because there was a significant decrease 

in the power required to propel 

the passively actuated fin the energy 

economy increased. 


The authors would like to thank 

Daphne Rein-Weston, Dan Quinn, 

and Dr. Melissa Green for their aid 

in developing the low-friction carriage 

experiment. We would also like to 

thank Professor Frank Fish for correspondence 

regarding manta rays in 

nature. The authors would like to acknowledge 

funding from the Office 


of Naval Research through the 

MURI program on Biologically- 

Inspired Autonomous Sea Vehicles 

(grant N0001408-1-0642), the 

David and Lucille Packard Foundation, 

the National Science Foundation 

(grant CMS-0384884), and the Virginia 

Space Grant Consortium. 

Corresponding Author: 




Princeton, NJ 08544 

Email: asmits@princeton.edu 

References 

Anderson, J.M., Streitlien, K., Barrett, D.S., 

& Triantafyllou, M.S. 1998. Oscillating 

foils of high propulsive efficiency. 


S0022112097008392. 

Bandyopadhyay, P.R. 2005. Trends in 

biorobotic autonomous undersea vehicles. 


doi: 10.1109/JOE.2005.843748. 

Borazjani, I., & Sotiropoulos, F. 2008. 

Numerical investigation of the hydrodynamics 

of carangiform swimming in the transitional 

and inertial flow regimes. J Exp Biol. 

211:1541-58. doi: 10.1242/jeb.015644. 

Borazjani, I., & Sotiropoulos, F. 2009. 

Numerical investigation of the hydrodynamics 

of anguilliform swimming in the transitional 

and inertial flow regimes. J Exp Biol. 

212:576-92. doi: 10.1242/jeb.025007. 

Buchholz, J.H.J., & Smits, A.J. 2008. 

The wake structure and thrust performance 

of a rigid low-aspect-ratio pitching panel. 


S0022112008000906. 

Clark, R.P., & Smits, A.J. 2006. Thrust 

production and wake structure of a 

batoid-inspired oscillating fin. J Fluid 

Mech. 562:415-29. doi: 10.1017/ 

S0022112006001297. 

Colgate, J.E., & Lynch, K.M. 2004. Mechanics 

and control of swimming: A review. 


doi: 10.1109/JOE.2004.833208. 

Dewey, P.A., Carriou, A., & Smits, A.J. 

2011. On the relationship between efficiency 

and wake structures of a batoid-inspired 

oscillating fin. J Fluid Mech. Under review. 

Dong, H., Mittal, R., & Najjar, F.M. 2006. 

Wake topology and hydrodynamic performance 

of low-aspect-raio flapping foils. 


S002211200600190X. 

Fish, F.E. 2010. Private Communication. 

Fish, F.E., Fegely, J.F., & Xanthopoulos, 

C.J. 1991. Burst-and-coast swimming in 

schooling fish (Notemigonus crysoleucas) with 

implications for energy economy. Comp 

Biochem Phys A Phys. 100:633-7. 

doi: 10.1016/0300-9629(91)90382-M. 

Fish, F.E., Haj-Hariri, H., Smits, A.J., 

Bart-Smith, H., & Iwasaki, T. 2011. 

Biomimetic swimmer inspired by the 

manta ray. In: Biomimetics: Nature-Based 

Innovation, ed. Bar-Cohen, Y., 495-523. 

Boca Raton, FL: CRC Press. 

Fish, F.E., Lauder, G.V., Mittal, R., Techet, 

A.H., Triantafyllou, M.S., Walker, J.A., & 

Webb, P.W. 2003. Conceptual design for the 

construction of a biorobotic AUV based on 

biological hydrodynamics. In: Proceedings 

of the 13th International Symposium on 

Unmanned Untethered Submersible 

Technology. Durham, NH: Autonomous 

Systems Undersea Institute. 

Heathcote, S., & Gursul, I. 2007. Jet 

switching phenomenon for a periodically 

plunging airfoil. Phys Fluids. 19(2):027104. 

doi: 10.1063/1.2565347. 

Heathcote, S., Wang, Z., & Gursul, I. 2006a. 

Effect of spanwise flexibility on flapping wing 

propulsion. In: 36th AIAA Fluid Dynamics 

Conference and Exhibit. Reston, VA: American 

Institute of Aeronautics and Astronautics Inc. 

Heathcote, S., Wang, Z., & Gursul, I. 

2006b. Flexible flapping airfoil propulsion 

at low Reynolds numbers. AIAA J. 45(5): 

1066-79. doi: 10.2514/1.25431. 

Heine, C. 1992. Mechanics of flapping fin 

locomotion in the cownose ray, Rhinoptera 

bonasus (Elasmobranchii, Myliobatidae). 

Ph.D. Thesis Duke University. 

Jones, K.D., Dohring, C.M., & Platzer, 

M.F. 1998. Experimental and computational 

investigation of the KnollerBetz effect. 

Zool Anz. 173:1240-6. 

Klausewitz, W. 1964. Der lokomotionsmodus 

der Flugelrochen (Myliobatoidei). AIAA J. 

37:1240-6. 

Leftwich, M.C., & Smits, A.J. 2010. Thrust 

production by a mechanical swimming 

lamprey. Exp Fluids. 50(5):1349-55. 

doi 10.1007/s00348-010-0994-x. 

Liao, J.C. 2004. Neuromuscular control 

of trout swimming in a vortex street: Implications 

for energy economy during the 

Karman gait. J Exp Biol. 207:3495-506. 

doi: 10.1242/jeb.01125. 

Liao, J.C., Beal, D.N., Lauder, G.V., & 

Triantafyllou, M.S. 2003. Fish exploiting 

vortices decrease muscle activity. Science. 

302:1566-9. doi: 10.1126/science.1088295. 



oscillation. J Exp Biol. 204(2):379-94. 

Schmidt-Nielsen, K. 1972. Locomotion: 

Energy cost of swimming, flying, and running. 

Science. 177:222-8. doi: 10.1126/science. 

177.4045.222. 

Taylor, G.K., Nudds, R.L., & Thomas, A.L.R. 

2003. Flying and swimming animals cruise 

at a Strouhal number tuned for high power 

efficiency. Nature. 425(6959):707-11. 

doi: 10.1038/nature02000. 


PAPER 

Inspired by Sharks: A Biomimetic Skeleton 

for the Flapping, Propulsive Tail of an 

Aquatic Robot 

AUTHORS 

John H. Long, Jr. 


Vassar College 

Tom Koob 

MiMedx Group, Inc. 

Justin Schaefer 

David Geffen School of Medicine, 

University of California 

Adam Summers 

Friday Harbor Labs, 

University of Washington 

Kurt Bantilan 



Sindre Grotmol 


University of Bergen 

Marianne Porter 



Propulsive Functions of 

Vertebral Columns 

I 

n sharks and other fish, the 

body’s primary skeleton is the vertebral 

column, which runs from the 

head to the caudal fin (Summers & 

Long, 2006). The vertebral column is 

a jointed framework to which muscles 

attach and on which the muscles pull 

to create the traveling waves of flexure 

that transfer momentum from the 

body to the surrounding fluid. The 

vertebral column is composed of rigid 

ABSTRACT 

The vertebral column is the primary stiffening element of the body of fish. This 

serially jointed axial support system offers mechanical control of body bending 

through kinematic constraint and viscoelastic behavior. Because of the functional 

importance of the vertebral column in the body undulations that power swimming, 

we targeted the vertebral column of cartilaginous fishes—sharks, skates, and rays— 

for biomimetic replication. We examined the anatomy and mechanical properties of 

shark vertebral columns. Based on the vertebral anatomy, we built two classes of 

biomimetic vertebral column (BVC): (1) one in which the shape of the vertebrae 

varied and all else was held constant and (2) one in which the axial length of the 

invertebral joint varied and all else was held constant. Viscoelastic properties of 

the BVCs were compared to those of sharks at physiological bending frequencies. 

The BVCs with variable joint lengths were then used to build a propulsive tail, consisting 

of the BVC, a vertical septum, and a rigid caudal fin. The tail, in turn, was 

used as the propeller in a surface-swimming robot that was itself modeled after a 

biological system. As the BVC becomes stiffer, swimming speed of the robot increases, 

all else being equal. In addition, stiffer BVCs give the robot a longer stride 

length, the distance traveled in one cycle of the flapping tail. 

Keywords: biomimetics, robot, vertebral column, propulsion 

elements, called vertebrae, connected 

by flexible intervertebral joints 

(Grotmol et al., 2003; Koob & 

Long, 2000). The joints and their 

adjoining vertebrae limit the body’s kinematic 

degrees of freedom, constraining 

bending primarily to the lateral 

direction in response to loads imposed 

by muscle, inertia, and external fluid 

forces (Grotmol et al., 2006; Porter 

et al., 2009; Symmons, 1979; Schmitz, 

1995). In this way, the vertebral 

column functions to control dynamic 

reconfigurations of the self-propelling 

body. 

In roll-stable sharks and fish, lateral 

body bending is characterized by the 

overlay of harmonic and transient motions 

that range from small traveling 

flexures to large-amplitude standing 

waves (Long et al., 2010). These bending 

motions produce the propulsive 

forces that create forward swimming, 

turning maneuvers, and rapid accelerations. 

Across many different kinds of 

fish, the stiffness of the bending joints, 

measured as the apparent Young’smodulus, 

E, rangesfrom0.1to8MPa 

(Long et al., 2002). The E is a mechanical 

property, sometimes called 

the “material stiffness,” that measures 

the contribution of the material, independent 

of its geometric arrangement 

in the structure, to the structure’s resistance 

to changing shape when an external 

load is applied to it. 


Joints with any appreciable stiffness 

at all may at first seem to be a paradox: 

why not have low-stiffness joints that 

cost little, in terms of mechanical 

work, to bend The answer seems to 

be two-fold: (1) Stiff joints increase 

their resistance, in terms of the absolute 

bending moment, M (in units of 

Nm), in proportion to the magnitude 

of bending curvature, κ (m −1 ). This resistance, 

which can grow nonlinearly 

with κ, serves as a brake, limiting 

lateral bending (Long et al., 2002). 

(2) Stiff joints store and release more 

mechanical work, so-called “elastic 

energy,” than flexible joints (Long, 

1992). The amount of work released 

in elastic recoil is also in proportion 

to E and to the square of κ. Hence, 

the vertebral column functions best 

as a spring when muscles have reached 

their functional limits, at the end of a 

FIGURE 1 

large-κ bend when connective tissues 

are stiffest. 

To explore the mechanical design 

space of vertebral columns, we created 

biomimetic vertebral columns (BVCs). 

The BVCs are modeled after the vertebral 

column of sharks. We chose 

sharks’ vertebral columns since they 

are structurally simple, compared to 

those of bony fish, consisting of cylindrical 

centra, small neural and hemal 

arches, and thin intervertebral joints 

(Figure 1). Together, a centrum and 

its arches are called a vertebra, and in 

the cartilaginous sharks, skates, and 

rays, the vertebrae (plural form of 

‘vertebra’) are composed of mineralized 

cartilage. The compressive stiffness 

of these vertebrae, measured by 

E, ranges from 25 to 500 MPa, overlapping 

the lower range of E for bone 

(Porter et al., 2006). Compared to the 

Vertebral columns of two species of sharks. For scale, each centrum shown here is between 4- and 

6-mm long in the head-to-tail direction. Head is to the left; tail is to the right. 

bone of mammals, for a given value of E 

the vertebrae of sharks are stronger, where 

strength is measured in terms of breaking 

stress (Porter & Long, 2010). Thus, 

in some ways, vertebral columns made of 

mineralized cartilage perform better than 

vertebral columns made of bone. 

In summary, vertebral columns serve 

at least three important propulsive functions 

during swimming: (1) they control 

dynamic reconfigurations of the body 

by limiting the kinematic degrees of 

freedom, (2) they brake high-amplitude 

bends by virtue of their stiffness, and 

(3) they integrate muscle work over 

time by recoiling elastically. 

To build a BVC that can function as 

part of an aquatic propulsion system, 

we (1) characterized the morphology 

(size and shape) of the vertebral columns 

of sharks, (2) measured the mechanical 

properties of those vertebral 

columns as they underwent sinusoidal 

bending, (3) used that information 

about morphology and mechanical 

properties to design BVCs, (4) tested 

BVCs as they underwent the dynamic 

bending characteristic of swimming 

and propulsion, and (5) tested the 

BVCs as the primary skeleton in the 

flapping tail of an aquatic robot. 

Morphology of Sharks’ 

Vertebral Columns 

In three individuals of the blacktip 

shark, Carcharinus limbatus, and 

the bonnethead shark, Sphryna 

tiburo, we measured, from radiographs, 

the following features from 

the head to the beginning of the caudal 

fin (Figure 2): (1) the length of centrum, 

c, (2), the diameter of the centrum, 

d, (3) the length of the intervertebral 

joint, j, and (4) the cone angle, Ξ, 

of the capsule of the joint. Blacktip 

sharks, members of the family 

Carcharhinidae, were chosen because 


FIGURE 2 

Measuring vertebral morphology of blacktip and bonnethead sharks. Representative X-rays 

show the heavily mineralized vertebral centra, which possess an “X” shapeinthistwo-dimensional 

view that is from cone-shaped joint capsules. The dark space between vertebrae is the intervertebral 

joint. The morphology of each vertebra and intervertebral joint was measured from digitized landmarks 

(blue dots). (Color versions of figures available online at: http://www.ingentaconnect.com/ 

content/mts/mtsj/2011/00000045/00000004.) 

they are known to be fast swimming 

predators of fish. In contrast, 

bonnethead sharks, members of the 

hammerhead family Sphyrnidae, are 

known less for their speed and more 

for their maneuverability and ability 

to find and eat crustaceans. Of similar 

adult body size, the two species 

represent contrasting swimming styles 

and ecologies. Three individuals of 

each species were used for this study. 

In the bonnethead shark, all the 

morphological features, except d, 

increased in size from the head to the 

end of the abdomen and then decreased 

towards the caudal fin (Figure 3). In 

blacktip shark, only Ξ and c varied 

from head to caudal fin. The significance 

(α = 0.05) of the morphological 

variation was determined with a 

multivariate analysis of covariance 

(MANCOVA), with species and position 

as main effects and individual as 

the covariate (JMP 8.0.2., SAS Institute, 

Cary, NC). Following an identity 

model MANCOVA, univariate 

ANCOVAs were also run. 

The variation in the morphology of 

these vertebral columns was used to 

guide the construction of the BVCs. 

For each of the sharks’ morphological 

features, we indicated which dimensions 

were used (see black arrows on 

the ordinates, Figure 3). 

Mechanical Properties of 

Sharks’ Vertebral Columns 

Using the same freshly dissected 

vertebral columns from which morphology 

was measured, we conducted 

3-point dynamic bending tests using 

an MTS model Mini Bionix 858 

(Eden Prairie, MN) with a 500 N 

load cell. For blacktip sharks, each 

vertebral column was cut into five 

segments of 19 vertebrae each. For 

bonnethead sharks, each vertebral 

column was cut into five segments of 

14 vertebrae each. The number of segments 

in each species was varied to 

keep the absolute length of each test 

segment approximately equal. Each 

segment was subjected to sinusoidal 

bending at a frequency, f (Hz), and 

maximum curvature, κ (m −1 ), varied 

to hold constant the time rate of 

change of κ, which is equivalent to 

the strain rate (actuator displacement 

amplitude of 2 mm s −1 ). 

To characterize the viscoelastic 

properties of the vertebral column during 

bending, the apparent storage and 

loss moduli, E′ and E″ (MPa), respectively, 

were measured at each combination 

of two species, five segment 

positions, and three κ. TheE′ measures 

the purely elastic component of 

the stiffness; it is the force proportional 

to the magnitude of the bending of the 

vertebral column. The E″ measures the 

purely viscous component of the stiffness; 

it is the force proportional to the 

velocity of the bending of the vertebral 

column. These properties were calculated 

from the following formulae: 

E ′ ¼ E*cos δ and E ″ ¼ E*sin δ, 

wherein E* ¼ F maxL 3 

48Iy max 

and δ is the 

phase lag (radians) between the displacement 

and load signals. Moreover, 

F max is the force (N) measured at the 

load cell, L is the gauge length of the 

specimen (m), I is the specimen’s 


FIGURE 3 

Vertebral morphology of sharks. Four dimensions were used to characterize the size and shape of 

the vertebral centra and the intervertebral joints. The means of three individuals for each species 

are shown; individuals ranged from 0.59 to 0.91 m in overall body length. The error bars indicate 

the standard error of the mean. Black arrows show the specific dimensions represented in our 

BVCs. MANCOVA, using the identity method, calculated a significant Wilkes λ (p < 0.0001), 

with a significant interaction of species and position and significant main effect of species; the 

covariate, individual, was also significant. Partial correlations among the response variables 

ranged from a low of 0.38 between j and d toahighof0.79betweenΞ and c. Significance 

level is indicated (*p < 0.05, **p < 0.01, ***p < 0.001). 

second moment of area (m 4 ), and y max 

(m) is the distance from the presumed 

neutral plane of bending (transverse 

center of specimen) and the lateralmost 

fibers of the specimen. 

In blacktip sharks, the E′ and E″ 

values were of greater magnitude 

( p < 0.05) than those of bonnethead 

sharks (Figure 4). In both species, 

E′ increased towards the tail, an effect 

that is amplified at higher values of κ, 

as indicated by a significant (p < 0.05) 

interaction term. The significance of 

the variation in E′ and E″ was determined 

using ANOVA, with species, 

position, and κ as main effects (JMP 

8.0.2., SAS Institute, Cary, NC). 

Since the data blacktip and bonnethead 

sharks were taken at a single amplitude 

of strain rate (2.0 mm s −1 ), we 

sought additional information about 

how E′ and E″ vary with changes in 

strain rate. We also wanted to test the 

hypothesis that the intervertebral capsule, 

which contains liquid under 

above-ambient pressure, uses its internal 

fluid pressure to alter the apparent 

E′ and E″ of the vertebral column. Because 

blacktip and bonnethead sharks 

were not available for these tests, 

spiny dogfish, Squalus acanthias, were 

used. Fresh 10-vertebrae segments 

were removed from the region of the 

first dorsal fin inthreedogfish. Each 

segment was pressure-clamped at the 

terminal vertebrae and end-loaded 

with bending moments, M (for experimental 

configuration, see Long et al., 

2011). The bending motion was delivered 

via moment arms attached to a 

single-axis linear actuator using an 

MTS model Tytron 250 (Eden Prairie, 

MN) and a 50-N load cell. To test the 

effects of both f and κ on E′ and E″, 

each segment was bent sinusoidally at 

each combination of five f values and 

three κ values. In addition, to test the 

effects of the integrity of the fluid-filled 

intervertebral joint capsule on E′ and 

E″, we repeated this suite of tests 

after (a) puncturing a single joint capsule 

located in the middle of the segment 

and (b) puncturing three joint 

capsules, including the first one punctured 

and two adjoining capsules. 

Increases in f increased only E′ 

( p < 0.05) while increases in κ 

increased both E′ and E″ (Figure 5). 

The only significant effect of puncturing 

the intervertebral capsule was 

when three capsules were punctured, 

and even then only E″ increased. The 


FIGURE 4 

Mechanical properties of the vertebral columns of sharks in sinusoidal bending. Points are means 

from three individuals. Error bars are the standard error of the mean. Size of the symbol indicates 

the relative magnitude of the curvature, κ. Significance level is indicated (n.s. = not significant, 

*p < 0.05, **p < 0.01, ***p < 0.001). 

significance of the variation in E′ and 

E″ was determined using ANCOVA, 

with puncture, f, and κ as main effects 

and individual as the covariate (JMP 

8.0.2., SAS Institute, Cary, NC). 

In summary, the vertebral columns 

of sharks have mechanical properties 

that are highly variable. As species 

and anatomical position change, so, 

too, do E′ and E″. Within a given vertebral 

segment, the apparent storage 

modulus, E′, and the apparent loss 

modulus, E″, can be altered by the 

bending that they undergo. Increasing 

the segment’s curvature,κ, increases 

both E′ and E″; increasing the segment’s 

frequency of bending, f, increases the 

E′. Knowing the mechanical behavior 

of shark vertebral columns under realistic 

bending conditions creates specifications 

for BVCs. 

Designing BVCs 

To begin to understand how to 

control the mechanical behavior of 

BVCs, we built two classes of sharkinspired 

BVC: (1) BVC with variable 

cone angle, Ξ (BVC Ξ ): vertebrae 

were created with variable Ξ and the 

BVC had constant joint length, j, 

and (2) BVC with variable joint length, 

j (BVC j ): vertebrae were created with 

aconstantΞ and the BVC had variable 

j. In addition to exploring the effects 

of the structures Ξ and j, we 

also varied the amount of cross-linking 

of the hydrogel material forming the 

joint. Thus, we explored the BVC 

“morphospace,” the variety of designs 

described by three dimensions: Ξ, j, 

and cross-linking. Part of this exploration 

involved the challenge of making 

composite structures that concatenate 

flexible and rigid elements. After fabrication 

and mechanical testing of both 

classes of BVC, we selected a single 

class, the BVC j , for performance testing 

in a tail-flapping aquatic robot. 

In the BVC Ξ , vertebrae were designed 

in software (SolidWorks, 

Dassault Systèmes SolidWorks Corp., 

Concord, MA) to have the following 

values of Ξ: 15°, 30°, and 45° (Figure 

6). These values correspond to 

low, medium, and high values of Ξ 

measured in sharks (see Figure 3). 

The diameter, d, andaxiallength,c, 

of the vertebrae were fixed at 1 cm 

for both. The j of the column was 

fixed at 0.25 cm. 

Vertebrae were fabricated with a 

rapid prototyper (Z-Corp, model 

310), which produced a porous, plaster 

part that was subsequently infiltrated 

with cyanoacrylate (EZ bond 

5cps, K&R International, Diamond 

Bar, CA). This process yielded vertebrae 

with mean compressive moduli, 

E (MPa) of 43, 50, and 61 for vertebrae 

with values of Ξ at 15°, 30°, 

and 45°, respectively. These values of 

E are within the range measured for 

shark vertebrae (Porter et al., 2006). 

Vertebrae of a given Ξ were assembled 

into a BVC Ξ in two stages. First, 

seven vertebrae were linked together, 

spaced at the fixed j, with eight horse 

hairs (E in tension of 900 MPa) 

arrayedaxiallyandaffixed to the 

outer circumference of the vertebrae. 

These horse hairs served as first 


FIGURE 5 

Mechanical properties of the vertebral column vary as a function of cycle frequency, f, and the integrity 

of the intervertebral joint in the spiny dogfish, Squalus acanthias. Points are means from 

three individuals. Error bars are the standard error of the mean. Size of the symbol indicates 

the relative magnitude of the curvature, κ. Significance level is indicated (n.s. = not significant, 

*p < 0.05, **p < 0.01, ***p < 0.001). 

10% porcine gelatin fixed in 2.5% 

glutaraldehyde (Long et al., 2006). 

Vertebrae were slid onto the hydrogel, 

spaced evenly at the desired j, and affixed 

to the hydrogel with cyanoacrylate 

adhesive. A total of 12 different 

types of BVC j were produced, one 

type for each of the 12 different values 

of j. Three replicates of each type were 

produced and tested. Please note that 

in the BVC j horse hairs were omitted 

because at all but the smallest values 

of j, the hairs cut into the hydrogel during 

bending. 

approximations of the intervertebral 

ligaments found in the vertebral columns 

of sharks. Second, a 10% porcine 

gelatin solution was injected in 

between the vertebrae; the gelatin was 

solidified at 4° C. Once solidified, each 

BVC Ξ was then subjected to one 

of three fixation treatments: 0, 1%, 

or 5 % glutaraldehyde, a chemical 

agent that cross-links the collagen in 

the hydrogel. A total of nine different 

types of BVC Ξ were produced, with 

each possible pairwise combination of 

Ξ and glutaraldehyde concentration. 

In the BVC j , vertebrae were designed 

to have a Ξ of 90°, which created 

ring-shaped vertebrae (Figure 7). 

The d and c of the vertebrae were 

fixed at 0.5 and 1.0 cm, respectively. 

The overall length of the BVC j was 

fixed at 8.4 cm. As the number of 

nonterminal vertebrae were varied 

from 0 to 11, j varied from 720 to 

0.5 mm. These values of j created a 

range that extended below and above 

the range of j measured in sharks (see 

Figure 3). 

Vertebrae were milled from 

Delrin, a polyoxymethylene thermoplastic. 

Delrin has a compressive E 

of 3.1 GPa (Delrin Design Guide, 

Module III, from DuPont), which 

liesinthemiddleoftherangeof 

E values reported for shark vertebrae 

(Porter et al., 2006). 

The ring vertebrae had an inner diameter 

of 0.8 cm, which matched the 

outer diameter of hydrogels made from 

Mechanical Properties 

of BVCs 

The E′ and E″ of the BVCs were 

measured in two different kinds of 

sinusoidal bending test, which corresponded 

to the tests performed on 

sharks’ vertebral columns. In the 

BVC Ξ , 3-point bending tests were 

conducted in a manner identical with 

those on the blacktip and bonnethead 

sharks. In the BVC j , end-loaded bending 

tests were conducted in a manner 

identical with those on the spiny dogfish 

sharks. The E′ and E″ data for the 

BVC j have been analyzed previously 

(Long et al., 2011). In the analysis 

here, the data have been reanalyzed 

to calculate the mechanical work required 

to bend the BVC j and the mechanical 

work recovered as recoil. 

In the BVC Ξ , both E′ and E″ increased 

as the glutaraldehyde concentration 

increased, E′ and E″ decreased 

as the Ξ increased, and E′ increased 

and E″ decreased as κ increased (Figure8).Thesignificance 

of the variation 

in E′ and E ″ was determined 

using ANOVA, with glutaraldehyde 

concentration, Ξ, andκ. asmain 

effects (JMP 8.0.2., SAS Institute, 

Cary, NC). 


FIGURE 6 

BVCs (BVC Ξ ) with variable cone angles, Ξ, and constant joint length, j. 

FIGURE 7 

BVCs (BVC j ) with variable joint lengths, j, and constant cone angle, Ξ. 

Compared to the mechanical properties 

of shark vertebral columns, the 

BVC Ξ have values of E′ that have a 

wider range, overlapping the lower 

values and exceeding the sharks’ higher 

values by an order of magnitude (compare 

Figures 8 and 4). In contrast, the 

E″ values of the BVC Ξ overlap only 

with those of the bonnethead shark; 

the BVC Ξ has much lower values of 

E″ than either the blacktip or spiny 

dogfish shark. Moreover, the E′ for 

BVC Ξ decreases as κ increases; we 

measured the opposite trend in sharks 

(see Figures 4 and 5). Hence, the 

BVC Ξ is not a good biological model 

in this sense. Our hypothesis as to the 

source of this strain softening is that 

the horse hairs force the column to 

bend primarily by compression, rather 

than by a combination of tension and 

compression. 

In the BVC j ,bothE′ and E″ decreased 

nonlinearly as j increased 

(Figure 9). Compared to the mechanical 

properties of dogfish vertebral 

columns, the BVC j span a nearly identical 

range of E′ and E″ values. The 

greatest sensitivity to changes in j occurred 

at the smallest values of j (Figure 

9) in the region that corresponds 

to the j measured in the vertebral columns 

of sharks (Figure 3). In data 

shown elsewhere (Long et al., 2011), 

E ′ and E ″ of the BVC j increased 

with increasing κ, just as in sharks 

(see Figure 5 herein). Moreover the E′ 

increased with increasing f, aslikewise 

seen in sharks (Figure 5). 

The mechanical work to bend the 

BVC j increased with increasing κ and 

increasing E′ (Figure 9). The mechanical 

work recovered as elastic recoil, 

W recoil , is a function of the resilience, 

R, which, over all the testing 

conditions and sizes of joints, averaged 

76%. 

BVCs in Aquatic Vehicles 

The flexible skeletons of sharks and 

fish are inspiring the design of novel 

propulsive systems and aquatic vehicles 

(for review, see Fish, 2006; Long, 

2007, 2011). Fins with life-like flexibility 

have been built to propel a 0.7 m 

long robotic turtle (Long et al., 2006) 

anda0.4mlongroboticknifefish 

(Curet et al., 2011). Bodies with 

life-like flexibility have been built to 


FIGURE 8 

The mechanical properties of the BVCs (BVC Ξ ) with variable cone angles, Ξ, and constant joint 

length, j. Horizontal bars indicate the median, the lower and upper limits of the box indicate the 

25th and 75th percentiles, respectively, and the whiskers indicate the range. 

FIGURE 9 

The mechanical properties of the BVCs (BVC j ) with variable joint length, j, and constant cone 

angle, Ξ. Top row: points represent the means of E′ and E″ pooled across f and κ; error bars 

are one standard error of the mean. Bottom row: points are not pooled. 

propel a 0.7-m long robotic electric 

ray (Krishnamurthy et al., 2010), 

a 0.5-m long robotic trout (Kruusmaa 

et al., 2011), a 0.12-m long mechanical 

sunfish (McHenry et al., 1995), 

and a 0.5-m long mechanical pickerel 

(Conte et al., 2010). Of these 

self-propelled aquatic vehicles, only 

the mechanical pickerel has anything 

resembling a vertebral column: 

a piece of spring steel designed to 

release mechanical work to power 

accelerations. 

The BVC j presented here was invented 

to propel a surface swimming, 

0.3-m long tadpole robot (Long 

et al., 2006; Doorly et al., 2009), 

known as Tadro4 (Figure 10). BVC j 

were attached to a servo motor that 

created a sinusoidally varying pitching 

motion of a tail. That pitch bent the 

BVC j , creating a bending moment 

that propagated down the length of 

the BVC j in a traveling wave that, in 

turn, oscillated the terminal caudal 

fin. In this configuration, without 

distributed muscles, the BVC j acts 

as both a transmission system, transferring 

momentum from the servo motor 

to the caudal fin, and as a propeller, 

directly transferring momentum to 

the surrounding fluid. 

Since Tadro4 was built to behave 

reactively, with sensorimotor feedback 

systems creating foraging and predator 

avoidance, we needed a version that 

could be programmed to swim straight 

using a constant flapping frequency of 

the tail, f, and lateral amplitude of the 

caudal fin. That modified version of 

Tadro4 was called MARMT (Mobile 

Autonomous Robot for Mechanical 

Testing), and it had a hull length of 

17 cm and a tail length of 10 cm 

(Long et al., 2011). 

Outfitted with a given BVC j , 

MARMT’s steady swimming performance 

was measured as swimming 


FIGURE 10 

The aquatic robot, Tadro4, is propelled by a BVC (BVC j ). Tadro4 is a fully-autonomous surfaceswimmer 

with a flattened circular body and propulsive undulatory tail. It is modeled after fish like 

the extinct Drepanaspis and the living electric ray, Narcine. Using sensory input from photoresistors 

and IR proximity detectors, Tadro4 searches for and swims up light gradients while avoiding 

collisions. Tadro4 is propelled by its submerged BVC j , which is wrapped in a thin membrane, attached 

to a caudal fin, and actuated by an oscillating servo motor. Tadro4 was developed by Doorly 

et al. (2009). Photo of Drepanaspis specimen 8462, American Museum of Natural History. Photo 

of adult Narcine is courtesy of Dr. Steve Kajiura. 

of E′ (Figure 11, top panel). The stride 

length of MARMT increased initially, 

doubling as E′ doubled, before tapering 

off. 

When the BVC j operates in the 

flapping tail, MARMT’s swimming 

performance is clearly linked to the 

mechanical properties of the BVC j , E′ 

in the case shown here. Those mechanical 

properties are, in turn, under 

the control of the structure of the 

BVC j . Thus these experiments, taken 

together, demonstrate the functional 

relationship between the structure of 

the BVC j and the performance of a 

self-propelled aquatic robot. 

speed, U, and stride length, the slope 

ofthelineofU regressed onto f, 

which measures the distance the 

robot travels over one period of the 

flapping tail (Figure 11). As f increased 

for any BVC j , so, too, did the U. The 

BVC j with greater values of E′ produced 

a more rapid increase in U, 

over the same range of change in f, 

compared to BVC j , with smaller values 

Summary 

Using the morphology and mechanical 

properties of the vertebral 

columns of sharks as our biological 

target, we built and tested a series of 

BVC. The mechanical behavior of 

the BVCs, measured by the storage 

and loss moduli over a range of bending 

frequencies and curvatures, can be 

altered by changing (1) the material 

properties of the hydrogel that makes 

up the intervertebral joint, (2) the 

length of the intervertebral joint, or 

(3) the shape of the vertebrae. BVCs 

are sufficient to function as propulsive 

elements in swimming aquatic robots: 

in Tadro4 and MARMT the 

BVC converts a simple pitch oscillation 

from a servo motor into a wave 

of bending that drives the caudal fin 

laterally. 

Having identified variables that influence 

the mechanical behavior of 

BVCs, we offer a few observations for 

those wishing to build jointed, flexible 

biomimetic skeletons for use in flexible, 

flapping propulsive systems: 

(1) Design of biomimetic systems: 

Engineered systems that are much 


FIGURE 11 

Swimming performance of a surface-swimming robot, MARMT, propelled by a tail with the BVC j as 

the primary skeleton. MARMT is a version of Tadro4 (Figure 10) modified for mechanical testing 

over a range of flapping frequencies of the tail, f (Hz). For all types of BVC j tested, swimming speed 

of MARMT, U, increased linearly with increases in f (all R 2 values > 0.92). The rate of change of 

U with respect to f is the stride length (distance traveled per period of the flapping cycle); it was 

greatest with BVC j having larger storage moduli, E′. Three replicates of each kind of BVC j were 

tested (N = 36). Means (N = 12) are shown here. Data reanalyzed from Long et al., 2011. 

functional morphology. Next, test 

the morphology’s mechanical behavior 

under physiologically relevant testing 

conditions. Finally, build and test 

simple biomimetic models of the system 

that change just a single structural 

variable over a wide range. Repeat this 

process with different variables, ceteris 

paribus, until the designer knows 

which variables permit the natural system 

and its operational range to be 

mimicked or extended in biomimetic 

form. 


We thank Carl Bertsche, Nicole 

Doorly, Carina Frias, Andres Gutierrez, 

Jonathan Hirokawa, Kira Irving, Doug 

Pringle, Foster Ranney, Hannah 

Rosenblum, Hassan Sahktah, Sonia 

Roberts, Elise Stickles, Josh Sturm, 

and Janese Trimaldi for their help in 

designing, building, and testing vertebral 

columns, BVCs, and the aquatic 

robots. This work was supported by 

the National Science Foundation 

of the USA (DBI-0442269 and 

IOS-0922605). 

simpler than the targeted biological 

system can match and extend 

the targeted range of mechanical 

behaviors. 

(2) Control of mechanical properties: 

The spacing of rigid elements 

in a flexible matrix is more important 

than the shape of the rigid elements 

or the material properties of 

the flexible material. 

(3) Control of reconfiguration: Because 

of the strain- and strain-ratedependence 

of viscoelastic materials, 

no passive, flexible propulsive system, 

if its E′ and E″ matches that 

of the vertebral column of sharks, 

will produce constant motions over 

a wide range of motor inputs. 

This work is a straight-forward example 

of one method of biomimetic 

design (Fish, 2006; Long, 2007): 

describe, test, build, and test. Start 

by identifying a specific operational 

context—aquatic undulatory propulsion 

in this case. Then describe, 

quantitatively, the biological system’s 

Lead Author: 

John H. Long, Jr. 



124 Raymond Avenue, 

Poughkeepsie, NY 12604-0513 

Email: jolong@vassar.edu 

References 

Conte, J., Modarres-Sadeghi, Y., Watts, M.N., 

Hover, F.S., & Triantafyllou, M.S. 2010. A 

fast-starting mechanical fish that accelerates 

at 40 m s-2. Bioinspir Biomim. 5(3):1-9. 

doi: 10.1088/1748-3182/5/3/035004. 






8(60):1041-50. (10.1098/rsif.2010.0493). 

Doorly, N., Irving, K., McArthur, G., 

Combie, K., Engel, V., Sakhtah, H., … 

Long, J.H., Jr. 2009. Biomimetic evolutionary 

analysis: Robotically-simulated vertebrates in 

a predator-prey ecology. In: Proc. 2009 

IEEE Symp. Artificial Life. 147-54. Nashville, 

TN: IEEE (Institute of Electrical and 

Electronic Engineers). doi: 10.1109/ 

ALIFE.2009.4937706. 

Fish, F.E. 2006. Limits of nature and 

advances of technology: What does biomimetics 

have to offer to aquatic robots Appl. 

Bionics Biomech. 3(1):49-60. doi: 10.1533/ 

abbi.2004.0028. 

Grotmol, S., Kryvi, H., Keynes, R., Krossoy, 

C., Nordvik, K., & Totland, G.K. 2006. 

Stepwise reinforcement of the notochord 

and its intersection with the myoseptum: 

An evolutionary path leading to the development 

of the vertebra J Anat. 209:339-57. 

doi: 10.1111/j.1469-7580.2006.00618.x. 

Grotmol, S., Kryvi, H., Nordvik, K., & 

Totland, G.K. 2003. Notochord segmentation 

may lay down the pathway for the 

development of the vertebral bodies in the 

Atlantic salmon. Anat Embryol. 207:263-72. 

doi: 10.1007/s00429-003-0349-y. 

Koob, T.J., & Long, J.H., Jr. 2000. The 

vertebrate body axis: Evolution and mechanical 

function. Am Zool. 40(1):1-18. 

doi: 10.1668/0003-1569(2000)040[0001: 

TVBAEA]2.0.CO;2. 

Krishnamurthy, P., Khorrami, F., de Leeuw, 

J., Porter, M., Livingston, K., & Long, 

J.H., Jr. 2010. An electric ray inspired 

biomimetic autonomous underwater vehicle. 

Am Control Conf. 2010:5224-9. 

Kruusmaa, M., Salumäe, T., Toming, G., 

Ernits, A., & Ježov, J. 2011. Swimming speed 

control and on-board flow sensing of an 

artificial trout. In: Proceedings of IEEE 

International Conference of Robotics and 

Automation (IEEE ICRA 2011). Shanghai, 

China: IEEE (Institute of Electrical and 

Electronic Engineers). 

Long, J.H., Jr. 1992. Stiffness and damping 

forces in the intervertebral joints of blue marlin, 

(Makaira nigricans). J Exp Biol. 162:131-55. 

Long, J.H., Jr. 2007. Biomimetic robotics: 

Building autonomous, physical models to test 

biological hypotheses. Proc. Inst. Mech. Eng, 

C, J. Mech. Eng. Sci. 221:1193-200. 

Long, J.H., Jr. 2011. Biomimetics: Robotics 

based on fish swimming. In: Encyclopedia of 

Fish Physiology: From Genome to Environment, 

ed. Farrell, A.P., 1, 603-12. San Diego: 


Long, J.H., Jr., Koob-Emunds, M., Sinwell, 

B., & Koob, T.J. 2002. The notochord of 

hagfish, Myxine glutinosa: Viscoelastic properties 

and mechanical functions during steady 

swimming. J Exp Biol. 205:3819-31. 

Long, J.H., Jr., Koob, T.J., Irving, K., 

Combie, K., Engel, V., Livingston, N., … 

Schumacher, J. 2006. Biomimetic evolutionary 

analysis: Testing the adaptive value of vertebrate 

tail stiffness in autonomous swimming 

robots. J Exp Biol. 209(23):4732-46. 

doi: 10.1242/jeb.02559. 

Long, J.H., Jr., Krenitsky, N., Roberts, S., 

Hirokawa, J., de Leeuw, J., & Porter, M.E. 

2011. Testing biomimetic structures in 

bioinspired robots: How vertebrae control 

the stiffness of the body and the behavior 

of fish-like swimmers. Integr Comp Biol. 

51(1):158-75. doi: 10.1093/icb/icr020. 

Long, J.H., Jr., Porter, M.E., Liew, C.W., & 

Root, R.G. 2010. Go reconfigure: How fish 

change shape as they swim and evolve. Integr 

Comp Biol. 50(6):1120-39. doi: 10.1093/icb/ 

icq066. 

McHenry, M.J., Pell, C.A., & Long, J.H., Jr. 

1995. Mechanical control of swimming speed: 

Stiffness and axial wave form in an undulatory 

fish model. J Exp Biol. 198:2293-305. 

Porter, M.E., Beltran, J.L., Koob, T.J., & 

Summers, A.P. 2006. Material properties 

and biochemical composition of mineralized 

vertebral cartilage in seven elasmobranch 

species (Chondrichthyes). J Exp Biol. 

209:2920-8. doi: 10.1242/jeb.02325. 

Porter, M.E., & Long, J.H., Jr. 2010. 

Vertebrae in compression: Mechanical 

behavior of arches and centra in the gray 

smooth-hound shark (Mustelus californicus). 

J Morph. 271:366-75. 

Porter, M.E., Roque, C.M., & Long, J.H., Jr. 

2009. Turning maneuvers in sharks: Predicting 

body curvature from body and vertebral 

morphology. J Morph. 270:954-65. 

doi: 10.1002/jmor.10732. 

Schmitz, R.J. 1995. Ultrastructure and 

function of cellular components of the 

intercentral joint in the percoid vertebral 

column. J Morph. 226(1):1-24. 

doi: 10.1002/jmor.1052260102. 

Summers, A.P, & Long, J.H., Jr. 2006. Skin 

and bones, sinew and gristle: The mechanical 

behavior of fish skeletal tissues. In: Fish 

Biomechanics, eds. Shadwick, R.E., & Lauder, 

G.V., 141-77. San Diego: Academic Press. 

Symmons, S. 1979. Notochordal and elastic 

components of the axial skeleton of fishes 

and their functions in locomotion. J Zool. 

189(2):157-206. doi: 10.1111/j.1469- 

7998.1979.tb03958.x. 


PAPER 

Lateral-Line-Inspired Sensor Arrays for 

Navigation and Object Identification 

AUTHORS 

Vicente I. Fernandez 

Audrey Maertens 


Engineering, Massachusetts 

Institute of Technology (MIT) 

Frank M. Yaul 

Department of Electrical 

Engineering and Computer 

Science, Massachusetts 

Institute of Technology 

Jason Dahl 




Jeffrey H. Lang 

Department of Electrical 

Engineering and Computer 

Science, Massachusetts 


Michael S. Triantafyllou 




Introduction 

The lateral line organ is a unique 

sensory mechanism in fish, enabling 

complex behaviors based on the interpretation 

of local fluid mechanics. For 

example, the blind Mexican cavefish 

(Astyanax fasciatus) is able to navigate 

new environments at high speed 

without collision and to identify and 

remember features of the environment 

(Montgomery et al., 2001; von 

Campenhausen et al., 1981). This surprising 

feat is accomplished relying 

primarily on its lateral line organ for 

ABSTRACT 

The lateral line is a critical component of fish sensory systems, found to affect 

numerous aspects of behavior, including maneuvering in complex fluid environments 

with poor visibility. This sensory organ has no analog in modern ocean vehicles, 

despite its utility and ubiquity in nature, and could fill the gap left by sonar 

and vision systems in turbid, cluttered environments. 

To emulate the lateral line and characterize its object-tracking and shape recognition 

capabilities, a linear array of pressure sensors is used along with analytic 

models of the fluid in order to determine position, shape, and size of various objects 

in both passive and active sensing schemes. We find that based on pressure information, 

tracking a moving cylinder can be effectively achieved via a particle filter. 

Using principal component analysis, we are also able to reliably distinguish between 

cylinders of different cross section and identify the critical flow signature information 

that leads to the shape identification. In a second application, we employ pressure 

measurements on an artificial fish and an unscented Kalman filter to successfully 

identify the shape of an arbitrary static cylinder. 

Based on the experiments, we conclude that a linear pressure sensor array for 

identifying small objects should have a sensor-to-sensor spacing of less than 0.03 

(relative to the length of the sensing body) and resolve pressure differences of at 

least 10 Pa. These criteria are used in the development of an artificial lateral line 

adaptable to the curved hull of an underwater vehicle, employing conductive polymer 

technologies to form a flexible array of small pressure sensors. 

Keywords: underwater sensing, artificial lateral line, pressure sensor arrays 

sensory feedback. All fish have this 

organ, although not all use it to the 

extent of the blind cavefish (see 

Montgomery et al., 2001). Many fundamental 

behaviors in fish have been 

identified by biologists to be lateralline-mediated, 

including tracking 

prey by their wake (Pohlmann et al., 

2004) and recognizing nearby physical 

objects (von Campenhausen et al., 

1981). Although the lateral line is a biological 

organ, its functionality would 

translate well to needs in underwater 

vehicle design, such as with object detection, 

navigation, and flow sensing. 

The lateral line organ consists of 

two subsystems responding separately 

to velocity and to pressure gradients 

on the surface of the fish, both using 

the same underlying sensory element, 

the neuromast, which responds directly 

to flow velocity. In the case of the 

pressure gradient measurements, these 

velocity sensors are embedded underneath 

the skin in canals periodically 

opening via pores to the external flow 

(van Netten, 2006). Biological studies 

have demonstrated the ability of the 

fish to use their lateral line to interrogate 

their environment through both 

active and passive sensing. In active 

sensing, a fish uses the flow generated 

by repeatedly gliding near new objects 

at a short distance (von Campenhausen 


et al., 1981) to interrogate their shape. 

In particular, blind cave fish were found 

to detect and discriminate between 

stationary objects or openings of different 

geometries in still water (von 

Campenhausen et al., 1981; Weissert 

& von Campenhausen, 1981; Burt de 

Perera, 2004). During passive sensing, a 

moving object generates a flow field that 

is detected by the stationary lateral line of 

astillfish. Vogel and Bleckmann (2000) 

demonstrated that goldfish use their 

lateral line to passively detect and discriminate 

the size, velocity and shape of 

passing rods in still water. One key element 

emerging from these studies is that 

in both active and passive settings, the 

object identification behavior is tied to 

the pressure gradient measurements of 

the lateral line and appears independent 

of the velocity measurements. 

While the lateral line organ’s role in 

many fish behaviors is becoming progressively 

better understood, there 

still remain many questions about the 

level of information detail available via 

the lateral line. The recent advances in 

understanding the central processing 

of the lateral line have all been with respect 

to the oscillating dipole stimulus 

(Curcic-Blake & van Netten, 2006, 

Goulet et al., 2007), which is inapplicable 

to both passive and active object 

sensing. In the passive case, a dipole 

model neglects the wake that forms 

about a moving object. In the active 

sensing situation, the interaction betweenthetwobodiesinstillwater 

leads to very different pressure distributions. 

Due to the difficulty in studying 

the neurological aspects of how a 

fish utilizes the stimulus of the lateral 

line, an artificial representation of the 

lateral line can give insight on the use 

of pressure sensing for biologically inspired, 

engineered applications. This 

paper investigates the ability of pressure 

sensor arrays, emulating the lateral 

line organ, to distinguish the shapes of 

physical objects through both active 

and passive sensing, while also identifying 

physical requirements for a constructed 

artificial lateral line to be used 

for underwater vehicle navigation applications. 

The results demonstrate 

the promise of a lateral-line-like sensor 

for autonomous underwater vehicles 

(AUVs) in severe environments. 

Previous approaches with artificial 

lateral lines have taken a biomimetic 

approach of reproducing the lateral 

line at the structural level, by recreating 

a sophisticated neuromast-like 

cantilevered sensors and using a canal 

system similar to that of the fish in 

order to obtain pressure gradient 

measurements (Chen et al., 2006; 

Yang et al., 2008). Yet, at its core, 

the portion of the biological lateral 

line associated with the behaviors of 

interest can be viewed as a more elementary 

processing unit, taking distributed 

pressure measurements as 

inputs and, via some processing that 

remains to be understood, extracting 

information about the environment. 

Thus, the current work instead takes 

a bioinspired approach in which the 

lateral line is abstracted by a linear 

pressure sensor array and the resulting 

information is processed relying on 

modern inference algorithms. In addition 

to providing potential clues as to 

the fundamental processes taking 

place in the biological sensory system, 

the present abstracted approach, by directly 

focusing on the desired functionality 

of the artificial lateral line, 

bears significant engineering benefits. 

As a sparse artificial lateral line can be 

readily implemented, making use of 

diaphragm-based pressure sensors, the 

initial focus is able to shift from sensor 

fabrication to the inference and processing. 

In turn, insights from the inference 

results are then used to develop 

specifications for optimal sensor design. 

As another benefit of using pressure 

sensors to approximate the lateral 

line, note that the available pressure 

information includes, but is not limited 

to, discrete pressure gradients (as in the 

lateral line canal system). 

First, we present two experiments 

demonstrating passive object detection 

through the use of an artificial lateral 

line. In this study, an artificial lateral 

line is used to track the size and location 

of a moving cylinder and separately to 

identify the shape of a moving cylinder 

between known possibilities. Second, 

we present an experiment demonstrating 

active object detection, where a 

moving artificial lateral line is used to 

identify the completely unknown 

shape of an object. In this study, the 

sensor arrangement is studied with 

relation to the applied problem in 

order to identify constraints of sensor 

spacing. Finally, we discuss the construction 

and testing of a prototype artificial 

lateral line using conductive 

polymer materials for use in object detection 

and navigation applications. 

Passive Cylinder 

Identification 

In passive object identification, an 

interaction between the object and the 

environment generates a pressure field 

that stimulates the sensor array. In the 

examples considered here, this interaction 

is between a moving cylinder 

and still water. The sensor array is also 

stationary in the water. The problem of 

identifying a moving cylinder via the 

pressure along a lateral-line-like sensor 

is addressed in two stages. First, we discuss 

an approach for determining the 

position and size of a cylinder, focusing 

on a single shape, and subsequently, we 

consider the question of distinguishing 

between different shapes. 


Experimental Setup 

Two experiments were used to analyze 

passive object detection techniques. 

In the first experimental setup 

(see Figure 1A), the position tracking 

ofacircularcylinderwasexperimentally 

tested using off-the-shelf pressure 

sensors (Honeywell 19C015PG4K) 

arranged in a linear array. Seven sensors 

were embedded along a 46-cm 

square flat plate spaced 1.9 cm 

(0.75 inch) apart, as shown in Figure 

1A. Using a mechanical stage 

with millimeter accuracy in the 

x-y plane, a 32-mm diameter circular 

cylinder was passed in front of the sensor 

array at constant velocity and 

known position. These experiments 

were performed in a 3.6 m × 1.2 m × 

1.2 m water tank at the Singapore- 

MIT Alliance for Research and Technology 

(SMART) Centre. Pressure 

signals were amplified by a factor of 

1000 immediately adjacent to the sensors 

using AD620 instrumentation 

amplifiers and were next sampled 

using an NI USB-6210 analog-todigital 

converter. The sensors were 

calibrated using static pressure. 

In the second passive detection experimental 

setup, as with the first, a 

verticallymountedcylinderpassesin 

front of a stationary linear pressure 

sensor array at constant velocity (Figure 

1B). In this experiment, four 

Honeywell 242PC15M pressure sensors 

were enclosed in a streamlined 

cylindrical enclosure. The sensors 

were mounted rigidly with a small diameter 

outlet in order to minimize 

noise. Due to onboard amplification, 

the sensors were only amplified by a 

factor of 10 before data acquisition 

(NI USB 6210). These experiments 

were performed in the MIT towing 

tank facility (36.6 m × 2.4 m × 

1.2 m). As the goal of these experiments 

and subsequent analysis is to 

FIGURE 1 

Schematics of two experimental apparatuses for passive object identification and sample pressure 

data from one apparatus. (A) A moving circular cylinder that is towed past an array of seven stationary 

pressure sensors on a flat wall. (B) An array of four pressure sensors mounted on a stationary streamlined 

body with a square cylinder towed past the body. (C) A representative data set corresponding to 

the layout in part B, filtered with a cut-off frequency of 100 Hz. (Color versions of figures available 

online at: http://www.ingentaconnect.com/content/mts/mtsj/2011/00000045/00000004.) 

distinguish between two cylinder 

cross-section shapes known apriori, 

both square and round cross-section 

cylinders were towed past the sensor 

array. In order to provide a reliable 

common denominator for shape classification, 

the data was gathered from 

a variety of speeds, cylinder sizes, and 

distances from the sensor array as 

shown in Table 1. Calibration of the 

pressure sensors was completed in situ 

by comparing the amplitude of pressure 

oscillations due to regularly generated 

waves, using a wavemaker, to the 

theoretical amplitude based on linearized 

surface wave theory. 

Passive Cylinder Tracking 

Tracking the position and size of a 

circular cylinder can be interpreted as 

the first step or lowest level of shape 

identification. This task is complicated 

by the wake that forms when a cylinder 

moves. In order to track the position 


and size, an accurate and preferably 

simple model for relating the cylinder 

state to the pressure measurements is 

needed as well as a technique for estimating 

the state in real time. Potential 

flow solutions are an immediate candidate 

for modeling an object in a fluid 

since they provide an analytic solution 

with relatively few parameters to define 

the complete flow field. 

The base model chosen for tracking 

a circular cylinder in the present experimentsisaRankinehalf-bodyin 

potential flow. This well-known structure 

is described by the superposition 

of a source and a free stream. A mirror 

image of the half-body provides the 

necessary boundary conditions to 

define the wall in the model. This 

model was found to approximate the 

pressure field of a cylinder with a 

wake when compared with pressure 

measurements obtained through a viscous 

numerical simulation. The radius 

of the cylinder was taken as the distance 

from the location of the source 

to the nearby forward stagnation 

point in the flow. It is important to 

note that the dipole model frequently 

used in a vibratory setting in studies 

with fish (e.g., Coombs & Janssen, 

1990) does not model the flow well 

about a steadily translating body due 

to the absence of the wake in the 

model. It is not surprising that a potential 

flow model like the Rankine halfbody 

approximates the experimental 

situation, since the flow outside the 

separated region in the wake can be 

reasonably considered as potential 

flow. Although the real situation is 

not steady and the pressure far from 

the cylinder was found to converge 

more slowly than in numerical simulations, 

a Rankine model captures the 

amplitude and shape of the pressure response 

well in the immediate vicinity 

of the cylinder. Using the velocity potential 

from this model, the pressure at 

the sensor locations is calculated using 

Bernoulli’s equation. 

Although the pressure sensors used 

in these experiments measure pressure 

with respect to a fix reference value, 

one key element in the analysis for 

tracking a cylinder is to consider the 

difference in pressure between adjacent 

pressure sensors instead of the absolute 

pressure. Long-wavelength, 

small-amplitude disturbances at the 

free surface of the tank, caused by the 

motion of test cylinder, were found to 

cause pressure fluctuations on the 

order of 20 Pa, significant compared 

to the 100 Pa pressure signals from 

the cylinder. Due to the long wavelength, 

taking the difference in pressure 

signals effectively removes this contamination. 

As mentioned in the Introduction, 

the portion of the lateral line that 

has been associated with object identification 

responds to pressure differences 

(Coombs, 2001), so it is interesting 

that the use of pressure differences is 

necessary even when absolute pressure 

measurements are available. 

For the purpose of tracking the 

position and size of a cylinder, the forward 

model relating the hidden parameters 

of interest to the available 

measurements has been defined by 

the Rankine half-body representation 

of the cylinder and wake, in conjunction 

with the Bernoulli equation to 

obtain the pressure at the sensors. To 

tackle the inverse problem and estimate 

the cylinder state based on the 

pressure measurements, several additional 

issues require consideration. 

First, the variables of interest are the 

position (x and y as labeled in Figure 

1A) and radius (R) of the cylinder. 

In order to cast the problem in the 

form of a hidden Markov model (see, 

for example, Cappe et al., 2005), in 

which the state at each timestep depends 

solely on the previous iteration, 

the velocity (u) must to be included in 

the state vector. Using this hidden 

Markov model formulation allows for 

the use of well-known estimation algorithms 

for solving the inverse problem, 

such as the Kalman filter extensions for 

nonlinear models (Gelb, 1974; Julier 

& Uhlmann, 2004). Second, the forward 

model requires a strict constraint 

on two of the state variables, R and y, in 

order to correspond to a physical realization: 

0 < R < y. Unfortunately, there 

is no natural boundary in the measurement 

model that corresponds to these 

physical constraints, as the potential 

flow model works uniformly well and 

there is no sudden shift in the predicted 

pressure. As a consequence of this, an estimation 

technique such as the extended 

Kalman filter fails when it is applied to 

naturally noisy experimental data. 

The successful tracking of the position 

and size of a cylinder was accomplished 

instead using a particle filter 

(see Branko et al., 2004). This general 

inverse problem technique operates by 

using a number of samples, which represent 

possible states. To each sample, 

there is a corresponding weight that 

roughly tracks the quality of that sample. 

At each time step, each sample is 

updated to a new value randomly chosen 

from a distribution based on the 

previous sample, and the weight of 

that sample is updated based on how 

likely it is to have generated the measurements. 

The key advantage of this 

approach is that, in specifying the 

noise distributions that govern the update 

of the samples, it is possible to 

limit the range of state vectors to 

those that correspond to a physical system. 

In the present case, the x position 

and the velocity u are assumed to have 

Gaussian noise distributions, which is 

the typical assumption for an unconstrained 

variable where the noise is 


generated by the accumulation of 

many small random influences. The 

noise distribution associated with the 

y position is assumed to be a lognormal 

distribution, and for the radius 

it is assumed to be a Gaussian distribution 

that is truncated at 0 and y and renormalized. 

These distributions satisfy 

the constraints between the variables 

and have intuitive limiting behaviors. 

When the mean is far from zero with 

respect to the standard deviation, the 

lognormal is approximately symmetric 

about the mean. In the case of the 

truncated Gaussian distribution, if 

the previous value of the radius is 

greater than the new y value, then the 

distribution is weighted heavily to 

larger values and has a maximum 

likelihood of y. Themaincostto 

using a particle filter is that it frequently 

requires a large number of 

samples to generate repeatable accurate 

results. For this application, it was 

found that approximately 150 particles 

are sufficient. 

A typical result of the particle filter 

implementation on experimental data 

is shown in Figure 2. In the figure, 

the black line corresponds to the estimate 

of the variable being output 

from the filter, and the red line corresponds 

to the true value. Note that 

the filter is initialized with a uniform 

distribution and does not begin to 

converge until there are significant 

pressure differences. As seen from the 

results, the particle filter estimate generally 

tracks the x position and radius 

well but underestimates the y position. 

The velocity is also underestimated in 

this example, although both overestimates 

and underestimates of the velocity 

were recorded in general. These 

results demonstrate that using a steady 

potential flow model, the radius and 

aspects of the position can be accurately 

tracked. With further refinements of 

the model, the underestimation of 

the y position could likely be overcome. 

The use of a simple analytical 

model here makes it more likely that 

equivalent information is also available 

to a fish via its lateral line, and similarly 

likely that it could be adapted for 

quickly tracking objects near a hull. 

TABLE 1 

Matrix of experiments for passive cylinder shape classification. 

Passive Cylinder 

Shape Classification 

The second set of experiments is 

concerned with the more complicated 

problemofshapeclassification. Fundamentally, 

we consider whether it is 

possible to distinguish between two 

similar shapes after a single pass of an 

object past the sensor array, using a 

robust test for the decision. The test 

in question must identify whether 

a passing cylinder has a square or 

round cross section. Since the decision 

is based on a single pass, the stochastic 

nature of the flow in the wake of the 

cylinder must be overcome. 

The large data set of pressure obtained 

from the experimental setup in 

Figure 1B, corresponding to cylinders 

of different sizes, velocities, distances, 

and shapes provides a broad range of 

parameters for verifying the ability of 

any decision rule. In addition, the 

large number of runs for each point 

in the test matrix allows for an accurate 

comparison of the mean pressure responses. 

As shown in Figure 3, there 

are distinct differences between the 

average pressure measured due to the 

circular and square cross sections; analogous 

differences exist for comparisons 

in all the tested sizes and velocities. It is 

notable that the most easily distinguished 

differences in the pressure signals 

occur after the zero crossing point 

(circled). 

The analysis presented here develops 

a test to determine, after the fact, 

whether a square or circular cylinder 

has passed the sensors. Based on the 

observations of the mean pressure 

traces, it is tempting to choose features 

such as the location of the minimum 

pressure to classify the shape of the 

cylinder. This approach is limited in 

that the features would be localized 

in the pressure trace and, therefore, 

highly susceptible to the type of noise 

observed in Figure 1C. Instead, we use 

a principal component analysis (PCA) 

to identify a limited number of features 

in the form of weighted sums over the 

full pressure data from a single run. 

This type of feature depends on all the 

data and is less susceptible to localized 

perturbations such as those in the 

wake. By carefully applying the PCA 

to a training data subset, the results 

can be directed towards a decision rule 

in classifying the shape of the cylinders. 

In implementation, PCA works as a 

singular value decomposition of the 

7.62 cm Diameter 5.08 cm Diameter 

0.5 m/s 0.75 m/s 0.5 m/s 0.75 m/s 

Square 100 runs 100 runs 100 runs 100 runs 

Round 100 runs 100 runs 100 runs 100 runs 

Each of the experiments listed corresponds to the cylinder passing at a distance of 5.1 mm from the 

sensors. Additional data, not listed in the table, were collected with the 5.08-cm-diameter cylinders passing 

1.27 cm from the sensors at 0.75 m/s. 


FIGURE 2 

Results of a cylinder tracking experiment using a particle filter. In the first four parts, each vertical 

slice in the surface corresponds to the probability density function for that variable at the time of 

the slice, given the previous pressure measurements. In these sections, the red line represents the 

true value of the parameter at each time instant, and the black line corresponds to the expected 

value of the distribution. The final part shows the corresponding pressure difference measurements, 

filtered at 60 Hz for anti-aliasing. 

sample covariance matrix of the data 

(Jolliffe, 2002; Jackson, 2003). One important 

detail in the implementation is 

that the mean of each initial feature (a 

sample point in the pressure trace) 

must be removed. This mean is taken 

across all training data, not over each 

class. With the mean removed, the 

sample covariance is straightforward 

to calculate. Each resulting principal 

component is a vector that accounts 

for the maximum possible variance, 

subject to having unit area and being 

uncorrelated with all the previous principal 

components. 

Since only four sensors were available 

for this experiment (as opposed 

to seven in the previous cylinder tracking 

experiments) and the cylinder 

stimuli are translating at constant velocity, 

an equivalence between the 

pressure time history and the spatial 

pressure field was used in the analysis. 

This approximation is very accurate in 

regions in front of the cylinder, but less 

so in the unsteady wake areas. In order 

to compare data sets with different cylinder 

velocities on the same spatial 

scale, the 0.5 m/s velocity data is 

down-sampled appropriately. In addition, 

the data is aligned by the point 

at which the pressure crosses zero 

(marked in Figure 3), which forms a 

reliable internal placement marker. 

Finally, the PCA approach does not 

naturally produce features which capture 

the cylinder shape. The decomposition 

of the covariance matrix 

generates principal components that 

capture the majority of the data variance 

in the first few components. 

Therefore, the resulting principal components 

will be most useful if the primary 

cause of differences in the data is 

due to the shape of the cylinders. 

While the variation in the data due to 

the flow separation and the random 

phase of the wake is impossible to 


FIGURE 3 

A comparison of the mean pressure measured at one experimental configuration reveals clear 

differences between the two cylinder shapes. The standard deviation is also plotted, offset by 

−2.0 kPa. 

remove, the velocity of the cylinder 

strongly correlates with the pressure 

signal amplitude and masks the response 

to shape. Normalizing each 

pressure trace by the maximum pressure 

effectively removes this masking 

effect since the pressure amplitude is 

consistently proportional to the velocity 

squared. 

Using the first three principal components 

based on a training set of data 

from a single sensor and covering all of 

the experimental parameters, a highly 

successful classification test is obtained. 

These three principal components 

form a space in which the data 

points generated from experimental 

runs are grouped into two roughly 

ellipsoidal clouds, which can be optimally 

separated by a decision plane, 

minimizing the sum of squared error 

on the training data. Since the projections 

of the data points on each axis are 

based on a linear combination of the 

pressure data with the corresponding 

FIGURE 4 

principal component, a single vector 

of weights can be found that corresponds 

to an axis perpendicular to 

the decision plane (Figure 4B). This 

results in a single linear combination 

using these weights with the 

normalized pressure measurements 

to determine the score of an experimental 

run, with a positive value 

implying a square cross section. When 

applied to the test data (excluding the 

training data), this test produced 

a very small misclassification rate 

of 1.2%. 

While the decision test encapsulated 

in the first two parts of Figure 4 

results in a surprisingly high accuracy, 

the difficulty with using principal 

components is that there is generally 

little corresponding intuition for why 

it works so well. The key element of 

the decision test is given by the decision 

weights, illustrated in Figure 4B. 

There are three distinct main sections 

The elements of the PCA-based classification test (A and B), and the accuracy of the test using 

different subsections of the full decision coefficients (C). 


to the decision weights based on the 

PCA results, labeled II, III, and IV in 

Figure 4B. Region II is the smallest in 

magnitude and corresponds to the region 

between the maximum pressure 

and the zero crossing. Region III corresponds 

to the area of largest difference 

between the two shapes shown in Figure 

3. Region IV weighs an area of the 

pressure response that is directly influenced 

by the wake of the cylinder. This 

region has considerable variation in the 

data, but it is unclear whether it is related 

to the cylinder shape. Regions I 

and IV on either side are very lightly 

weighted. By zeroing out regions, it is 

possible to examine the importance of 

the remaining portion of the decision 

weights to the accuracy in classifying 

the shape. In this zeroing process (Figure 

4C), we find that data on the rear 

side of the zero-crossing point is the 

most important in classifying cylinder 

shape. In fact, the data before the zerocrossing 

point appears to add almost 

no new information for classifying 

the shape. 

The emphasis on region IV implies 

that the wake of the cylinder contains a 

substantial portion of the information 

being used to classify the shape. Initially, 

this may appear to be through 

vortex spacing, governed by the Strouhal 

number. Although this could help 

in classifying the shape if particular 

sizes were being compared, in these experiments 

the large round cylinder and 

the small square cylinder had nearly 

equal vortex spacing, meaning that 

half of the data would be difficult to 

separate by this trait. This observation is 

inconsistent with the 1% error rate. 

The results based on principal components 

demonstrate that it is possible 

to distinguish between two relatively 

similar cross sections of moving cylinders 

based on an artificial lateral line 

sensor. This has been hinted at based 

on experiments with fish, but the 

clear identification of shape without 

regard to changes in velocity or size 

has not been demonstrated. Given the 

importance of the wake in identifying 

the shape in the experiment, it is possible 

that the ability of the lateral line, 

and any artificial analogs, to identify 

shapes does not extend to arbitrary 

shapes. In particular, the bluntness 

of the leading face of the cylinder 

has a strong impact on the location 

and flow direction at the point of 

flow separation. If two shapes with 

similar sharp flow separation are employed, 

it may be much harder to distinguish 

them. 

Active Cylinder 

Identification 

In active object identification, we 

use pressure measurements from an artificial 

lateral line sensor array to locate 

and identify stationary objects in still 

water without prior information on 

their shape. The problem is considered 

from a two-dimensional perspective 

in which a moving fish-like 

body glides at constant speed past a 

column-like stationary object of 

unknown location, size, and crosssection 

geometry. The moving body 

is equipped with a finite number of 

pressure sensors distributed over its 

surface. The viscous effects are 

ignored in the model used to calculate 

the pressure and the limits of this assumption 

are discussed. 

Hassan (1985, 1992) showed that 

in the case of a fish gliding towards a 

cylinder or a wall, the pressure increases 

noticeably in the front region 

when the fish gets close to the obstacle. 

He also showed (Hassan, 1985) that 

for a fish gliding past a cylinder, there 

is a univocal relationship between 

spacio-temporal hydrodynamic signature 

and size and distance to the cylinder, 

an indication that an artificial 

lateral line may be able to distinguish 

these features. More recent studies 

(Sichert et al., 2009; Bouffanais et al., 

2011) have considered parameterizations 

of the shape of an object relevant 

to potential flow models, giving an 

analytic representation of the flow 

field that can be easily related to pressure 

and velocity measurements. 


Experiments were conducted in the 

SMART Centre water tank. The complete 

setup is shown in Figure 5A. The 

FIGURE 5 

Active sensing experimental layout. (A) A picture 

of the experimental set-up used for active object 

identification (the foil is moving towards 

the photographer). (B) A schematic of the 

cross-section of the same experiment showing 

the location of the pressure ports. 


moving body used to both generate the flow and sense pressure was a NACA 

0018 foil (chord c = 15 cm and span s = 60 cm) cast with internal 0.318 cm PVC 

tubing to transmit pressure from taps at the foil’s midspan to the top. Honeywell 

19C015PG4K pressure sensors were mounted on top of the foil, and measurements 

were collected at a sampling rate of 500 Hz via a NI USB-6289 DAQ. 

The location of the sensor ports is shown in Figure 5B. The foil was dragged at 

velocity v = 0.5 m/s past a static cylinder of elliptical cross section oriented at various 

angles. At its closest point, the foil was 5-10 mm away from the cylinder. 

The Forward and Inverse Problem 

It is believed that blind cave fish can encode separately the distance, size and 

shape of objects (Burt de Perera, 2004). Therefore, a convenient and potentially 

biologically relevant way to parameterize the problem is to characterize an object 

by two parameters accounting for its position, one size parameter and several shape 

parameters. Another desirable feature of this parameterization is that the number 

of shape parameters needed to account for the pressure decreases with the 

distance to the object. Bouffanais et al. (2011) proposed such a characterization: 

 

 

SðθÞ ¼ a þ R e iθ ∞ 

þ ∑ μk e ikθ 

k¼1 

¼ x þ iy þ R e iθ þ ∑ 

∞ jμk je ik ð θ 

k¼1 

α kÞ 

 

; θ ∈ ½0; 2πŠ 

wherein a (a = x + iy) refers to the location of the object, R to its size and each μ k 

(μ k =|μ k |e ikα k 

)termisassociatedwitha(k +1)-gonaltypeofperturbationofthe 

shape from that of a circle. As k increases, the impact of the μ k term on the pressure 

field decays very quickly with the distance from the cylinder. For an ellipse, 

as in these experiments, only the first shape parameter μ 1 is non-zero and therefore 

the subscripts will be dropped. 

Given the moving object, its trajectory, and the location, size and shape (a, 

R, μ) of the stationary object, the velocity potential anywhere in the flow field 

can be expressed in terms of a singularity distribution over the surface of the objects. 

The problem is solved numerically using a source panel method: the surface 

of each object is broken up into line segments of constant source strength (364 line 

segments were used for the moving body and 150 for the stationary object). The 

pressure at the sensor locations on the surface of the moving body can then be 

expressed in terms of the potential at these points using the unsteady Bernoulli 

equation. The combination of the parameterization of the object shape and 

the numerical potential flow model defines the forward methodology, relating 

the pressure measurements to the variables of interest (location, size, and 

shape). In experiments, the pressure measured by the sensors is corrupted 

by noise. The two main sources of noise are laboratory noise (electrical and 

mechanical) and the background noise of the fluid flow (which includes viscous 

effects). 

For proper inversion, the technique must be (1) robust to noise, (2) capable of 

handling non-linearity since the pressure does not depend linearly on the characterizing 

parameters of the stationary object, and (3) dynamic, in order to be used 

for navigation of underwater vehicles. 

A particularly suitable algorithm for 

such inversion is the unscented Kalman 

filter (UKF) ( Julier & Uhlmann, 

2004), which is a robust dynamic 

probabilistic signal filtering technique 

for highly nonlinear systems. The 

UKF is more accurate than the more 

traditional extended Kalman filter for 

highly nonlinear problems and does 

not require the computation of derivatives 

for which no analytic expressions 

are available. It also propagates the statistics 

with fewer samples than the 

more powerful particle filter, which 

makes it better suited for real time 

applications. 

Subtle modifications to the UKF 

are necessary before applying it to our 

problem. Non-physical configurations, 

such as bodies intersecting each 

other cannot be identified by an UKF 

assuming Gaussian distributions. This 

can skew statistics, producing unusable 

results. To avoid this problem, if after 

updating the location of the moving 

object, the estimated mean configuration 

is not valid, the estimated position 

is shifted ‘out of the way.’ If the mean 

configuration is valid, the parameter α 

that determines the spread of the samples 

around the mean in the unscented 

transform (Wan & van der Merwe, 

2001) is chosen small enough (for 

each time step) that all the samples 

are valid. 

TheUKFandtheforwardmodel 

are combined to solve the inverse problem: 

locating and identifying a cylinder 

using pressure measurements. In the 

results and analysis discussed here, 

the steady pressure due to the constant 

velocity of the foil has been subtracted 

from the pressure signal. The measurement 

covariance matrix used in the 

UKF was calculated for each run 

based on the pressure measured 0.5- 

0.3 s before the characteristic drop 


of pressure at the second sensor (see 

Figure 7C). 

FIGURE 6 

Results of a cylinder detection and identification in two passes (A and B) using an UKF. The colored 

dashed lines show the true value of the parameters. (a–f) The corresponding shape estimate (red 

circle or ellipse), the actual cylinder (green dotted ellipse), and the position of the foil at various 

times. (Color versions of figures available online at: http://www.ingentaconnect.com/content/mts/ 

mtsj/2011/00000045/00000004.) 

Active Sensing Results 

and Analysis 

Due to the amount of noise in the 

experiments and the fact that it was 

highly correlated, all attempts to simultaneously 

fully identify the location 

and geometry of the cylinder 

were unsuccessful. However, fish 

have been observed to pass several 

times in front of new objects (von 

Campenhausen et al., 1981), and it 

seems reasonable to assume that they 

first locate the objects and estimate 

their size before refining their estimate 

of the shape. A similar approach is used 

here: the first pass is used to get a first 

estimate of the position (x and y) and 

size (R) of the cylinder. A second pass 

using the same data refines the first estimates 

and estimates the shape parameters 

(|μ| andα). This approach is 

consistent with the dynamic hierarchical 

access associated with the shape 

characterization parameters used by 

Bouffanais et al. (2011). An example 

of such a process of cylinder geometry 

reconstruction is shown in Figure 6 for 

a cylinder with shape parameters R = 

3.81 cm and μ = 0.2i and the foil passing 

6.5 mm away from it. 

Both behavioral studies (Weissert 

& von Campenhausen, 1981) and 

mathematical modeling (Hassan, 

1985) have shown that the presence 

of the static object in still water cannot 

be detected until the fish (or here the 

foil) is very close to it (on the order 

of the width of the moving body). As 

canbeseeninFigure6A,assoonas 

the foil is close enough to the cylinder, 

the position (x and y) and size (R) estimates 

converge steadily towards the 

true value of the parameters. The second 

run allows for very good orientation 

(α) estimation (Figure 6B). The 

aspect ratio of the ellipse (|μ|) is the 

more difficult feature to reconstruct. 

The estimated parameter |μ| first wanders 

around the actual value of the 

parameter before decreasing towards 

a lower value. As can be seen in Figure 

6F even though the shape of the 

ellipse has not been exactly reconstructed, 

the estimate of the half of 

the ellipse that is closest to the foil is 

reasonably accurate. This observation 

confirms that object identification 

based on pressure sensing becomes 

less reliable for features further from 

the sensors. 

The results demonstrate that object 

localization and object recognition are 

possible with experimental pressure 

measurements; however, there are limitations 

to the proposed method. The 

main limitation is the use of potential 

flow models to establish the governing 

equations for the filter and the simulations 

(the importance of viscosity on 

the stimulus to the lateral line system 

of fish is also discussed in Windsor & 

McHenry, 2009). A fourth pressure 

sensor placed near the tail of the foil 

measured an oscillating pressure signal 

characteristic of the wake, unlike the 

first three sensors. Particle image 

velocimetry was used to visualize the 

flow and compare it to the flow predicted 

by the inviscid model. As can 

be seen in Figures 7A and 7B, the 

two flows are nearly identical almost 

everywhere, but as the foil passes the 

cylinder, separation occurs over a 

small region on the foil (in the orange 

ellipse). Comparing the pressure measurements 

and simulations (Figure 

7C), we can see that the potential 

flow model gives very good approximations 

of the pressure measurements 

until the second sensor passes the cylinder 

(black dotted line). After that 

point at which separation occurs, the 

inviscid assumption is violated and 

the model is no longer valid. Therefore, 

only the measurements before 

the black dotted line on the figure 


FIGURE 7 

Comparison between simulated (A) and experimental (B) flow field. The green and orange ellipses 

show where the flows differ. (C) The pressure data (plain line) is filtered with a cut-off frequency of 

100 Hz. (Color versions of figures available online at: http://www.ingentaconnect.com/content/ 

mts/mtsj/2011/00000045/00000004.) 

have been used to locate and identify 

the cylinder. 

some of the experiments that still provided 

successful tracking results. The 

pressure data for the active sensing experiments 

was of a similar scale. Any 

sensors used for an artificial lateral 

line application must provide the sensitivity 

and correspondingly low noise 

floor to distinguish small changes in 

pressure of at least 10 Pa. 

The experiments in active sensing 

ofobjectsprovideaguidetotheoptimal 

density of sensors needed for 

similar applications, since the curved 

surface of the foil and the self-generated 

flow increase in the importance 

of the instantaneous measurements 

of the spatial pressure distribution. 

Simulations were performed to examine 

the affect of sensor density on the 

ability to identify a stationary object. 

The simulations consisted of the 

same foil described earlier gliding at 

0.5 m/s and passing 7 mm away from 

a cylinder with the geometry parameters 

R 0 = 1.5 cm and μ =0.2i. 

White Gaussian noise of standard 

deviation 2 Pa was added to the 

simulated pressure measurements. Between 

10 and 70 pressure measurement 

points were evenly distributed 

along the front three quarters of the 

foil (see Figure 8C), and the sampling 

Sensor Array Constraints 

The use of off-the-shelf sensors severely 

limits the density which can be 

achieved in the sensor array while 

maintaining the sensitivity to deviations 

from the static pressure needed 

for engineering applications. From 

the perspective of sensor sensitivity, 

the weakest signals are those associated 

with tracking the position and size of a 

moving cylinder. The maximum absolute 

pressure was as small as 30 Pa in 

FIGURE 8 

Convergence time (A) and final error (B) of the object identification as a function of sensor 

spacing. (C) The location of the sensors for a spacing of 6 mm. 


ate was chosen such that f = (number of sensors) / 20,000. The error was calculated 

as E ¼ 1 ∇R 

5 R 0 

þ ∇x 

R 0 

þ ∇y 

R 0 

þ ∇μ x 

þ ∇μ y 

. For each simulation, the convergence 

time (time elapsed before E < 0.2) and the final error were calculated. 

Each case was simulated 12 times, and the mean and standard deviations of the 

time of convergence and final error were computed (Figures 8A and 8B). The 

goal of this procedure was to identify a sensor density below which the lack of 

spatial density cannot be compensated by a higher sampling frequency. 

Both the convergence time and final error plots suggest that, at least for the 

configuration considered, the performance of the object identification decreases 

as the sensor spacing exceeds 5 mm, which corresponds to a spacing of 0.03 relative 

to the length of the foil. The optimal spacing that these observations suggest 

scales favorably with the actual spacing of lateral line canal neuromasts in the 

trunk canal of the blind Mexican cave fish, which is roughly 0.02 body lengths 

(measured from image in Windsor & McHenry, 2009). This optimal spacing 

derived from simulations is specific to identifying the shape of cylindrical objects 

in which the cross section size is on the same order or slightly smaller than the 

body length. In general, the optimal spacing for measuring the pressure distribution 

may scale with the size of the object being detected as well as the body 

length. In addition, based on the observed separation that occurs on the surface 

of the sensing body, the pressure sensors should be distributed only over roughly 

the forward half of the body in order to maximize the effectiveness of the sensor 

array without waste. 

A New Approach for a Flexible Pressure Sensor Array 

In order to achieve the optimal spatial distributions and sensitivity of pressure 

sensors in an effective artificial lateral line, it is necessary to develop sensor 

arrays on smaller scales than those possible using off-the-shelf technologies. In 

addition to the spatial and sensitivity constraints, it would be necessary that the 

sensor be applied directly on a surface in order for a lateral line sensor to be 

practical for ocean vehicles, instead of fabricating the sensing body around 

the sensor as was necessary in the experiments reported here. To address 

these problems, we have begun to develop a thin, flexible, one-dimensional 

array of pressure sensors to meet the pressure and spatial resolution requirements. 

The array is designed to conform to an AUV’s hull without protruding 

significantly. 

While MEMS pressure sensors are typically made with a silicon substrate 

(Senturia, 2001), the flexible sensor array described here is made entirely of a 

silicone elastomer material, allowing it to be rugged, waterproof, and flexible. 

A conductive polymer is used as the pressure sensitive element in order to function 

while maintaining mechanical compatibility with the rest of the flexible 

structure. 

The conductive polymer is composed of the silicone Polydimethylsiloxane 

(PDMS) doped with conductive carbon black particles (Ding et al., 2007). This 

material has been used for chemical sensors (Andreadis et al., 2007) but not for a 

high-resolution pressure sensor. Prior work involving large pressure sensing arrays 

has concentrated on tactile pressure sensing, which requires reduced sensitivity 

and greater dynamic range (Someya et al., 2004; Harsanyi, 2000). In contrast, 

the carbon black pressure sensors are designed specifically for the lateral line 

application, where small pressure variations 

in the tens of Pascals are of 

interest. 

Pressure Sensor Design 

and Fabrication 

An individual pressure sensing cell 

is depicted and diagrammed in Figure 

9. Its active components consist 

of a resistive strain gauge patterned 

on the surface of a 10-mm-wide, 

1-mm-thick elastomer membrane, 

as shown in Figures 1A and 1B. 

PDMS was chosen as the membrane 

material due to its low tensile modulus 

(Schneider et al., 2008), which 

improves sensitivity. A differential 

pressure across the surfaces of the 

membrane causes the membrane to 

deflect. For small pressures, the deflection 

is linear (Senturia, 2001). The deflection 

then induces strain in the 

resistive strain gauge. The resulting resistance 

change is measured using the 

four terminals of the strain gauge, 

as depicted in Figures 9B and 9C. A 

constant current is applied through 

the outer terminals, and the voltage is 

measured across the inner voltage-tap 

terminals. This four-point probe measurement 

desensitizes the device to variations 

in contact resistance between 

the wires and the strain gauge. The 

voltage taps are positioned to capture 

the greatest resistance change at the 

central edge of the membrane where 

the greatest strain occurs (Senturia, 

2001). 

For a lateral line application on an 

underwater vehicle, the differential 

pressure between sensors in the array 

is of interest, but the depth of the 

aquatic vehicle, which corresponds to 

the absolute pressure, is not. The 

slow-varying absolute pressure may 

be cancelled out by low frequency 

pressure equilibration using the air 


FIGURE 9 

Diagrams and photo of a single pressure sensing cell with a 10-mm square membrane. The device 

is 3-mm thick. 

channel diagrammed in Figure 9A. 

This equilibration prevents damage 

to the sensors caused by the water 

depth at which the vehicle operates. 

It also relaxes the design requirements 

on the sensor, allowing it to have a 

thin membrane that is sensitive while 

not requiring it to withstand large 

pressures. 

In order to verify the performance 

of the carbon black sensing element 

in the context of a flexible pressure sensor, 

a single pressure sensing cell was 

fabricated (Figure 9C). PDMS was 

cast in a mold to form the membrane 

structure overhanging a cavity as 

showninFigure9A.Toformthe 

strain gauge, conductive carbon black 

particles approximately 1 μm in diameter 

were uniformly mixed into 

PDMS. Patterning of the strain gauge 

was accomplished with a stencil mask, 

which yielded a 100-μm-thick layer. 

The carbon black strain gauge terminals 

were bonded to aluminum wire 

with conductive epoxy. For testing 

purposes, a glass slide was used to simulate 

the rigid hull. 

Sensor Performance 

The dynamic response of the single 

test pressure sensor was characterized 

using a manual pressure source independently 

measured using a Honeywell 

pressure sensor. The resistance of 

the test sensor’s carbon black strain 

gauge was recorded as the pressure 

varied. Figures 10A and 10B show 

the resistance of the strain gauge as a 

function of time, compared to the 

pressure measured by the off-the-shelf 

sensor. 

From the dynamic response data in 

Figure 10, the sensitivity can be estimated 

as 0.55% change in resistance 

over 100 Pa. An amplification circuit 

was used to cancel the DC offset voltage 

and amplify the small resistance 

changes by 100, resulting in a roughly 

55% change in voltage over 100 Pa. 

This is more than sufficient for detecting 

pressure variations on the order of 

10 Pa, so it is adequate for the lateral 

line application. 

Figure 10C depicts the applied 

pressure plotted against the measured 

resistance for the dynamic tests in Figures 

10A and 10B, as well as a static 

test. The static test was performed by 

applying a series of pressure steps 

with a 1-min hold time and recording 

the resistance at the end. The pressure 

was stepped up, down, and up again to 

characterize sensor hysteresis. The 

slope of the curves represents the sensitivity 

of the sensor, and the asymmetry 

between the up and down steps 

is indicative of hysteresis. This is expected 

because the PDMS membrane 

is a viscoelastic material that experiences 

both creep and stress relaxation 

(Schneider et al., 2008). Both behaviors 

are byproducts of its low tensile modulus 

but are well understood and can be 

modeled and accounted for using signal 

processing techniques. The creep 

causes increased strain for a given pressure, 

resulting in the static response 

having a greater slope than the dynamic 

response. The dynamic responses are 

more linear and consistent from cycle 

to cycle than the static response, because 

the creep is not significant on 

the timescale of these tests. 

In addition to the polymer creep, 

there is the time-dependent relaxation 

behavior inherent in the conductive 

polymer. From the data in Figures 

10A and 10B, there is a very 

short time constant in the carbon 

black strain gauge resistance data 

when the pressure is being stepped 

up, but a much longer time constant 

when the pressure is being stepped 

down. This is thought to be a product 

of the way the carbon black conductive 

polymer responds to strain (Ding et al., 

2007). However, as shown in dynamic 

response transfer curves of Figure 10C, 


FIGURE 10 

Dynamic and static responses of the single cell pressure sensor, normalized by the strain gauge 

resistance at atmospheric pressure R initial ∼ 20 kΩ. Parts A and B compare the output of an offthe-shelf 

Honeywell sensor with the output of the carbon black pressure sensor when the same 

pressure is applied to both. Part C presents dynamic transfer curves using the dynamic data 

from parts A and B as well as static data obtained by applying a constant pressure to the sensor 

and holding it for 1 min at each data point. 

simulations. Ultimately, the membrane 

thickness and the width of the 

strain gauge patterns will be the limiting 

factors. The carbon black–based 

flexible sensor array is a promising 

technology for the lateral line application 

because it meets the pressure and 

spatial resolution criteria, it has repeatable 

and predictable performance, and 

it can conform to the hull of an aquatic 

vehicle without protruding. 

this behavior is repeatable and consistent 

during cyclic loading, so future 

work will consist of using signal processing 

to compensate for this effect. 

Creating a Sensor Array 

Extending the single carbon black 

sensor cell into an array involves both 

miniaturization of each sensor cell 

and routing of the carbon black strain 

gauges. A prototype array of four pressure 

sensing cells is shown in Figure 

11. The four sensors share two 

common current terminals, and each 

individual sensor has two voltage 

taps, reducing the necessary wiring. 

Each cell has a 5-mm-wide, 0.5-mmthick 

square membrane. The dimensions 

of the cells have been scaled 

down by a factor of 2 from the single 

test pressure sensor. The spatial resolution 

of the array is determined by the 

sensor cell spacing, which is 7-mm 

center-to-center for this array. 

Further miniaturization is possible 

by reducing the dimensions of the 

membrane and strain gauge pattern 

in order to achieve the sub-5 mm spacing 

found optimal in the active sensing 

Conclusion 

This paper has considered several 

aspects of object identification based 

entirely on a lateral-line-like pressure 

sensor array, with the aim of examining 

its potential for application 

in underwater vehicles. Object identification 

was divided into passive object 

identification, where an external flow 

interacts with the object to stimulate 

the sensor array, and active object 

identification, where self-generated 

flow is used to interrogate an object 

in still water. In each case, experiments 

were used to examine the problem of 

estimating the position, size, and 

shape of a cylinder based on a linear 

array of pressure sensors in a realistic 

noise environment. Constraints with 

off-the-shelf sensors in the experimental 

implementation of an artificial lateral 

line, have led to the development 

of a flexible artificial lateral line that 

makes use of a new polymer sensing 

technology. 

The passive sensing of a moving 

cylinder separated the identification 

of the position and size from the 

shape into different experiments. The 

wake generated by the motion of 

the cylinder was a critical component 

in the success of each part. For the cylinder 

tracking, a simple steady potential 

model of the wake was sufficient 

to allow a particle filter to track the 


FIGURE 11 

Diagram and photo of an array of four pressure sensors, each with a 5-mm transparent square 

membrane. The sensors share common current terminals. Each individual sensor has two voltage 

tap terminals to measure the resistance of the part of the carbon black strain gauge located on the 

corresponding membrane. 

position and size in real time. In order 

to distinguish between a square and 

circular cylinder, a decision rule derived 

from PCA was highly successful. 

This rule demonstrated robustness in 

dealing with cylinders of different 

sizes, distances, and speeds. In analyzing 

the decision rule, it was found that 

pressure data in the highly variable 

region near and aft of the flow separation 

is vital in determining the correct 

shape. This implies limitations 

in the types of shapes that can be distinguished 

passively, as dissimilar 

shapes with similar wakes may cause 

difficulty. 

By recreating an active sensing encounter 

with a foil standing in for the 

vehicle or fish, we demonstrated the 

possibility of identifying the location 

and shape of a cylinder without prior 

knowledge of its shape. A potential 

flow model was chosen for simplicity 

and to avoid heavy calculations that 

would make it impossible to run the algorithm 

in real time. Using this model 

with an UKF allowed us to obtain a 

very good object location estimate 

and reasonable ellipse identification 

using three pressure sensors. The results 

stress the importance of the 

head lateral line of the fish (that had 

been emphasized in the case of the 

fish moving towards an obstacle by 

Hassan, 1986) when it passes an object, 

due to the timing and location 

of flow separation. To make more use 

of a biomimetic trunk lateral line, the 

next step is to develop a model that 

takes the separation into account to 

be able to extract information from 

the second half of the data. 

For both active and passive sensing, 

we have found that information about 

thelocationandsizeofacylinderis 

available via an artificial lateral line. 

Beyond this point, however, there are 

strong differences in the two scenarios. 

In the case of passive sensing, the flow 

separation and wake are integral components 

to the pressure distribution 

and must be accounted for from the 

beginning. In fact, it appears the 

shape information may largely be available 

through these components. In 

contrast, for active sensing flow separation 

does not occur immediately, 

allowing it to be ignored in the initial 

shape estimation as done in this 

paper. Using pre-separation data limits 

the number of measurements 

available but allows the use of a general 

model for identifying arbitrary 

shapes. 

Simulations based on the active 

sensing experiments have shown that 

increasing the sensor spacing of 

approximately 0.03 body lengths 

achieves the fastest and most accurate 

object identification for cylindrical objects 

with a diameter on the same scale 

as the sensing body. Coupled with the 

sensitivity (less than 10 Pa) required to 

suitably measure the pressure signals in 

the experiments, this defines the specifications 

for an artificial lateral line 

that could be used to identify or locate 

similar objects for an underwater 

vehicle. Of significant additional concern 

is the need for a sensor that can 

be easily mounted to a hull without 

significantly disturbing the flow. 

Based on the design and test results 

shown, the carbon black-based flexible 

sensor array is a promising technology 

for the lateral line application. Used as 

a sensing element, it can meet the pressure 

and spatial resolution criteria, it 

has repeatable and predictable performance, 

and it is able to conform to 

the smooth curves of a hull while protruding 

only 3 mm. 



The authors gratefully acknowledge 

the support of the Singapore-MIT 

Alliance for Research and Technology 

(SMART) program’s Center for Environmental 

Sensing and Modeling 

and that of the National Oceanic and 

Atmospheric Administration’s Sea 

Grant program under project number 

R/RT-2/RCM-17. 

Lead Author: 

Vicente I. Fernandez 

Department of 


Massachusetts Institute of Technology 

5-424, 77 Massachusetts Avenue 

Cambridge, MA 02139 

Email: vicentef@mit.edu 

References 

Andreadis, N., Chatzandroulis, S., Goustouridis, 

D.,Kosma,V.,Beltsios,K.,&Raptis,I.2007. 

Fabrication of conductometric chemical sensors 

by photolithography of conductive polymer 

composites. J Microelec Eng. 84(5-8): 

1211-4. doi: 10.1016/j.mee.2007.01.050. 

Bouffanais, R., Weymouth, G.D., & Yue, 

D.K.P. 2011. Hydrodynamic object recognition 

using pressure sensing. P R Soc A. 467(2125): 

19-38. doi: 10.1098/rspa.2010.0095. 

Branko, R., Arulampalam, S., & Gordon, N. 

2004. Beyond the Kalman Filter: Particle 

Filters for Tracking Applications. Boston: 

Artech House. 318 pp. 

Burt de Perera, T. 2004. Spatial parameters 

encoded in the spatial map of the blind 

Mexican cave fish, Astyanax fasciatus. 

Anim Behav. 68(2):291-5. doi: 10.1016/ 

j.anbehav.2003.11.009. 

Cappe, O.,Moulines,E.,&Ryden,T.2005. 

InferenceinHiddenMarkovModels.NewYork: 

Springer Science+Business Media, Inc. 652 pp. 

Chen, J., Engel, J., Chen, N., Pandya, S., 

Coombs, S., & Liu, C. 2006. Artificial lateral 

line and hydrodynamic object tracking. In: 

Micro Electro Mechanical Systems 2006, 

MEMS 2006 Istanbul. 19th IEEE International 

Conference on Micro Electro Mechanical 

Systems. pp. 694-7. Istanbul: IEEE. 

doi: 10.1109/MEMSYS.2006.1627894. 

Coombs, S. 2001. Smart skins: Information 

processing by lateral line flow sensors. Auton 

Rob. 11(3):255-61. 

Coombs, S., & Janssen, J. 1990. Behavioral 

and neurophysiological assessment of the lateral 

line sensitivity in the mottled sculpin, 

Cottus bairdi. J Comp Physiol A. 167(4): 

557-67. doi: 10.1007/BF00190827. 

Curcic-Blake, B., & van Netten, S. 2006. 

Source location encoding in the fish lateral 

line canal. J Exp Biol. 209(8):1548-59. 

doi: 10.1242/jeb.02140. 

Ding, T., Wang, L., & Wang, P. 2007. 

Changes in electrical resistance of carbonblack-filled 

silicone rubber composite during 

compression. J Polym Sci Part B Polym 

Phys. 45(19):2700-6. doi: 10.1002/ 

polb.21272. 

Gelb, A. 1974. Applied Optimal Estimation. 

Massachusetts: MIT Press. 382 pp. 

Goulet, J., Engelmann, J., Chagnaud, B., 

Franosch, J., Suttner, M., & Hemmen, J. 

2007. Object localization through the lateral 

line system of fish: Theory and experiment. 

J Comp Phys A 194(1):1-17. doi: 10.1007/ 

s00359-007-0275-1. 

Harsanyi, G. 2000. Polymer films in sensor 

applications: A review of present uses and 

future possibilities. Sens Rev. 20(2):98-105. 

doi: 10.1108/02602280010319169. 

Hassan, E.S. 1985. Mathematical analysis 

of the stimulus for the lateral line organ. 

Biol Cybern. 52(1):23-36. doi: 10.1007/ 

BF00336932. 

Hassan, E.S. 1986. On the discrimination 

of spatial intervals by the blind cave fish 

(Anoptichthys jordani). J Comp Physiol A. 

159(5):701-10. doi: 10.1007/BF00612042. 

Hassan, E.S. 1992. Mathematical description 

of the stimuli to the lateral line system derived 

from a three dimensional flow field analysis: 

I. The case of moving in open water and of 

gliding toward a plane surface. Biol Cybern. 

66:443-52. doi: 10.1007/BF00197725. 

Jackson, J.E. 2003. A User’s Guide to 

Principal Components. New Jersey: Wiley. 

592 pp. 

Jolliffe, I.T. 2002. Principal Component 

Analysis. New York: Springer. 502 pp. 

Julier, S.J., & Uhlmann, J.K. 2004. Unscented 

filtering and nonlinear estimation. 

Proc IEEE. 92(3):401-22. doi: 10.1109/ 

JPROC.2003.823141. 

Montgomery, J.C., Coombs, S., & Baker, 

C.F. 2001. The mechanosensory lateral line 

system of the hypogean form of Astyanax 

fasciatus. Eviron Biol Fishes. 62(1-3):87-96. 

doi: 10.1023/A:1011873111454. 

Pohlmann, K., Atema, J., & Breithaupt, T. 

2004. The importance of the lateral line 

in nocturnal predation of piscivorous 

catfish. J Exp Biol. 207(17):9271-8. 

doi: 10.1242/jeb.01129. 

Schneider, F., Fellner, T., Wilde, J., & 

Wallrabe, U. 2008. Mechanical properties 

of silicones for MEMS. J Micromech 

Microeng. 18(6):065008. doi: 10.1088/ 

0960-1317/18/6/065008. 

Senturia, S.D. 2001. Microsystem Design. 

Boston: Kluwer Academic Publishers. 

720 pp. 

Sichert, A., Bamler, R., & van Hemmen, J. 

2009. Hydrodynamic object recognition: 

When multipoles count. Phys Rev Lett. 

102(5):058104. doi: 10.1103/ 

PhysRevLett.102.058104. 

Someya, T., Sekitani, T., Iba, S., Kato, Y., 

Kawaguchi, H., & Sakurai, T.A. 2004. 

Large-area, flexible pressure sensor matrix 

with organic field-effect transistors for artificial 

skin applications. P Natl Acad Sci. 

101(27):9966-70. doi: 10.1073/pnas. 

0401918101. 

van Netten, S. 2006. Hydrodynamic detection 

by cupulae in a lateral line canal: 

functional relations between physics and 


physiology. Biol Cybern. 94(1):67-85. 

doi: 10.1007/s00422-005-0032-x. 

Vogel, D., & Bleckmann, H. 2000. Behavioral 

discrimination of water motions caused 

by moving objects. J Comp Physiol A. 186(12): 

1107-17. doi: 10.1007/s003590000158. 

von Campenhausen, C., Riess, I., & 

Weissert, R. 1981. Detection of stationary 

objects by the blind cave fish Anoptichthys 

jordani (Characidae). J Comp Physiol A Sens 

Neural Behav Physiol. 143(3):369-74. 

doi: 10.1007/BF00611175. 

Wan, E., & van der Merwe, R. 2001. 

The unscented Kalman filter. In: Kalman 

filtering and neural networks, ed. Haykin, S., 

221-80. New York: Wiley. doi: 10.1002/ 

0471221546.ch7. 

Weissert, R., & von Campenhausen, C. 

1981. Discrimination between stationary 

objects by the blind cave fish Anoptichthys 

jordani (characidae). J Comp Physiol A 

Sens Neural Behav Physiol. 143(3):375-81. 

doi: 10.1007/BF00611176. 

Windsor, S.P., & McHenry, M.J. 2009. The 

influence of viscous hydrodynamics on the 

fish lateral-line system. Integr Comp Biol. 

49(6):691-701. doi: 10.1093/icb/icp084. 

Yang, Y., Chen, N., Tucker, C., Pandya, S., 

Jones, D., & Liu, C. 2008. Biomimetic flow 

sensing using artificial lateral lines. ASME 

Conf Proc. 2007(43025):1331-8. 


PAPER 

A Conserved Neural Circuit-Based 

Architecture for Ambulatory and 

Undulatory Biomimetic Robots 

AUTHORS 

Joseph Ayers 

Anthony Westphal 

Daniel Blustein 

Department of Biology and 

Marine Science Center, 

Northeastern University 

Introduction 

The innate behavior of underwater 

animals provides an effective model for 

the adaptive behavior of unmanned 

underwater vehicles (Ayers, 2004). 

Underwater animals must respond to 

a broad variety of environmental challenges 

including turbidity, hydrodynamic 

flow, heterogeneous and 

highly structured bottom types and 

impediment. Their relatively neutral 

buoyancy renders them especially susceptible 

to hydrodynamic perturbation. 

As a result, they have evolved a 

behavioral set that includes a broad 

variety of compensatory responses to 

perturbation. This behavioral set results 

from layered exteroceptive reflexes 

responding to exteroceptive 

sensor input resulting from changes 

in orientation relative to gravity, impediment, 

chemical cues, and hydrodynamic 

and optical flow (Ayers, 

2004; Blustein & Ayers, 2010). 

These layered exteroceptive reflexes can 

form taxic responses to point sources 

of sound or chemicals (Westphal 

et al., 2011). As the point sources 

form motivational cues for goal achieving 

behavioral sequences, they can 

ABSTRACT 

The adaptive capabilities of underwater organisms result from layered exteroceptive 

reflexes responding to gravity, impediment, and hydrodynamic and optical 

flow. In combination with taxic responses to point sources of sound or chemicals, 

these reflexes allow reactive autonomy in the most challenging of environments. 

We are developing a new generation of lobster and lamprey-based robots that operate 

under control by synaptic networks rather than algorithms. The networks, 

based on the command neuron, coordinating neuron, and central pattern generator 

architecture, code sensor input as labeled lines and activate shape memory alloybased 

artificial muscles through a simple interface that couples excitation to contraction. 

We have completed the lamprey-based robot and are adapting this sensor, 

board, and actuator architecture to a new generation of the lobster-based robot. The 

networks are constructed from discrete time map-based neurons and synapses and 

are instantiated on the digital signal processing chip. A sensor board integrates inputs 

from a short baseline sonar array (for beacon tracking and supervisory control), 

accelerometer, a compass, antennae, and optionally chemosensors. Actuator 

control is mediated by pulse-width duty cycle coding generated by the electronic 

motor neurons and a comparator and power field-effect transistor (FET) system 

housed on low- and high-current driver boards. These circular boards are stacked 

in a tubular hull with the processor and batteries. This system can readily mimic the 

biomechanics of the model organisms by the addition of hydrodynamic control surfaces. 

The behavioral set results from chaining sequences of exteroceptive reflexes 

released by sensory feedback from the environment. 

Keywords: biomimetic, robot, UUV, lobster, lamprey 

guide reactive autonomy in the most 

challenging of environments. The 

task is to capture these performance 

advantages in engineered devices. 

The Biological Model 

We are developing a new generation 

of lobster and lamprey-based 

robots that operate under control by 

synaptic networks rather than algorithms. 

Previous generations of these 

vehicles were controlled by finite 

state machines that were organized 

around the elements of the corresponding 

neurobiological models (Ayers 

et al., 2000; Ayers & Witting, 2007). 

The neuronal circuits that control 

our current generation of vehicles are 

based on the command neuron, coordinating 

neuron, central pattern generator 

(CCCPG) architecture (Figures 1a 

and 1b) of innate animal behavior 

(Kennedy & Davis, 1977; Stein, 1978; 

Pearson, 1993). The networks are organized 

into segmental central pattern 

generators (CPGs) that control appendages 

or axial body musculature in the 


FIGURE 1 

CCCPG architecture with exteroceptive reflex. Labeled circles represent neurons; synapses are 

shown as connecting lines with triangular (excitatory) or circular (inhibitory) endpoints. (a) Neuronal 

circuit-based controller for a walking lobster robot. The effector organs of each body segment 

are controlled by CPGs that contain a neuronal oscillator, a pattern generator and sets of 

motor neuron pools. The CPGs are coordinated among themselves by a set of coordinating neurons 

(CoN) that provide information about the activity status of a governing oscillator to a governed 

oscillator. The CPGs are brought into operation by a set of command neurons (CN) that 

initiates their operation and controls their average frequency and amplitude. (b) Neuronal 

circuit-based controller for a swimming lamprey robot. Slight modification of the CCCPG architecture 

and effectors transforms the system’s motor output from walking to swimming. (c) The 

CNs are organized into exteroceptive reflexes that are released by neuronally coded sensor information 

(rounded rectangles: heading from a compass, target orientation from SBA) through sensory 

interneurons, which mediate in place rotation and yaw during locomotion. 

animal models and the robots (Ayers 

et al., 2010). The CPGs are coordinated 

among themselves by a category 

of neurons called coordinating neurons 

that pass status information 

from a governing CPG to a governed 

CPG that alters its period to remain 

coordinated at a particular phase, 

depending on the ratio of intrinsic frequencies 

of the governing and governed 

CPGs (Selverston & Ayers, 

2006). This temporal resetting occurs 

on a cycle-by-cycle basis to entrain the 

CPGs in a particular gait in the case of 

walking or to ensure propagation of a 

wave of flexion down the body during 

undulation (Figure 1). 

The CPGs are brought into operation 

and modulated by a category of 

neurons called command neurons 

(Kupfermann & Weiss, 1978). Command 

neurons generally constitute the 

locus at which the decision to evoke a 

behavioral act is made and project 

from the brain through the central nervous 

system to bring the segmental 

CPGs into operation. They typically 

perform this process through the 

mechanism of neuromodulation 

through second messengers that alter 

both the cellular properties and synaptic 

connectivity within the CPG 

(Dickinson, 2006). By this mechanism 

or through direct synaptic 

modulation, the same CPG can often 

produce variations on a behavioral act 

in response to different commands 

(Selverston & Ayers, 2006). 

The sensors we employ are configured 

to encode sensory input as a labeled 

line code (Bullock, 1968). In this 

form of coding, each sensory neuron is 

a unique source of information. The 

information consists of (1) the nature 

of the sensory stimulus (light, optical 

flow, chemicals, bumps, etc.), (2) the 

receptive field or position of the stimulus 

on the body and (3) the magnitude 

of the stimulus coded as an 

action potential train where the frequency 

of the action potentials is 

proportional to the logarithm of the 

stimulus intensity. We configure 

these sensory elements in networks 

that filter out features of the environment 

through lateral inhibition, 

range fractionation and motion detection. 

These filtered outputs provide 

input to the command neurons to 

release behavior. 

We have completed the lampreybased 

robot and are adapting this sensor, 

board, and actuator architecture to 

a new generation of the lobster-based 

robot (Figure 2). The lamprey robot 

features an electronic nervous system 

that we are adapting to the new 

lobster-based robot. The key feature 

of this architecture is that it is generalizable 

between all animal models. 

Electronic Nervous Systems 

Thediscretetimemap-based 

(DTM) neuron and synapse equations 

(Rulkov, 2002) phenomenologically 

modelneuronalactivity.Themodel 

has two state variables x and y, two 

FIGURE 2 

Biomimetic robots. (a) Fourth generation lobsterbased 

robot. (b) Second generation lampreybased 

robot. 


control parameters α and σ, anda 

parameter β for integrated synaptic 

input. Variations in α and σ can configure 

neurons into a silent type, a 

spiking type, a bursting type and a 

chaotic type (Ayers & Rulkov, 2007). 

Similar control parameters for the synapse 

instruments determine the synaptic 

strength, relaxation rate, release 

threshold, and reversal potential that 

determine whether the synapse is excitatory 

or inhibitory. The electronic nervous 

systems are first prototyped in 

the National Instruments LabVIEW 

software. Neuron and synapse instruments 

are configured with different 

properties, and the modeled 

neurons and synapses are wired together 

in LabVIEW. 

Figure 3 demonstrates a simple 

CPG circuit configured to illustrate 

operation of the four types of neurons 

in our CPGs. Here, a command neuron 

(1) initiates an oscillation between 

a bursting neuron (2) and a spiking 

follower (3) using a slow modulatory 

synapse. A fourth coordinating neuron 

(4) can be activated in bursts to entrain 

the bursting pattern evoked by the 

command. In contrast to coordinating 

neurons that reset the timing of the oscillation 

on a cycle-by-cycle basis by 

perturbation, command neurons initiate 

operation of the circuits and modulate 

their average frequency and amplitude 

as parameters (Figure 3b–c). 

FIGURE 3 

DTM network integration. (a) The modeled neuronal circuit: (1) command neuron, (2) bursting 

neuron, (3) spiking neuron, (4) coordinating neuron. (b) Parametric modulation of 2 and 3 generates 

an antagonistic bursting pattern. Voltage vs. simulation iteration traces are shown for each 

component of the network. The trace below neuron 1 shows the current injected to initiate activity. 

(c) Perturbation of the bursting pattern in 2 by a coordinating neuron (3) to entrain the evoked 

rhythm. The trace below 4 shows the injected current into that neuron. Adapted from Ayers and 

Rulkov (2007). 

FIGURE 4 

Configuration of robots. (a) Current implementation of board set of the lamprey-based robot. The 

sonar stack processes the analog hydrophone signals from the short baseline array (SBA). The 

electronic nervous system is instantiated on a Texas Instruments TMS320C6727 chip on the CNS 

DSP board. Sensors and the SBA processor are housed on the sensor array board. Low- and highcurrent 

drivers provide the current pulse trains that activate the nitinol actuators. The robot operates 

on a 12-V, 4.5-Ah NiMH battery pack. (b) Configuration of the lamprey robot. A hinge in the 

pitch module allows the undulator to alter its pitch relative to the hull to dive or climb. (c) Configuration 

of the lobster robot. A tubular hull similar to the lamprey robot houses the electronics 

and batteries. Anterior claw-like and posterior abdomen-like hydrodynamic control surfaces provide 

a thrust vector into the substrate to increase traction. Externally mounted sensors include 

optical flow detectors, hydrodynamic flow sensors on the antennae, and sonar transducers for 

beacon tracking and supervisory input. 

Board Architecture 

The networks that control the robots 

are constructed from DTM neurons 

and synapses in procedural C and 

are instantiated on a Texas Instruments 

digital signal processing (DSP) 

chip. A common board architecture 

is used to control both the swimming 

and walking robots (Figure 4a). 

The board set consists of four types: 


(1) The DSP board houses the DSP 

chip and interconnects to sensor and 

actuator boards. (2) A sensor array 

houses a compass, inclinometers, accelerometers 

and a processor board to 

derive azimuth and inclination deviation 

signals from the short baseline 

array stack. (3) A low-current driver 

board receives logic signals from the 

motor neuron output from the DSP 

chip and in turn controls (4) a highcurrent 

driver that applies current to 

heat the individual nitinol actuators. 

Artificial muscles constructed from 

the shape memory alloy nitinol move 

both the walking legs and the undulatory 

body axis. The nitinol is operated 

on a thermal cycle. When cooled by 

seawater, it can be deformed into martensite 

state that is associated with 

about a 5% length increase. When 

heated by electrical current, it transforms 

into the austenite state and contracts 

rapidly. Increases in the length 

of one muscle are produced by the 

contraction of its antagonist. Both 

the amplitude and velocity of the contractions 

can be graded by pulse-width 

duty cycle modulation of the drive 

pulses. Excitation-contraction coupling 

with the motor neurons is mediated 

by a comparator circuit on the low 

current boards that thresholds the 

action potentials to generate a square 

wave pulse that controls a power on 

the high-current boards to activate 

the actuators. Changes in the firing frequency 

of the motor neurons provide 

the duty cycle modulation. 

A common DSP board (Figure 4a) 

interfaces the sensor array board to 

the current driver boards. A separate 

regulator board provides the load to 

these boards via a 12 V NiMh battery 

pack. Feed-through connectors in the 

end caps lead the current conductors 

to the actuators. The boards are 

stacked in a tubular hull whose length 

can be varied to accommodate a variety 

of mission packages. 

Behavioral Set 

This system can readily mimic the 

biomechanics of the model organisms 

by the addition of hydrodynamic control 

surfaces. Turns in the undulatory 

robot are mediated by modulation 

of the amplitude of the flexions to 

thetwosidesasintheanimalmodel 

(Ayers, 1989). The direction of propagation 

of the flexion waves along the 

body axis can be reversed to mediate 

backward swimming. Dives and 

climbs can be mediated by alteration 

of the pitch of the hull relative to the 

undulator (Figure 4b). Dorsal flexion 

of the hull generates a low-pressure 

area above the hull to mediate a climb 

while ventral flexion generates a lowpressure 

area below the hull to mediate 

diving. 

The primary response to hydrodynamic 

flow in the lobster is to orient 

into the flow, lower the anterior control 

surfaces and elevate the posterior 

control surfaces. As the lobster is only 

slightly negatively buoyant, this creates 

a thrust vector into the substrate and 

increases the traction of the legs on 

the bottom. The three degree of freedom 

walking legs of the lobster robot 

allow the vehicle to walk in all directions 

(Ayers & Witting, 2007). Alterations 

in the degree of depression can 

regulate the height above the substrate, 

while variations along the long body 

axis regulates pitch. Biasing the depression 

on the two sides can correspondingly 

regulate roll to maintain primary 

orientation on tilted substrates. 

The behavioral set of both robots 

is organized around exteroceptive reflexes 

(Kennedy & Davis, 1977). An 

innate releasing mechanism composed 

of sensory neurons and interneurons 

filters incoming information to extract 

relevant features of the environment 

such as bumps, tilt, hydrodynamic 

and optical flow (Figures 1c and 5a). 

These sensory releasers are coded in interneurons 

that, in turn, activate command 

systems. The interneurons use 

lateral inhibition from low-threshold 

to high-threshold elements to produce 

range fractionation so that different 

ranges of a scalar input are coded by 

different sensory neurons, providing 

the capability for detailed circuit logic. 

An example of such exteroceptive 

reflexes are those involved in the mediation 

of the yaw plane responses to 

hydrodynamic flow that occur during 

rheotaxis. Lobsters typically walk 

with their antenna projected to the 

front (Figure 5a, I). If wave surge 

occurs from the side, it bends the 

upstream antenna medially and the 

downstream antenna laterally (Figure 

5a, II). Our hypothesis is that 

this perturbation activates a rheotaxic 

interneuron that activates the backward 

walking command on the upstream 

side and the forward walking 

command on the downstream side. 

This would cause the animal/vehicle 

to rotate in place into the flow. While 

orienting into flow, the animals project 

their antenna laterally, which would 

switch control to another bilateral 

pair of surge interneurons (Figure 5a, 

III). As the most upstream antennae 

would be bent more than the more 

downstream antenna, and these interneurons 

project to the contralateral 

forward walking commands, the 

animal/vehicle would continue to 

yaw into the flow until current to the 

two antenna is balanced ensuring 

proper orientation into the flow for 

maximal hydrodynamic stability. The 

hydrodynamic control surfaces can 

then ensure proper traction to overcome 

the perturbation. 


FIGURE 5 

Neural simulation of rheotaxis. (a) Epochs of a lobster’s rheotaxic behavioral response to water 

surge shown with the predominant active neural reflex circuit. In the network diagrams, top circles 

represent sensory neurons corresponding to high (H), medium (M), or low (L) antennal bending in 

the lateral or medial direction. Black ovals represent interneurons that project to bilateral commands 

for forward (F) or backward (B) walking. In I, as a lobster walks forward, bilaterally balanced 

low lateral bending of the antennae is observed, which serves to sustain forward locomotion. In II, 

left-to-right water surge (blue arrows) causes a high medial bend of the ipsilateral antenna and a 

high lateral bend of the contralateral antenna eliciting rheotaxis. In III, bilaterally asymmetrical 

lateral antennal bending mediates yawing upstream during forward walking. (b) Voltage vs. 

time traces for the neurons of the rheotaxis circuit. Dashed lines distinguish the behavioral epochs 

shown in a. 

alter locomotory outputs of a network, 

as we have shown to produce both 

swimming and walking. Even hybrid 

systems can be achieved, such as the 

gait of an alligator resulting from a 

combination of the lamprey and lobster 

CPG networks. Sensor packages 

can be adapted with little restructuring. 

Layered exteroceptive reflex 

networks provide capabilities for navigation, 

investigation and obstacle negotiation 

in unpredictable near-shore 

marine environments with a minimum 

of supervisory control. 

Lead Author: 

Joseph Ayers 

Department of Biology and 

Marine Science Center 

Northeastern University, 

East Point, Nahant MA 01908 

Email: lobster@neu.edu 

This overall control scheme applies 

to a variety of environmental circumstances 

and perturbations in the yaw, 

pitch, and roll planes. Many exteroceptive 

reflexes form taxic systems. 

For example, the three hydrophone 

short baseline sonar array (SBA) on 

the lamprey robot reports the deviation 

of the sonar beacon relative to 

the hull orientation in terms of inclination 

and azimuth (Westphal et al., 

2011). The azimuthal signal modulates 

swim command systems to 

cause the vehicle to yaw toward the 

beacon (Figure 1c) while the inclination 

signal modulates the pitch system 

to cause the vehicle to climb/dive 

toward the beacon. Taken together 

these layered reflexes will cause the 

vehicle to home on a sonar beacon. A 

similar 2-D SBA is planned for the 

lobster robot to control yaw taxis. 

The sonar transducers also provide 

a capability for supervisory control. 

The vehicles will be sent supervisory 

commands that specify a heading and 

odometry information for distance. 

The command will include a propensity 

to negotiate or investigate obstacles 

depending on the mission. At the 

end of the search vector, the vehicle 

would annunciate its location to a 

long baseline sonar array and be sent 

a new search vector. By this mechanism 

an operator could supervise several 

robots simultaneously. 

Conclusion 

The neuronal mechanisms of 

innate behavior can be applied to a 

broad variety of biomimetic underwater 

robots. Minimal modification 

of neuronal components can easily 

References 

Ayers, J. 1989. Recovery of oscillator function 

following spinal regeneration in the Sea 

Lamprey. In: Neuronal and Cellular Oscillators, 

ed. Jacklett, J., 371-406. New York: Marcel 

Dekker Inc. 

Ayers, J. 2004. Underwater walking. Arthropod 

Struct Dev. 33(3):347-60. doi: 10.1016/ 

j.asd.2004.06.001. 

Ayers, J., & Rulkov, N. 2007. Controlling 

biomimetic underwater robots with electronic 

nervous systems. In: Bio-mechanisms of 

Animals in Swimming and Flying, eds. 

Kato, N., & Kamimura, S., 295-306. 

Tokyo: Springer-Verlag. 

Ayers, J., Rulkov, N., Knudsen, D., Kim, 

Y-B., Volkovskii, A., & Selverston, A. 2010. 

Controlling underwater robots with electronic 

nervous systems. Appl Bionics Biomech. 

7:57-67. doi: 10.1080/11762320903244843. 

Ayers, J., Wilbur, C., & Olcott, C. 2000. 

Lamprey robots. In: Proceedings of the 

International Symposium on Aqua 


Biomechanisms, eds. Wu, T., & Kato, N., 

Tokai University. 

Ayers, J., & Witting, J. 2007. Biomimetic 

approaches to the control of underwater 

walking machines. Philos T R Soc Lond A. 

365:273-95. doi: 10.1098/rsta.2006.1910. 

Westphal, A., Rulkov, N., Ayers, J., Brady, D., 

& Hunt, M. 2011. Controlling a lampreybased 

robot with an electronic nervous system. 

Smart Struct Syst. 7(6):471-84. 

Blustein, D., & Ayers, J. 2010. A conserved 

network for control of arthropod exteroceptive 

optical flow reflexes during locomotion. Lect 

Notes Artif Int. 6226:72-81. 

Bullock, T.H. 1968. Representation of information 

in neurons and sites for molecular 

participation. P Natl Acad Sci. 60:1058-68. 

doi: 10.1073/pnas.60.4.1058. 

Dickinson, P.S. 2006. Neuromodulation of 

central pattern generators in invertebrates and 

vertebrates. Curr Opin Neurobiol. 16(6): 

604-14. doi: 10.1016/j.conb.2006.10.007. 

Kennedy, D., & Davis, W.J. 1977. Organization 

of invertebrate motor systems. In: 

Handbook of Physiology, eds. Geiger, S.R., 

Kandel, E.R., Brookhart, J.M., & Mountcastle, 

V.B., Section 1, Vol. 1, Pt. 2:1023-87. 

Bethesda, MD: American Physiological 

Society. Section 1, Part 2. 

Kupfermann, I., & Weiss, K. 1978. The 

command neuron concept. Behav Brain 

Sciences. 1:3-39. 

Pearson, K.G. 1993. Common principles of 

motor control in vertebrates and invertebrates. 

Annu Rev Neurosci. 16:265-97. doi: 10.1146/ 

annurev.ne.16.030193.001405. 

Rulkov, N.F. 2002. Modeling of spikingbursting 

neural behavior using two-dimensional 

map. Phys Rev E. 65:041922. doi: 10.1103/ 

PhysRevE.65.041922. 

Selverston, A., & Ayers, J. 2006. Oscillations 

and oscillatory behavior in small neural circuits. 

Biol Cybern. 95:537-54. doi: 10.1007/ 

s00422-006-0125-1. 

Stein, P.S.G. 1978. Motor systems, with 

specific reference to the control of locomotion. 

Annu Rev Neurosci. 1:61-81. doi: 10.1146/ 

annurev.ne.01.030178.000425. 


PAPER 

A Hybrid Class Underwater Vehicle: 

Bioinspired Propulsion, Embedded 

System, and Acoustic Communication 

and Localization System 

AUTHORS 

Michael Krieg 

Peter Klein 

Robert Hodgkinson 


and Aerospace Engineering, 

University of Florida 

Kamran Mohseni 1 

Departments of Mechanical 

and Aerospace Engineering 

and Computer Engineering, 

Institute for Cyber Autonomous 

Systems, University of Florida 

Introduction 

Traditionally unmanned underwater 

vehicles fall into one of two categories. 

One class of vehicles (torpedo 

like) are built to travel long distances 

with minimal energy and are usually 

characterized by a long slender body, 

a rear propeller for propulsion and a 

set of fins to provide maneuvering 

forces. This type of vehicle is poorly 

suited for missions requiring a high degree 

of positioning accuracy because 

the control surfaces provide little to 

no maneuvering force at low forward 

velocity. The other class of vehicle 

similar to remotely operated vehicles 

(ROVs) is designed to operate in 

these situations, which do require 

1 This work was started while the group was at 

the University of Colorado. 

ABSTRACT 

Inspired by the natural locomotion of jellyfish and squid, a series of compact 

thrusters series is developed for propulsion and maneuvering of underwater vehicles. 

These thrusters successively ingest and expel jets of water in a controlled manner 

at high frequencies to generate propulsive forces. The parameters controlling 

the performance of the thrusters are reviewed and investigated to achieve higher 

thrust levels. The thrusters are compact and can be placed completely inside a vehicle 

hull providing the desired maneuvering capability without sacrificing a sleek 

hydrodynamic shape for efficient cruising. The system design of a prototype hybrid 

vehicle, called CephaloBot, utilizing these thrusters, is also presented. A compact 

and custom-developed embedded system is also designed for the CephaloBot. Key 

features of the system include a base set of navigational sensors, an acoustic 

system for localization and underwater communication, Xbee RF transceiver for 

communication above water, and a LabVIEW programmed processing board. 

Keywords: AUV, thruster, bioinspired, communication 

high positioning accuracy, and incorporate 

several thrusters at various locations 

to provide maneuvering forces 

in all directions. However, this class 

of vehicle typically has a very high 

drag coefficient due to the abundance 

of external thrusters and cannot travel 

to remote locations without additional 

support. 

The abundance of remote marine 

research sites requiring high positioning 

accuracy for inspection, as well as 

the desire to create fully autonomous 

vehicle sensor networks, has inspired 

significant research in a hybrid class 

of vehicles with the efficient cruising 

characteristics of the torpedo class 

and the maneuvering abilities of the 

ROV class. Some take a mechanical 

approach moving the maneuvering 

propellers into tunnels which run 

through the hull of the vehicle 

(Mclean, 1991; Torsiello, 1994) or 

into the fins themselves (Dunbabin 

et al., 2005). Others observe that nature’s 

swimmers have a healthy balance 

of long-distance endurance and highaccuracy 

low-speed maneuvering. Vehicles 

have been designed to use fins 

for both high-speed maneuvering as 

well as mimic the low-speed flapping 

of turtles and marine mammals (Licht 

et al., 2004; Licht, 2008; Kato, 

2011); and some use tail fins as a primary 

means of propulsion (Barrett 

et al., 1999). Our inspiration comes 

from the cyclical jet propulsion seen 

in jellyfish, scallops, octopus, squid 

and other cephalopod. Squid jet propulsion 

produces the fastest swimming 


velocities seen in aquatic invertebrates 

(O’Dor & Webber, 1991; Anderson 

& Grosenbaugh, 2005). 

Jetting locomotion begins when 

the squid inhales seawater through a 

pair of ostia behind the head, filling 

the mantle cavity (see Figure 1). The 

mantle then contracts forcing seawater 

out through the funnel that rolls into a 

high-momentum vortex ring and imparts 

the necessary propulsive force 

(Anderson & Grosenbaugh, 2005). 

The versatility of the system permits 

two distinct gaits, cruising and escape 

jetting (Bartol et al., 2008). During 

cruising, squid swim at nominal 

speed with a greater efficiency than escape 

jetting, which involves a hyperinflation 

of the mantle followed by a 

fast powerful contraction to impart 

significant acceleration at the cost of 

both muscular and fluid dynamic 

losses. Bartol et al. (2009) report cruising 

mode efficiency at 69% (±14%) 

averaged over several species and 

swimming speeds and 59% (±14%) 

for escape jetting. Additionally, propulsive 

efficiency was seen to rise as 

high as 78% in adult L. brevis swimming 

at high velocities and averaged 

87% (±6.5%) for paralarvae (Bartol 

et al., 2008), challenging the notion 

that a low-volume high-velocity jet 

inherently negates a high propulsive 

efficiency. 

The locomotion of jellyfish tends 

to be very similar to that of squid 

with some key differences, primarily 

that the refilling phase of jellyfish 

FIGURE 1 

Diagram of squid layout and locomotion. 

swimming uses the same bell opening 

as the jetting phase. Despite the fact 

that squid do not use the funnel during 

refilling, the inlet vents are still 

on the anterior side of the mantle cavity, 

meaning that locomotion for both 

organisms is quite different from traditional 

pumping mechanisms. Jellyfish 

use the cyclic jetting process for 

feeding as well as locomotion as is 

evidenced by Lagrangian coherent 

structures and particle tracer analysis 

(Lipinski & Mohseni, 2009; Wilson 

et al., 2009); in addition, both squid 

and jellyfish utilize jetting for respiration, 

taking advantage of the large 

fluid flow rates. Both of these factors 

can make it difficult to determine 

which swimming behaviors are optimized 

for propulsion versus secondary 

functions. Similar to the different gaits 

seen in squid locomotion, different 

species of jellyfish generally fall into 

two categories of swimmers based on 

the ‘quality’ of vortex ring they produce. 

Jellyfish like moon jellyfish 

have a very large bell opening, and 

the jetting motion is similar to a paddling 

type motion. Box jellyfish and 

otherfasterswimmingjellyfish have 

smaller bell openings with nozzle-like 

flapsandhaveamuchmoredistinct 

jet. Jellyfish morphology during swimming 

has been digitally captured from 

experiment, and the body motions 

were imported into numerical simulations 

to predict body forces on the 

swimming jellyfish, determining drastically 

different swimming efficiencies. 

Froude propulsive efficiency of jellyfish 

was directly calculated by Sahin 

and Mohseni (2008, 2009; Sahin et al., 

2009) to be 37% for Aeqorea victoria 

and 17% for Sarsia tubulosa. It should 

be noted that both species of jellyfish 

most likely do not use vortex 

generation for the sole purpose of 

locomotion. Aeqorea victoria uses vortex 

generation for feeding and Sarsia 

tubulosa as an escape mechanism. Empirical 

data gathered through digital 

particle image velocimetry (DPIV) 

measurements of several species shows 

similar efficiency characteristics for 

the different swimming patterns 

(Dabiri et al., 2010). 

The general concept of propelling 

water craft by ejecting a high-velocity 

water jet is centuries old, was hypothesized 

by both Bernoulli and Benjamin 

Franklin, and was utilized in a rudimentary 

sense in one of the first steam 

boat designs by James Rumsey (Allen, 

2010). Continuously pumped jets are 

usedforpropulsioninmodernwater 

craft-like jet skis and bow thrusters 

of motorboats; however, this type of 

jet propulsion is inherently different 

from the propulsion of squid and jellyfish, 

which create distinct vortex rings. 

The thrusters of this paper also produce 

finite jets, which form arrays of 

vortex rings, and should be considered 

fundamentally different than continuous 

jet thrusters. 

This paper showcases a complete 

hybrid class vehicle that demonstrates 

added maneuvering capabilities utilizing 

a set of bio-inspired jet thrusters. 

The manuscript will focus on three 

primary systems of the vehicle: a bioinspired 

thruster system and fundamentals 

of thruster mechanics, an 

acoustic system, which serves the 

dual purpose of communication and 

localization, and a compact embedded 

control system. The manuscript is 


organized as follows. The mechanics 

ofthethrusteraswellasthethrust 

dynamics are described in ‘Vortex 

Ring Thrusters’ section. A brief history 

of thruster and vehicle prototypes is 

given in ‘Thruster and CephaloBot 

Evolution’. The‘Hybrid Vehicle Description’ 

section gives basic requirements 

for a hybrid class vehicle, and the 

subsystem components of CephaloBot 

are described in more details in ‘Communication 

Localization System’, 

‘Embedded System’, and‘Sensors’ 

sections. 

Vortex Ring Thrusters 

Ourthrusterinspiredbyjellyfish 

and squid propulsion consists of an internal 

fluid cavity, with a semi-flexible 

plunger used to drive fluid motion, 

and a small circular orifice exposed to 

the external fluid. See Figure 2 for a 

diagram of the thruster layout. The 

cavity of the thruster provides the 

same functionality as the squid mantle 

or the jellyfish bell, expanding and 

contracting to cycle water in and out 

of the circular orifice (functionally 

similar to the squid funnel/siphon). 

Since the thruster generates propulsion 

by creating energetic vortex rings, it 

FIGURE 2 

Conceptual diagram of the thruster key components. 

Jet shown as hypothetical slug of 

fluid. 

is termed the ‘vortex ring thruster’ 

(VRT). 

Since VRTs are contained internal 

to the vehicle (with only a small opening 

on the vehicle surface), they do not 

significantly affect the vehicle’s forward 

drag profile, which means that a 

vehicle equipped with a set of VRTs 

for low-speed maneuvering and a rear 

propeller for primary propulsion will 

have a sleek aerodynamic shape allowing 

fast efficient cruising to a site of 

interest but still maintain full maneuverability 

(even at zero forward speed) 

upon reaching that site of interest. See 

Krieg and Mohseni (2010) for ‘parallel 

parking’ capability of an earlier version 

of our vehicle. Additionally, since 

the VRT only needs a single opening 

(unlike tunnel thrusters or traditional 

pumps, which extend from one end 

of the hull to the other), it allows for 

a greater degree of freedom for internal 

system arrangement. 

The impulse generated by this type 

of device can be modeled as if the jet 

acts like a solid slug of fluid with a uniform 

velocity across the nozzle opening 

(Mohseni, 2004, 2006; Krieg & 

Mohseni, 2008). Properties of fluid 

slug have been investigated by several 

groups (Glezer, 1988; Gharib et al., 

1998; Shariff & Leonard, 1992; 

Mohseni & Gharib, 1998; Krieg & 

Mohseni, 2008). The total impulse 

generated for a single pulsation under 

slug assumptions is 

I slug ðÞ¼ρπ=4∫ t 

t 0 u2 ðτÞD 2 dτ; 

ð1Þ 

Krueger and Gharib (2003) showed 

that the impulse created by a cylinder 

piston type vortex generator was consistently 

higher than the impulse predicted 

by the slug model. This added 

impulse was attributed to a pressure 

gradient at the nozzle exit plane, 

referred to as ‘nozzle overpressure.’ 

Adding I p (the impulse due to overpressure), 

we get an equation for the 

impulse, in terms of the nozzle pressure, 

p (which is a function of time 

and radial position), and the stagnation 

pressure, p ∞ . 

IðÞ¼I t slug ðÞþI t p ðÞ t 

I p t 

ðÞ¼∫ t 0∫ A pr; τ 

½ ð Þ p ∞ ŠdAdτ ð2Þ 

We performed initial testing on a 

thruster which periodically ingested 

and expelled jets with a sinusoidal velocity 

program; which was chosen for 

simplicity of fabrication. Assuming 

that there is no net momentum transfer 

during the ingestion phase (fluid 

being taken into the cavity starts at 

rest outside of the thruster and ends 

at rest inside the cavity) and ignoring 

the pressure impulse (which will be 

accounted for with a coefficient term 

post analysis), then the average thrust 

produced over a full cycle can be calculated 

in terms of the thruster frequency 

to be (Krieg & Mohseni, 2008) 

 

‐ 

T ¼ ρπ3 L 2 

16 D4 f 2 ð3Þ 

D 

where u is the piston velocity (mass 

flux across thruster opening divided 

by nozzle area), D is the nozzle Diameter, 

ρ is the fluid density, t is the time 

at which the impulse is evaluated, and 

τ is a dummy variable for time initialized 

at the beginning of pulsation. 

Here f is the frequency of actuation, 

and the term L=D is the jet stroke 

ratio. If the jet maintained its shape as 

a solid cylinder the stroke ratio would 

be the ratio of length to diameter of 

that cylinder (see Figure 2). The stroke 

ratio has also been called the formation 


FIGURE 3 

Thruster static testing environment. 

time since it is equivalent to the time 

since initiation of the jet flow scaled 

by jet velocity and nozzle diameter, 

t* ¼ L=D ¼ ∫ t 0 u ð τ Þdt=D. Theformation 

time is closely related to the 

jet formation dynamics. When a jet is 

expelled into a stationary fluid, the viscous 

forces cause the initial portion of 

the jet to roll into a tightly wound vortex 

ring. As more fluid is expelled, it 

feeds the growing vortex ring until 

a critical saturation point is reached 

and the vortex ring can no longer support 

the added circulation. At this 

point, the vortex ring separates from 

the remaining shear flow. The formation 

time when the jet has achieved the 

same circulation as the final vortex ring 

is known as the formation number. 

Gharib et al. (1998) demonstrated 

that impulsively started vortex rings 

have a universal formation number 

(≈3.6-4.2) independent of jet velocity 

and diameter. However, numerical 

studies have shown the formation 

number to be drastically lower for 

jets created with a parabolic velocity 

profile (Rosenfeld et al., 1998) or a 

2-D jet velocity like those produced 

in conical nozzles (Rosenfeld et al., 

2009). 

The slug model predicts that the 

average thrust is proportional to both 

the square of the actuation frequency 

and the square of the stroke ratio/ 

formation time. To test this assertion, 

the thruster was placed in a static fluid 

reservoir and suspended from a load 

cell, and the thrust output was measured 

directly. This testing setup is 

shown in Figure 3 (Krieg & Mohseni, 

2008). It should be noted that if the 

vehicle is moving during thruster 

pulsation there will inherently be a 

non-zero momentum transfer during 

ingestion; however, as was previously 

mentioned these thrusters are primarily 

intended for low-speed maneuvering 

(vehicle velocities well below 

jetting velocities), so that the momentum 

transfer during ingestion will still 

be negligible compared to the momentum 

transfer of the jetting phase. In 

addition, the expelled jet rolls into an 

isolated vortex ring which inherently 

produces a large momentum transfer 

associated with the fluid impulse 

of the vortex ring itself. During ingestion, 

the small internal cavity severely 

limits vortex ring formation as well as 

momentum transfer associated with it. 

The thruster was tested over a wide 

range of stroke ratios and actuation frequencies. 

As can be seen in Figure 4, 

the thrust shows a square proportionality 

to frequency, within a certain frequency 

range. When producing jets 

with low stroke ratio, this range this 

is the entire sub-cavitation frequency 

range. However, when the jet stroke 

ratio goes above the formation number, 

the thruster exhibits a parabolic 

dependence on the actuation frequency 

after a short range of square dependence. 

To more clearly show this 

trend, we define a scale factor, which 

FIGURE 4 

Thrust versus frequency. Each set of markers 

shows thrust data for a different stroke ratio. 

The error bars plotted along with the thrust 

relationship represent a single standard deviation 

of the thrust data. 

is a measure of the accuracy of the 

slug model for various operating conditions 

α = T Exp / ‐ T (Krieg & Mohseni, 

2008) where ‐ T is the slug model predicted 

average thrust from equation (3). 

This scale factor is plotted with respect 

to frequency for stroke ratios below 

the formation number in Figure 5a 

and for stroke ratios above the formation 

number in Figure 5b. Note that 

the thruster of this study has a nozzle 

which is essentially a flat plate with a 

circular orifice in the middle. This 

type of nozzle produces a 2-D jet 

flow similar to a conical nozzle, meaning 

that the jet formation number is 

closer to 3 (Rosenfeld et al., 2009). 

For a more in depth analysis of the 

shift in formation number due to the 

2-D aspect of the jet, see Krieg and 

Mohseni (2011). 

First, consider the thrust response 

of the actuator operating below the 

formation number (Figure 5a). In the 

low-frequency regime, the scale factor 

is higher than that predicted by the 

slug model due to the nozzle overpressure, 

reaching 1.4 times the predicted 

value. Krueger and Gharib (2003) 

showed that the pressure impulse can 

reach as much as 40% of the total 


FIGURE 5 

Scale factor (slug model accuracy) versus frequency for stroke ratios below (a) and above 

(b) the formation number. Error bars shown on data points indicate a standard deviation of 

measured values at that frequency (also taken in the scaled space). 

FIGURE 6 

Successive frames of jet flow showing the 

thruster re-ingesting wake flow. 

impulse for low stroke ratio jets corresponding 

to a total impulse 1.6 times 

the predicted slug model impulse. 

However, as the frequency increases, 

the total thrust settles on the value predicted 

by the slug model, meaning that 

in this frequency range the impulse 

due to overpressure during expulsion 

is equal to the impulse due to “underpressure” 

during refilling so that the 

net impulse transfer is that predicted 

by the slug model. Now consider 

the thruster response when operating 

above the formation number, shown 

in Figure 5b. Again the low-frequency 

ranges exhibit an added impulse due 

to the nozzle overpressure; albeit to a 

lower extent as observed by Krueger and 

Gharib (2003). But the high-frequency 

range exhibits an added loss in thrust 

with respect to the slug model prediction. 

This relative loss is seen to 

increase monotonically with both actuation 

frequency and stroke ratio. 

This suggests that another assumption 

made in the slug model is no longer 

valid when operating above the formation 

number. We assume that this loss 

in model accuracy is tied into the assumption 

made that all fluid being 

ingested between pulsations is at rest 

outside of the thruster. When a jet is 

ejected with a stroke ratio above the 

formation number, some of the shear 

flow is left behind in the trailing 

wake of the leading vortex ring. The 

trailing wake has a lower momentum 

than the leading vortex ring and travels 

at a much lower induced velocity but 

still has a forward momentum substantially 

larger than the surrounding resting 

fluid. Therefore, the loss in slug 

model accuracy could be explained 

by the thruster ingesting some of 

the trailing wake during the refilling 

phase. Figure 6 shows successive 

frames from a video of the thruster’s 

forming jet (at a high stroke ratio) 

where some of the trailing wake is ingested 

back into the thruster. 

It should also be noted that the 

scale factor results are only presented 

above an actuation frequency of 4 Hz, 

because the thruster was designed to 

operate cyclically rather than generate 

individual pulsations. The 2-D nature 

of the jet created by this type of 

thruster (orifice nozzle) has an added 

effectonthenozzleoverpressurenot 

seen in the frequency ranges presented. 

This effect is fully explained (along 

with a more in depth description of 

impulse generation) for a single pulsation 

with constant jet velocity in Krieg 

and Mohseni (2011). Despite the 

relative magnitude of the overpressure 

impulse (with respect to the momentum 

impulse, I slug ), it is only observed 

in low actuation frequencies, which are 

coupled with low thrust output. Often 

vehicle mission scenarios will necessitate 

a large magnitude thrust output 

(higher actuation frequencies), where 

the overpressure is cancelled out, and 

the slug model provides an accurate 

thrust measurement. 

The vortex ring formation phenomenon 

plays a key role in the jet 

locomotion process in squid as well. 

Bartol et al. (2009) observed that squid 

have two distinct swimming gaits. In 

the efficient cruising gate jets are expelled 

below the formation number, 

so that the majority of the jet rolls into 

the primary vortex ring. Alternatively, 

in threatening situations the squid employs 

a swimming technique referred 

to as escape jetting; which begins with 

the hyperinflation of the mantle followed 

by a fast contraction expelling a 

jet well above the formation number. 


Presumably this behavior indicates that 

this type of jet propulsion is most efficient 

when expelling jets below the formation 

number and that jetting above 

the formation number can achieve higher 

thrust at the expense of fluid losses. 

Therefore, all subsequent vehicle 

thrusters have been designed with a 

set diameter resulting in a stroke ratio 

near the formation number; to achieve 

a maximum level of thrust, while still 

being accurately described by the slug 

model. It should be noted that the formation 

number for a jet created on a 

moving vehicle will not be the same 

as the formation number of the jet in 

the static setup. Krueger et al. (2006) 

showed that vortex rings formed in 

the presence of a uniform background 

co-flow have a lower formation number 

than vortex rings formed in a resting 

fluid. Since a moving vehicle will 

inherently induce a co-flow with the 

jetting direction, this effect must be 

taken into consideration. However, 

the reduction in formation number 

is proportional to the ratio between 

co-flow velocity and jet velocity; therefore, 

the low-frequency (low jet velocity) 

pulsation will experience pinch off 

at an earlier formation time, but the 

lower-frequency pulsation is also less 

effected by the dynamics of cyclic vortex 

ring formation. 

One advantage of the VRT is that 

it produces a desired level of thrust 

almost instantaneously (Krieg & 

Mohseni, 2010). Propeller style thrusters 

suffer from a rise time associated 

with reaching the static thrust level 

after initiating rotation. This rise time 

is inversely proportional to the desired 

level of thrust and can be on the order 

of several seconds for low thrust levels 

(Fossen, 1991; Yoerger et al., 1990). 

VRTs also have a rise time associated 

with reaching the desired level of thrust, 

which is inversely proportional to the 

level of thrust. However, this rise time 

is an order of magnitude smaller for 

VRTs. The exact thrust program as a 

function of time is sinusoidal, due to 

the nature of the thruster; however, 

using several thrust data sets to average 

out the dynamic component and fitting 

the average thrust to a basic logarithmic 

curve the mean rise time can be observed. 

The fitted curves for several operational 

frequencies, and a stroke ratio 

of 4.3 is shown in Figure 7. 

FIGURE 7 

Mean thrust produced at various frequencies 

(static desired thrust level) versus time. Rise 

time inversely proportional to thrust level. 

Along with a minimal rise time, the 

VRT is also immune to a thrust lag seen 

in tunnel thrusters (Mclean, 1991), 

where thrust continues to be exerted 

on the vehicle after the thruster has 

been terminated. 

Thruster and CephaloBot 

Evolution 

The compact thrusters used in vehicle 

testbeds to provide validation of 

static testing have taken a wide variety 

of forms. The first generation utilized 

a solenoid driving mechanism and a 

flexing diaphragm to expel fluid (Figure 

8a). The solenoid driving mechanism 

suffered from reduced stroke 

at higher actuation frequencies as the 

solenoid stroke was load dependent. 

All of the subsequent iterations have 

used mechanical driving mechanisms 

for better flexibility in adjusting the operation 

conditions. Gen. 2 (Figure 8b) 

used a completely flexible cavity in a 

fitted mold, whereas Gen. 3 and 4 

switched to a semi-flexible cavity 

which was reinforced to ensure constant 

diameter but allow compression 

in height. Gen. 3 (Figure 8c) used a 

complicated encompassing cylindrical 

tube cam to drive cavity compression, 

but mechanical complexities and reliability 

issues caused us to simplify to 

a basic crank shaft design for Gen. 4 

(Figure 8d). In the figure, this thruster 

is shown with an acrylic casing to allow 

the components to be seen. The vehicle 

ready thruster uses an aluminum 

casing (Clark et al., 2009). 

Along with the thrusters themselves 

the vehicle testbeds housing 

the thrusters have evolved rapidly. Figure 

9 shows the evolution of vehicle 

testbeds used to demonstrate the feasibility 

of maneuvering using VRTs. 

Starting with the oldest vehicle at the 

bottom and successive generations 

upwards, the first vehicle only had 

fins to provide a standard for maneuvering 

capabilities, 2nd, 3rd and 4th 

generation vehicles contain 1st, 2nd 

and 3rd generation thrusters, respectively, 

and increasing levels of autonomy. 

The lessons learned from these 

vehicles directly lead to the development 

of the most recent hybrid class 

vehicle described in the following 

section. 

Hybrid Vehicle Description 

The newest generation of vehicle 

is intended to be used in autonomous 

sensor network applications. Therefore, 

the vehicle must be able to travel 

on long range missions collecting data 


FIGURE 8 

Successive generations of thrusters (a) utilize solenoid driver (7.5 cm/3-inch diameter), (b) utilize cavity and mold (10 cm/4-inch casing diameter), 

(c) have encompassing driving mechanism (12.5 cm/5-inch plate diameter) and (d) utilize simple crank shaft and semi-flexible ducting (10 cm/4-inch 

plate diameter). 

butmustalsobecapableofautonomously 

docking with permanent 

support structures. These structures 

will be responsible for downloading 

the vehicle’s mission data, changing 

mission objectives and recharging vehicle 

batteries. 

Autonomous docking with such a 

structure is a complicated problem, 

requiring not only high-accuracy maneuvering, 

but equally high-accuracy 

localization and attitude determination. 

In addition vehicles in this type of environment 

need to be capable of communicating 

with each other at short 

FIGURE 9 

The first four generations of our vehicle test 

beds. Oldest vehicle at bottom, and successive 

vehicles placed in ascending order. 

FIGURE 10 

Fifth generation hybrid vehicle CephaloBot. 

distances for coordinated missions. In 

general reduction in cost and internal 

structure are also desirable to allow 

for more vehicles in the network with 

a wider range of payload options. 

The base vehicle (Figure 10) is separated 

into three hull sections. The 

front and back sections house all of 

the actuators. Each section has two 

VRTs and an active buoyancy control 

device (BCD). Each thruster has an 

overall diameter of 7.6 cm (3 inches), 

a nozzle diameter of 1.8 cm (0.6 inch), 

has a stroke ratio of 4.5, and produces 

close to 2N of thrust at 30 Hz before 

cavitation starts to occur in the cavity. 

The back section also has the rear propeller 

motor. The primary batteries 

and all the electronics except motor 

control are housed in the center section. 

An additional payload section 

may be added between front and center 

to include additional sensors, 

devices, and/or batteries. Regulated 

power and digital communication 

lines interface the payload to the center 

section. When put together, the 

0.15 m (6 inches) diameter, 0.92 m 

(36 inches) long vehicle weights a neutrally 

buoyant 16 kg (36 lb). The 

BCDs can change the buoyancy by 

±1% to dive or surface, and 2 kg of 

internal ballast are adjustable to balance 

the pitch of the vehicle. 

Communication/ 

Localization System 

Due to the physical properties of 

water, RF communication methods 


are not practical for small unmanned 

underwater vehicles. CephaloBot has 

a joint acoustic communication and 

localization system which is ideally 

suited for underwater sensor network 

applications. The communication 

technique is based on binary frequency 

modulation whereas the localization 

methodology is based on a time delay 

of arrival technique. The system consists 

of a specialized hydrophone array 

(fabricated in house as described later 

in this section), which interpret acoustic 

signals for both information and 

directional content. The receiving hydrophone 

array consists of three piezo 

electric ceramics spaced in a triangular 

arrangement parallel to the vehicle 

principle axis plane (Figure 11). Two 

piezo electric ceramics are placed on a 

line parallel to the pitching axis, and 

the 3rd is extended along the roll axis 

from the midpoint of the other two. 

The entire array is encased in urethane 

rubber with an acoustic impedance 

similar to water. The transmitting 

node consists of a single high-power 

piezo electric ceramic encased in the 

same type of urethane rubber as the 

FIGURE 11 

Definition of principle submarine axes. 

hydrophone array. Each vehicle is 

equippedwithbothareceivingand 

transmitting node on the underside 

of the vehicle (see Figure 12). 

FIGURE 12 

Transducer nodes on the vehicle. Receiving 

node on the right and transmitting node on 

the left. 

All three hydrophones in the receiving 

node receive a signal from a 

transmitting node (either on another 

vehicle or a docking station). The 

phase lag (ϕ) between the signal coming 

from the hydrophone on the roll 

axis and the signals coming from the 

two hydrophones on the pitching axis 

is measured and correlated with the 

signal frequency ( f ) to determine the 

time between when the hydrophones 

received the source signal Δt i = ϕ i /f. 

The time lag is then multiplied by 

the speed of sound in water to get 

relative source distances in the vehicle 

frame and transformed into the inertial 

frame to get the azimuth and elevation 

of the receiving node with respect to 

the source node. 

The hybrid localization/communication 

system is comprised of three 

main phases: data sending, data receiving 

and localization. Data receiving 

and localization both use the 

same incoming acoustic wave, the 

first 1000 cycles are dedicated to localization 

and the remaining cycles are 

frequency modulated. The acoustic 

wave propagates from a sending transducer 

located at some other node and 

propagates towards the vehicle. Once 

this wave is received by the hydrophone 

array, it is conditioned by filtering 

and amplification circuitry. 

The conditioning circuitry contains a 

17-dB gain pre-amplifier followed by 

two stages of filtering and two stages 

of amplification to bring the signal 

voltage to 5 V. The electrical voltage 

is then passed through exclusive or 

gates with one of the other signals (in 

the case of hydrophone 0, the signal is 

also passed to demodulation circuitry). 

During the localization portion of receiving 

an onboard microcontroller 

reads and stores the output of each of 

the three exclusive-or gates and also 

determines which of the three signals 

arrives first. The duty cycle of the 

exclusive-or gates directly relates to 

the phase difference of the signals. The 

localization hardware calculates and 

passes the azimuth and elevation angles 

(which defineaconeofpossiblevehicle 

locations with the source node at the 

vertex) to the main navigation processor 

which uses secondary positioning sensors 

(depth sensor and electronic compass) 

to determine a unique solution to 

the position of the receiving node with 

respect to the transmitting node. 

After the localization portion of 

receiving is complete the output of 

the exclusive-or gates are ignored and 

the microcontroller reads and stores the 

DC voltage level from a frequency to 

voltage converter which directly represents 

the frequency of the incoming 

frequency modulated signal. 

In order for the localization methodology 

to function properly the hydrophones 

must be placed no more 

than λ/2 apart where the wavelength 

λ is equal to the speed of propagation 

divided by the frequency (λ =c/f ). 

This leads to a maximum spacing of 

3 cm for a frequency of 25 kHz and 

1.875 cm for a frequency of 40 kHz. 


Miniature hydrophones which are approximately 

1 cm in diameter are commercially 

available from Reson but are 

on the order of $1000 per unit. Halfinch 

diameter cylindrical piezo electric 

ceramics (SMC14H12111) are available 

from Steminc for less than $10/ 

each. Placement of the three halfinch 

diameter piezo electric ceramics 

in a triangle pattern yields a maximum 

navigational frequency of 25 kHz. 

The sending and receiving transducers 

were made in house using a method 

similar to that in Li et al. (2010). 

The sending piezoceramic was chosen 

primarily based on its resonant frequency 

of 22 kHz and cylindrical 

shape to provide an omni-directional 

signal. The receiving hydrophone 

array and transmitter are shown placed 

on the belly of the CephaloBot in 

Figure 12. 

Overall this customized communication 

localization system places a 

minimal load on the vehicle. The circuitry 

has a very small footprint (Figure 

13) compared to typical commercial 

acoustic modems. In a two-way communication 

mode, the power draw is 

less than 1 W, and in simple listening 

mode, the power draw is less than 

0.25 W. The transducers are fabricated 

in house so they can be customized to 

FIGURE 13 

Communication and localization hardware. 

any shape and placed at any location 

on the vehicle to improve vehicle drag 

characteristics. The entire system is 

fabricated for under $300 in materials, 

making it a suitable option for sensor 

networks requiring several low 

cost vehicles. 

FIGURE 14 

Embedded System 

The embedded system was custom 

designed for the CephaloBot. The vehicle 

will be used by researchers for 

various underwater sensor networking 

applications and multi-vehicle 

coordination. In order to enable this 

underwater network to interact with 

a potential aerial sensor network, 

CephaloBot is also equipped with RF 

communication capabilities, spare 

computation power, and the ability 

to quickly add new sensors. Each vehicle 

must be robust and easy to handle 

and operate. The embedded system is 

separated into multiple printed circuit 

boards (PCB). A power distribution 

board handles voltage regulation and 

battery charging. An interface board 

connects the onboard devices and sensors 

with the power board and processing 

device. Smaller PCBs are located 

throughout the vehicle to provide 

specific functionality such as motor 

control, user interface, or simply wire 

routing. Figure 14 shows the electronics 

located in the center section, where 

everything except the motor controllers 

and user interface is located. 

Processing 

The primary processing device on 

the vehicle is a National Instruments 

Single-board RIO (sbRIO). It has an 

on-board 400 MHz processor running 

real-time LabVIEW software, 128 MB 

of RAM, 256 MB of flash, and a 

40-MHz 2 M gate FPGA (field programmable 

gate array), also programmed 

in LabVIEW, providing 

110 digital I/O pins. The combination 

of microprocessor and FPGA was 

proven effective in the 4th generation 

vehicle where a Compact RIO was 

used. All of the low-level communication, 

interface, and control tasks are 

handled on the FPGA, leaving the 

microprocessor open to perform high 

level mission control. The core vehicle 

software is located on the FPGA. The 

primary benefit is that the mission critical 

software will always run at full 

speed and safety checks will be implemented 

that a vehicle user cannot easily 

override. In this way, even if the 

algorithms being tested fail, the vehicle 

will remain safe to itself and surroundings. 

The 400-MHz real-time processor 

is little used by the core system and 

therefore provides significant computational 

power to the researcher. An 

Hybrid vehicle embedded system mounted on battery pack. 


additional 2 GB of flash storage is 

added with a serial data logger and 

can be used for mission review. 

Batteries 

The power source is a four-cell lithium 

polymer pack. The nominal voltage 

of 14.8 V ranges from 10 to 16.8 V, 

depending on charge level. The lithium 

polymer chemistry was selected because 

of its high-power density and 

excellent discharge characteristics. A 

single four-cell battery pack was chosen 

because, during most of its discharge 

cycle, it will have a voltage 

between 14 and 16 V, which minimizes 

the voltage change of the highcurrent 

devices. The availability of 

integrated circuits and components 

for four-cell batteries was found to be 

better than an eight-cell approach, 

which would have a maximum voltage 

of 33.6 V and require larger 35 V tolerant 

components. Overall current required 

dictates some trace thicknesses 

of approximately 8 mm. The capacity 

of each cell is 21 Ah, which provides a 

total of 310 Wh for the pack. A commercial 

frontend battery circuit board 

protects the batteries from overcharging, 

over discharge, and short 

circuit. It also balances the cells to increase 

overall service life. The circuit 

has a very low resistance to have minimal 

impact on the efficiency. The 

charger built into the power distribution 

can charge the batteries in approximately 

12 h. The high capacity of the 

batteries allows them to be discharged 

at a rate lower than 0.5 C, which increases 

battery runtime and overall 

lifetime (Murphy et al., 1990). 

Power Distribution 

The power distribution is the most 

complex custom-designed circuit board 

in the vehicle made into a four-layer 

10 × 17 cm PCB. It regulates and 

monitors the voltages to power 

the rest of the electronics on the 

submarine. 

The power distribution module 

uses a Microchip PIC18F45K22 as 

a microcontroller supervisor. This 

microcontroller monitors and controls 

voltages and current draws and parts or 

the whole system can be shut off if 

excessive current is drawn or voltages 

are not in an acceptable range. The 

presence of power sources is also monitored, 

and the microcontroller correctly 

decides which one to use and 

whether a battery requires charging. 

The sbRIO can signal the supervisor 

to enable or disable certain regulators 

on the vehicle to save power (i.e., wireless 

bridge is off when submerged). A 

function of the microcontroller also 

allows the vehicle to enter a “sleep” 

mode where everything except the supervisor 

microcontroller is turned off 

for a pre-determined period of time 

which reduces the total power usage 

to about 1 W. The microcontroller 

has an onboard analog-to-digital converter 

and a 16-channel multiplexer 

is used to read all of the required 

voltages and currents. Voltages are 

measured using a resistor-divider network 

to reduce the voltage into the 

multiplexer, and therefore the microcontroller 

to between 0 and 5 V with 

some error margin in case the voltage 

rises to more than intended. Allegro 

ACS714 Hall effect current sensors 

are used to provide very low loss method 

current measurements (0.06 W at 5 A). 

Sensors 

CephaloBot incorporates a minimum 

set of sensors needed to maintain 

a heading and depth underwater. 

Acoustics provide the vehicle with 

a relative position to a static pinger. 

Intervehicle communication prevents 

collisions between vehicles and the 

walls of the pool. The vehicle is robust 

enough to withstand a collision if it 

does occur. When the vehicles are 

deployed to an ocean environment, 

payload sensors may be added heterogeneously 

to the vehicles and shared so 

that each submarine has all required 

data to successfully navigate its environment. 

To simplify the design, 

CephaloBot uses an all-in-one IMU 

solution from VectorNav. The device 

has an onboard three-axis accelerometer, 

gyroscope, and magnetometer. It 

performs Kalman filtering and outputs 

quaternions, Euler angles, and the raw 

sensor data to the sbRIO FPGA. The 

onboard filtering eliminates the need 

to develop or perform the computations 

on the sbRIO. Stated accuracies 

are less than 2°, and because a magnetometer 

provides an absolute reference, 

this accuracy will not degrade 

with time. A Honeywell pressure sensor 

provides fine resolution (0.01 m) 

measurements to 10-m depths. The 

device outputs an analog voltage. The 

test pool for the vehicle is 5-m deep, 

and so a higher resolution device as 

chosen over one that may be used to 

a deeper depth. 

Conclusion 

The CephaloBot provides an ideal 

low cost option for underwater sensor 

networking and hybrid vehicle applications. 

The vehicle has maneuvering 

capabilities at zero forward velocity necessary 

for docking and high-resolution 

sensing. This capability is provided by 

an array of novel squid and jellyfish 

inspired thrusters. The thrusters are 

located internal to the hull with only 

asmallorifice exposed to the outer 

flow minimizing the effect on forward 

drag and allowing for efficient 


high-speed transit. Additionally these 

thrusters only require the single 

opening, allowing for greater system 

freedom internal to the vehicle inbetween 

thrusters. 

The embedded controller system 

designed for the vehicle has a compact 

modular design allowing for a wide 

variety of possible mission objectives. 

The microcontroller, which is operated 

on an easily adaptable LabVIEW 

platform, includes several open connections 

for future mission operations, 

on top of the base level vehicle operation 

input/outputs. The vehicle also has 

a wide variety of communication options 

including a low cost and in house 

developed acoustic communication/ 

localization system for communication 

between underwater vehicles and 

support structures. The vehicle has an 

RF system for communication with 

aerial vehicles while on the surface, 

and a WIFI bridge for communication 

with testers and data loggers while in 

controlled laboratory environments. 



S. Lawrence-Simon, Tyler Thomas, 

Ryan Delgizzi, Dan Ambrosio, Colin 

Miller, Mikhail Kosna, and Matt 

Rhode for their hours of working on 

design and fabrication of the vehicle. 

We would also like to thank the Office 

of Naval Research (code 34) for funding 

this research project. 

Corresponding Author: 

Kamran Mohseni 

231 MAE-A, Department of 


Engineering 

University of Florida, Gainesville, FL 

Email: mohseni@ufl.edu 

References 

Allen, D.G. 2010. James Rumsey: Steamboat 

inventor or pioneer in steam navigation. 

Magazine of the Jefferson County Historical 

Society. 96:48-54. 

Anderson, E., & Grosenbaugh, M. 2005. 

Jet flow in steadily swimming adult squid. 

J Exp Biol. 208:1125-46. doi: 10.1242/ 

jeb.01507. 

Barrett, D., Triantafyllou, M.S., Yue, D.K.P., 

Grosenbaugh, M.A., & Wolfgang, M.J. 1999. 

Drag reduction in fish-like locomotion. 


S0022112099005455. 

Bartol, I.K., Krueger, P.S., Stewart, W.J., & 

Thompson, J.T. 2009. Hydrodynamics of 

pulsed jetting in juvenile and adult brief squid 

Lolliguncula brevis: Evidence of multiple jet 

‘modes’ and their implications for propulsive 

efficiency. J Exp Biol. 212:1889-903. 

doi: 10.1242/jeb.027771. 

Bartol, I.K., Krueger, P.S., Thompson, J.T., 

& Stewart, W.J. 2008. Swimming dynamics 

and propulsive efficiency of squids throughout 

ontogeny. Integr Comp Biol. 48(6):720-33. 

Clark, T., Klein, P., Lake, G., Lawrence- 

Simon, S., Moore, J., Rhea-Carver, B., … 

Wu, A. 2009. Kraken: Kinematically roving 

autonomously kontrolled electro-nautic. 

In: Aerospace Sciences Meeting. Orlando, FL: 

AIAA. 

Dabiri, J.O., Colin, S.P., Katija, K., & 

Costello, J.H. 2010. A wake based correlate of 

swimming performance and foraging behavior 

in seven co-occurring jellyfish species. J Exp 

Biol. 213:1217-75. doi: 10.1242/jeb.034660. 

Dunbabin, M., Roberts, J., Usher, K., 

Winstanley, G., & Corke, P. 2005. A hybrid 

auv design for shallow water reef navigation. 

In: Proceedings of the International Conference 

on Robotics and Automation. pp. 2105-10. 

Barcelona, Spain: IEEE. doi: 10.1109/ 

ROBOT.2005.1570424. 

Fossen, T.I. 1991. Nonlinear modeling and 

control of underwater vehicles. Ph.D. thesis, 

Norwegian Inst. Technol., Trondheim, 

Norway. 

Gharib, M., Rambod, E., & Shariff, K. 1998. 

A universal time scale for vortex ring formation. 


S0022112097008410. 

Glezer, A. 1988. The formation of vortex rings. 

Phys Fluids. 31(12):3532-42. doi: 10.1063/ 

1.866920. 

Kato, N. 2011. Swimming and walking of 

an amphibious robot with fin actuators. 

Mar Technol Soc J. 45(4):181-97. 


characterization of pulsatile vortex ring 

generators for locomotion of underwater 

robots. IEEE J Oceanic Eng. 33(2):123-32. 

doi: 10.1109/JOE.2008.920171. 

Krieg, M., & Mohseni, K. 2010. Dynamic 

modeling and control of biologically inspired 

vortex ring thrusters for underwater robot 

locomotion. IEEE T Robot. 26(3):542-54. 

doi: 10.1109/TRO.2010.2046069. 


enhancement from radial velocity in squid 

inspired thrusters. OCEANS 2011, in press. 

Kona, HI: MTS/IEEE. 

Krueger, P.S., Dabiri, J., & Gharib, M. 2006. 

The formation number of vortex rings formed 

in a uniform background co-flow. J Fluid 

Mech. 556(1):147-66. doi: 10.1017/ 

S0022112006009347. 

Krueger, P.S., & Gharib, M. 2003. The 

significance of vortex ring formation to the 

impulse and thrust of a starting jet. Phys 

Fluids. 15(5):1271-81. doi: 10.1063/ 

1.1564600. 

Li, Y., Kastner, R., Faunce, B., Domond, K., 

Kimball, D., Schurgers, C., & Benson, B. 

2010. Design of a low-cost underwater 

acoustic modem. IEEE ESL. 2:58-61. 

doi: 10.1109/LES.2010.2050191. 

Licht, S. 2008. Biomimetic oscillating foil 

propulsion to enhance underwater vehicle 

agility and maneuverability. Ph.D. thesis, 

Massachusetts Institute of Technology, 

Cambridge, Massachusetts. 

Licht, S., Polidoro, V., Flores, M., Hover, F., 

& Triantafyllou, M. 2004. Design and 

projected performance of a flapping foil 


auv. IEEE J Oceanic Eng. 29(3):786-68. 

doi: 10.1109/JOE.2004.833126. 

Lipinski, D., & Mohseni, K. 2009. A numerical 

investigation of flow structures and 

fluid transport with applications to feeding 

for the hydromedusae Aequorea victoria and 

Sarsiatubulosa. J Exp Biol. 212:2436-47. 

doi: 10.1242/jeb.026740. 

Mclean, M. 1991. Dynamic performance of 

small diameter tunnel thrusters. Ph.D. thesis, 

Naval Postgraduate School, Monterey, 

California. 

Murphy, T., Cason-Smith, D., James, S., & 

Smith, P. 1990. Characterization of aa size 

lithium rechargeable cells. In: Power Sources 

Symposium, Proceedings of the 34th International. 

pp. 176-80. Cherry Hill, NJ: IEEE. 

Mohseni, K. 2004. Zero-mass pulsatile jets 

for unmanned underwater vehicle maneuvering. 

In: 3rd “Unmanned Unlimited” Technical 

Conference, Workshop and Exhibit. 

pp. 2004-3686. Chicago, IL: AIAA. 

Mohseni, K. 2006. Pulsatile vortex generators 

for low-speed maneuvering of small underwater 

vehicles. Ocean Eng. 33(16):2209-23. 

Mohseni, K., & Gharib, M. 1998. A model 

for universal time scale of vortex ring formation. 

Phys Fluids. 10:2436-8. doi: 10.1063/1.869785. 

O’Dor, R.K., & Webber, D. 1991. Invertebrate 

athletes: Trade-offs between transport 

efficiency and power density in cephalopod 

evolution. J Exp Biol. 160:93-112. 

Sahin, M., & Mohseni, K. 2009. An arbitrary 

Lagrangian-Eulerian formulation for the numerical 

simulation of flow patterns generated 

by the hydromedusa Aequorea Victoria. 

J Comp Physiol. 228:4588-605. doi: 10.1016/ 

j.jcp.2009.03.027. 

Sahin, M., Mohseni, K., & Colins, S. 2009. 

The numerical comparison of flow patterns 

and propulsive performances for the hydromedusae 

Sarsia tubulosa and Aequorea Victoria. 

J Exp Biol. 212:2656-67. doi: 10.1242/ 

jeb.025536. 

Shariff, K., & Leonard, A. 1992. Vortex 

rings. Annu Rev Fluid Mech. 34:235-79. 

doi: 10.1146/annurev.fl.24.010192.001315. 

Torsiello, K. 1994. Acoustic positioning of 

the NPS autonomous underwater vehicle 

(AUV II) during hover conditions. Master’s 

thesis, Naval Postgraduate School, Monterey, 

California. 

Wilson, M.M., Peng, J., Dabiri, J.O., & 

Eldridge, J.D. 2009. Lagrangian coherent 

structures in low Reynolds number swimming. 

J Phys-Condens Mat. 21. Art. 204105. 

doi: 10.1088/0953-8984/21/20/204105. 

Yoerger, D., Cooke, J., & Slotine, J.J. 1990. 

The influence of thruster dynamics on underwater 

vehicle behavior and their incorporation 

into control system design. IEEE J Oceanic 

Eng. 15(3):167-78. doi: 10.1109/48.107145. 

Rosenfeld, M., Katija, K., & Dabiri, J.O. 

2009. Circulation generation and vortex ring 

formation by conical nozzles. J Fluids Eng. 

131, 091204:1-8. 

Rosenfeld, M., Rambod, E., & Gharib, M. 

1998. Circulation and formation number 

of laminar vortex rings. J Fluid Mech. 

376:297-318. 

Sahin, M., & Mohseni, K. 2008. The numerical 

simulation of flow patterns generated 

by the hydromedusa Aequorea victoria using 

an arbitrary Lagrangian- Eulerian formulation. 

In: 38th Fluid Dynamics Conference and 

Exhibit. pp. 2008-3715. Seattle: AIAA. 


PAPER 

Modeling of Artificial Aurelia aurita 

Bell Deformation 

AUTHORS 

Keyur B. Joshi 

Alex Villanueva 

Colin F. Smith 

Shashank Priya 

Center for Energy Harvesting 

Materials and Systems, 

Center for Intelligent Material 

Systems and Structure, 

Virginia Polytechnic Institute 

and State University 

ABSTRACT 

Recently, there has been significant interest in developing underwater vehicles 

inspired by jellyfish. One of these notable efforts includes the artificial Aurelia aurita 

(Robojelly). The artificial A. aurita is able to swim with similar proficiency to the 

A. aurita species of jellyfish even though its deformation profile does not completely 

match the natural animal. In order to overcome this problem, we provide a systematic 

finite element model (FEM) to simulate the transient behavior of the artificial 

A. aurita vehicle utilizing bio-inspired shape memory alloy composite (BISMAC) actuators. 

The finite element simulation model accurately captures the hyperelastic 

behavior of EcoFlex (Shore hardness-0010) room temperature vulcanizing silicone 

by invoking a three-parameter Mooney-Rivlin model. Furthermore, the FEM incorporates 

experimental temperature transformation curves of shape memory alloy 

wires by introducing negative thermal coefficient of expansion and considers the 

effect of gravity and fluid buoyancy forces to accurately predict the transient deformation 

of the vehicle. The actual power cycle used to drive artificial A. aurita vehicle 

was used in the model. The overall profile error between FEM and the vehicle profile 

is mainly due to the difference in initial relaxed profiles. 

Keywords: autonomous undersea vehicle, Aurelia aurita, BISMAC, finite element 

analysis, transient dynamics 

Introduction 

J 

ellyfish have been in existence 

for millions of years and are the earliestknownmetazoansthatusemuscles 

for swimming (Valentine, 2004). 

They are found at various ocean 

depths and possess the ability to survive 

under hostile ocean environment. 

They exhibit colonial behavior and 

have the ability to maintain certain 

depth and certain distance from the 

ocean shore (Albert, 2009). Jellyfish 

have relatively simple biological form 

and muscle architecture, lacking advanced 

sensors and a complex neural 

network possessed by many oceanic 

creatures (Chapman, 1974; Gladfelter, 

1972, 1973). but they are still able 

to survive and adapt in hostile environments. 

It maintains territorial existence 

by swimming on minimal energy 

intake. These abilities have created 

tremendous interest in the scientific 

community to discover their structureproperty-performance 

relationships 

and apply the learning towards creating 

a jellyfish-inspired swimming vehicle 

to perform various surveillance 

and monitoring tasks. 

There have been various efforts 

in literature on developing jellyfishinspired 

robots by using smart materialbased 

actuators. Inspired by the jetter 

class of jellyfish that swim by creating 

a jet of water by forcing it out of the 

bell, Villanueva et al. (2009) developed 

the JETSUM. This prototype used 

shape memory alloy (SMA) wires and 

created an actuating stroke of bell 

segments attached to a passive neutrally 

buoyant bell structure. Yang 

et al. (2007) used flappers with control 

surfaces made of ionomeric polymer 

metal composites (IPMC) to 

control directionality of the vehicle. 

Tadesse et al. (2010a) used polypyrrole– 

polyvinylidene difluoride composites 

to achieve bending actuation to create 

a jellyfish robot. Larger jellyfish typically 

use “rowing” locomotion (Colin 

& Costello, 2002) and are characterized 

by formation of counter rotating 

starting and stopping vortex rings. Interaction 

of these vortex rings reduces 

energy lost in the wake and lends 

rowerstheirsuperiorswimmingefficiencies 

relative to jetters (Colin & 

Costello, 2002). Thus, with regard to 

energy efficiency, rowers provide a better 

platform for larger vehicles. Yeom 

and Oh (2009) proposed an entire 

jellyfish made from IPMC actuators 

with segments cut such that, on 

contraction, all the segments close to 

form a contracted bell shape. Recently, 

asignificant breakthrough was made 

by Villanueva et al. (2010b), who proposed 

the high-energy density bioinspired 

shape memory alloy composite 


(BISMAC) actuator that opened the 

possibility of converting high-force generation 

capability of SMA wires into 

high displacements. Using BISMAC, 

the design and implementation of 

biomimetic rowers became feasible. 

In this study, we investigate the 

bell deformation of BISMAC-based 

jellyfish robots using a finite element 

model (FEM, conducted using 

ANSYS) and identify the correlation 

with the natural species. In order to 

do so, the first major challenge was a 

precise implementation of SMA in 

FEM due to their giant aspect ratio 

and hysteretic temperature transformation.Wewereabletosuccessfully 

demonstrate the deformation of SMA 

using ANSYS by optimizing the meshing 

technique, identifying the variability 

in thermal coefficient of expansion, 

and separating the total deformation 

cycle into individual heating and cooling 

curves. Building upon this success, 

we implemented the SMA in BISMAC 

structure and investigated its mechanics 

to optimize the BISMAC configuration 

for mimicking the jellyfish 

profile. The FEM model provides the 

understanding of mechanism for bending 

strain amplification in BISMAC 

actuators and clearly delineates the 

effect of structural and thermal variables.UsingtheFEMresults,we 

were able to identify the inaccuracies 

that can occur due to variability in 

prototyping of BISMAC. Furthermore, 

our results provide important 

insight towards the development of 

feedback controller based on resistance 

changes. Next, we introduce the design 

of bell geometry in FEM for artificial 

Aurelia aurita (later referred to as, 

A. aurita for convenience in this work) 

using radial arrangement of BISMAC 

actuators. The objective was to develop 

the proper joint geometry that allows 

BISMACs to provide maximum deformation. 

Next, we describe the method for fabricating artificial A. aurita and 

experimental characterization. Lastly, using the FEM simulations, we present 

a comparative analysis between the biological A. aurita (swimming profile) 

and artificial A. aurita and FE simulation. 

BISMAC Actuator 

The BISMAC actuator is a composite of an incompressible flexible metal strip 

and SMA wires separated by distance d and embedded in silicone rubber. Thermal 

transformation from martensite phase into austenite upon Joule heating induces 

the contraction of SMA wires, which is resisted by the incompressible metal 

strip. This introduces a tensile force f in the SMA that opposes contraction and 

reduces the strain in SMA wire by a small amount. 

Figure 1 helps in explaining the mechanics of BISMAC bending deformation 

(Figure 1(a)). Under Joule heating, as SMA transforms from martensite 

to austenite phase and tends to contract, but due to the structure of the BISMAC, 

the metal strip resists this contraction and induces tensile force f in the SMA wire 

and compressive force f of the same magnitude in the metal strip. This force 

couple being distance d apart generates effective moment in the BISMAC actuator 

causing it to bend. Figure 1(b) shows the deformed BISMAC geometry. The 

mechanics of BISMAC has been discussed in detail by Smith et al. (2011). By 

using constitutive relation for SMA, we can express the generated stress as 

σ σ 0 ¼ E SMA ðɛ ɛ 0 ÞþΩðζ ζ 0 ÞþΘðT T 0 Þ ð1Þ 

where σ is the stress, ɛ is the strain, E SMA is the effective Young’s modulus of 

SMA, Ω is the transformation coefficient, ζ is the martensite fraction, Θ is the 

thermoelastic coefficient, and T is the temperature. Subscript zero denotes the 

initial state. Since transformation here is from fully martensite to fully austenite 

phase, ζ = 0, and ζ 0 = 1. Pure thermoelastic expansion is negligible, and initial 

stress and strain states are zero. Using E SMA = E austenite , since SMA is completely 

in austenite phase, Eq. (1) transforms into 

σ ¼ E austenite ɛ Ω ð2Þ 

FIGURE 1 

(a) Schematic of the force and moment in BISMAC. (b) Geometry of beam curvature. 


If there was no BISMAC structure to resist the SMA contraction, σ = 0 and ɛ = ɛ l 

transformation strain, providing 

Ω ¼ E austenite ɛ l 

If there is a uniform tensile force applied onto the SMA wire while transformation 

takes place, 

σ ¼ E austenite ɛ þ E austenite ɛ l 

Since SMA wires are thin with negligible bending stiffness, it is safe to assume 

uniform distribution of axial stress. Thus, axial stress can be written as 

σ ¼ 

f 

A SMA 

¼ ɛ f E austenite ⇒ ɛ f ¼ ɛ þ ɛ l 

where f is the force generated by the SMA wire, A SMA is the total area of SMA 

cross section, and ɛ f is the force induced tensile strain. If SMA length after contraction 

is reduced from L to L′, from kinematics consideration, we can write 

L 0 

L ¼ 1 þ ɛ ¼ R 

d 

R 

where R is the radius of curvature of the metal strip and d is the distance between 

SMA wire and the metal strip. Solving Eq.(5) and (6), 

1 þ ɛ f ɛ l ¼ 1 

R ¼ 

d 

R 

d 

ɛ l ɛ f 

ð8Þ 

Using Euler beam theory, 

M 

I ¼ E R ⇒ fd I ¼ E ɛ l 

d 

ɛ f E austenite A SMA d 

I 

ɛ f ¼ 

¼ E ɛ l 

d 

EI ɛ l 

E austenite A SMA d 2 þ EI 

ɛ f 

 

ɛ f 

 

⇒ ɛ f ¼ EI ɛ l ɛ f 

 

E austenite A SMA d 2 

ð3Þ 

ð4Þ 

ð5Þ 

ð6Þ 

ð7Þ 

ð9Þ 

ð10Þ 

ð11Þ 

where EI is the total bending stiffness of the composite beam. For the particular 

configuration of the BISMAC used by Smith et al. (2011), Figure 2(a) shows the 

relationship between the radius of 

curvature for the BISMAC and the 

distance between SMA wires and flexible 

metal strip. Figure 2(b) reveals 

that as the distance d decreases, the 

tensile force induced in SMA wires 

increases dramatically and attains a limiting 

value of 80 g force for 100-μmthick 

SMA wire (BioMetal Fiber, 

TOKI Corporation) beyond which 

high stresses cause austenite phase to 

transform into stress-induced martensite 

resulting in loss of transformation 

strain and thus loss of performance, 

which is not accounted by Eqs. (1)-(11). 

BISMAC Customization 

and Bell Geometry 

A. aurita curvature profiles in its 

relaxed and contracted states are 

showninFigure3(a)(Dabirietal., 

2005). Before any actuation, the bell 

is fully expanded and said to be in 

the relaxed position. Fully contracted 

state refers to the state corresponding 

to complete contraction of subumbrellar 

muscles and minimum bell volume. 

After actuation, the bell passively 

regains its original (relaxed) position. 

In earlier work (Villanueva et al., 

2010b), we have presented the methodology 

used to customize BISMAC 

configurations to mimic the natural 

A. aurita curvature profile. For a 

curved beam, we can write 

M ¼ EI ðÞ s 

dðΔθÞ 

ds 

ð12Þ 

where M is the moment, EI(s) isthe 

local bending stiffness at location s, 

Δθ(s) =θ(s) − θ 0 (s) is the change in 

slope at location s, ands is location 

on curved profile length. Since the 

force generated by SMA wires is uniform,weensureaconstantmoment 

M by maintaining the fixed distance 


FIGURE 2 

(a) Radius of curvature vs. distance d. (b) Tensile force in the SMA wires vs. distance d. 

FIGURE 3 

(a) A. aurita profile in relaxed and contracted conditions. (b) Curvature comparison of BISMAC 

muscle with A. aurita after customization. (c) Schematic of the BISMAC placement in the artificial 

A. aurita. 

between SMA wires and the metal 

strip. The bending stiffness was varied 

(i) to match jellyfish exumbrella and 

subumbrella profiles in the relaxed 

state and (ii) to manipulate the bending 

stiffness EI(s) such that upon SMA 

actuation, dðΔθÞ 

ds 

matches that of real 

A. aurita in order to ensure that we 

have a good match between natural 

and artificial vehicles in the contracted 

state. The resultant BISMAC achieved 

close similarity with A. aurita profile as 

illustrated in Figure 3(b). The inflexion 

point in contracted A. aurita profile 

at ∼2.5 cm represents the fact that 

at the center bell thickens and the profile 

becomes a little convex near the 

center and changes to concave at inflexion 

point. Towards the end of the 

profile, the BISMAC shows reduction 

in curvature due to passive material at 

the tip end for protection from water. 

Similar reductions in contracted 

A. aurita curvature profile towards 

the bell margin lead to discovery of 

the passive flap in A. aurita near the 

bell margin. The passive flap was 

shown to improve the swimming performance 

of the artificial A. aurita significantly 

(Villanueva et al., 2010a). 

Figure 3(c) shows a schematic of artificial 

A. aurita consisting of eight 

BISMACs radially distributed around 

the bell, which is made of soft silicone. 

Artificial A. aurita has shown bell deformation 

and kinematics as well as a 

swimming performance comparable 

to that of natural animal (Villanueva 

et al., 2010a). Since most of the available 

engineering materials have lower 

compliance compared to that of natural 

animal, the artificial A. aurita design 

includes joint structures between 

BISMAC actuators to localize the material 

folding upon contraction. These 

joint structures modify the original 

axisymmetric A. aurita bell shape and 

increases similarity to the Cyanea capillata 

bell shape. The joints are wedgelike 

cavities and are found on natural 

jellyfish. The joint structure is described 

by Smith and Priya (2010) 

and is copied from the Polyorchis montereyensis. 

Since the cross-sectional 

shape of the joints varies with bell 

height, we take the equation below to 

represent the profile at midbell. The 

joint structure can be described as a 

piecewise function representing two 

symmetric sides of a single function, 

mirrored about the y-axis, 

y ¼ δjk ð 

δk 

xÞ 

jx 

ð13Þ 

This function has been formed to describe 

the joint shape across a 2-D 

plane with height j, half-widthk, 

and curvature δ as parameters. In this 

way, the joint structure from a wide 

variety of species can be described 

with one basic equation by changing 


FIGURE 4 

(a) Original axisymmetric A. aurita bell (b). Schematic of joint geometry (Smith & Priya, 2010). 

(c) Top and side view showing location in bell for which joint calculations were made. (d) Final 

artificial A. aurita bell shape. 

−2.75. The final shape was evolved 

by sweeping circular section at bell 

tip into the curve at mid-bell height 

joint section for a smooth blending. 

Figure 4 shows the evolution of 

original axisymmetric A. aurita bell 

shape into the final bell design of artificial 

A. aurita. Figure 5 shows the details 

of the unigraphics model that was 

used in finite element (FE) simulation. 

the constants. Artificial A. aurita’s 

joint structure was produced by using 

a program that could manipulate the 

constants in Mathematica. The values 

to match the shape of joints in P. montereyensis 

were found to be height = 

2.15, half-width = 3, and curvature = 

FIGURE 5 

Unigraphics model of the artificial A. aurita. 


Artificial A. aurita consists of a 

central mount that houses the electrical 

circuitry and clamps the BISMAC 

actuators together (see Figure 6). The 

radius of this hub was 25% of the bell 

diameter and covers a region where 

minimal deformation is expected to 

occur in both the natural and artificial 

A. aurita. Since the distance between 

the SMA wires and the metal strip is 

very crucial parameter in BISMAC design, 

the SMA wires and the metal 

strip are slide in position in small 

acrylic supports, designed to maintain 

this distance. The supports are then 

placed in the mold and silicone is 

poured in to settle (Villanueva et al., 

2010a). The actual vehicle has a 

small portion (3-4 mm) of metal strip 

and SMA wires protruding out of the 

bell geometry. This is neglected in this 

work for model simplification. The 

simplification is justified by negligible 

bending moment contribution from 

the protruded part towards the bell 

deformation. The experimental data 

used for the deformation comparison 

were acquired by using the same artificial 

A. aurita and experimental setup as 

FIGURE 6 

Artificial A. aurita with uniform bell and flap, 

in the relaxed configuration. 


described in previous work (Villanueva 

et al., 2010a). Artificial A. aurita was 

submerged underwater and clamped 

down by supports pressing on the top 

and bottom of the bell. The contacting 

area between supports and bell covered 

the region of the internal central 

mount. This region is meant to undergo 

negligible deformation since 

BISMAC actuators do not directly deform 

the bell at that location. The bell 

was contracted by heating the SMA 

wires using a rapid heating control algorithm 

(Villanueva & Priya, 2010a). 

This controller uses SMA resistance 

feedback to monitor the bell state of 

deformation. It sends high-current 

pulses for rapid contraction and low 

current to maintain deformation 

allowing fast contraction while minimizing 

power consumption. The controller 

was developed in LabView 

(National Instruments), and the measurements 

were made using a NI 

cDAQ 9172 with NI-9215 and 

NI-9263 analog input and output 

cards, respectively. A NF HAS 4052 

power amplifier was used to amplify 

the DAQ output and actuate the 

robot. The bell profile was recorded 

during the first actuation cycle. The 

deformation was captured using an 

IN250 high-speed camera from Fastec 

Imaging. The bell deformation was 

tracked by placing reflective beads 

along the profile and by processing 

the images manually using ImageJ. 

This process included an error on the 

order of ±1 cm. 

FEM Setup and Solution 

Material Properties 

To ensure accuracy of the model in 

adequately representing the behavior 

of various materials, we determined 

the properties experimentally. There 

are three materials critical to the simulation: 

silicone matrix, flexible but incompressible metal strip and SMA wires 

(BioMetal Fiber). 

Silicone Rubber 

Room temperature vulcanizing (RTV) silicone is a candidate material for bell 

mesoglea. In addition to forming the main jellyfish body, it is also responsible for 

maintaining the required distance between spring steel and the SMA wires. We 

have tested several silicone rubbers with different shore hardness to evaluate their 

mechanical properties. Ecoflex TM (Smooth-On) with initial tensile Young’s modulus 

of the order of 10,580 Pa was selected to construct the Artificial A. aurita 

because it offered much less resistance to the BISMAC deflection. Figure 7(a) 

shows the tensile test, on a dog bone–shaped sample with gauge length of 7 mm 

and cross-sectional area of 2.45 × 2.8 mm 2 for several cycles at room temperature 

at 1 mm/s displacement rate up to 20 mm extension to ensure silicone properties 

do not change with multiple loading cycles. Hysteresis is evident in the figure, and 

we used the average of the two curves to generate our model. Figure 7(b) represents 

completely defined stress-strain behavior of silicone including compression 

test data carried out on 16- mm-thick 25.4-mm diameter cylindrical specimen. 

We generated a three-parameter Mooney-Rivlin model for silicone from test 

data using ANSYS’s curve fitting tool, as a more conventional two-parameter 

Mooney-Rivlin model failed to provide a good fit to the experimental data. 

The Mooney-Rivlin model for Ecoflex TM is given by Eq. (14), 

W ¼ c 10 ðI 1 3Þþc 01 ðI 2 3Þþc 11 ðI 1 3ÞðI 2 3Þþ 1 d ðJ 1Þ2 

ð14Þ 

where W is the strain energy function, I 1 , I 2 , I 3 are the stretch invariants, J is the 

determinant of deformation gradient tensor, and c 10 = 2307.1 Pa, c 01 = −223.76 Pa, 

c 11 =142.83Pa,d =0Pa −1 (compressibility parameter). Silicone thermal conductivity 

was measured to be 0.22 W/m K. Silicone properties used in the 

FEM are tabulated in Table 1. 

SMA Wire 

We selected 100-μm diameter BioMetal fibers due to their superior performance 

(Tadesse et al., 2010b). The temperature transformation of the wires was 

measured experimentally as shown in Figure 8(a). Figure 8(b) represents the 

FIGURE 7 

(a) Tensile test for several cycles showing hysteresis. (b) Stress-strain curve used to model 

Ecoflex TM . 


TABLE 1 

Silicone properties used in the FEM. 

Property 

Value 

Thermal Conductivity 

0.22 W/m K 

Elastic modulus 

Mooney-Rivlin model 

Poisson’s ratio 0.49 

Density 982 kg/m 3 

Specific heat 

300 J/kg K 

TABLE 2 

SMA properties used in the FEM. 

Property Martensite Austenite 

Thermal Conductivity 8 W/m K 18 W/m K 

Elastic modulus 28 MPa 75 MPa 

Poisson’s ratio 0.33 0.33 

Density 6450 kg/m 3 6450 kg/m 3 

Specific heat 837.36 J/kg K 837.36 J/kg K 

stress-strain relationship as measured 

by tensile test to confirm the manufacturer’s 

claims (Toki Corporation) that 

the wires can easily take 400 MPa 

stress. 

Other properties that were used in 

the simulation are tabulated in Table 2. 

For temperatures other than transformation 

temperatures, the properties 

were interpolated according to martensite 

fraction in the SMA. Martensite 

fraction was calculated based on the 

temperature-strain curve. 

In order to model the temperature 

transformation hysteresis of SMA 

wires, we defined two separate material 

curves, one for the heating profile and 

other for the cooling profile. Accordingly, 

the elastic modulus EX and thermal 

conductivity KXX also followed 

two separate profiles as depicted in Figures 

9(a) and 9(b). Figure 9(c) shows 

the artificially defined negative thermal 

coefficient of expansion to achieve the 

transformation strains with increase in 

temperature during heating and cooling, 

respectively. 

Metal Strip 

For choosing the metal strip, two 

criteria should be met: (i) incompressibility 

for BISMAC mechanics to work 

well and (ii) flexibility (low bending 

stiffness) to obtain maximum 

deformation with a small actuation 

moment. We selected standard spring 

steel (low carbon steel) as the suitable 

metal strip material with properties 

tabulated in Table 3. 

Transient Heat Transfer Model 

Meshing, Boundary Conditions, 

and Loads 

We chose ANSYS as our simulation 

package. To simulate transient 

heat transfer, we built the model 

by meshing the SMA, metal strip 

and one element thick silicone layer 

around them with SOLID70 (8-node 

brick element). The rest of the silicone 

matrix was meshed with SOLID87 

(10-node tetrahedrons) and SOLID90 

(20-node brick elements) was used as 

transition element between SOLID70 

FIGURE 8 

(a) BioMetal Fiber temperature transformation curve (A s = Austenite start temperature, A f = Austenite finish temperature, M s = Martensite start 

temperature, M f = Martensite finish temperature). (b) Stress-strain relationship of Martensite phase. 


FIGURE 9 

(a) Variation of SMA Young’s modulus with temperature. (b) Variation of SMA thermal conductivity 

with temperature. (c) Variation of thermal coefficient of expansion with temperature. 

and SOLID87 element meshes as shown 

in Figure 10(a). All these elements have 

TEMP degree of freedom. 

We took advantage of the circular 

symmetry and modeled only 1/8th of 

the bell segment with no loss of physics. 

To reduce the meshing efforts 

we chose to mesh only half of the 

1/8th segment of the bell, reflected 

the mesh about the plane of symmetry 

and finally merged the nodes to create 

the complete model. The boundary 

FIGURE 10 

(a) Mesh detail for transient thermal model. (b) Boundary conditions. 

conditions are depicted in Figure 10(b). 

Due to symmetry, the circular symmetric 

sides (Figure 10(b), violet colored 

faces) of the model do not have 

any temperature gradient; thus, they 

are modeled as insulated boundaries. 

Exumbrella and subumbrella surfaces 

(Figure 10(b), yellow colored faces) 

are in contact with surrounding water 

and conveys heat into the fluid, and 

thus, were modeled as convective 

boundaries. The heat transfer coefficient 

of both these surfaces were 

taken to be 20 W/m 2 K (Baker, 1972), 

and the ambient temperature was 

fixed at 25°C. The thermal loading 

resulting from electrical heating was 

applied using Villanueva et al.’s resistance 

feedback control algorithm 

(Villanueva & Priya, 2010a) to reduce 

power requirement. The internal heating 

load was applied on the SMA wire 

as shown in Figure 11, which is consistently 

1/8th of total power consumed 

by the vehicle’s eight segments. It consists 

of rapid heating pulses of high 

TABLE 3 

Metal strip properties used in the FEM. 

Property 

Value 

Thermal Conductivity 47 W/m K 

Elastic modulus 210 GPa 

Poisson’s ratio 0.3 

Density 7860 kg/m 3 

Specific heat 

510 J/kg K 


current followed by a constant current 

regime and finally reducing the magnitude 

to idling minimum current governed 

by need for resistance feedback 

measurement. The three initial spikes 

seen in the curve correspond to the 

high-current impulse sent for rapid heating. 

Multiple pulses are usually needed 

during the first few actuation cycles 

since the low current is not enough to 

maintain the deformation. The material 

surrounding the SMA warms up and 

eventually the low current input is 

enough to maintain deformation. 

Solution: Transient Thermal Analysis 

We first find the solution with 

SMA wires having material properties 

corresponding to heating curve. To 

capture initial sharp temperature rise 

accurately over the first 0.12 s, small 

time steps of 0.01 s were used to provide 

sufficient time resolution. The 

low current period from 0.12 to 0.70 s 

was solved in 100 uniform time steps. 

After solving two more transient current 

time steps at 0.71 and 0.72 s in 

one time step each, we switched the 

SMA wire material to the one with 

properties corresponding to cooling 

curve. Temperatures of SMA wires 

corresponding to t =0.72sbeingfar 

beyond A f ensures properties of both 

the curves are same at this point. Finally, 

relaxation phase of A. aurita 

jellyfish vehicle from 0.72 to 2.00 s 

was solved in 200 uniform time steps. 

FIGURE 11 

Variation of internal heating load on SMA wire with time. 

FIGURE 12 

Transient Structural 

Deformation Model 

Meshing, Boundary Conditions, 

and Loads 

To simulate transient structural 

deformation, we reused the mesh 

from transient heat transfer model 

for rapid model building. SOLID70 

(8-node brick element) of the SMA, 

metal strip and one element thick silicone 

layer around them were transformed 

into SOLID185 (8-node 

brick element) with structural degree 

of freedom UX, UY, UZ. Similarly, 

SOLID87 (10-node tetrahedrons) 

was transformed into SOLID187 

(10-node tetrahedrons) elements and 

transition elements SOLID90 (20- 

node brick elements) were transformed 

into SOLID186 (20-node 

brick elements) as shown in Figure 

12(a). Figure 12(b) displays the 

displacement boundary condition for 

the model. The boundary condition 

was applied in cylindrical coordinate 

system defined using jellyfish vehicle 

centralaxisasthez-axis of the cylindrical 

coordinate system (CSYS = 5). 

(a) Transient structural deformation model mesh. (b) Displacement boundary conditions. (c) Temperature 

and acceleration boundary conditions. 


Both sides of 1/8th segment being circular symmetric were constrained from 

moving in hoop direction (UY = 0). The line on the axis of the vehicle was 

also constrained from moving in radial direction (UX = 0). Also, to constrain 

any rigid body translation, we chose to constrain one node on rigid central hub 

lying on the axis of the vehicle from moving in axial direction (UZ = 0). 

We accounted for gravitation and buoyancy force of the water by applying 

inertial acceleration corresponding to reduced gravitation under buoyancy from 

Eq. (15), 

 

ACEL ‐ 

Y ¼ ΣN mat¼1 V matρ mat ρ water Σ N mat¼1 V 

mat 

Σ N mat¼1 V 9:81 m 

matρ mat 

s 2 ð15Þ 

where ACEL_Y is the inertial acceleration for the vehicle, mat is the material index, 

ρ is the density, and V is the volume. Earth’s gravitation acceleration was taken to 

be 9.81 m/s 2 . Temperatures were applied as body forces and were read from previously 

solved transient heat transfer model result. To account for the fluid drag 

we have introduced damping by defining BETAD = 0.03. This creates a damping 

matrix [C] =BETAD[K ], where [K ] = stiffness matrix of the FEM. 

The overall FE system equation can be written as 

½M 

ŠfÜ 

gþ ½CŠ 

: 

U þ ½K 

ŠfU 

g ¼ fFg 

ð16Þ 

Here, [M], [C]and [K ] are displacement dependent on the mass matrix, the 

damping matrix and the stiffness matrix, respectively, {U} is the displacement 

vector and {F } is effective force vector. The system defined by Eq. (16) is nonlinear 

due to variable [M ], [C], and [K ] matrices. 

Solution: Transient Structural Deformation Analysis 

As in transient heat transfer model, we start the modeling with SMA wires having 

properties corresponding to heating curve (martensite to austenite temperature 

transformation) for contraction phase of the cycle and during relaxation we 

use material corresponding to cooling curve (austenite to martensite temperature 

transformation). In simulation, this was achieved by defining two separate SMA 

material property curves corresponding to (martensite (M) to austenite (A) and austenite 

to martensite) and switching from material with M → A curve to material with 

A → M curve after completion of contraction phase. As SMA wire is well above austenite 

finish temperature A f , for which both the material curves (cooling and heating) 

have identical properties there’s no abrupt jump between these curves. We used the 

time intervals as used in transient heat transfer model. 

Results and Discussion 

Transient Heat Transfer Analysis 

Figure 13 summarizes all major results of transient heat transfer analysis. Figure 

13(a) shows overall temperature distribution at the end of heating phase. It 

suggests that practical region of interest is concentrated near SMA wire, and for 

most part, it does not change along SMA wire length. Near the edge of the bell, 

however, model predicts a little more temperature rise in the SMA wire compared 

to rest of the area due to very thin silicone 

layer as same amount of generated 

heat is absorbed by lesser silicone 

available. Since rate of heat conduction 

into silicone is higher compared to the 

rate of heat being convected away at 

the subumbrella surface, silicone temperature 

rises faster and temperature 

gradient between SMA wire and silicone 

reduces, inhibiting heat conduction 

from SMA wire to subumbrella 

surface. Figure 13(b) shows temperature 

history at a typical location at 

SMA center and in silicone 1 element 

away from SMA wire surface. It shows 

typical response of first order system to 

a given excitation. Initial high-current 

pulses till t = 0.12 s result in fast temperature 

rise at SMA center which 

keep increasing at logarithmic rate 

from t = 0.12 s to t = 0.70 s. Temperature 

at SMA center decreases exponentially 

from t = 0.70 s to t = 2.00 s 

but fails to return to starting temperature 

of 25°C at the end of the cycle 

(T =45°C at t = 2.00 s) suggesting 

net heat accumulation in the model. 

It also suggests that we may be adding 

heat unnecessarily at higher rate 

during low current constant heating 

cycle as the temperature at the end of 

high current pulsed heating is well 

beyond austenite finish temperature 

A f . This additional heating also aggravates 

heat accumulation problem that 

eventually results in partial loss of performance 

of the actuator as insufficient 

cooling results into incomplete transformation 

into martensite phase. Temperature 

in silicone 1 element away 

from SMA surface shows quite similar 

trend but has reduced temperature rise 

peaks during high-current pulse cycle 

as low conductivity dampens out the 

sharp peaks. Cooling cycle starts with 

significant temperature gradient between 

SMA center and the aforementioned 

silicone location but diminishes 


FIGURE 13 

(a) Overall temperature distribution at t = 0.70 s. (b) Temperature time history at SMA wire 

center and in silicone 1 element away from SMA surface. (c) Typical temperature distribution 

along SMA wire length. 

exponentially fast as is evident from Figure 

13(b). Figure 13(c) compares temperature 

distribution near SMA length 

at typical location for key time points 

t = 0.12 s (end of high-current pulse 

heating), t = 0.70 s (end of constant 

low current heating), and t = 2.00 s 

(end of cooling cycle). 

Transient Structural 

Deformation Analysis 

Overall Bell Deformation 

Since, gravity and buoyancy forces 

are accounted for, in the two equilibrium 

positions (relaxed and contracted), 

where fluid pressure differential does 

not exist across subumbrella and 

exumbrella due to the FEM and the 

artificial A. aurita being held at the 

bell center; we can predict deformation 

at these positions accurately 

at this location without worrying 

about approximations involved in 

fluid drag. Figure 14(a) shows various 

views of FE simulation result 

for contracted state at t =0.12s. 

Original undeformed model is shown 

with only black edges superimposed 

on deformed model for comparison. 

Figure 14(a.1) represents bottom 

view of the contracted artificial 

A. aurita bell model, which qualitatively 

matches the contracted bell 

segments of C. capillata species (Figure 

14(d)) and experimental artificial 

FIGURE 14 

(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 

location, and (a.4) side view. (b) Biological A. aurita contraction. (c) Artificial A. aurita experimental contraction. (d) C. capillata bell segment 

contraction bottom view. (e) Artificial A. aurita experimental contraction bottom view (Villanueva, submitted). 


A. aurita contraction (Figure 14(e)) 

that we have intended to model in 

our FE simulation. Note that due to 

manufacturing imperfections not all 

the bell segments deform the same 

amount (Figure 14(e)). The comparison 

of contracted profile with that of 

C. capillata is justified as our artificial 

A. aurita possess joint structure and 

radial muscles that were inspired by 

C. capillata possessing similar radial 

muscle arrangement (Gladfelter, 

1973). Figure 14(a.2) depicts isometric 

view of deformed model for 

better visualization. Figure 14(a.3) 

shows cross-sectional view taken along 

BISMAC center line. Figure 14(a.4) 

exhibits side view of the contracted 

FEM also matching well with natural 

A. aurita shown in Figure 14(b) and 

experimental artificial A. aurita (Figure 

14(c)). It is evident that the bell 

curves inwards at the BISMAC locations 

radially and axially well and 

confirms that the joint design provided 

by Smith and Priya (2010) is 

effective in assisting the bell deformation 

at the BISMAC locations. 

This behavior was also confirmed in 

experimental A. aurita deformation 

(Figure 14(c)). 

Comparison of Deformation 

at the BISMAC Location 

and at the Joint Location 

Figures 15(a)-15(f) reveal radial 

displacements in the bell at the 

BISMAC cross-section and in the joint 

cross-section, respectively, at the key 

time points (t = 0.12 s, t = 0.70 s and 

t = 2.0 s); each group uses a common 

color legend for ease of comparison. 

It is obvious that the deformation is 

negligible near the centre of the jellyfish 

bell and maximum at the tip of 

the bell. At joint location, the bell is 

practically undeformed. 

Figure 16(b) plots the time history 

of radial and axial tip displacement at 

BISMAC location, which conforms 

to second order system response 

to step excitation. In fact, the highpower 

pulse heating is done at much 

higher frequency than the bell is able 

to respond, thus inertia of the bell 

acts as low-pass filter for the pulsed 

heating excitation. But as discussed 

in transient heat transfer results, SMA 

goes through complete martensite to 

austenite transformation in this period, 

contracting due to phase transformation 

strain and achieves maximum 

deformation of U radial =1.31mm 

and U axial =7.87mmatt =0.12s. 

At the maximum deformation point, 

the inertial forces and elastic restoring 

forces are not in equilibrium. The bell 

has a slight overshoot above equilibrium 

position due to inertia. During 

moderate constant current excitation 

from t = 0.12s to t =0.70sastemperature 

keeps rising, without any 

additional benefit toSMAtransformation 

strain the structure relaxes 

under stored strain energy during 

rapid contraction and after a small oscillation 

around equilibrium condition 

finally settles to U radial = 10.87 mm 

and U axial = 6.04 mm. During relaxation 

phase from t = 0.70 s to t = 2.0 s, 

SMA goes through austenite to 

martensite transition as temperature 

drops rapidly and returns to original 

configuration after little oscillations as 

FIGURE 15 

(a-c) Deformed artificial A. aurita bell at BISMAC location at different time t = 0.12 s, 0.70 s, and 2.0 s (d-f) Deformed artificial A. aurita bell at joint 

location at different time t = 0.12 s, 0.70 s and 2.0 s. 


FIGURE 16 

(a) Comparison of the FE results with artificial A. aurita bell deformation and natural A. aurita. 

(b) Time history of A. aurita tip displacement at BISMAC location. 

FIGURE 17 

is evident from Figures 16(b), 15(a)- 

15(b), and 17(a). 

Figure 16(a) compares the relaxed 

and contracted position of the 

simulated A. aurita bell with natural 

A. aurita and robotic A. aurita. The 

natural A. aurita with its muscle contracting 

in excess of 40% clearly 

achieves the highest deformation, 

not being fully matched by either 

artificial A. aurita or FE simulation. 

Experimental deformation on artificial 

jellyfish was measured through 

image processing thus only exumbrella 

points are available for comparison. 

Experimental model, however, 

has changed its form during manufacturing, 

which was designed to match 

FEM in relaxed state. This results in 

apparent mismatch in FE prediction 

from the experimental deformation. 

Inspite of this mismatch, the overall 

deformations are comparable. Figure 

16(b) shows comparison of time 

response of the transient tip displacements 

between FE simulation and artificial 

and natural A. aurita. It should 

be noted that the natural A. aurita 

Trace of tip displacement at BISMAC location by (a) FE simulation and (b) artificial A. aurita 

experiment and (c) natural A. aurita. 

swims freely in water moving forward, 

while artificial (experimental) A. aurita 

and the model are held at the bell center 

and does not move it water. Also, 

for the particular cycle for which data 

is obtained has cycle time of 1.7 s instead 

of 2 s. Natural A. aurita contracts 

and relaxes very smoothly and 

does not have any steady equilibrium 

position. Artificial A. aurita achieves 

about twice as much radial displacement 

as predicted by simulation but 

follows similar trend of sharp rise 

(fall), overshooting above the equilibrium 

position beyond t =0.12s,relaxing 

to equilibrium position during 

t = 0.12 s to t = 0.70 s due to stored 

strain energy in the bell during contraction 

cycle. In relaxation cycle, as 

SMA cools down and undergoes austenite 

to martensite transition, the 

bell returns back to its original location 

slightly overshooting beyond 

equilibrium position before returning 

back to relaxed configuration. FE simulation 

captures this entire physics 

perfectly. Artificial A. aurita returns 

to relaxed position slower than predicted 

by model, suggesting that 

heat gets conducted away from SMA 

more slowly than we predict. However, 

artificial A. aurita axial displacement 

being too small; gets affected by 

finite resolution of image processing 

and shows inconclusive trend. FE 

simulation predicts axial tip displacement 

time history similar to that of 

radial tip displacement but reduced 

in magnitude. 

Figure 17(a) exhibits the trace of 

the bell tip at BISMAC location and 

suggests that bell tip does not follow 

the same path while relaxing to its original 

configuration than it did during 

contraction. It also illustrates the 

overshoot around the equilibrium positions 

(relaxed and contracted). Tip 

displacement for the artificial A. aurita 


(Figure 17(b)) shows different trace patterns during contraction and relaxation as 

predicted by the model. Since artificial A. aurita begins with different relaxation 

configuration, this is expected. Interaction with surrounding fluid would cause 

the tip displacement trace to change which are not accommodated in the 

model. Oscillations in the trace of the experimental results are partly due to tracking 

error. The magnitude of the oscillations is close to the predicted error from the 

tracking method. Figure 17(c) represents trace of tip displacement of natural 

A. aurita scaled to the same bell diameter. Natural A. aurita is freely moving in 

water. The direction of contraction and relaxation is quite similar to the one predicted 

by the model (Figure 17(a)); however, relative location of the path during 

contraction and relaxation is reversed. We believe that this is caused by the fluid 

forces and the freely forward motion present in natural A. aurita. 

For a more quantitative comparison, we calculated the curvature of the profiles 

by calculating radius of circle passing through three consecutive points, 

RP ð 2 Þ ¼ 

ρðP 2 Þ ¼ 1 

RP ð 2 Þ 

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 

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 

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 

ð17Þ 

ð18Þ 

where R(P 2 ) is the radius at point P 2 (x 2, y 2 ), having previous point P 1 (x 1 , y 1 ) 

and next P 3 (x 3 , y 3 ). Except first and last point, radius and curvature of all points 

can be calculated by using Eq. (17) and (18). Equal interval of all the points is 

very essential for this method, thus additional or fewer points were generated 

from available FE nodes, experimental traces and A. aurita profile. 

Figures 18(a) and 18(b) compare the curvature of the profile along the 

length of the curved exumbrella surface in relaxed and contracted condition between 

FE simulation, experimental and natural A. aurita. In relaxed condition, 

FEM curvature matches that of A. aurita. Experimentally as evident from Figure 

16(a) and Figure 18(a) it does not match A. aurita profile in the relaxed state. 

Figure 18(b) emphasizes that curvatures of FEM, experimental as well as natural 

animal, increases as the fish contracts. Artificial A. aurita does not have the same 

relaxed state as the natural animal, but in contracted state it follows natural 

A. aurita exumbrella deformation. It outperforms the curvature of the natural 

A. aurita by achieving a maximum curvature of 67 m −1 at about 80% length. 

FEM and natural A. aurita achieve maximum curvature of 50 m −1 at 67% and 

45 m −1 at 85%, respectively, and show sharp increase at the tip, suggesting dynamic 

overshoot of the passive flap that is hypothesized to help swimming performance. 

Figure 18(c) displays profile errors between various profiles and 

conditions, 

ɛ FER ðÞ¼ s 

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 

ðÞ s x ExpBR ðÞ s 

2 

þ yFEBR ðÞ s y ExpBR ðÞ s 

2 

x FEBR 

ð19Þ 

The profile error ɛ FER is defined as the 

distance between corresponding 

points (x(s), y(s)) (subscript BR corresponds 

to the BISMAC location and 

relaxed condition) of FEM and the artificial 

A. aurita vehicle relaxed profiles 

at location s. This was calculated by 

Eq. (19), and we observe that profile 

errors between the FE simulation and 

artificial A. aurita experiment in relaxed 

and contracted states, ɛ FER and 

ɛ FEC , have similar trends (Figure 18(c)). 

This indicates that the difference between 

FE simulation and artificial 

A. aurita is mainly due to the initial difference 

in relaxed profiles. We hypothesize 

that if artificial A. aurita were 

re-designed to match the natual 

A. aurita’s relaxed state, the simulation 

would match it more closely. The profile 

error does not increase beyond 

5.8mm which corresponds to 22% 

of the initial profile error in relaxed 

configuration. 

The model developed in this study 

gives better understanding of the temperature 

rise in the SMA wires and 

transient heat distribution in and 

around it during entire cycle. It reveals 

that in spite of complex geometry, heat 

distribution is practically uniform 

along SMA length and can be effectively 

captured by 1st order unsteady heat 

conduction equation. It explains degradation 

of performance of artificial 

A. aurita, over number of cycle due 

to heat accumulation resulting from 

insufficient cooling and suggests 

opportunity of making the bioinspired 

A. aurita vehicle more energy 

efficient by reducing oversupply of 

heat to SMA wire and effectively controlling 

SMA temperature. The model 

predicts transient deformation behavior 

of the artificial A. aurita qualitatively 

at BISMAC and fold locations, 

confirming effectiveness of the join 

design. It reveals that the transient tip 


FIGURE 18 

Curvature comparison at BISMAC and fold location between ANSYS, experiment, and biological 

A. aurita (a) relaxed condition, (b) contracted condition, (c) profile errors at BISMAC location- 

FER: between FE and experiment relaxed profile, FEC:FE and experiment contracted profile, AEC: 

A. aurita and experimental contracted profile, FAC:FE and A. aurita contracted profile. 

displacement response can be modeled 

as a second-order system; however, due 

to mismatch in initial profile, it predicts 

the radial tip displacement response 

∼44% of the experimental curve. The 

model captures these physical aspects 

of bell deformation accurately and is a 

useful tool to evaluate effectiveness of 

design changes on performance of the 

vehicle without building one. 

Conclusion 

We introduce a customization procedure 

for BISMAC actuators to 

be used as radial muscles in artificial 

A. aurita such that we can match 

the relaxed and contracted profiles of 

A. aurita at the BISMAC locations. 

By accurately modeling the hyperelastic 

behavior of EcoFlex (Shore 

00-10) RTV silicone and using experimentally 

obtained temperature-strain 

transformation curves for SMA wires, 

we provide a high-fidelity model that 

captures most essential physics of 

artificial A. aurita deformation. The 

model suggests a unique approach 

to overcome ANSYS limitation in 

modeling SMA temperature transformation 

by using negative thermal coefficient 

of expansion and two separate 

material curves for heating and cooling. 

This approach is generic enough 

to be used in numerical modeling 

of the SMA temperature-dependent 

transformation in any design. The 

transient heat transfer model provides 

better understanding of temperature 

rise in SMA wire and heat distribution 

around it during an entire contractionrelaxation 

cycle. This information is 

crucial in designing and evaluating 

SMA heating algorithm and cannot 

be obtained from the prototype directly. 

The model also reveals that in 

spite of the complex geometry, the 

temperature distribution along the 

length of SMA wires is uniform and 

we could use a simple first order heat 

transfer model to study temperature 

distribution in and around SMA 

wires. It also exposes the reason for 

degradation of the performance of artificial 

A. aurita over a number of cycles 

due to heat accumulation and reveals 

an opportunity to make the artificial 

A. aurita more energy-efficient by reducing 

the amount of heat supplied 

above the austenite’s finish temperature 

A f that does not contribute toward 

bell deformation. Transient structural 

model accounts for the gravity and 

the buoyancy forces and thus in equilibrium 

conditions (fully contracted 

and fully relaxed states), where fluid 

pressure differentials across subumbrella 

and exumbrella do not 

exist, we can predict the bell deformations 

fairly accurately. Transient 

structural analysis captures all essential 

behavior of artificial A. aurita at 

BISMAC and fold locations and confirms 

effectiveness of the joint design, 

though an exact match was not 

achieved due to initial profile mismatch. 

The model reveals that the 

transient tip displacement response 

can be modeled as a second-order system; 

however, due to initial profile 

mismatch; the model predicts the radial 

tip displacement response at 

∼44% of the experimental curve. Profile 

error analysis suggests that initial 

error in simulation and artificial 

A. aurita relaxed profiles is a major 

cause for the mismatch. The profile 

error increases by a small amount 

with an average of 0.00015 m (0.58% 

of ɛ FER ) and reaches a maximum of 

0.0058 m (22% of ɛ FER ). The model 

captures these physical aspects of bell 

deformation physics accurately and 

isausefultooltoevaluateeffectiveness 

of design changes on performance 

of the vehicle without the need for 


time-consuming physical construction. 

The modeling methodology is 

generic and could be used to model 

other bio-inspired robots using similar 

construction technique. 


This research is sponsored by the 

Office of Naval Research through contract 

N00014-08-1-0654. 

Lead Authors: 

Keyur B. Joshi and Shashank Priya 

Center for Energy Harvesting 

Materials and Systems 

Center for Intelligent Material 

Systems and Structure 

Virginia Polytechnic Institute 

and State University 

310 Durham Hall, Blacksburg, 

VA 24061 

Email: key4josh@vt.edu; 

spriya@mse.vt.edu 

References 

Albert, D.J. 2009. Aurelia labiata (Scyphozoa) 

jellyfish in Roscoe Bay: Their spatial distribution 

varies with population size and their 

behaviour changes with water depth. J Sea 

Res. 61(3):140-3. doi: 10.1016/j.seares. 

2008.11.001. 

Baker, E. 1972. Liquid cooling of microelectronic 

devices by free and forced convection. 

Microelectron Reliab. 11(2):213-22. 

doi: 10.1016/0026-2714(72)90704-4. 

Chapman, D.M. 1974. Cnidarian histology. 

In Coelenterate Biology, eds. Muscatine, L., 

& Lenhoff, H.M., 2-92. New York: 


Colin, S.P., & Costello, J.H. 2002. Morphology, 

swimming performance and propulsive 

mode of six co-occurring hydromedusae. 

J Exp Biol. 205:427. 

Dabiri, J.P., Colin, S.P., & Costello, J.H. 

2005. Flow patterns generated by oblate 

medusan jellyfish: Field measurements and 

laboratory analyses. J Exp Biol. 208(7):1257. 

doi: 10.1242/jeb.01519. 

Gladfelter, W.G. 1972. Structure and function 

of the locomotory system of Polyorchis montereyensis 

(Cnidaria, Hydrozoa). Helgoland Wiss 

Meer. 23(1):38. doi: 10.1007/BF01616310. 

Gladfelter, W.G. 1973. Structure and function 

of the locomotory system of the Scyphomedusa 

Cyanea capillata. Mar Biol. 14(2):150. 

doi: 10.1007/BF00373214. 

ImageJ. Image Processing and Analysis in 

Java. http://rsbweb.nih.gov/ij/. (accessed 12 

July 2011). 

National Instruments. LabView. http:// 

www.ni.com/labview/. (accessed 12 July 2011). 

Smith, C., Villanueva, A., Joshi, K., Tadesse, 

Y., & Priya, S. 2011. Working principle of 

bio-inspired shape memory alloy composite 

actuators. Smart Mater Struct. 20(1):012001. 

doi: 10.1088/0964-1726/20/1/012001. 

Smith, C.F., & Priya, S. 2010. Bio-inspired 

Unmanned Undersea Vehicle. Behavior and 

Mechanics of Multifunctional Materials and 

Composites 2010, March 8, 2010 - March 

11, 2010, The Society of Photo-Optical 

Instrumentation Engineers (SPIE); American 

Society of Mechanical Engineers. San Diego, 

CA: SPIE. 

Tadesse, Y., Brennan, J., Smith, C., Long, 

T.E., & Priya, S. 2010a. Synthesis and 

characterization of polypyrrole composite 

actuator for jellyfish unmanned undersea 

vehicle. 764222-11. San Diego, CA: SPIE. 

Tadesse, Y., Thayer, N., & Priya, S. 2010b. 

Tailoring the response time of shape memory 

alloy wires through active cooling and prestress. 

J Intel Mat Syst Str. 21(1):19-40. 

doi: 10.1177/1045389X09352814. 

Toki Corp. Biometal Fiber. http://www.toki. 

co.jp/biometal/english/qanda.php. (accessed 

12 July 2011). 

Valentine, J.W. 2004. On the Origin of 

Phyla. Chicago: University of Chicago Press. 

Villanueva, A., Bresser, S., Chung, S., 

Tadesse, Y., & Priya, S. 2009. Jellyfish 

inspired underwater unmanned vehicle. 

Proc SPIE. 7287:72871G. doi: 10.1117/ 

12.815754. 

Villanueva, A., & Priya, S. 2010. BISMAC 

control using SMA resistance feedback. 

In: SPIE Conference Proceedings, SPIE 

NDE Conference. 76421Z-12. San Diego, CA: 

SPIE. 

Villanueva, A., Priya, S., Anna, C., & Smith, 

C. 2010a. Robojelly bell kinematics and 

resistance feedback control. In: IEEE 

International Conference on Robotics and 

Biomimetics (ROBIO). 1124-9. Tianjin, 

China: IEEE (Institute of Electrical and 

Electronics Engineers). 

Villanueva, A., Smith, C., & Priya, S. 

submitted. Under review, Paper Ref no. 

“BB/378339/PAP/255146”. Biomimetic robotic 

jellyfish (Robojelly) using shape memory 

alloy. IOP Bioinspir Biomim. 

Villanueva, A.A., Joshi, K.B., Blottman, J.B., 

& Priya, S. 2010b. A bio-inspired shape 

memory alloy composite (BISMAC) actuator. 

Smart Mater Struct. 19(2):025013. 

doi: 10.1088/0964-1726/19/2/025013. 

Yang, Y., Ye, X., & Guo, S. 2007. A new 

type of jellyfish-like microrobot. In: IEEE 

International Conference on Integration 

Technology. 673-8. Rourkela, India: IEEE 

(Institute of Electrical and Electronics 

Engineers). doi: 10.1109/ICITECHNOLOGY. 

2007.4290404. 

Yeom, S.-W., & Oh, I.-K. 2009. A biomimetic 

jellyfish robot based on ionic polymer 

metal composite actuators. Smart Mater 

Struct. 18(8):085002. doi: 10.1088/0964- 

1726/18/8/085002. 


PAPER 

Swimming and Walking of an Amphibious 

Robot With Fin Actuators 

AUTHOR 

Naomi Kato 

Graduate School of Engineering, 

Osaka University 

Introduction 

I 

t has recently been clarified that 

natural coastline areas and tidal flats 

play an important role in preserving 

ocean environments. To protect 

coastal environments, regular monitoring 

of these areas is important. 

Monitoring has previously been 

done by humans on foot or using 

boats, but this work can be dangerous 

because of breaking waves and rip 

currents. Monitoring on foot is limited 

because it cannot be carried out 

in deep waters, while monitoring by 

boat is limited to the areas accessible 

by water. Automatic monitoring by 

an amphibious robot is therefore expected 

to eliminate the safety threats 

to human monitors and improve operational 

efficiency. However, if an 

amphibious robot moves by gaining 

traction with screws and caterpillars, 

it will not be able to move about in 

areas such as marshes and will damage 

the environments of the areas in 

which it moves. An environmentally 

friendly amphibious robot is thus 

needed. 

Several studies have reported the 

development of amphibious robots. 

An amphibious snake-like robot, the 

ACM-R5 (Yamada et al., 2005), can 

operate both on ground and in water 

by undulating its long body. The 

ACM-R5 uses special paddles and 

ABSTRACT 

With the goal of automatic monitoring of environments along natural coastal 

areas and tidal flats, researchers designed and developed an amphibious robot 

equipped with fin actuators called “RT-I” that mimics the locomotion of both a tortoise 

and a sea turtle. Experiments were carried out using a forearm with 4 degrees 

of freedom, which can reproduce the walking motions of tortoises and sea turtles on 

sand, to evaluate the walking performances of a robotic tortoise and a robotic sea 

turtle. It was clarified that the arm for a robotic tortoise is more suitable for use on 

soil compared with the arm for a robotic sea turtle. The advantages of both sea 

turtles and tortoises were adopted in a robotic turtle, namely, the lift-based swimming 

mode sea turtles use and the quadrupedal locomotion tortoises use. The present 

amphibious robot consists of four main components: (i) leg units, (ii) a control 

unit pressure hull, (iii) a buoyancy adjusting device, and (iv) a fairing cover. To realize 

not only swimming motion with the combination of flapping, rowing, and 

feathering, but also tortoise-like walking motion, three motors were set up at the 

acromioclavicular joint using a differential gear mechanism to independently produce 

the three types of motion, and one motor was set up to produce elbow joint 

motion. A buoyancy-adjusting device was installed to realize walking on land and in 

water as well as swimming in shallow water. The swimming and walking performances 

of the amphibious robot in water were evaluated by measuring the forward 

swimming speed, backward swimming speed, speed of turning, and speed of descending 

vertically as the indexes of the maneuverability of the robotic turtle, and 

the walking speed and propulsive efficiency with the crawl gait for various walking 

patterns in still water and in waves. 

Keywords: locomotion, tortoise, sea turtle, experiment 

wheelsmountedarounditsbodyto 

propel itself through water and over 

ground in a snake-like fashion, generating 

propulsive force that allows it to 

glide freely in the tangential direction. 

The biomimetic amphibious soft cord 

robot (Wakimoto et al., 2006), which 

is made of Mckibben actuators and 

plastic plates, can move both on the 

ground and in water, undulating its 

long body. A spinal cord model and 

its implementation in an amphibious 

salamander robot (Ijspeert et al., 

2007) were studied to demonstrate 

how a primitive neural circuit for 

swimming can be extended using 

phylogenetically more recent limb oscillatory 

centers to explain the ability 

of salamanders to switch between 

swimming and walking. The AQUA 

(Dudek et al., 2007), an amphibious 

robot, can swim and walk along the 

shore and on the bottom of the ocean 

by moving its fins. The AQUA uses 

six paddles, which act as control surfaces 

during swimming and as legs 

while walking. An amphibious walking 

robot was developed by Tanaka 

and Shirai (2006) to perform a shoreline 

survey. It has six legs, each of 


which has three joints. It was successful 

in obtaining the distribution of 

ground levels from land to shallow 

water. An amphibious robotic turtle 

was built by Low et al. (2007) to imitate 

the locomotion of Cheloniidae, 

both in water and on land, to perform 

various operations. The crawling and 

lift-based swimming gaits were analyzed 

and implemented in the prototype. 

However, all of these robots are 

not operated in practice to monitor 

the coastal environment on land and 

in water by using the multiple functions 

of walking and swimming. 

For the field operation of an amphibious 

robot, a rigid fuselage with 

an adequate payload is necessary so 

that a control system and sensors for 

monitoring the environment can be 

installed. We have been studying a biomimetic 

underwater robot equipped 

with mechanical pectoral fins from 

the viewpoint of high maneuverability 

under disturbances such as waves and 

water currents (Kato et al., 2006; 

Suzuki & Kato, 2005; Kato & Liu, 

2003). Taking the field operation 

and application of our experiences on 

the biomimetic underwater robot 

into account, this study focuses on an 

amphibious robotic turtle, which not 

only can swim in the sea but also 

walk on the land to perform environmental 

monitoring of natural coast 

and tidal flat areas. 

Turtles that can walk and swim are 

generally categorized intoseaturtles 

and tortoises. Sea turtles have good 

swimming ability but poor walking 

ability because they drag their bodies 

on the land, which causes friction 

against the sand. Tortoises, on the 

other hand, cannot swim smoothly, 

but they can walk better than sea 

turtles can because of their quadrupedal 

locomotion capability. In this 

study, we attempted to adopt the advantages 

of these two turtles into a robotic 

turtle. 

This paper presents (1) a description 

of the walking performance of 

an arm with 4 degrees of freedom 

(DOF) that can reproduce the walking 

motionsofseaturtlesandtortoises 

from the viewpoints of mobility and 

terrain trafficability, (2) details of the 

design and development of an amphibious 

robot with fin actuators, 

and (3) an evaluation of the walking 

and swimming performance of the 

robot in a laboratory environment. 

Here, mobility is defined as the walking 

performance of a vehicle depending 

on motor torque and vehicle 

configuration. Trafficability is the 

soil-bearing capacity of a vehicle. 

Locomotion of Tortoises 

and Sea Turtles 

Terrestrial Locomotion 

Walker (1971) studied the walking 

of a tortoise, Chrysemys picta, using 

cinephotography and X-rays and 

clarified the relationship between the 

structure of the skeleton and the movements 

of the fore and hind limbs. The 

structure of the skeleton of the forelimb 

is the same as that of the hind 

limb. The forelimb consists of an 

acromioclavicular joint with 3 DOF, 

ahumerus,anelbowwith1DOF,a 

FIGURE 1 

Top view and side view of tortoise Chinemys reevesii. 

forearm, a wrist with 2 DOF, and a 

hand. Wyneken (1997) explained 

that in the locomotion of sea turtles 

on land, clutching movements are 

seen in adult Chelonia mydas, Natator 

depressus, andDermochelys coriacea, 

while adults of other cheloniid species 

employ quadrupedal gaits in which diagonally 

opposite feet move as a pair. 

Sea turtles support themselves on the 

carpus and the anterior edge of the 

hand rather than on the palmar 

surface. 

To design a robotic turtle, we analyzed 

quantitative information on the 

movement of the forelimb joints of a 

captured tortoise, Chinemys reevesii, 

with the following dimensions: length 

of shell × length of forelimb × length 

of hindlimb = 230 mm × 40 mm × 

40 mm. The markers and body-fixed 

coordinates (x, y, z) on the tortoise 

were set up as shown in Figure 1. 

The origin of the body-fixed coordinates 

was set on the acromioclavicular 

joint. First, movies of the walking 

motions of the tortoise were taken 

using two CCD cameras. The movies 

from the top view and side view were 

taken at the same time. Second, the 

marked points were tracked using 

software that computed the twodimensional 

coordinates of the points. 

Third, the three-dimensional coordinates 

of the elbow and wrist on 

the body-fixed coordinates were 


FIGURE 2 

Trajectories of the elbow and the wrist in x-y plane. 

computed. Figures 2 and 3 show the 

two-dimensional tracks in the x-y 

plane and x-z plane of the motion of 

the wrist and the elbow, respectively, 

during walking in the case of a walking 

speed of 0.13 m/s, a period of walking 

of 1.3 s, and a stance phase of 0.78. 

Here, the stance phase denotes the fraction 

of time during which the forelimb 

is set on the ground during the walking 

period. 

If the forelimb is considered to be 

an arm, the joint angles of the forelimb 

FIGURE 3 

Trajectories of the elbow and the wrist in x-z plane. 

derived from inverse kinematics of the 

arm can be obtained. Here, an arm is 

assumed to consist of an acromioclavicular 

joint with 3 DOF, a humerus, an 

elbow with 1 DOF, a forearm, and a 

wrist. The (x, y, z) coordinates are 

fixed at the acromioclavicular joint, as 

shown in Figure 4. The rowing motion 

is defined as rotational motion around 

the z axis. The (x′, y′, z′) coordinates 

are defined as coordinates rotated 

around the z axis by the rowing motion. 

The feathering motion is defined 

as rotational motion around the y′ axis. 

The (x″, y″, z″) coordinates are defined 

as the coordinates rotated around the 

y′ axis by the feathering motion. The 

flapping motion is defined as rotational 

motion around the x″ axis. 

The coordinate system of the arm 

was set up as shown in Figure 4, 

where n, s, anda, denotingunitvectors 

fixed on the forearm, were set parallel 

to x, y, and z, respectively, when 

all of the joint angles were zero. The 

joint angles θ 1 , θ 2 , θ 3 , θ 4 are defined 

as the angle of rowing motion, angle 

of feathering motion, angle of flapping 

motion, and angle of bending of forearm, 

respectively. Lengths of the arms 

l 1 , l 2 are defined as the length of the 

humerus and the length of the forearm, 

respectively. 

Figure 5 shows time variations of 

thejointangles.Wecanseethatthe 

feathering angle varies from 5° to 49° 

during the power stroke from 0 to 

0.6 s, and the bending angle of the 

forelimb varies between −75° and 

−115° during the power stroke, 

which indicates that the forelimb produces 

forward thrust by kicking the 

ground and positioning the forearm almost 

vertically on the ground. During 

the recovery stroke from 0.6 to 1.3 s, 

the bending angle of the forelimb 

reaches −20°, which indicates that the 

forearm is raised from the ground. The 

flapping angle does not vary much 

during the entire stroke. 

Aquatic Locomotion 

Sea turtles swim in water using 

their forelimbs to provide thrust. The 

synchronous sweeping of the flippers 

introduces a stable heading direction. 

Wyneken (1997) explains lift-based 

mechanisms of thrust production in 

which the locomotor apparatus of a 

sea turtle acts as a wing to generate 


FIGURE 4 

Coordinate system and definitions. 

minimize surface area as they are 

brought forward. 

Experiment on the 

Walking Performance 

of an Arm 

We constructed an arm that could 

reproduce the walking motion of a tortoise 

and a sea turtle to analyze the 

walking performance from the viewpoints 

of mobility and trafficability. 

lift forces during large portions of the 

powerstroke. Isobe et al. (2010) clarified 

that the feathering motion of 

fore flippers of a sea turtle influences 

the thrust production by measuring 

the 3-D motion of fore flippers of a 

sea turtle in a water circulating tank 

and conducting numerical simulations 

FIGURE 5 

Time variations of joint angles. 

based on quasi-steady wing element 

theory. 

Wyneken (1997) showed dragbased 

propulsion by a swimming semiaquatic 

turtle. Diagonally opposite 

limbs are protracted, then retracted together, 

and act as paddles. The distal 

elements of the limbs are flexed to 

Arm 

The robotic arm we developed 

makes the motions of rowing, feathering, 

and flapping, and the forearm can 

also be bent. The arm moves on rails to 

make these motions on sand. The angles 

of the motions of rowing, feathering, 

flapping, and bending were 

measured by four potentiometers, the 

distance of the movement along the 

rails was measured by a potentiometer, 

and the forces that were applied to the 

hand were measured by a six-axes force 

sensor, as shown in Figure 6. The humerus 

and forearm are each 150 mm 

long. The rowing angle, feathering 

angle, flapping angle, and bending 

angle of the forearm vary within the 

following ranges, respectively: ±70°, 

FIGURE 6 

Picture of manipulator. 


±70°, ±50°, and 0° to −110°. The rowing motion, feathering motion, flapping motion, and bending motion were produced by 

four motors independently. Because frictional force works between the rails and the arm, a force corresponding to the static 

frictional force was applied to the arm along the rail using a pulley and a weight. In the experiments on the walking performance 

of a sea turtle, an amount of force was subtracted from the static frictional force to simulate the friction a sea turtle 

generates on sand. The arm was connected to the weight by a string. Toyoura sand, which has almost constant particle 

diameter and known physical parameters, was used in the experiments. 

Figure 7 shows the hand shapes. The shape of the end of the hand of the model simulating a tortoise (T ) is a 70 mm × 

70 mm square. That of the model simulating a sea turtle (S) is a 150 mm × 125 mm rectangle. 

Kinetic Relations Between External Force and Joint Torque 

To discuss the propulsive efficiency in terms of walking performance of the arm, we need to know the torque of each joint 

of the arm. We then consider the kinetic relations between external force and joint torque. 

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 

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 

forearm (see Figure 6), ( P : : : : 

r ðx; 

y; zÞ 

T ), and rotational velocities of the end of the forearm, ( Φ : r ð ϕ : x ; ϕ : y ; ϕ : z Þ T ), are described as 

follows: 

: 

P : r 

¼ J θ 

: Φ r 

ð1Þ 

 

J ¼ s 

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 

s 1 s 2 s 3 s 4 

ð2Þ 

where J, P i ,ands i denote the Jacobian matrix, position vector, and vector of rotational direction, respectively (refer to 

Figure 8). 

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 , and 

joint torque, Q(q 1 , q 2 , q 3 , q 4 ). The moment, M i , that is applied to each joint is 

M i ¼ ðP r P i Þ×Fþ M ð3Þ 

and thus the torque of each joint is 

q i ¼ s i ⋅ M i 

¼ s i ⋅ðP r 

P i 

Þ× F þ s i ⋅M 

¼ fs i ×P ð r P i Þg⋅F þ s i ⋅M 

ð4Þ 

and the relation between eternal forces, moments, and joint torque is represented using a transposed Jacobian matrix as 

follows: 

2 3 2 

3 

q 1 s 1 × ðP r P 1 Þ s 1 

q 

Q ¼ 6 2 

7 

4 q 3 

5 ¼ s 2 × ðP r P 2 Þ s 2 

6 

7 

4 s 3 × ðP r P 3 Þ s 3 

5 F 

¼ J T F 

M M 

q 4 s 4 × ðP r P 4 Þ s 4 

ð5Þ 


FIGURE 7 

Hand models. 

In the present study, we discuss 

propulsive efficiency from the viewpoints 

of mobility and trafficability. 

We define propulsive efficiency,η, as 

follows: 

FIGURE 9 

Trajectories of wrist and elbow of type (a). 

motion of a robotic tortoise were configured 

with reference to the experimental 

motion analysis of a tortoise 

in locomotion of tortoises and sea 

turtles. Type (a) motion includes 

movement of the arm vertically during 

the power stroke. It has a symmetric 

trajectory in the anteroposterior direction 

(see Figure 9). Type (b) motion is 

based on the pulling motion during 

the power stroke. It has a larger anterior 

trajectory component within the 

whole trajectory, which is analogous 

to the observed trajectory of the tortoise. 

Type (c) motion is based on 

the kicking motion during the power 

stroke. It has a larger posterior trajectory 

component within the whole trajectory. 

Figures 10, 11, and 12 show 

the time variations of the joint angles 

of type (a), type (b), and type (c), respectively. 

The range of movement of 

η ¼ 

∫ T F x vdt 

4 : 

∑ ∫ T q i θi dt 

i¼1 

ð6Þ 

Here, v denotes the translational speed 

along the x axis. The numerator is the 

value of the work that the arm applies 

to the sand. The denominator is the 

value of the total input work by the 

joints. 

Comparison of Walking 

Performance of an Arm 

Between a Robotic Tortoise 

and a Robotic Sea Turtle 

The walking performance of an 

arm for a robotic tortoise using the 

hand models (T ) was measured. The 

following three types of walking 

FIGURE 8 

Vectors of coordinates. 


FIGURE 10 

Time variations of joint angles of type (a). 

FIGURE 12 

Time variations of joint angles of type (c). 

FIGURE 11 

Time variations of joint angles of type (b). 

the wrist along the x axis and that along 

the z axis were taken as 162 and 

70 mm, respectively. The range of 

movement of the wrist along the 

y axis was taken as 103 mm. The period 

of motion was changed from 7.5 

to 15.0 s with an interval of 2.5 s. 

The sinkage of the end of the forearm 

below the surface of the sand was also 

varied; values of 10, 15, and 20 mm 

were used. Although the sinkage of a 

robotic turtle changes with time according 

to the weight of the body, 

the type of arm motion and the condition 

of the soil, it was treated as an independent 

parameter affecting the 

walking performance of the arm in 

this study so we could evaluate it together 

with other parameters from 

the viewpoint of fundamental walking 

performance. It was controlled by 

a feedback controller located in the 

arm controller using the measured angles 

of the motions of rowing, feathering, 

flapping, and bending by four 

potentiometers, after the initial attitude 

of the arm was set up in relation 

to the surface of the sand. 

Figure 13 shows the averaged propulsive 

forces, Fx, in the direction of 

the x axis during one period for the 

three types of arm motion with the 

motion period of 10 s. Figure 14 

shows the averaged vertical forces, Fz, 

in the direction of the z axis during one 

period for the three types of arm motion 

with the motion period of 10 s. 

The fact that the averaged vertical 

forces, Fz, for type (b) motion are negative 

can be attributed to the posture of 

the forearm in the first half of the time 

period. Namely, the flat plate at the 

end of the forearm digs into the sand 

by positioning the forearm anteriorly 

in the firsthalfofthetimeperiod. 

This action leads it to carry sand on 

the flat plate, which causes a negative 

vertical force during one cycle. On 

the other hand, the forearm in type 

(a) motion is positioned almost vertically 

on the sand during one period, 

and the forearm in type (c) motion is 

positioned posteriorly so as to kick 

sand. Although it is difficult to discriminate 

among the three motion 

types in terms of the propulsive efficiencies 

beyond the sinkage of 15 mm, 

type (a) shows the largest values for 

the averaged propulsive force. The 


FIGURE 13 

Averaged propulsive force Fx in the direction of the x-axis during one 

period for the three types of arm motions against sinkage of the end 

of the forearm below the surface of the sand using the hand model (T). 

FIGURE 15 

Time variations of joint angles of type (d). 

same relation among the three motion 

types can be applied to the averaged 

vertical forces. Type (a) arm motion is 

suitable not only from the viewpoint 

of mobility, but also trafficability. 

Because sea turtles use the anterior 

edge of the hand for locomotion on 

land, we used type (d) of walking 

FIGURE 14 

motion, in which the flipper moves inclined 

at an angle of 45°, in the design 

of the robotic sea turtle. Figure 15 

shows the time variations of the joint 

angles of type (d) with the motion period 

of 10 s. The sinkage of the bottom 

of the flipper below the surface of the 

sand was varied among three different 

Averaged vertical force Fz in the direction of the z-axis during one period for the three types of arm 

motions against sinkage of the end of the forearm below the surface of the sand using the hand 

model (T). 

levels:10mm,15mm,and20mm. 

The walking motion of a sea turtle generates 

friction on the sand because the 

turtle drags its body on the land as it 

walks. To model this situation, friction 

on the carriage of the manipulator was 

added. We compared the walking performance 

of these two types of motion 

under the conditions including friction 

on the carriage of the manipulator 

set at the following levels: 2.94, 4.90, 

and 6.86 N. 

Figure 16 shows the averaged propulsive 

force, Fx, in the direction of the 

x axis during one period against sinkage 

of the end of the forearm below 

the surface of the sand and friction 

on the carriage of the manipulator 

(the key to the symbols expresses friction 

on the carriage of the manipulator). 

As the friction on the rail increases, the 

averaged propulsive force, Fx, also increases. 

Figure 17 shows the average 

vertical force during one period against 

sinkage. As the sinkage increases, the 

maximum vertical force, Fz, also 

increases. The dependency of the 

maximum vertical force, Fz, on the 


FIGURE 16 

Averaged propulsive force Fx in the direction of x-axis during one period 

for type (d) arm motion against sinkage of the end of the forearm 

below the surface of sand using the hand model (S). 

FIGURE 17 

Averaged vertical force Fz in the direction of z-axis during one period 

for type (d) of arm motion against sinkage of the end of the forearm 

from the surface of sand using the hand model (S). 

friction on the rail is clear. When we 

consider the walking motion of a robotic 

sea turtle on the beach, friction 

between the sand and the body depends 

on the vertical force subtracting 

the vertically upward force from the 

weight of the robotic sea turtle. 

When comparing type (a) walking 

motion of a robotic tortoise, which 

showed the best performance among 

the three types, with type (d) walking 

motion of a robotic sea turtle, it can be 

seen that the propulsive efficiency of 

type (d) is influenced by friction and 

the averaged vertical force acting on 

the hand plate in the case of the arm 

for a robotic tortoise is larger than in 

the case of the arm for a robotic sea turtle. 

An arm for a robotic tortoise is suitable 

for moving heavy payloads over a 

variety of terrains. To discuss the trafficability 

of the arm for a sea turtle, we 

define M as the mass of the body, g as 

gravitational acceleration, k as the frictional 

coefficient (Setouchi & Shinjo, 

2001) between the soil and the ventral 

surface of the robotic sea turtle, Fxm as 

the mean value of propulsive force, and 

Fzm as the mean value of vertical force 

against the soil. Assuming that a pair of 

fore flippers and a pair of rear flippers 

exert thrust forces simultaneously, we 

obtain the following equation of static 

equilibrium for walking on soil: 

ðM⋅g − 4⋅FzmÞ⋅k ¼ 4⋅Fxm ð7Þ 

IfweassumeFxm=Fzmforthe 

type (d) arm motion based on Figures 

16 and 17, we obtain the following 

relation: 

k 

Fzm ¼ M⋅g⋅ < M⋅g=4 ð8Þ 

41þ ð kÞ If we set the length of the humerus as l 1 

and the length between the elbow joint 

and center of the hand plate as l 2 using 

the arm structure shown in Figures 6 

and 8, the torque around the x axis 

and that around the z axis are Fzm∙ 

(l 1 + l 2 ). For a robotic tortoise using a 

crawling gait with slow speed, we obtain 

Fzm = M · g/4 during the time when 

the 4 feet are placed on ground. If we 

assume that the robotic turtle uses 

type (a) arm motion and that Fxm is 

nearly equal to Fzm/25 based on Figures 

13 and 14, we obtain the torque 

around the x axis of Fzm∙l 1 and a 

torque around the z axis of Fzm · l 1 /25. 

The total torque on the arm of the 

sea turtle robot, Q S , is expressed as 

follows: 

k 

Q S ¼ 2⋅M⋅g⋅ 

41þ ð kÞ ⋅ ð l 1 þ l 2 Þ: ð9Þ 

The total torque on the arm of the 

tortoise robot, Q L , is expressed as 

follows: 

Q L ¼ 26=25⋅M⋅g=4⋅l 1 

The ratio of Q S to Q L γ is 

γ ¼ 25 

13 ⋅ k 

1 þ k ⋅ 1 þ l2 

l1 

ð10Þ 

ð11Þ 

Because the friction ratio of sand 

(Setouchi & Shinjo, 2001) is in the 


ange of 0.4-0.6 and l 2 /l 1 is larger than 

1.0 for a sea turtle, we find that γ tends 

to approach 1.0 if k becomes smaller, 

and that γ tends to become larger 

than 1.0 if k becomes larger. Because 

smaller torques acting around the 

joints are needed from the viewpoint 

of the mechanical design of a 

turtle robot, the arm for the robotic 

tortoise is suitable on soil from the 

viewpoint of trafficability. 

Design and Development 

of an Amphibious Robot 

With Fin Actuators 

A previously designed mechanical 

pectoral fin (Kato & Liu, 2003) has a 

drag-based swimming mode and a liftbased 

swimming mode. The former is 

characterized by the rowing action 

forming a high angle to the horizontal 

axis of the body, while the latter is 

characterized by the flapping action 

forming a small angle to the horizontal 

axis. It was revealed through the optimization 

of motion of the mechanical 

pectoral fin that the lift-based swimming 

mode rather than the dragbased 

swimming mode is suitable for 

generation of propulsive force in uniform 

flow, while the drag-based swimming 

mode rather than the lift-based 

swimming mode is suitable for generation 

of propulsive force in still water. 

To realize both the drag-based swimming 

mode and the lift-based swimming 

mode, a combination of rowing 

motion, flapping motion, and feathering 

motion is needed. The hydrodynamic 

characteristics of the drag-based 

swimming mode and the lift-based 

swimming mode are discussed in details 

elsewhere (Suzuki et al., 2007). 

Walking locomotion using fin actuators 

is of two types, imitating the motionofatortoiseandthatofasea 

turtle. Sea turtle-like walking motion 

has the following characteristics: 

It is dynamically stable. 

Terrain condition strongly affects 

the attitude of the body. 

Friction between the soil and the 

body necessitates additional thrust 

force. 

Tortoise-like walking motion has the 

following characteristics: 

It is dynamically more unstable 

compared with the sea turtle-like 

walking motion. 

The attitude of the body has more 

DOF compared with the sea turtlelike 

walking motion. 

The body weight is vertically applied 

to the foot. 

In this study, we adopted the 

tortoise-like walking motion, which 

has better trafficability on soil than 

the sea turtle-like walking motion, as 

discussed in the previous section. 

FIGURE 18 

Components of Robotic Turtle (RT-I) and fin actuator with 4 DOF. 

The amphibious robot we designed 

consists of the following four main 

components (see Figure 18): 

(1) leg units, 

(2) a control unit in a pressure hull, 

(3) a buoyancy adjusting device in a 

pressure hull, and 

(4) a fairing cover 

Each pressure hull was designed to resist 

the pressure at the water depth of 

10 m. We used “Solidworks” for 3-D 

CAD software to design the hardware 

and “LabVIEW” to develop 

the control software. Four DOFs are 

needed to realize not only three 

types of swimming motion but also 

the tortoise-like walking motion. 

Therefore, three motors were set up 

at the acromioclavicular joint using a 

differential gear mechanism to independently 

produce flapping motion, 

rowing motion, and feathering motion, 

and one motor was set up at 


the elbow joint to produce bending 

motion of the forearm (see Figure 18). 

Motors and reduction gears were 

selected according to the simulation 

results on walking. An open dynamic 

engine of the kinetics calculation library 

with open sources was used for 

the simulation. A fuselage with the dimensions 

(W × H × L = 0.50 × 0.20 × 

0.80 m) was used. The length of the 

humerus was set at 0.25 m, and the 

length of the forearm was set at 

0.15 m. There are various walking 

gaits for quadrupedal locomotion. 

We selected a crawling gait for low 

speed where the legs are lifted up one 

by one. The fin actuators of this 

robot have two functions, walking and 

swimming, so we had to design the 

shape of the fin withbothwalking 

and swimming in mind. For swimming, 

it is desirable to have a large 

fin area. However, the fin mayinterfere 

with the base of the leg unit 

during walking. The fin shapeisdesigned 

to minimize the interference. 

Figure 19 shows the form of the fin 

with the root chord of 0.195 m, the 

span of 0.15 m, and the maximum 

thickness of 0.32 m. 

Figure 20 shows the outline of 

the control unit’s electric circuit. The 

control unit consists of a CPU, 

motor drivers, an azimuth sensor, 

FIGURE 19 

Form of fin. 

FIGURE 20 

Electric circuit of the control unit. 

three-axes rate gyros, a pressure sensor, 

a GPS, and related minor parts. Batteries 

are used separately for motors 

and the CPU with sensors. Nickel 

hydride batteries with the capacity of 

13.2 V, 8.0 Ah are used for motors 

by arranging a series circuit of two 

and a parallel circuit of two. It is also 

possible to provide the control unit 

with electric power from outside. 

The amphibious robot has to realize 

walking in which the buoyancy is 

less than the weight and swimming 

in which the buoyancy is equal to or 

greater than the weight, because the 

mission of the robot includes travel 

in shallow water with breaking waves. 

To realize these capabilities, the robot 

must be equipped with a buoyancy 

adjusting device. The buoyancy adjusting 

device with the buoyancy capacity 

of ±0.35 kg was designed using 

a pair of pistons arranged symmetrically 

in the longitudinal direction so 

as not to have an effect on the attitude 

of the robot. 

It is desirable to cover the body 

of the robot with streamlined fairing. 

The general-purpose computer fluid 

dynamics software “FLUENT” was 

used to estimate the hydrodynamic 

drag on the fairing cover and to design 

a fairing cover with low hydrodynamic 

drag. The cover was made of glass 

fiber-reinforced plastics. 

Figure 21 shows a photograph of 

the amphibious robot equipped with 

fin actuators, named “RT-I.” The 

principal dimensions are shown in 

Table 1. 


FIGURE 21 

Photograph of the RT-I amphibious robot 

equipped with fin actuators. 

TABLE 1 

Specification of RT-I. 

Total mass [kg] (without buoyancy control device and battery) 90 

Depth of pressure resistant [m] 10 

Walking speed in the water [m/s] (experimental value) 0.025 

Swimming speed [m/s] (experimental value) 0.168 

Dimension [m] Body width 0.73 

Body height 0.55 

Body length 1.68 

Forearm length 0.26 

Humerus length 0.2 

Swimming and Walking 

Performance by the 

Robotic Turtle RT-I 

Swimming Performance 

A towing tank test was carried out 

to measure drag forces acting on the 

body of the robotic turtle in the towing 

tank of Osaka University. The forces 

applied to the body were measured 

by the force sensor at various towing 

speeds. From this experiment, the relationship 

between the drag forces and 

the fluid velocities was obtained to estimate 

the thrust forces produced by 

the fins. The swimming speeds in 

free swimming condition of the robotic 

turtle were measured in the 

same towing tank for the estimation 

of the swimming performance. The 

lift-based swimming mode was 

adopted in the experiments. It uses 

the flapping motion and the feathering 

motion in horizontal plane of the 

robot; on the other hand, it uses the 

rowing motion and the feathering motion 

in vertical motion of the robot. 

The speeds of the towing carriage 

were determined by operating it parallel 

to the robot for several values of 

amplitudes of flapping motion and 

feathering motion. From the experimental 

result of the swimming speeds, 

the thrust forces were estimated by 

using the relationship between drag 

forces and swimming speeds. Figure 

22 shows the thrust forces for the 

amplitudes of feathering motion of 

30° and 45° and the amplitudes of flapping 

motion of 20° and 30°. From this 

figure, we can see that the thrust force 

increases as the amplitude of the 

flapping motion increases. Figure 23 

shows the thrust forces for amplitudes 

of a flapping motion of 30° versus amplitudes 

of the feathering motion. 

From this figure, we can see that the 

thrust force becomes the largest at the 

FIGURE 22 

amplitude of the feathering motion of 

30°. Judging from these results, the 

largest thrust force is made at the amplitudes 

of flapping motion and feathering 

motion of 30°. 

The swimming speeds of forward 

motion, backward motion, turning 

motion, and vertically descending motion 

were measured as performances of 

the maneuverability of the robotic turtle. 

Table 2 shows the swimming 

speeds of these motions with fixed amplitudes 

of joint angle. Unlike the forward 

motion, the swimming speed in 

backward motion with the amplitudes 

of feathering motion of 45° is larger 

than the case of 30°. 

Thrust forces in swimming motion against amplitudes of flapping motion and feathering motion. 


FIGURE 23 

Thrust forces in swimming motion against amplitudes of feathering motion. 

Walking Performance 

in Still Water 

TABLE 2 

The swimming speeds for various swimming motions. 

The crawl gait was adopted as the 

walking gait of the robotic turtle because 

it has a high level of stability. 

During the walking motion using the 

crawl gait, the robot moves each arm 

individually. This means the robot 

supports the body with at least three 

arms at all times. In operation of the 

robot in sea water, it is supposed that 

the seabed is not flat and there exist 

disturbances such as waves and water 

currents. Therefore, the robotic turtle 

needs more stability under such conditions. 

For that reason, we introduced a 

modified walking motion by considering 

the movement of the center of 

gravity in the normal crawl gait, as 

showninFigure24.Themovement 

of the center of gravity makes the stability 

margin larger, which is defined as 

the distance from the side of the support 

polygon to the projection of the 

center of gravity on land, as shown in 

Figure 25. 

Two types of walking patterns were 

made. The first walking pattern consists 

of four steps to move an arm, for 

a total of 16 steps to move four arms 

during a cycle. The trajectory of the 

end of an arm in this pattern forms a 

rectangle. The second walking pattern 

consists of three steps to move an arm, 

for a total of 12 steps to move four 

Swimming Pattern 

Flapping (Rowing) 

Amplitude (°) 

Feathering 

Amplitude (°) 

Swimming 

Speed 

Forward motion 30 30 0.168 

Forward motion 30 45 0.15 

Backward motion 30 30 0.088 m/s 

Backward motion 30 45 0.102 m/s 

Turning motion 30 30 9.33°/s 

Turning motion 30 45 10.30°/s 

Descending motion 30 (rowing) 45 0.05 m/s 

arms during a cycle (Figure 26). The 

trajectory of the end of the arm of 

the second walking pattern forms a 

right triangle. In this case, it is expected 

that walking will be faster because of 

the reduction in the number of walking 

steps and that the energy consumption 

will be smaller because of the 

reduction of total motor speed. 

The walking speed and the motor 

currents for each walking pattern 

were measured while the robot walked 

on the bottom of the pool (L × B × H = 

4.5 m × 2.0 m × 0.8 m) (see Figure 27) 

to evaluate the walking performances 

in still water. Here the length of the 

stride of each walking pattern was set 

at 0.2 m. The joint torque was derived 

from the value of the motor current 

measured by the experiment, the 

torque constant, the reduction ratio, 

and the transmission efficiency, as 

showninEq.(12).Theconsumption 

energy was derived from the 

value of the derived joint torque, as 

follows: 

q ij 

¼ I ij ⋅K t ⋅i⋅η′ 

ð12Þ 

Here, I ij denotes the motor current of 

the joint P ij ,K t is the torque constant, 

i is the reduction ratio, and η′ is the 

transmission efficiency. Figure 28 

shows a comparison of walking speeds 

between measurement and simulation 

by the walking simulator for the two 

walking patterns. From this figure, 

we can see that the walking speed for 

the second walking pattern is greater. 

This is attributed to the small period 

of the second walking pattern for one 

cycle because of the reduction in walking 

steps. We can see that the experimental 

value of the walking speed for 

each walking pattern is smaller than 

the calculated value. This is because 

the end of the arm slips on the bottom 


FIGURE 24 

Schematic view of crawl gait and modified crawl gait. 

FIGURE 27 

Picture of Robotic Turtle in walking in a pool. 

FIGURE 25 

Definition of stability margin. 

FIGURE 26 

Trajectories of the arm end of the first and second walking patterns. 

of the pool when the robot walks in 

still water. 

Walking Performance in Waves 

We estimated the walking performance 

of the robotic turtle in waves 

through a series of experiments. Figure 

29 shows a schematic view of the 

experiment for estimating the walking 

performance in waves. Two types of 

experiments were carried out in the 

towing tank of Osaka University. 

The first type measured the walking 

performance by changing the walking 

patterns in waves and in still water. 

The second type measured the walking 

performances of the second walking 

pattern in waves by changing the 

wave conditions, namely, the wave 

height, the wavelength, and the 

depth of the virtual ground. For the 

first type of experiment, the walking 

speeds of the first and second normal 

walking patterns were measured for 

thestridelengths(L)of0.2mand 

0.3 m under the wave height of 0.1 m, 

the wavelength of 2 m, and the water 

depth of 0.9 m. The walking speeds 

of the first and second modified walking 

patterns were also measured. Figure 

30 shows a comparison of the 

walking speeds for various walking patterns 

in waves and in still water. From 

this figure, we can see that the walking 

speeds for all cases decreased in waves 

compared to the performance in still 

water. The walking speed reaches its 

maximum value in the case of the 

length of the stride of 0.3 m and the 

second walking pattern, regardless of 

crawl gait. Figure 31 shows a comparison 

of the consumption energy for 

various walking patterns in still water 

and in waves. From this figure, we 

can see that the consumption energy 

of the first walking pattern in waves 

is larger than that in still water in the 

caseofthelengthofthestrideof 


FIGURE 28 

Comparison of walking speed in water. 

FIGURE 29 

Schematic view of experiment for walking performance in waves. 

FIGURE 30 

Comparison of walking speeds for various walking patterns in still water and in waves. 

0.3 m, although that of the second 

walking pattern in waves is smaller 

than that in still water. Regarding 

that the walking speeds of the first 

and second walking patters in waves 

become smaller than those in still 

water, we find that the propulsive efficiency 

of the first walking pattern in 

waves becomes worse than that of the 

second walking pattern in waves in 

the case of the length of the stride of 

0.3 m. This may be attributed to the 

interaction between the hydrodynamic 

forces acting on the arms and the paths 

of the arms, namely, larger energy is 

consumed in the first walking pattern 

during one cycle of the arm motion 

in waves by the hydrodynamic forces 

than in the second walking. 

Figure 32 shows the variation of 

walking speed against the wave heights 

from 0 m to 0.15 m with the wavelength 

of 2 m, the water depths of 

0.6mand0.9m,andnormaland 

crawl modified gaits for the second 

walking pattern with the length of 

the stride of 0.3 m. From this figure, 

we can see that the walking speed is decreased 

as the wave height increases 

and that the walking speed becomes 

smaller and the rates of the decrease become 

larger as the water depth decreases 

from 0.9 m to 0.6 m. Figure 33 shows 

the variation of consumption energy 

against the wave heights under the 

same condition for Figure 30. From 

this figure,wecanseethattheconsumption 

energy is almost constant 

against the wave height, although the 

consumption energies are smaller 

than those in still water, and that the 

propulsive efficiency in waves become 

worse if the water depth decreases because 

the rate of decrease of walking 

speed for d = 0.6m is larger than for 

d = 0.9 m. This is attributed to larger 

water current induced by waves in 

shallow water, which produces larger 


FIGURE 31 

Comparison of consumption energy for various walking patterns in still water and in waves. 

FIGURE 32 

Variation of walking speed against wave height. 

hydrodynamic drag force on the 

vehicle. 

Summary 

In this paper we discussed the 

mechanisms of swimming and walking 

of turtles and examined how those 

mechanisms could be applied to an 

amphibious robot, with the goal of 

designing a robot that can perform 

automatic monitoring of environments 

along natural coastal areas and 

tidal flats. Using a manipulator with 

4 DOF, we discussed the characteristics 

of walking of sea turtles and tortoises 

from the viewpoints of mobility 

and trafficability. The advantages of 

sea turtles in swimming and tortoises 

in walking were adopted in the robotic 

turtle. The experiments in a water tank 

revealed that the mobility and the trafficability 

of the amphibious robot at 

the bottom of the water were greatly 

affected by waves in shallow water. 

The author will investigate the tracking 

performance of the amphibious 

robot on tidal flats, moving from 

land to sea and vice versa. 

FIGURE 33 

Variation of consumption energy against wave height. 


This research was funded for 

3 years beginning in 2008 by the Ministry 

of Education, Culture, Sports and 

Technology, Japan (grant 20246130) 

under the project title “Establishment 

of basic technology of a biomimetic 

underwater vehicle and its applications.” 

Author: 

Naomi Kato 

Graduate School of 

Engineering, Osaka University 

Yamadaoka 2-1, Suita, 

Osaka 565-0871, Japan 

Email: kato@naoe.eng.osaka-u.ac.jp 


References 

Dudek, G., Giguere, P., Prahacs, C., 

Saunderson, S., Sattar, J., Torres-Mendez, 

L.-A., … Georgiades, C. 2007. Aqua: An 

amphibious autonomous robot. IEEE 

Comput Mag. 40(1):46-53. 

Ijspeert, A.J., Crespi, A., Ryczko, D., & 

Cabelguen, J-M. 2007. From swimming to 

walking with a salamander robot driven by 

a spinal cord model. Science. 315(5817): 

1416-20. doi: 10.1126/science.1138353. 

Isobe, Y., Kato, N., Suzuku, H., Senga, H., 

Matsuda, Y., Kawamura, K., … Oshika, T. 

2010. Motion analysis of sea turtle with 

prosthetic flippers. In: The 20th (2010) Int. 

Offshore and Polar Eng. Conf. pp. 335-42. 

Beijing: International Society of Polar and 

Offshore Eng. 

Kato, N., Ando, Y., Shigetomi, T., & 

Katayama, T. 2006. Biology-inspired precision 

maneuvering of underwater vehicles (Part 4). 

Int J Offshore Polar. 16(3):195-201. 

Kato, N., & Liu, H. 2003. Optimization 

of motion of a mechanical pectoral fin. 

JSME Int J C. 47(4):1356-62. doi: 10.1299/ 

jsmec.46.1356. 

amphibious walking robot—The design 

concepts and the 1st field experiment. 

In: paper presented at the International 

Symposium on Automation and Robotics 

in Construction 2006 (ISARC 2006). 46-51. 

Tokyo: IEEE. 

Wakimoto, S., Suzumori, K., & Kanda, T. 

2006. A bio-mimetic amphibious soft cord 

robot. JSME Int J C. 72(714):171-7. 

Walker, W.F., Jr. 1971. A structual and 

functional analysis of walking in the turtle 

Chrysemys picta marginata. J Morphol. 

134:195-213. doi: 10.1002/jmor.1051340205. 

Wyneken, J. 1997. Sea turtle locomotion: 

Mechanisms, behavior, and energetics. In: 

The Biology of Sea Turtles, eds. Lutz, P.L., 

& Musick, J.A., 165-98. New York: CRC 

Press. 

Yamada, H., Chigisaki, S., Mori, M., Takita, 

K., Ogami, K., & Hirose, S. 2005. Development 

of amphibious snake-like robot ACM-R5. 

In: paper presented at the 36th International 

Symposium on Robotics(ISR 2005). Tokyo: 

Japan Robot Association. 

Low, K.H., Zhou, C., Ong, T.W., & Yu, J. 

2007. Modular design and initial gait study 

of an amphibious robotic turtle. In: paper 

presented at the International Conference 

on Robotics and Biomimetics, 2007. 535-40. 

Sanya, China: IEEE. 

Setouchi, H., & Shinjo, T. 2001. The effect 

of water on frictional characteristics between 

calcareous sands and a steel plate. In: Report 

of Faculty of Agriculture. University of the 

Ryukyus. 48:103-11. 

Suzuki, H., & Kato, N. 2005. A numerical 

study on unsteady flow around a mechanical 

pectoral fin. Int J Offshore Polar. 15(3):161-7. 

Suzuki, H., Kato, N., & Suzumori, K. 2007. 

Load characteristics of mechanical pectoral 

fin. Exp Fluids. 44(5):759-71. doi: 10.1007/ 

s00348-007-0432-x. 

Tanaka, T., & Shirai, T. 2006. Development 

of automated shoreline surveying system using 


PAPER 

Marine Applications of the Biomimetic 

Humpback Whale Flipper 

AUTHORS 

Frank E. Fish 


Paul W. Weber 

Applied Research Associates, Inc. 

Mark M. Murray 

United States Naval Academy 

Laurens E. Howle 

Duke University 

Introduction 

Life began with water. The highdensity 

and viscous nature of water 

has imposed a strong evolutionary selection 

pressure on the design of animals 

that move through this aqueous 

medium. Over the course of millions 

of years, different phylogenetic lines 

of animals have, in effect, experimented 

with various combinations 

of morphologies and behaviors to enhance 

locomotor performance. The 

great diversity of body shapes, surface 

textures, and propulsive mechanisms 

exhibited by aquatic animals has produced 

a variety of biomechanical solutions 

for the reduction of drag, increase 

in thrust production and efficiency, 

maintenance of stability, and enhancement 

of maneuverability. By emulating 

these biological characteristics in 

those instances where animal performance 

is superior to manufactured devices,theperformanceofengineered 

marine systems may be improved 

through the field of biomimetics. 

The cetaceans (whales, dolphins, 

porpoises) have been the focus of inspiration 

for technological development in 

ABSTRACT 

The biomimetic approach seeks technological advancement through a transfer 

of technology from natural technologies to engineered systems. The morphology of 

the wing-like flipper of the humpback whale has potential for marine applications. 

As opposed to the straight leading edge of conventional hydrofoils, the humpback 

whale flipper has a number of sinusoid-like rounded bumps, called tubercles, which 

are arranged periodically along the leading edge. The presence of the tubercles 

modifies the water flow over the wing-like surface, creating regions of vortex generation 

between the tubercles. These vortices interact with the flow over the tubercle 

and accelerate that flow, helping to maintain a partially attached boundary layer. 

This hydrodynamic effect can delay stall to higher angles of attack, increases lift, 

and reduces drag compared to the post-stall condition of conventional wings. As 

the humpback whale functions in the marine environment in a Reynolds regime 

similar to some engineered marine systems, the use of tubercles has the potential 

to enhance the performance of wing-like structures. Specific applications of the tubercles 

for marine technology include sailboat masts, fans, propellers, turbines, and 

control surfaces, such as rudders, dive planes, stabilizers, spoilers, and keels. 

Keywords: tubercles, delayed stall, Megaptera novaeangliae, leading edge, 

bio-inspired design 

the marine environment. The cetacean 

lineage dates back 55 million years. 

The intense selection pressures for a 

fast swimming, maneuverable marine 

predator have culminated in a highly 

streamlined body with advanced sensory 

capabilities that is propelled by a 

highly efficient propulsion mechanism. 

There are a number of examples 

where cetaceans have been the inspiration 

for the development or improvements 

of marine technology. The 

cetacean body shape was used by 

Cayley (circa 1800) as a solid of leastresistance 

for the development of airplane 

fuselage and boat hull designs 

(Gibbs-Smith, 1962). The famous 

but erroneous “Gray’s paradox” led 

to examination of special drag reduction 

mechanisms (Gray, 1936; Fish 

& Rohr, 1999; Fish, 2006), including 

the biomimetic development of compliant 

coatings for viscous dampening 

(Kramer, 1960; Riley et al., 1988; 

Carpenter & Pedley, 2003). The compactness 

and high resolution of the 

echolocation system of dolphins provides 

a benchmark for the improvement 

of SONAR systems (Au, 1993). 

The thrust performance of oscillating, 

wing-like systems, such as the flukes of 

dolphins, has been considered superior 

to screw propellers (Peterson, 1925; 

Liu & Bose, 1993; Triantafyllou & 

Triantafyllou, 1995). The flexibility 

of the oscillatory flukes can allow operation 

at high efficiency over an extended 

speed range without cavitation 

(Fish & Lauder, 2006; Iosilevskii & 

Weihs, 2007). 

This report focuses on a unique morphology 

of a highly derived aquatic 


mammal that has general biomimetic 

applications for marine systems. The 

flipper of the humpback whale and 

its bumpy leading edge provide a 

novel approach to enhance the hydrodynamic 

performance of wing-like 

structures for operation in water. A 

principal attribute of using the whale 

as a model to construct a biomimetic 

system is the scale. As the whale is of 

a large size and swims at speeds that 

compliment the scale and operation 

of engineered marine systems, application 

for marine technologies can be 

readily undertaken. 

Humpback Whale 

as Inspiration 

The humpback whale (Megaptera 

novaeangliae) hasthelongestflipper 

of any cetacean (i.e., whale, dolphin, 

porpoise), with regard to both absolute 

and relative size (Figure 1; Fish 

& Battle, 1995). The flippers are involved 

with the underwater maneuvers 

performed by the species that are associated 

with their mode of feeding 

(Friedlaender et al., 2009; Hazen 

et al., 2009). Humpback whales are 

the only baleen whales (e.g., blue 

whale, fin whale, minke whale, right 

whale) that rely on tight, rapid turns 

to capture prey (Fish & Battle, 

1995). The humpback whales use 

their flippers as biological hydroplanes 

to achieve tight circles to corral and 

engulf prey (Fish et al., 2011). 

FIGURE 1 

Flippers on humpback whale (left), showing 

scalloped pattern of tubercles (center), and flipper 

cross-section (right) (from Fish & Lauder, 

2006). 

The humpback whale flippers are 

unique because of the presence of 

large tubercles along the leading edge, 

which gives this surface a scalloped 

appearance (Figure 1). The distances 

between tubercles decrease distally, 

although these distances remain relatively 

constant at 7-9% of span over 

the midspan of the flipper (Fish & 

Battle, 1995). 

The planform and cross-sectional 

views of the humpback whale flipper 

are shown in Figure 1. Whereas typical 

wing-like structures have a straight 

leading edge without the presence of 

irregularities or perturbations, the 

humpback flipper defies convention 

with prominent rounded bumps that 

are regularly spaced along the leading 

edge of its high-aspect ratio flipper 

(Fish et al., 2008). Humpback 

whale flippers closely resemble the 

21% thick, low drag NACA 63 4 -021 

foil in cross section (Abbott & von 

Doenhoff, 1959; Fish & Battle, 

1995). Furthermore, the flippers have 

high mobility (Edel & Winn, 1978). 

The elongate flippers function as 

wings to generate the forces necessary 

for turning maneuvers (Fish et al., 

2011). Turning is important in the 

capture of elusive prey. Humpback 

whales feed on shoals of small fish 

and krill. The preys are forced into a 

tight ball by the whales striking the 

water surface with their flukes or circling 

the prey from underneath while 

emitting bubbles. The bubbles rise to 

corral the prey as a bubble net. In either 

case, the whale maneuvers under 

the prey for engulfment by executing 

a rapid turn. Lift generated by the flippers 

is used to produce a centripetal 

force for the turn. The tubercles produce 

vortical flows over the surface of 

the flipper and control lift characteristics 

at high angles of attack and delay 

stall (Fish & Battle, 1995; Miklosovic 

et al., 2004; Fish & Lauder, 2006; Fish 

et al., 2011). 

Hydrodynamic Effect 

of Tubercles 

The prominence of the tubercles 

on the leading edge of the humpback 

whale flippers and the swimming pattern 

of the whale suggests that these 

novel structures have a distinct hydrodynamic 

effect (Figure 2). In addition, 

the pattern of barnacle distribution on 

the flippers (i.e., barnacles are confined 

to the peak of the tubercle and do 

not occur between tubercles; Fish & 

Battle, 1995) indicates that the flow 

over the flipper is affected by tubercles. 

This section reviews potential hydrodynamic 

advantages, flow control, 

and limitations of the tubercles on 

wing-like structures. 

Hydrodynamic Advantages 

The presence of leading-edge tubercles 

on a wing-like structure can have 

a positive influence on the hydrodynamic 

performance. Wind tunnel 

tests showed that wings with tubercles 

improved maximum lift by over 6%, 

increased the ultimate stall angle by 

40%, and decreased drag by as much 

as 32% (Figure 3; Miklosovic et al., 

2004). The tubercles, when facing 

FIGURE 2 

Idealized humpback flipper with leading-edge 

tubercles. 


FIGURE 3 

Humpback whale flipper models and results of wind tunnel experiments. The models (left) with 

and without tubercles were machined from clear polycarbonate, based on a symmetrical NACA 

0020 foil section. Lift and drag data (right) for the flipper models were obtained from tests in a 

wind tunnel. The solid lines in a, b, and c show the average of the data for the flipper model without 

tubercles, and open triangles are for the model with tubercles. Lift coefficient C L (a), drag coefficient 

C D (b), and aerodynamic efficiency L/D (c) are plotted against angle of attack, α (from 

Miklosovic et al., 2004). 

intothefreestreamflow, alter the 

fluid flow over wing-like structures 

(Bushnell & Moore, 1991; Fish & 

Battle, 1995; Fish et al., 2011). Furthermore, 

the lift to drag (L/D) ratio, 

which represents the aerodynamic efficiency, 

displayed a greater peak L/D 

for a wing geometry with tubercles 

(Miklosovic et al., 2004, 2007; 

Hansen et al., 2009). 

The position and number of tubercles 

on the flipper suggest analogues 

with specialized leading edge control 

devices that improve the hydrodynamic 

performance of wings. The 

occurrence of “morphological complexities” 

on a lifting body could reduce 

or use pressure variation at the tip to decrease 

drag and improve lift to prevent 

tip stall (Bushnell & Moore, 1991). Alternatively, 

various biological wings 

utilize leading-edge control devices to 

maintain lift and avoid stall at high attack 

angles and low speeds. 

The function of the tubercles may 

be analogous to strakes used on aircraft. 

Strakes are large vortex generators 

that change the stall characteristics 

of a wing (Hoerner, 1965; 

Shevell, 1986; Bertin & Smith, 

1998). Stall is postponed because the 

vortices exchange momentum within 

the boundary layer to keep it partially 

attached over the wing surface. Lift is 

thus maintained at higher angles of attack 

with strakes compared to wings 

without strakes, although maximum 

lift is not increased by strakes (Shevell, 

1986). Another leading-edge device are 

slots or slats that delay stall or move 

the angle of attack of maximum lift 

to a lower value (Wegener, 1991; 

Bertin & Smith, 1998). The moveable 

slats create a space anterior of the fixed 

wing to allow higher-pressure fluid to 

rise from the underside of the wing. 

The movement of fluid from the 

high-pressure side to the low-pressure 

side of the wing improves mixing and 

helps to maintain the boundary layer 

and delay stall (Wegener, 1991). However, 

the tubercles have distinct advantages 

over slats. Tubercles are 

passive structures that have no drag 

penalty when designed onto wings 

(Miklosovicetal.,2004),whereas 

slats are actively deployed and incur increased 

drag (Hoerner, 1965). 

Vortex Generation 

The mechanism for enhanced hydrodynamic 

performance due to the 

presence of tubercles appears due to 

the specific pattern of vortex generation 

over the surface of the flipper. 

Flow visualization experiments on 

wavy bluff bodies showed periodic 

variation in the wake width across the 

span (Owen et al., 2000). A wide wake 

with two simultaneous vortices occurred 

where the body protruded 

downstream and a narrow wake occurred 

where the body protruded upstream. 

The flowinthewakeofthe 

wavy body was different from the 

wake of a straight cylinder, which exhibited 

a typical von Karman Vortex 

Street of alternating vortex pairs. A 

bluff body with a spanwise sinusoidal 

form could reduce drag by at least 

30%, compared to equivalent straight 

bodies (Bearman & Owen, 1998). 

The vortices produced by a wing 

section with tubercles are shown in 

Figure 4. The tubercles generate 

separated, chordwise vortices in the 

troughs at high angles of attack. 

These vortices are formed as the flow 

strikes the leading edge of the trough. 

As the flow does not strike the leading 

edge normally, the flow is sheared into 

the trough’s center to generate the 

vortices. These vortices are convected 

along the chord. The spanwise arrangement 

of the vortices is in a pair 

on each side of the tubercle crest with 

opposite spins (Hansen et al., 2010). 

The flow directly over the tubercle interacts 

with the vortices located downstream 

and lateral to the tubercle crest. 


FIGURE 4 

Pressure contours and streamlines at α = 10° for NACA 63-021 with straight leading edge (left) and 

with tubercles (right). An unsteady Reynolds-averaged Navier-Stokes (RANS) simulation was 

used. A separation line is shown on the wing section without tubercles. For the wing with tubercles, 

large vortices are formed posterior of the troughs along the leading edge and flow posterior of 

the tubercles is shown as straight streamlines without separation. Images courtesy of E. Paterson. 

The tangential velocities of the inward 

facing flows of the pair of vortices are 

directed toward the trailing edge of 

the wing section. The flow from the 

tubercle peak is accelerated posteriorly 

due to the interaction with the vortex 

pair. These effects prevent the local 

boundary layer downstream of the tubercles 

from separating and push the 

stall line further posterior toward the 

trailing edge. When integrated over 

the entire structure, the wing with tubercles 

will stall at a higher angle of 

attack than a wing without tubercles. 

Flow experiments conducted on a 

model wing section with leading-edge 

tubercles at low speeds showed flow 

separation from the troughs between 

adjacent tubercles but attached flow 

on the tubercles (Johari et al., 2007). 

Flow separation pattern and surface 

pressure was dramatically altered by 

the tubercles. For regions downstream 

of tubercle crest, separation was delayed 

almost to the trailing edge. Although 

this flow pattern did not 

result in improved lift generation, 

drag reduction or delay in the stall 

angle of attack, the post-stall characteristics 

were greatly smoothed (Johari 

et al., 2007; Saadat et al., 2010), 

which could provide important performance 

benefits for systems that routinely 

operate beyond the stall point. 

The vortices produced from the 

tubercles re-energize the boundary 

layer by carrying high-momentum 

flow close to the flipper’s surface 

(Figure 5; Wu et al., 1991; Pedro & 

FIGURE 5 

Kobayashi, 2008; Hansen et al., 

2010). The flow dynamics are improved 

also by confining separation 

to the tip region. Tubercles delay stall 

by causing a greater portion of the flow 

to remain attached on a wing, with the 

attached flow localized behind the tubercle 

crests (Weber et al., 2011). 

Thesizeandfrequencyofthetubercles 

along the leading edge influences 

the performance of a wing. 

Small amplitude tubercles show the 

best performance with regard to lift 

and stall characteristics (Johari et al., 

2007; Hansen et al., 2009, 2011). 

The wavelength and thus frequency 

of tubercles was found, however, to 

have little effect on performance 

(Johari et al., 2007). 

The tubercle effect is further enhanced 

when sweepback is added to 

the wings (Murray et al., 2005). 

Wings with sweep angles of 15° and 

30°requiredhigheranglesofattack 

to achieve stall than nonswept wings 

and showed superior drag performance 

over most of the range of α compared 

to models without tubercles (Murray 

et al., 2005). Flow tests on delta 

Vorticity computed from detached eddy simulation (DES) for flippers with (a) and without (b) tubercles 

at an angle of attack of 15°. Vortices re-energize boundary layer to delay separation and 

stall. Images courtesy of H. T. C. Pedro and M. H. Kobayashi. 


wings with a sweep of 50° showed that 

at high angles of attack large-scale 

three-dimensional separation occurred 

for the wing with a straight 

leading edge (Goruney & Rockwell, 

2009). However when tubercles are 

added, the flow is radically transformed. 

Tubercles with amplitude of 

4% of wing chord can completely 

eradicate the negative effect of the 

separation and foster re-attachment. 

Experiments performed on flapping 

wings with tubercles (Figure 6) 

showed an affect on the spanwise 

flow (Ozen & Rockwell, 2010). Typically 

a straight wing, whether flapping 

FIGURE 6 

or static, will develop a spanwise flow 

due to the pressure differential that develops 

between the upper and lower 

surfaces. This spanwise flow becomes 

manifest as a wing tip vortex, which increases 

the drag and reduces the efficiency 

of a wing. A flapping wing 

with tubercles does not produce a pronounced 

region of spanwise flow, but 

the wing tip vortex generation is unaffected 

(Ozen & Rockwell, 2010). 

Limitations of Tubercles 

Enhanced performance due to the 

presence of the tubercles is not universal. 

There exist limitations to the 

Comparison of flapping plate at Reynolds number of 1300 at angle of attack of 8° with and without 

tubercles from Ozen and Rockwell (2010). The pattern of the flow structure for the plate with tubercles 

produces a series of spanwise vortices that limit spanwise flow compared with the flapping 

plate without tubercles. Image courtesy of D. Rockwell. 

advantages of the tubercles. Tubercles 

improve performance when in concert 

with wing geometries that are characterized 

by the combination of finite 

span, swept wing, tapered planform, 

and thick foil. 

Foil sections with no wing tip that 

emulate infinite wings do not demonstrate 

reduced drag and increased lift, although 

stall is still delayed (Miklosovic 

et al., 2004; Johari et al., 2007; Van 

Nierop et al., 2008). Tip effects occur 

as a consequence of lift generation 

when a fully three-dimensional wing 

is canted at an angle of attack to an incident 

flow. Induced drag is produced 

in lift generation from kinetic energy 

imparted to the fluid from pressure differences 

between the two surfaces of 

the wing as there is leakage of fluid 

from high pressure to low pressure 

around the distal tip of a lifting surface 

resulting in spanwise flow and the formation 

of tip vortices (Vogel, 1981). 

The flow pattern set up by the tubercles 

helps to maintain a chordwise 

flow and reduce the induced drag due 

to tip vortices. This effect is only realized 

for finite wings. 

The tubercles require wings with 

thick sectional geometries to function. 

The section must have a prominent 

nose radius. In part, this is due to the 

necessity to contour the tubercles threedimensionally 

into the leading edge 

to avoid any flat surfaces. Leadingedge 

tubercles were tested on foils 

based on NACA 0021, 63 4 -021 and 

65-021 sections (Miklosovic et al., 

2004; Johari et al., 2007; Custodio 

et al., 2010; Hansen et al., 2009, 

2011). These designs approach the 

cross-sectional geometry of the humpback 

whale flipper (Fish & Battle, 

1995). 

It is necessary to have a relatively 

steady flow to maintain the pattern of 

thevorticesandincurthehydrodynamic 


advantages (Stanway, 2008). Foils 

with tubercles that were oscillated in 

roll and pitch demonstrated that the 

tubercles did not improve hydrodynamic 

performance (Stanway, 2008). 

Flapping degraded performance by 

the redirection of energy to tuberclegenerated 

vortices from the vortices 

of the wake, which are necessary for 

thrust production during flapping. 

Furthermore, if the period of oscillations 

is too rapid, there may be 

insufficient time to allow the full development 

of the vortices over a wing. It is 

necessary to have a relatively steady 

flow to maintain the pattern of the 

vortices and incur the hydrodynamic 

advantages (Stanway, 2008). 

Tubercle Technologies 

Application of natural technologies 

into biomimetic-engineered systems 

has a number of problems that are inherent 

due to differences in biological 

and engineered systems (Fish, 2006). 

Engineered systems are relatively large 

in size, are composed of dry rigid 

materials, including metals and ceramics, 

use rotation motors, and are 

controlled by computational systems 

with limited sensory feedback. Biological 

structures associated with animal 

systems are relatively small in size, are 

composed of wet compliant materials, 

including composites of ceramics, 

polysaccharides and proteins, move 

by translational displacements generated 

by muscles, and are controlled 

by complex neural networks with multiple 

sensory inputs and fine scale 

motor outputs. 

The humpback whale tubercles 

provide an ideal solution for application 

to engineered marine systems. 

The size of the whale and its flippers 

operate near or at the same scale as 

some marine vehicles. The mature 

whales have a maximum length of 

17 m and weigh 40,000 kg (Clapham 

& Mead, 1999). Flipper length is approximately 

one-third the length of 

the animal and can be over 5 m. The 

whale cruises between 1.1 and 4.0 m/s 

and is able to burst to a speed of 7.5 m/s 

(Fish & Rohr, 1999). The flow experienced 

by and modified by the tubercles 

is within the same Reynolds regime 

that coincides with a large array of 

engineered applications. The Reynolds 

number of a flipper is 1.6 × 10 6 when 

the whale is lunge feeding (2.6 m/s). 

Thus, the flipper and tubercles are operating 

in a turbulent flow regime, 

which is the standard operating condition 

for most marine systems. Furthermore, 

the tubercle functions passively 

to modify flow and maintain favorable 

hydrodynamic conditions. Therefore, 

control systems can be simplified. 

Control Surfaces 

There are few other passive means 

of altering fluid flow around a winglike 

structure that can delay stall and 

both increase lift and reduce drag at 

the same time. As a result, the application 

of leading-edge tubercles for passive 

flow control has potential in the 

design of marine technologies. The 

ubiquity of wing-like structures with 

marine applications for stability and 

maneuverability presents an opportunity 

to enlist tubercles to improve performance. 

Included in such structures 

are fixed surfaces, such as keels, fins 

and skegs, and mobile control surfaces, 

such as rudders and dive planes. 

Delay of stall by tubercles on both 

fixed and mobile control surfaces 

provides a benefit in tight turning situations. 

To produce the required centripetal 

force to effect a turn, a lift force 

that is directed toward the center of the 

turn is generated by the control surface. 

The magnitude of the lift is 

directly dependent on the angle of attack 

with a higher angle providing a 

greater centripetal force. As stall can 

be delayed with tubercles, the control 

surface can operate at higher angles of 

attack producing a tighter turn radius 

with more control. If stall were to 

occur, the control surface would not 

be able to generate the centripetal 

force to maintain the turn. In effect, 

itwouldbelikedrivingacaralonga 

curved road and slipping on a patch 

of ice. The reduced friction and centripetal 

force between the ice and the 

tires would cause the car to drive off 

the road tangential to the curve, rather 

than following the original curved 

trajectory. 

A low-aspect ratio rudder with 

tubercles (Figure 7) and an unswept 

leading edge generated more lift at angles 

of attack above 22° compared to a 

smooth rudder at a Reynolds number 

of 200,000 (Weber et al., 2010). At 

higher Reynolds numbers, this effect 

diminishes and the tubercles accelerate 

the onset of cavitation. A humanpowered 

submarine, Umpty Squash, 

utilized tubercled dive planes and rudders 

(Figure 8). Students of the Sussex 

FIGURE 7 

Rudder with leading-edge tubercles. 


FIGURE 8 

Human-powered submarine (left) with rudder and dive planes with tubercles (right). Courtesy of 

Chris Land and the Sussex County Technical School. 

FIGURE 11 

Iceboat with leading-edge tubercles on the 

mast supporting a sail. 

County Technical High School, 

Sparta, NJ, constructed the submarine. 

In 2005, the submarine competed 

in the International Submarine 

RacesheldattheDavidTaylor 

Model Basin in Bethesda, MD. The 

submarine was capable of making a 

90° turn within 25 feet. The designers 

at Feadship De Voogt developed 

Breathe, a concept superyacht that 

incorporates biomimicry into the 

design (www.feadship.nl). The 

stabilizers and steering fins were 

based on the humpback whale flipper 

with tubercles (Figure 9). 

Commercially, the company Fluid 

Earth markets a surfboard skeg with 

leading-edge tubercles (Figure 10; 

Anders, 2009). The addition of tubercles 

would provide enhanced control 

during a cutback, when a surfer rapidly 

changes direction by 180° to maneuver 

the surfboard opposite to the direction 

of the wave’s braking motion. 

Application of tubercles to the mast 

of a sailboat (Figure 11) could be 

useful during close reach maneuvers. 

A close reach would have the sail set 

FIGURE 10 

Surfboard skeg with leading-edge tubercles. 

with a high angle of attack to the 

apparent wind. The lack of a thick 

cross-section by the sail itself may preclude 

any advantage in lift and stall. 

However, the presence of tubercles 

on the mast, representing a bluff 

body, could have advantages in terms 

of drag. Bluff bodies, like cylinders, 

FIGURE 9 

Conceptual yacht incorporating biomimetic 

structures. Stabilizers are shaped like the flippers 

of the humpback whale. Image courtesy 

of Feadship. 


FIGURE 12 

Two views of marine propeller designs with 

tubercles. 

can experience lowered drag when the 

leading edge has a sinusoidal design 

(Bearman & Owen, 1998). 

Propellers and Turbines 

The use of tubercles can effectively 

be employed in the generation of 

power by wing-like structures. Propellers 

operating in a marine system have 

the potential to be improved by the 

addition of tubercles (Figure 12). 

The effective angle of attack of a propeller 

blade can be increased by increasing 

the blade angle (Larrabee, 

1980). A higher angle of attack can 

produce more lift to derive greater 

thrust and increase the effective pitch 

of the propeller. Prevention of stall 

by the tubercle effect would reduce 

flow-induced vibrations. Furthermore, 

suppression of tonal noise is possible 

by the addition of tubercles to a propeller 

(Hansen et al., 2010). Tonal 

noise is most effectively reduced by 

large amplitude and smaller wavelength 

tubercles. As propellers produce 

a particular noise signature, reduction 

of noise would be advantageous for 

stealth in naval operations. Similarly, 

a reduction of noise pollution by commercial 

marine traffic could be beneficial 

to marine organisms, although the 

use of radiated noise by whales to avoid 

collisions with ships may be negatively 

impacted. 

Tubercle modified blades were also 

found to be effective in power generationofamarinetidalturbineatlow 

flow speeds (Murray et al., 2010). Tubercles 

were placed on the distal 40% 

of the three turbine blades. Compared 

to blades with smooth leading edges, 

blades with leading-edge tubercles 

demonstrated enhanced performance. 

The marine tidal turbine is analogous 

to wind turbines. A variable pitch 

wind turbine with retrofitted blades 

with tubercles demonstrated increased 

electrical generation at moderate wind 

speeds compared to unmodified blades 

(Howle, 2009; Wind Energy Institute 

of Canada, 2008). 

Conclusions 

A passive means of altering fluid 

flow around a wing-like structure that 

can delay stall and both increase lift 

and reduce drag simultaneously is 

highly novel. This performance by 

leading-edge tubercles, therefore, has 

potential application for passive flow 

control in the design of various marine 

technologies. The flow is modified by 

the formation of paired vortices in the 

troughs between tubercles. These vortices 

interact with the flow over the tubercles 

to keep the flow attached to the 

wing surface and delay stall. The tubercles 

perform best when designed into 

tapered wings with finite span, swept 

planform, and with a thick foil section 

that have limited oscillatory movement. 

Such applications for marine technology 

include fins, rudders, dive planes, 

water turbines and propellers. The fusion 

of these marine devices and tubercles 

can produce biomimetic designs 

that can exhibit superior performance 

to conventional engineered systems. 


We thank the technical support 

staff of the United States Naval 

Academy. This work was supported 

by the National Science Foundation 

(IOS-0640185) to FEF and the National 

Defense Science and Engineering 

Graduate (NDSEG) Fellowship 

to PWW. 

Lead Author: 

Frank E. Fish 

Department of Biology 


West Chester, PA 19383 

Email: ffish@wcupa.edu 

References 

Abbott, I.H., & von Doenhoff, A.E. 1959. 

Theory of Wing Sections. New York: Dover. 

Anders, M. 2009. Fins I like. Surfer’s Path, 

April/May 2009:116. 

Au, W.W.L. 1993. The Sonar of Dolphins. 

New York: Springer-Verlag. 

Bearman, P.W., & Owen, J.C. 1998. 

Reduction of bluff body drag and suppression 

of vortex shedding by the introduction of 

wavy separation lines. J Fluid Struct. 

12:123-30. doi: 10.1006/jfls.1997.0128. 

Bertin, J.J., & Smith, M.L. 1998. Aerodynamics 

for Engineers. Upper Saddle River, 

NJ: Prentice Hall. 

Bushnell, D.M., & Moore, K.J. 1991. Drag 

reduction in nature. Annu Rev Fluid Mech. 

23:65-79. doi: 10.1146/annurev. 

fl.23.010191.000433. 

Carpenter, P.W., & Pedley, T.J. 2003. 

IUTAM Symposium on Flow Past Highly 

Compliant Boundaries and in Collapsible 

Tubes. Dordrecht: Kluwer. 

Clapham, P.J., & Mead, J.G. 1999. 

Megaptera novaeangliae. Mamm Spec. 

604:1-9. 

Custodio, D., Henoch, C., & Johari, H. 

2010. The performance of finite-span 

hydrofoils with humpback whale-like leading 

edge protuberances. 63 rd Ann Mtg APS Div 


Fluid Dyn, 55 (16), Nov. 21, 2010, Long 

Beach, CA. 

Edel, R.K., & Winn, H.E. 1978. Observations 

on underwater locomotion and flipper 

movement of the humpback whale Megaptera 

novaeangliae. Mar Biol. 48:279-87. 

doi: 10.1007/BF00397155. 

Fish, F.E. 2006. The myth and reality of 

Gray’s paradox: Implication of dolphin drag 

reduction for technology. Bioinspir Biomim. 

1:R17-25. doi: 10.1088/1748-3182/1/2/R01. 

Fish, F.E., & Battle, J.M. 1995. Hydrodynamic 

design of the humpback whale flipper. 

J Morphol. 225:51-60. doi: 10.1002/ 

jmor.1052250105. 

Fish, F.E., Howle, L.E., & Murray, M.M. 

2008. Hydrodynamic flow control in marine 

mammals. Integr Comp Biol. 211:1859-67. 

Fish, F.E., & Lauder, G.V. 2006. Passive 

and active flow control by swimming fishes 

and mammals. Annu Rev Fluid Mech. 

38:193-224. doi: 10.1146/annurev. 

fluid.38.050304.092201. 

Fish, F.E., & Rohr, J. 1999. Review of 

Dolphin Hydrodynamics and Swimming 

Performance. San Diego, CA: SPAWARS 

TR 1801. 


Howle, L.E. 2011. The humpback whale’s 

flipper: Application of bio-inspired tubercle 

technology. Integr Comp Biol. 51:203-13. 

doi: 10.1093/icb/icr016. 

Friedlaender, A.S., Hazen, E.L., Nowacek, 

D.P., Halpin, P.N., Ware, C., Weinrich, 

M.T., … Wiley, D. 2009. Diel changes in 

humpback whale Megaptera novaeangliae 

feeding behavior in response to sand lance 

Ammodytes spp. behavior and distribution. 

Mar Ecol Prog Ser. 395:91-100. doi: 

10.3354/meps08003. 

Gibbs-Smith, C.H. 1962. Sir George 

Cayley’s Aeronautics 1796-1855. London: 

Her Majesty’s Stationery Office. 

Goruney, T., & Rockwell, D. 2009. 

Flow past a delta wing with a sinusoidal 

leading edge: Near-surface topology and flow 

structure. Exp Fluids. 47:321-31. 

doi: 10.1007/s00348-009-0666-x. 

Gray, J. 1936. Studies in animal locomotion 

VI. The propulsive powers of the dolphin. 

J Exp Biol. 13:192-9. 

Hansen, K.L., Kelso, R.M., & Dally, B.B. 

2009. The effect of leading edge tubercle 

geometry on the performance of different 

airfoils. In: 7th World Conf Exp Heat Transfer, 

Fluid Mechanics and Thermodynamics. 

28 June-03 July 2009. Krakow, Poland. 

Hansen, K.L., Kelso, R.M., & Dally, B.B. 

2011. Performance variations of leading-edge 

tubercles for distinct airfoil profiles. AIAA 

J Aircraft. 49:185-94. 

Hansen, K.L., Kelso, R.M., & Doolan, C.J. 

2010. Reduction of flow induced tonal noise 

through leading edge tubercle modifications. 

In: 16th AIAA/CEAS Aeroacoustics Conf. 

7-9 June 2010. Stockholm, Sweden. 

Hazen, E.L., Friedlaender, A.S., Thompson, 

M.A., Ware, C.R., Weinrich, M.T., Halpin, 

P.N., & Wiley, D.N. 2009. Fine-scale 

prey aggregations and foraging ecology of 

humpback whales Megaptera novaeangliae. 

Mar Ecol Prog Ser. 395:75-89. 

doi: 10.3354/meps08108. 

Hoerner, S.F. 1965. Fluid-dynamic Drag. 

Published by author. Brick Town, NJ. 

Howle, L.E. 2009. WhalePower Wenvor 

blade. A report in the efficiency of a WhalePower 

Corp. 5 meter prototype wind turbine blade. 

BelleQuant Engineering, PLLC. 

Iosilevskii, G., & Weihs, D. 2007. Speed 

limits on swimming of fishes and cetaceans. 

J R Soc Interface. 5:329-38. doi: 10.1098/ 

rsif.2007.1073. 

Johari, H., Henoch, C., Custodio, D., & 

Levshin, A. 2007. Effects of leading-edge 

protuberances on airfoil performance. AIAA 

J. 45:2634-42. doi: 10.2514/1.28497. 

Kramer, M.O. 1960. The dolphin’s secret. 

New Sci. 7:1118-20. 

Larrabee, E.E. 1980. The screw propeller. 

Sci Am. 243:134-48. 

Liu, P., & Bose, N. 1993. Propulsive performance 

of three naturally occurring oscillating 

propeller planforms. Ocean Eng. 20:57-75. 

doi: 10.1016/0029-8018(93)90046-K. 

Miklosovic, D.S., Murray, M.M., & 

Howle, L.E. 2007. Experimental evaluation of 

sinusoidal leading edges. J Aircraft. 44:1404-7. 

doi: 10.2514/1.30303. 

Miklosovic, D.S., Murray, M.M., Howle, 

L.E., & Fish, F.E. 2004. Leading-edge 

tubercles delay stall on humpback whale 

(Megaptera novaeangliae) flippers. Phys 

Fluids. 16:L39-42. doi: 10.1063/1.1688341. 

Murray, M.M., Fish, F.E., Howle, L.E., & 

Miklosovic, D.S. 2005. Stall delay by leading 

edge tubercles on humpback whale flipper 

at various sweep angles. In 14 th Int Symp 

Unmanned Untethered Submersible Tech. 

Durham New Hampshire: Autonomous 

Undersea Systems Insitute. 

Murray, M., Gruber, T., & Fredriksson, D. 

2010. Effect of leading edge tubercles on 

marine tidal turbine blades. 63 rd Ann Mtg 

APS Div Fluid Dyn, 55(16), Nov. 21, 2010, 

Long Beach, CA. 

Owen, J.C., Szewczyk, A.A., & Bearman, 

P.W. 2000. Suppression of Karman vortex 

shedding. Phys Fluids. 12:S9. 

Ozen, C.A., & Rockwell, D. 2010. Control 

of vortical structures on a flapping wing 

via a sinusoidal leading-edge. Phys Fluids. 

22:021701. doi: 10.1063/1.3304539. 

Pedro, H.T.C., & Kobayashi, M.H. 

2008. Numerical study of stall delay on 

humpback whale flippers. In: 46th AIAA 

Aerospace Mtg Exhibit. 7-10 Jan. 2008. 

Reno, Nevada. 

Peterson, C.J.G. 1925. The motion of 

whales during swimming. Nature. 116:327-9. 

doi: 10.1038/116327a0. 

Riley, J.J., Gad-el-Hak, M., & Metcalfe, 

R.W. 1988. Compliant coatings. Annu Rev 

Fluid Mech. 20:393-420. doi: 10.1146/ 

annurev.fl.20.010188.002141. 

Saadat, M., Haj-Hariri, H., & Fish, F. 2010. 

Explanation of the effects of leading-edge 


tubercles on the aerodynamics of airfoils 

and finite wings. 63 rd Ann Mtg APS Div 

Fluid Dyn, 55(16), Nov 21, 2010, Long 

Beach, CA. 

Shevell, R.S. 1986. Aerodynamic anomalies: 

Can CFD prevent or correct them J Aircraft. 

23:641-9. doi: 10.2514/3.45356. 

Stanway, M.J. 2008. Hydrodynamic effects 

of leading-edge tubercles on control surfaces 

and in flapping foil propulsion. M.S. thesis, 

Massachusetts Institute of Technology. 

Triantafyllou, G.S., & Triantafyllou, M.S. 


Sci Amer. 272:64-70. doi: 10.1038/ 


Van Nierop, E.A., Alben, S., & Brenner, 

M.P. 2008. How bumps on whale flippers 

delay stall: An aerodynamic model. Phys 

Rev Lett. 100:054502. doi: 10.1103/ 

PhysRevLett.100.054502. 

Vogel, S. 1981. Life in Moving Fluids. 

Boston: Willard Grant Press. 

Weber, P.W., Howle, L.E., & Murray, M.M. 

2010. Lift, drag, and caviation onset on 

rudders with leading-edge tubercles. Marine 

Technol Sname N. 47(1):26. 

Weber, P.W., Howle, L.E., Murray, M.M., 

& Miklosovic, D.S. 2011. Computational 

evaluation of the performance of lifting 

surfaces with leading edge protuberances. 

J Aircraft. 48:591-600. doi: 10.2514/ 

1.C031163. 

Wegener, P.P. 1991. What Makes Airplanes 

Fly New York: Springer-Verlag. 

Wind Energy Institute of Canada. 2008. 

WhalePower Tubercle Blade Power Performance 

Test Report. North Cape, PE: 

Wind Energy Institute of Canada. 

Wu, J.Z., Vakili, A.D., & Wu, J.M. 1991. 

Review of the physics of enhancing 

vortex lift by unsteady excitation. 

Prog Aerosp Sci. 28:73-131. doi: 10.1016/ 

0376-0421(91)90001-K. 


PAPER 

Shark Skin Separation Control Mechanisms 

AUTHORS 

Amy Lang 

University of Alabama 

Philip Motta 

Maria Laura Habegger 

University of South Florida 

Robert Hueter 

Mote Marine Laboratory 

Farhana Afroz 


Introduction 

Natural selection, operating for 

hundreds of millions of years, has 

honed fast swimming marine organism 

forms to reduce energy expenditure 

through streamlining. Among 

the fastest swimming fishes, certain 

species of sharks have converged on 

a suite of adaptations to reduce drag. 

Consequently, fast swimming sharks 

have been studied for insight into potential 

drag-reducing mechanisms for 

well over three decades. Drag on a 

submerged, swimming body consists 

of three sources. These include (i) form 

drag due to a difference in pressure 

around the body, (ii) drag due to lift, 

and (iii) skin friction due to boundary 

layer formation (Bushnell & Moore, 

1991). At low Reynolds numbers 

(Re) skin friction predominates, while 

at higher Re pressure drag can dominate 

if not minimized. It is not surprising 

then to consider the fact that 

aquatic organisms have evolved to 

minimize drag (Fish, 1998), with 

the primary decrease coming from 

a streamlined body shape to reduce 

flow separation and thus form 

drag. Aquatic organisms that swim at 

ABSTRACT 

Drag reduction by marine organisms has undergone millions of years of natural 

selection, and from these organisms biomimetic studies can derive new technologies. 

The shortfin mako (Isurus oxyrinchus), considered to be one of the fastest and 

most agile marine predators, is known to have highly flexible scales on certain locations 

of its body. This scale flexibility is theorized to provide a passive, flow-actuated 

mechanism for controlling flow separation and thereby decreasing drag. Recent biological 

observations have found that the shortfin mako has highly flexible scales, 

bristling to angles in excess of 50°, particularly on the sides of the body downstream 

of the gills. High “contragility,” which is explicitly definedhereastheabilityto 

change or move in a new or opposing direction while already in a turn, would 

occur if form drag were minimized. This would thus indicate the potential control 

of flow separation on body regions aft of the point of maximum girth or in regions of 

adverse pressure gradient. Thus results are consistent with the hypothesis that 

scale bristling controls flow separation. This scale flexibility appears to be a result 

of a reduction in the relative size of the base of the scales as well as a reorganization 

of the base shape as evidenced by histological examination of the 

skin and scales. Probable mechanisms leading to separation control are discussed. 

Keywords: shark skin, flow separation, drag reduction 

high Re (>10 3 )haveavarietyof 

shapes and structures to reduce drag, 

which we often attempt to duplicate 

(Vogel, 2003; Fish & Lauder, 2006). 

It has been deduced that a crescent tail 

design could decrease induced drag on 

the order of 8%, and not surprisingly 

this lunate tail design is found on 

many fast swimming marine animals 

(Fish, 1998; Donley et al., 2004). 

Several researchers (e.g., Anderson 

et al., 2001) have observed that swimming 

fish experience more friction 

drag than the same rigid body 

towed. This higher friction is attributed 

to motion of the body as it 

swims to produce thrust. The undulating 

body motion can result in measurably 

thinner boundary layers and 

thus higher skin friction (Fish, 2006). 

The argument has been made that 

because of this higher power output 

requirement, to overcome drag and 

maintain a certain speed, swimming 

drag reduction due to various morphological 

mechanisms is extremely 

probable (Schultz & Webb, 2002). 

Separation of the boundary layer 

from a body typically occurs in vicinities 

where the flow is decelerating 

along a curved body after the point of 

maximum thickness, resulting in an 

adverse pressure gradient. As a result 

separation typically occurs in areas 

posterior of the maximum body thickness. 

Incipient separation is characterized 

by regions of decreasing skin 

friction approaching zero, and consequent 

reversal of the flow at the surface 

(Doligalskietal.,1994).Swimming 

kinematics in thunniform fish such as 

tunas and mako sharks are characterized 

by cyclically repeating motions and 

small linear and angular accelerations 

(Blake, 2004). Most fast swimming 

sharks, such as the shortfin mako Isurus 


oxyrinchus, are thunniform swimmers 

where oscillations are for the most 

part limited to the posterior end of 

the body. It has been reported that incipient 

separation (inflected boundary 

layer profile) is often observed during 

swimming movements, and this motion 

may be tuned by the fish to take 

advantage of the lower shear stress in 

a nearly separating boundary layer; 

yet separation must be avoided to reduce 

form drag and increase thrust 

produced by the caudal fin (Anderson 

et al., 2001). 

The skin of sharks is covered by 

minute scales, called dermal denticles 

or placoid scales, which originally 

evolved as a hard, protective covering 

to the animal (Raschi & Tabit, 

1992). The bases of these hard scales 

are embedded in the superficial collagenous 

layer of the skin (dermis) termed 

the stratum laxum, with the crowns 

of the scales exposed to the water. It is 

the unique geometry observed on fast 

swimming sharks for these tooth-like 

scales that is of interest from a drag reduction 

standpoint. Beginning in the 

1970s researchers became interested 

in the small, streamwise ridges, 

or keels, located on the top of each 

crown (see scanning electron microscopy 

(SEM) pictures of shark scales 

shown in Figure 1). Now labeled as 

riblets, these ridges were found to result 

in a reduction in turbulent skin 

friction drag of up to 9.9% when 

sized correctly (Bechert et al., 1997). 

Even as early as the 1980s, riblets 

were utilized on boat hulls competing 

in the Olympics and America’s Cup 

but later were banned (Gad-el Hak, 

2000). 

However, the scales on some fast 

swimming sharks exhibit a different 

property that is flexibility or capability 

to bristle, and this is the focus of our 

current work. Previous work (Bechert 

FIGURE 1 

The average scale erection angles for 16 regions on the shortfin mako shark (Isurus oxyrinchus) 

together with scanning electron micrographs (top row) and histological sections (middle row) of 

the skin for three regions on the body. The flank region, including B2, has the most flexible scales, 

which are characterized by long backwardly projecting crowns and relatively short bases. The scale 

bases are embedded in the superficial part of the dermis. The surface of the scales has three riblets 

visible on the scanning electron micrographs. The scales in any region are oriented along the longitudinal 

axis of the shark, and anterior is to the left. The erection angles noted on the figure are the 

angles to which the scales in that region can be manually manipulated without damage to their 

attachment and which remain at that angle after release by the needle used to erect them. 

et al., 2000) investigating this aspect of 

the shark skin found no advantage 

from a skin friction reduction standpoint, 

and only greatly increased friction 

drag if the scales were allowed to 

remain bristled. Thus, a new, passive 

flow-actuated separation control 

mechanism is proposed that is inspired 

by the scale flexibility found on the 

shortfin mako shark. 

This investigation chose to focus 

on the skin of the shortfin mako 

based on several factors. First, of fast 

swimming pelagic sharks, the shortfin 

mako is considered by most to be the 

fastest and most agile (Stevens, 2009). 

It is also one of the more derived 

species of shark and is recognized as 

making its appearance roughly 55 million 

years ago in the line of shark 

evolution dating back more than 

400 million years (Naylor et al., 

1997). Next, it is one of two species 

of shark previously reported in literature, 

the other being the smooth hammerhead 

Sphyrna zygaena, ashaving 


flexible scales over large portions of its 

body (Bruse et al., 1993). Finally, the 

shortfin mako is readily obtainable off 

the Atlantic coast of the United States 

and is not currently listed as overfished 

or otherwise in a vulnerable state 

for conservation purposes (NMFS, 

2010). 

Materials and Methods 

To investigate scale structure and 

erection, we acquired two subadult 

shortfin makos (female: total length, 

192 cm; fork length, 171.5 cm; male: 

total length, 158 cm; fork length, 

150 cm) from commercial and recreational 

fishers in the coastal waters off 

Montauk, New York. The frozen 

specimens were shipped to the University 

of South Florida in Tampa. There, 

following Reif (1985), scales at 16 regions 

along the body were marked in 

order to sample the scales under a 

variety of flow regimes (Figure 1). 

Three 1 cm 2 samples from each location 

were removed, two for histological 

analysis and one for SEM. Because 

swimming sharks have superambient 

subcutaneous pressure ranging as 

high as 100-200 kPa increasing skin 

stiffness (Wainwright et al., 1978; 

Martinez et al., 2002), scale erection 

angles were recorded with and without 

subcutaneous pressure, as follows. 

We measured scales erection angle 

under a dissecting microscope at a 

magnification of 135×. In seven of 

the 16 regions (B1, B2, B4, B5, A1, 

A2, A3; Figure 1), an aneroid sphygmomanometer 

was placed under the 

skin and underlying muscle and the 

pressure elevated and held at 15 psi 

(103 kPa), the maximum possible 

without damaging the underlying 

muscle tissue. Scale erection was not 

noted with the increase in subcutaneous 

pressure, leading us to suspect a 

passive mechanism. While the skin 

was pressurized, we used a fine acupuncture 

needle to gently manipulate 

five haphazardly selected scales to 

their maximum erected position without 

tearing them from the skin. Releasing 

the scales, we allowed them to 

settle at an erected angle, which we 

then measured by calculating the 

change in length of the scale crown 

when viewed from above. The inverse 

cosine of the apparent crown length divided 

by the resting or true crown 

length provided the angle of erection. 

Because the individual scales could actually 

be easily erected past this resting 

angle, we were in essence calculating a 

minimal erection angle. The pressure 

was then released and the angle similarly 

calculated on the flaccid skin. 

Our aprioritest determined that 

stretching the skin in this manner did 

not affect the non-pressurized erection 

angle. 

At the other regions (H2, B3, B6, 

P1, P2, P3, C1, C2, C3; Figure 1) 

where subcutaneous pressure was not 

possible, such as in the fins and over 

the body cavity, the erection angles 

were measured by gently erecting the 

scales (if possible) with the acupuncture 

needle and measuring their 

pre- and post-erection crown length. 

Finally, to get a more global picture 

of scale flexibility over the entire 

body, 35 equidistant sampling locations 

encompassing the entire dorsal, 

left lateral, and ventral surfaces of 

each shark were marked and scales in 

each area manually erected as before 

without subcutaneous pressure. The 

erected scales were simply recorded as 

greater or less than 50°, an angle that 

appeared to the approximate maximum. 

We also measured the crown 

and base length of scales in select 

areas, as well as the spacing of the riblets 

on the scales. 

To understand the attachment of 

the scales to the skin, we prepared histological 

sections of the skin and scales, 

decalcified the scales, stained the samples 

to reveal the fibrous attachment, 

and examined the sections at 20× and 

40× with a compound microscope. Finally, 

surface pictures of the scales, at 

all studied regions, were prepared by 

examining the skin at 100× and 200× 

under a SEM. 

Findings 

The placoid scales of sharks have 

a pulp cavity and a hard enameloid 

covering over dentine and are anchoredatthebaseofthescaletothe 

stratum laxum collagenous layer of 

the dermis (Figures 1 and 2). The exposed 

crowns overlap each other on 

the shark’s surface, and the majority 

of scales have small riblets or keels on 

their surface, oriented in the streamwise 

direction of the flow (Figures 1 

and 2). On the fast swimming shortfin 

mako, the flank scales (e.g., areas B2, 

B5, and A2) have a crown length of approximately 

0.18 mm, and each crown 

typically has three keels each having a 

height of 0.012 mm and a spacing of 

0.041 mm. This differs from other 

slower swimming sharks such as the 

blacktip shark (Carcharhinus limbatus), 

whereby preliminary data indicate the 

flank scales are typically 0.32 mm 

in length with each crown typically 

having five keels with a height of 

0.029 mm and a spacing of 0.065 mm. 

When considering shark species as a 

whole, length of the scales is typically 

fixed for specific regionsofthebody 

within a species but differs among regions 

and species. Similarly, the number 

of keels per scale is also consistent 

per body location for a species. 

Scale flexibility on the shortfin 

mako varies considerably across the 


FIGURE 2 

Scanning electron micrographs of the scales and histological sections of the embedded scales 

from three regions of the pectoral fin, P1, P2, and P3 of the shortfin mako (Isurus oxyrinchus). 

The scales on the leading edge of the fin cannot erect and lack riblets, whereas those on the trailing 

edge can erect to greater angles than either the scales on the leading edge or the midregion of the 

fin. The bases of the scales (B) are anchored in the dermis (D), and the thin epidermis (E) is visible 

between the scales. The pulp cavity (PC) is visible in some of the scales, and not all scales are 

sectioned through their center, resulting in some crown (CR) lengths appearing shorter than 

others. Ceratotrichia or fin rays (CE) are visible in region P3 because this part of the fin is very 

thin. Anterior is to the left. 

body and fins, with average erection 

angles varying from 0° on the leading 

edge of the pectoral fin to approximately 

50° on the widest part of the 

body just behind the gill region (Figure 

3). However, only certain portions 

ofthebodyhaveveryflexible scales. 

The most flexible scales are found on 

the flank of the body extending behind 

the gills to the tail; here scales 

are found to be easily erected with slight 

manipulation on dead specimens to angles 

of approximately 50° or greater 

(Figure 3). The lateral scales (B2, B5, 

A2) had significantly greater erection 

angles (mean angle = 44° ± 1°) than 

both the dorsal (mean angle = 26° ± 

1° SE) and ventral regions (mean 

angle = 25° ± 2° SE). Erection angles 

for the dorsal region did not differ 

from that of the ventral region 

(Kruskal-Wallis one-way ANOVA, 

Tukey’s pairwise test; H = 53.173, 

FIGURE 3 

Outline of a representative shortfinmako,Isurus 

oxyrinchus, showing the approximate region 

(lines) on the flank with the most flexible scales 

capable of erection to at least 50°. This region 

approximates the region of flow separation. 

df =2,P ≤ 0.001). Highly flexible 

scales are also found at the trailing 

edge of the pectoral fins. The scales 

on the trailing edge of the pectoral 

(P3) had the highest erection angles 

compared to the leading edge (P1) 

and the central region of the fin (P2), 

and all the regions were significantly 

different from each other (Kruskal- 

Wallis one-way ANOVA, Tukey’s 

pairwise test; H = 26.289, df =2, 

P ≤ 0.001). Conversely, for the caudal 

fin the mean angles were not significantly 

different among the three regions 

(ANOVA; F = 0.0614, df =2, 

P = 0.941). 

At least two factors appear to control 

scale flexibility on the body. The 

first is a reduction in the length of 

the base relative to the length of the 

crown over certain regions of the 

body such as the flank in the shortfin 

mako (Table 1). The flank scales 

have a greater ratio of crown length 

to base length compared to the dorsal 

scales (Kruskal-Wallis one-way 

ANOVA, Tukey’s pairwisetest;H = 

18.104, df =2,P < 0.001) due to a significantly 

shorter base on the flank 

scales (ANOVA, Holm-Sidak method; 

F = 45.967, df =2,P =0.001).The 

flank scales have a relatively wide base 

compared to its length, whereas the 

base of the dorsal scales is more uniform 

in shape (Table 1, BL/BW ratio; 

Figure 4). The ventral scales of the 

shortfin mako are smaller than the 

dorsal and flank scales as they are 

shorter in crown (ANOVA, Holm- 

Sidak method; F =45.967,df =2, 

P < 0.001) and base length (ANOVA, 

Holm-Sidak method; F = 42.916, 

df =2,P < 0.001) overall. Secondly, relative 

changes in the length of the leading 

and trailing edges of the scale base 

affect its anchoring in the dermis, similar 

to the root system of a tree. More 

firmly anchored scales have a more 


TABLE 1 

Means and standard errors for scale crown length (CL), base length (BL), base width (BW), ratio of CL/BL, and ratio BL/BW for three body areas of 

shortfin mako (Isurus oxyrinchus). 

Body Position CL (mm) BL (mm) BW (mm) CL/BL BL/BW 

B1, B4, A1 Dorsum 0.173 ± 0.004 0.145 ± 0.005 0.161 ± 0.005 1.201 ± 0.029 0.938 

B2, B5, A2 Flank 0.179 ± 0.004 0.104 ± 0.003 0.161 ± 0.005 1.745 ± 0.076 0.625 

B3, B6, A3 Ventrum 0.128 ± 0.004 0.091 ± 0.005 0.125 ± 0.003 1.472 ± 0.102 0.692 

evenly distributed base like a tree with 

a broad but evenly distributed shallow 

root system (Figure 4B). The more 

erectable scales on the flank have a narrow 

base on the trailing edge (Figure 4A). 

In this manner, the scales can pivot up, 

analogous to a tree with a narrow root 

system on one side that is blown over 

away from this side. The scales return 

to their resting position with the help 

of elastic fibers that anchor its base 

(stained black in the histology sections). 

A sideview picture of the skin with the 

scales in the foreground manually 

bristled is shown in Figure 5. 

Effects on Fluid 

Flow Patterns 

The findings of scale flexibility and 

angle of erection on the shortfin mako 

FIGURE 4 

Representative scales of the shortfin mako 

(Isurus oxyrinchus) from(A)theflexible 

flankareaB5,and(B)thelessflexible area 

B4. The scales in (A) have a relatively short 

but wide base compared to those of (B) with 

a more uniformly shaped base. 

have led to a working hypothesis currently 

being investigated through 

hydrodynamic testing. Our discovery 

of highly flexible scales on the flank 

and trailing edges of the pectoral fin 

fits with the hypothesis that the scales 

act as a means of controlling flow separation. 

First, these results indicate that 

the erection of the scales is most likely 

initiated by a passive, flow-actuated 

mechanism, i.e., in a region of turbulent 

flow separation the flow consists of 

moments when the flow close to the 

wall is both in the main direction as 

well as reversed. Pressurizing the skin 

had no effect on scale erection. Flow 

reversal over a large region indicates 

FIGURE 5 

Coronal section through the shortfin mako 

skin showing the scales in the foreground 

that have been manually erected from location 

B2. Not all scales are erected to the same 

degree because of the individual manual erection. 

Flow would normally pass over the skin 

from left to right and reversed flow, as occurs 

during separation, is believed to cause bristling 

as shown. 

that global separation from the surface 

(body) occurs, resulting in increased 

pressure drag. In the case of the shark 

this would not only increase drag but 

also inhibit contragility or the ability 

to change direction quickly and to a 

large degree. As the shark swims and 

turns, the body is bent laterally with 

regions of greatest curvature occurring 

on the flank. From the nose to 

the point of maximum girth (around 

the location of the gills), the flow will 

experience a favorable pressure gradient, 

and unfavorable pressure gradient 

regions will be located downstream of 

this point. Thus, the findings that the 

most flexible scales are found on the 

flank of the body and downstream of 

the gills corroborate the hypothesis 

that these scales are working to control 

flow separation. Likewise, the delay of 

flow separation over the pectoral fin 

can lead to increased performance in 

their ability to act as lifting surfaces 

during high-speed swimming maneuvers. 

The generation of high lift 

by the pectoral fins is important for 

quick upward maneuvers while attacking 

prey, as has been observed in video 

evidence of a shortfin mako in pursuit 

of a towed baitfish. This same video 

evidence shows the shark’s ability to 

turn in one direction and then change 

direction before the body completes 

the initial turn. This type of turning 

behavior, also defined as contragility, 

requires not only large muscular effort 

but also low form drag (Frank Fish, 


personal communication, February 

18, 2008). 

Flow visualization images (Figure 

6) of particles illuminated with a 

laser sheet show characteristic flow scenarios 

found in a turbulent boundary 

layer undergoing separation. In this 

case, separation is induced on a flat 

surface via the presence of a rotating 

cylinder located above the wall on 

which a turbulent boundary layer is 

formed. The free stream flow moves 

FIGURE 6 

Flow visualization generated by particle streaking 

in a plane parallel to the free stream flow 

(16.5 cm/s) illuminated by a laser sheet. A rotating 

(40 RPM) cylinder (5.1 cm diameter) is 

located above the image with its center, 6.4 cm 

above the wall, to induce boundary layer separation. 

Boundary layer thickness prior to the 

test area was approximately 1 cm and an area 

of approximately 3 cm × 1.5 cm is imaged. 

Main flow moves left to right. Distinguishing 

moments in time for this unsteady flow are 

shown when (a) the flow is attached, (b) the 

separation process is initiated with reversed 

flow close to wall, and (c) separated flow characterizedbyvortexburstingawayfromthe 

wall is observed, with a large region of reversed 

flow near the wall. 

at 17 cm/s, giving a local Reynolds 

number, based on distance from the 

leading edge of 0.3 m, in the boundary 

layer of 5 × 10 4 . In comparison, a 

shark’s boundary layer will have a Re of 

∼10 6 -10 7 when swimming at 10 m/s. 

However, the characteristics of the turbulent 

flow will be similar; the flow on 

a shark will be faster and the boundary 

layer thinner resulting in shorter time 

and length scales than the water tunnel 

experiments. Under these conditions, 

there are moments when the flow is 

for the most part attached (Figure 6a), 

developing into a large region of separation 

with reversed flow near the wall 

(Figure 6b), and finally times when 

large vortex bursting occurs as the separated 

region becomes unstable resulting 

in a shedding of vortices (Figure 6c). 

A time trace of the velocity measured at 

a location adjacent to the wall where 

the flow is reversed about 50% of the 

timeisshowninFigure7.Here,the 

cyclic nature of a separating turbulent 

boundary is clearly evident. Regions in 

FIGURE 7 

theplotwherethevelocityisnegativeindicate 

moments of reversed flow, 

which would actuate scale bristling 

on the shark. It is transitions in the 

time trace of the velocity where the 

flow moves (Figure 7) from positive 

(u > 0) to negative (u < 0) (or from 

point a to point b labeled in Figure 7) 

when scale actuation would be initiated 

and potentially disrupt the evolving 

flow that leads to the formation 

of a separation bubble as shown in 

Figures 6b and 6c. 

Cassel et al. (1996) provide a description 

of the process leading to 

flow separation. Because of the suction 

pressure upstream (adverse pressure 

gradient) the region closest to 

the wall, where the flow has the lowest 

momentum, is where flow reversal 

is first initiated. This patch of fluid 

moves upstream and thickens, ultimately 

leading to large scale flow separation 

from the surface. It is our 

hypothesis that on the shark’s body 

in the region close to the wall where 

Velocity (u) measured as a function of time at a point close to the wall corresponding to approximately 

two-thirds the distance downstream in Figure 6. Measurements were made using timeresolved 

digital particle image velocimetry (TR-DPIV), which sampled the flow at 1 kHz. At this 

location for about 50% of the time, flow reversal is occurring, as can be seen by the regions where 

u < 0. This is a point in the vicinity where flow reversal is initiated in the turbulent boundary layer 

and develops into a large-scale region of reversed flow due to the presence of the adverse pressure 

gradient induced by the rotating cylinder above. Points labeled (a), (b), and (c) are 

moments in the flow typified correspondingly to those shown in Figure 6. 


flow reversal begins to occur, the scales 

are actuated by the flow to erect, thereby 

disrupting the unsteady flow separation 

process. Future experiments 

will investigate this hypothesis through 

hydrodynamic testing using models 

and real specimens of shark skin. 

This aspect of shark skin resulting in 

asurfacewithapreferredflow direction 

is likely key to its ability to control 

flow separation. 

Previousexperimentsoverabristled 

shark skin model confirmed the 

presence of embedded vortices forming 

between replicas of the scales 

(Lang et al., 2008). Thus, if flow is 

induced to form between the scales 

when bristled, there are two additional 

mechanisms that may aid to control 

the flow. The formation of embedded 

vortices, similar as occurs with dimples 

on a golf ball, would allow the flow to 

pass over the skin with a resulting partial 

slip condition, thereby leading to 

higher momentum adjacent to the surface. 

Secondly, with a turbulent flow 

forming in the boundary layer above 

the cavities, there may be additional 

momentum exchange whereby high 

momentum fluid typically located 

away from the surface is induced at a 

greater rate to move towards the surface 

and into the cavities. This latter 

mechanism, resulting in turbulence 

augmentation (Gad el-Hak, 2000), is 

another potential means to increase 

the momentum overall in the flow adjacent 

to the wall. These three mechanisms 

may be working in conjunction 

to inhibit flow separation over the 

surface of the shark. 

This method has advantages, above 

and beyond other methods currently 

in use to control flow separation, in 

that it is passive with no energy input 

required. Also it causes no additional 

drag penalty when not in use in that 

scale bristling would be controlled by 

on-demand erection of the scales induced 

by regions of flow reversal as 

occurs under conditions of incipient 

flow separation. This obviates the use 

of bristled scales to act as vortex generators 

as a means of separation control, 

as previously theorized by Bechert 

et al. (2000). Vortex generators, which 

consist of small, typically V-shaped 

protrusions, require careful placement 

and protrusion into the boundary layer 

flow upstream of the point of flow separation 

and work by mixing higher 

momentum flow down towards the 

surface (Lin, 2002). Vortex generators 

also result in a drag penalty due to their 

protrusion into the flow (Gad-el-Hak, 

2000). Our findings suggest that the 

scales are bristled passively and are activated 

in a region of flow reversal that 

occurs downstream of the point of separation 

and is thus a different methodology 

from that of vortex generators 

currently in use today. Finally, this 

new passive, flow-actuated mechanism 

may in fact go to the initial root cause 

of the separation, that of flow reversal 

adjacent to the wall, and disrupt it 

prior to growth into a fully separated 

region. Our ultimate aim is to gain a 

fundamental understanding of how 

the flow-actuated bristling of shark 

skin scales can control flow separation 

so that bio-inspired surfaces can be 

engineered for greater flow control in 

marine and other applications. 

Summary 

Fast swimming shortfin mako 

sharks Isurus oxyrinchus have highscale 

flexibility on the flank and 

trailing edges of the pectoral fin, with 

bristling angles up to a range of approximately 

50° on the flank. These regions 

correspond to those on the body 

where flow separation control is likely 

to be most beneficial. In the case of the 

flank region, high curvature of the 

body will result from the shark’s lateral 

swimming motion, and this is also 

in the vicinity of the point of maximum 

girth; both conditions indicate 

the presence of an adverse pressure 

gradient. In the case of the pectoral 

fin, it is hypothesized that the flexible 

scales can lead to the control of flow 

separation which indicates high drag 

and loss of lift. Control of the flow in 

both regions will lead to increased swimming 

speeds with high contragility for 

the shark. Increased scale flexibility in 

this shark appears to be due to a reduction 

in the relative size of the scale base 

compared to the crown and changes 

in the shape of the base where it is anchored 

into the dermis. Future work 

will lead to the manufacturing of bioinspired 

surfaces based on shark skin 

microgeometry whereby a passive, 

flow-actuated surface patterning can 

be used for applications where flow 

separation control is required. 


Funding for this work received 

through collaborative NSF grants 

(0932352, 0744670, and 0931787) 

to A. Lang, P. Motta, and R. Hueter 

to support both the engineering and 

biological work is gratefully acknowledged. 

We also thank Jessica Davis 

for assisting in the shark measurements 

and Candy Miranda for preparing the 

histological samples. Finally, we wish 

to express our gratitude to Paul and 

Jane Majeski and crew, Captain Mark 

Sampson, Captain Al VanWormer, 

Philip Pegley, Lisa Natanson, Jack 

Morris, and Mote Marine Laboratory 

for assistance in obtaining shark 

specimens. 


Lead Author: 

Amy Lang 

Department of Aerospace 

Engineering and Mechanics 


Box 870280 

Tuscaloosa, AL 35487 

Email: alang@eng.ua.edu 

References 

Anderson, E., McGillis, W., & Grosenbaugh, 

M. 2001. The boundary layer of swimming 

fish. J Exp Biol. 204:81-102. 

Bechert, D.W., Bruse, M., Hage, W., & 

Meyer, R. 2000. Fluid mechanics of biological 

surfaces and their technological application. 

Naturwissenschaften. 80:157-71. 

Bechert, D.W., Bruse, M., Hage, W., 

Van der Hoeven, J., & Hoppe, G. 1997. 

Experiments on drag-reducing surfaces and 

their optimization with an adjustable 

geometry. J Fluid Mech. 338:59-87. 

Blake, R. 2004. Fish functional design and 

swimming performance. J Fish Biol. 65:1193-222. 

Bruse, M., Bechert, D., van der Hoeven, J., 

Hage, W., & Hoppe, G. 1993. Experiments 

with conventional and with novel adjustable 

drag-reducing surfaces. In: Proc. of the Int. 

Cong. On Near-Wall Turbulent Flows. 

719-738. Tempe, AZ: Elsevier Science 

Publishers. 

Bushnell, D., & Moore, K. 1991. Drag 

reduction in nature. Annu Rev Fluid Mech. 

23:65-79. 

Cassel, K., Smith, F., & Walker, J. 1996. The 

onset of instability in unsteady boundary-layer 

separation. J Fluid Mech. 315:223-56. 

Doligalski, T., Smith, C., & Walker, J. 1994. 

Vortex interactions with walls. Annu Rev 

Fluid Mech. 26:573-616. 

Donley, J.M., Sepulveda, C.A., Konstantinidis, 

P., Gemballa, S., & Shadwick, R.E. 2004. 

Convergent evolution in mechanical design of 

lamnid sharks and tunas. Nature. 429:61-65. 

Fish, F.E. 1998. Imaginative solutions by 

marine organisms for drag reduction. In: 

Proceedings of the International Symposium 

on Seawater Drag Reduction, ed. Meng, J.C.S., 

443-50. Newport, Rhode Island. 

Fish, F. 2006. The myth and reality of 

Gray’s paradox: Implication of dolphin drag 

reduction for technology. Bioinspir Biomim. 

1:17-25. 

Fish, F., & Lauder, G. 2006. Passive and 

active flow control by swimming fishes and 

mammals. Annu Rev Fluid Mech. 38:193-224. 

Gad-el-Hak, M. 2000. Flow Control: Passive, 

Active and Reactive Flow Management. 

Cambridge, UK: Cambridge University 

Press. 421 pp. 

Lang, A., Hidalgo, P., Motta, P., & Westcott, 

M. 2008. Bristled shark skin: A microgeometry 

for boundary layer control 

Bioinspir Biomim. 3:046005. 

Lin, J. 2002. Review of research on lowprofile 

vortex generators to control boundarylayer 

separation. Prog Aerosp Sci. 38:389-420. 

Martinez, G., Drucker, E., & Summers, A. 

2002. Under pressure to swim fast. Integr 

Comp Biol. 42(6):1273-4. 

Naylor, G., Martin, A., Mattison, E., & Brown, 

W. 1997. The inter-relationships of lamniform 

sharks: Testing phylogenetic hypotheses with 

sequence data. In: Molecular Systematics of 

Fishes, eds. Kocher, T.D., & Stepien, C., 

199-217. New York: Academic Press. 

NMFS. 2010. Final Amendment 3 to the 

Consolidated Atlantic Highly Migratory 

Species Fishery Management Plan. Silver 

Spring, MD: National Oceanic and Atmospheric 

Administration, National Marine 

Fisheries Service, Office of Sustainable Fisheries, 

Highly Migratory Species Management 

Division. 632 pp. Public Document. 

Raschi, W., & Tabit, C. 1992. Functional 

aspects of placoid scales: A review and update. 

Aust J Mar Fresh Res. 43:123-147. 

Reif, W. 1985. Squamation and ecology of 

sharks, no. 78, Courier Forschungs-Institut 

Senckenberg, Frankfurt am Main. 

Schultz, W., & Webb, P. 2002. Power 

requirements of swimming: Do new methods 

resolve old questions Integr Comp Biol. 

42:1018-25. 

Stevens, J. 2009. The biology and ecology of 

the shortfin mako shark, Isurus oxyrinchus. 

In: Sharks of the Open Ocean: Biology, 

Fisheries and Conservation, eds. Camhi, 

M.D., Pikitch, E.K., Babcock, E.A., 87-94. 

Oxford, UK: Blackwell Publishing Ltd. 

Vogel, S. 2003. Comparative Biomechanics: 

Life’s Physical World. Princeton: Princeton 

University Press. 

Wainwright, S., Vosburgh, F., & Hebrank, J. 

1978. Shark skin: A function in locomotion. 

Science. 202:747-9. 


PAPER 

Can Biomimicry and Bioinspiration Provide 

Solutions for Fouling Control 

AUTHORS 

Emily Ralston 

Geoffrey Swain 


Introduction 

Antifouling is changing. The ban 

on tributyltin (TBT), arguably the 

most effective and yet environmentally 

damaging antifouling precipitated an 

increase in research to develop new 

technology (Swain, 1999; Omae, 

2003). The immediate response was 

to return to copper as the antifouling 

of choice. However, high levels of copper 

in many ports and harbors has led 

to concern and even the banning of 

copper based antifouling (i.e., San 

Diego, Washington State; Carson 

et al., 2009; Nehring, 2001; Qian 

et al., 2010; Thomas et al., 2001). 

The use of toxic coatings may also be 

linked to the transport of biocide tolerant 

non-indigenous species (Dafforn 

et al., 2008). Some users have switched 

to biocide-free silicone antifouling 

coatings for improved hydrodynamic 

and environmental performance. However, 

these coatings may foul, which 

necessitates cleaning, and there may 

be an increased risk of transporting 

non-native species. They are mechanically 

weak and can be damaged more 

easily than traditional coatings 

(Chambers et al., 2006; Nehring, 2001; 

Swain, 2010; Swain et al., 2007; Yebra 

et al., 2004). An ideal coating will be 

hydraulically smooth and control fouling 

for the lifetime of the vessel, be 

ABSTRACT 

Biomimicry, modeling biological systems to find engineering methods, and bioinspiration, 

improving upon or repurposing the biological model, may provide direction 

for the development of new antifouling solutions. Despite being subject to constant 

pressure from foulers, many organisms maintain a clean surface. The challenge lies 

in selecting the most effective and reproducible antifouling mechanisms from 

nature and mimicking or modifying them to provide a realistic engineered solution. 

Keywords: natural antifouling, marine coatings, biomimicry, bioinspiration, antifouling, 

foul release 

environmentally compliant, control 

invasive species, be easily applied, repaired 

and maintained, be compatible 

with materials and methods of hull 

construction and decommissioning, 

and be cost effective (Swain, 2010). 

In the decades since a ban on TBT 

was proposed, research has been directed 

towards new solutions to the 

fouling problem. Some of the discoveries 

have been in the form of new 

booster biocides to improve the performance 

of copper based antifouling 

against biofilms (slimes). These include 

Irgarol 1051, Seanine 211, diuron, 

pyrithiones, and many others. 

Unfortunately, some of these boosters 

may parallel TBT in terms of environmental 

impact. For example, Irgarol 

1051 is a photosynthesis inhibitor 

that reduces plants’ ability to create 

energy causing growth to slow, reproduction 

to stop and eventually death of 

the plant. It has a long half-life and 

does not partition into sediment so it 

is found primarily in the water column. 

It has been found to occur in estuaries, 

ports and harbors at elevated 

levels and is highly toxic to non-target 

marine plants (Hall et al., 1994; 

Thomas & Brooks, 2010; Thomas 

et al., 2001; Yebra et al., 2004). Recently, 

it has been detected in water 

samples collected from the Caribbean, 

Bermuda, Florida and Australia in the 

vicinity of seagrass beds and coral reefs 

where the effects could be devastating 

(Carbery et al., 2006; Owen et al., 

2002; Scarlett et al., 1999). Researchers 

must be careful to avoid any chemistry 

that may impact the environment 

and non-target organisms. One source 

of new technology is to understand 

how organisms prevent fouling and 

then mimic or draw inspiration from 

these biological models to create an 

engineering solution. 

Organisms that are unfouled or 

only lightly fouled provide insights 

into the mechanisms that have evolved 

to prevent surface colonization or epibiosis 

(Wahl, 1989). These “clean” organisms 

are the focus of research into 

biomimetic and bioinspired solutions 

to fouling on human structures. Biomimicry 

refers to the study of the 

structure and function of biological 

systems as models for the design of 

engineering solutions (dictionary.com) 

while bioinspiration expands on biomimicry 

by not only copying or imitating 

nature but also improving them or 


epurposing the biological model for 

an idealized engineering solution. 

This review will discuss the methods 

by which nature controls fouling 

and characterize natural antifouling 

and engineered solutions in terms of 

chemical, physical, mechanical, behavioral 

and combined mechanisms. 

Natural Antifouling 

Organisms have evolved several 

different strategies to prevent fouling. 

These natural antifouling methods include 

chemical, physical, mechanical, 

behavioral and a combination of 

more than one of the others (Bers & 

Wahl, 2004; Pawlik, 1992; Wahl, 

1989). 

Chemical 

Chemical antifouling has a long 

history of research and has been the 

subject of many reviews (Armstrong 

et al., 2000; Clare, 1996; Fusetani, 

2004; Omae, 2006; Pawlik, 1992; 

Qian et al., 2010; Raveendran & 

Mol, 2009; Rittschoff, 2000; and 

others). To date, thousands of active 

natural products have been identified 

(Pawlik, 1992). The activity of natural 

chemistries includes low pH, deterrents, 

anesthetics, attachment and 

metamorphosis inhibitors, or toxic 

chemicals. The chemicals may be surface 

bound or water soluble (Omae, 

2006; Rittschof, 2000; Wahl, 1989). 

Despite issues with the ecological relevance 

of some of the chemicals, active 

natural products have been isolated 

from an algae, sponges, soft corals 

and a limpet that are either available 

at the surface or released into the 

water column when the organism 

is disturbed (de Nys & Steinberg, 

2002; Fusetani, 2004; Hay, 1996; 

Hellio et al., 2002). The mucus of dolphins, 

echinoderms, fish and corals 

contain antifouling chemicals that 

have been found to dissolve glue, prevent 

attachment or act as antimicrobial 

toxins (Baum et al., 2002; Bavington 

et al., 2004; Ebran et al., 2000; Ritchie, 

2006; Shephard, 1994; Videler et al., 

1999). Many tunicates have acidic 

body pH and low epibiosis, especially 

in areas where density of acidic 

vacuoles is high (Hirose et al., 2001; 

Stoeker, 1980). Additionally, when 

looking at whole animal extracts, 

some tunicates contain chemicals that 

are cytotoxic, antimicrobial and antiviral 

(Davis & Wright, 1989). The 

eggs of many organisms, including 

fish and coral, are well protected with 

antimicrobial chemistries (Marquis 

et al., 2005, Ramasamy & Marugan, 

2007). Terrestrial plants have also 

yielded interesting chemistries such 

as tannins, pyrethroids and capsaicin 

(Feng et al., 2009; Perez et al., 2007; 

Xu et al., 2005). 

More recently, attention has turned 

to microorganisms. Antifouling metabolites 

that had been attributed in the 

past to algae, sponges, corals, etc., have 

been found on closer study, to be produced 

by surface associated bacteria 

and cyanobacteria (Armstrong et al., 

2000; Clare, 1996; Krug, 2006). 

These surface associated microorganisms 

are distinct from the communities 

in the water column, often found in 

higher densities than the water column 

community and are often highly 

pigmented (Dobretsov et al., 2005; 

Faimalietal.,2004;Holmstrom 

et al., 1992). Deterrent biofilms preventfoulingbytoxicordeterrent 

chemistries (Dobretsov et al., 2006; 

Pawlik, 1992). 

Physical 

The two primary physical means 

identified to prevent fouling are surface 

energy and surface texture. A surface 

energy range of 20-30 dynes/cm 

(Baier, 1972; Dexter, 1979) has been 

shown to minimize adhesion and 

favor the removal of epibionts. Such 

surface energy values have been measured 

on the surface of killer whales 

(Baier & Meyer, 1986), gorgonians 

(Vrolijk et al., 1990) and healthy 

teeth (Baier & Meyer, 1986; Glantz 

et al., 1991). Surface energy also effects 

settlement of some organisms, albeit 

in a species specific manner (Anderson 

et al., 2003; Meyer et al., 1988; Molino 

& Weatherbee, 2008; Rittschof & 

Costlow, 1989). For example, it has 

been found that barnacles and bryozoans 

prefer to settle on different surface 

energies (Dahlstrom et al., 2004; 

Rittschof & Costlow, 1989). Diatoms 

and the green algae Ulva have different 

adhesion strengths on surfaces with 

different wettability. Diatoms are more 

easily removed from hydrophilic surfaces, 

whereas Ulva releases more easily 

from hydrophobic surfaces (Finlay 

et al., 2002; Kirshnan et al., 2006). 

Low adhesion surfaces in nature are associated 

with waxes, oils, surfactants, 

mucuses or fluorinated or methylated 

compounds (Baum et al., 2003; Krug, 

2006; Shephard, 1994; Wahl, 1989). 

The antifouling properties of surface 

topography have received extensive 

attention and review (Scardino 

& de Nys, 2011; Scardino et al., 2008). 

The effectiveness of topography as 

antifouling appears to be the relationship 

of scale between the texture and 

the settling organism. An ultra-smooth 

surface offers no refuge from predation 

or hydrodynamic stresses and is therefore 

unattractive (Kohler et al., 1999; 

Walters & Wethey, 1996). Surfaces 

with textures that are smaller than 

the settler reduce settlement and/or 

attachment strength, the “attachment 

point theory” (Scardino et al., 2008). 

Textures that are the same size or 


slightly larger than the propagules offer 

the greatest number of attachment 

points, the strongest attachment and 

the best protection to settling organisms 

(Callow et al., 2002; Scardino 

et al., 2006). Organisms that have 

only one attachment point (barnacles, 

arborescent bryozoans, etc.) are more 

specific in searching for high quality 

pits than colonial organisms with multiple 

attachment points, likely because 

the colonial organisms outgrow the 

“refuge” of the pit quickly and can 

survive partial mortality (Walters & 

Wethey, 1996). The use of hairs in 

mussel spat (Dixon et al., 1995), spicules 

in gorgonian coral (Scardino & 

de Nys, 2011; Vrolijk et al., 1990) 

and spines in some colonial organisms 

(Dyrynda, 1986; Wahl, 1989) has 

been shown to prevent fouling and 

overgrowth of fouling organisms. It 

has been suggested that many other organisms 

including crabs, brittle stars, 

molluscs, marine mammals and sharks 

(adult skin and dogfish egg cases), use 

textured surfaces with one or more 

scales of complexity to prevent microand 

macro-fouling (Baum et al., 

2002; Bers & Wahl, 2004; Scardino 

& de Nys, 2004; Scardino & de Nys, 

2011). 

Mechanical 

Grooming is a common antifouling 

mechanism found in nature. 

Grooming involves specialized structuresthateitherpickorsweepan 

animals surface clean (Wahl, 1989). 

Decapod crustaceans have highly 

evolved brushes used to remove epibionts 

and parasites from specific 

parts of their bodies like the gills and 

carapace (Acosta & Poirrier, 1992; 

Batang & Suzuki, 2003; Bauer, 1981). 

Historically, echinoderms and bryozoans 

were thought to use specialized 

structures to clean their surfaces 

(Campbell & Rainbow, 1977; Dyrynda, 

1986), but recently this has been called 

into question as the pedicellaria of the 

crown of thorns starfish (Acanthaster 

planci) were found to be too unresponsive 

and widely placed to be effective in 

keeping their surfaces clean (Guenther 

et al., 2007). Many organisms use 

ciliary cleaning in conjunction with 

mucus to keep surfaces clean (Wahl 

et al., 1998). Symbiotic or mutualistic 

relationships such as fish visiting 

“cleaning stations” (Poulin & Grutter, 

1996), mutualistic grazing of snails 

within populations (Wahl & Sonnichsen, 

1992; Wahl et al., 1998) and branchiobdellid 

annelids that feed on epibionts 

in the gill chamber of crayfish 

(Brown et al., 2002) are examples of 

beneficial relationships that may prevent 

fouling. 

The other mechanical method of 

antifouling is surface renewal via shedding 

or molting of outer layers. Crustaceans, 

stone fish and algae all molt, 

either the entire surface simultaneously 

or in patches, which removes 

all attached fouling (Bakus et al., 

1986; Keats et al., 1997; Wahl, 1989). 

Additionally, many organisms use 

mucus as membrane to separate themselves 

from their environment. The 

mucus sloughs off removing foulers, 

makes adhesion difficult and fouls 

sensory and attachment apparatus of 

epibionts (Brown & Bythell, 2005; 

Davies & Hawkins, 1998; Denny, 

1989; Dyrynda, 1986; Shephard, 

1994; Wahl, 1989; Wahl et al., 1998). 

Behavioral 

Behavioral antifouling is the direct 

or indirect active avoidance of fouling 

organisms (Becker & Wahl, 1996). 

Burrowing into sediment, moving 

into the air, between fresh and salt 

water or into areas with very different 

oxygen contents and nocturnal activity 

or hiding in crevices are all mechanisms 

that remove less tolerant epibiotic 

organisms (Becker & Wahl, 1996; 

Wahl, 1989; Wahl et al., 1998). Organisms 

with a similar range of tolerances 

to their hosts will be unaffected 

by these behavioral methods (Brock 

et al., 1999). 

Combination 

Most organisms that are well studied 

use a combination of methods to 

prevent surface fouling. This was highlighted 

in reviews by Ralston and 

Swain (2009) and Scardino and 

de Nys (2011). Crustaceans groom 

and shed their shells and use behavioral 

mechanisms like burrowing and moving 

among habitats to ensure clean surfaces 

(Becker & Wahl, 1996; Wahl 

et al., 1998). Echinoderms groom, 

slough, excrete anti-adhesive mucus, 

have chemical antifoulants and may 

even use a strong negatively charged 

cuticle to prevent surface colonization 

(Bakus et al., 1986; Bavington et al., 

2004; Bryan et al., 1996; McKenzie 

& Grigolava, 1996). Corals use antibacterial 

mucus, select specific microbial 

colonists which in turn protect 

them from other microbes, slough 

mucus and surface layers and secrete 

secondary metabolites to keep their 

surfaces clean (Brown & Bythell, 

2005; Ritchie, 2006; Targett et al., 

1983). Dolphins and whales are hypothesized 

to use microtopography 

in conjunction with an enzymatically 

active zymogel to prevent attachment 

of macrofoulers (Baum et al., 2003; 

Meyer & Seegers, 2004) and surface 

sloughing, skin compliance and a critical 

surface tension in the preferred 

range for minimal adhesion combined 

with breaching may remove fouling at 

an early stage (Baum et al., 2003; Fish 

& Rohr, 1999; Scardino & de Nys, 

2011). Algae have provided many 


new chemical metabolites that prevent 

fouling, shed their outer layers and 

may remove settled epibionts by flexing 

beyond what their epibionts can 

withstand (Nylund & Pavia, 2005; 

Scardino & de Nys, 2011; Walters 

et al., 2003; Wikstrom & Pavia, 2004). 

Biomimetic and Bioinspired 

Engineering Solutions 

Chemical 

Due to our vast experience with incorporating 

chemicals into coatings, it 

is not surprising that natural products 

are the most investigated biological 

antifouling. SeaNine 211 is a booster 

biocide added to copper coatings to 

boost efficacy against fouling plants. 

It degrades quickly in water and sediment, 

binds strongly to sediment, has 

low environmental toxicity and has an 

excellent performance record from lab 

and field tests and ship trials (Thomas 

& Brooks, 2010; Yebra et al., 2004). 

SeaNine 211 is based on the natural 

product isothiozolone originally isolated 

in the 1980s from the soft coral 

Eunicea (Raveendran & Mol, 2009). 

Econea is a halogenated pyrrol that is 

the active ingredient in copper-free 

antifouling paints from manufacturers 

such as Petit, Interlux, Sea Hawk and 

others. Halogenated pyrrols are common 

secondary metabolites in bacteria 

and sponges (or possibly surface bacteria 

associated with sponges) and are 

potent settlement and metamorphosis 

inhibitors for barnacles and other 

animal foulers (Dahms et al., 2006; 

Omae, 2006). Because of its specificity 

against animal fouling, Econea is often 

combined with a booster like SeaNine 

to prevent plant fouling as well. 

Perhaps the most studied natural 

chemistry is the halogenated furanone 

originally isolated from the red algae 

Delisea pulchra. The chemical is present 

on the surface of the plant in concentrations 

that prevent fouling and 

the coverage of epibionts corresponds 

to concentrations of the furanone 

(de Nys & Steinberg, 2002). It has 

not yet been successfully incorporated 

into a long-lasting ship hull coating 

(Chambers et al., 2006); however, 

some have reported that it is available 

products called “Netsafe” and “Pearlsafe” 

marketed in Australia for use in 

commercial aquaculture (Raveendran 

& Mol, 2009). We were unable to 

find any record of these products for 

sale at this time so they may no longer 

be available. Many other natural 

chemicals from marine macroorganisms 

show promise for non-toxic or 

low-toxicity antifouling paints and 

are being investigated (see reviews by 

Armstrong et al., 2000; Fusetani, 

2004; Omae, 2006; Qian et al., 

2010; Raveendran & Mol, 2009). 

Terrestrial plants have also yielded 

promising chemistries for antifouling. 

These include products like tannins, 

pyrethroids and capsaicin (Feng et al., 

2009; Perez et al., 2007; Thomas & 

Brooks, 2010; Xu et al., 2005). Pyrethroids, 

synthetic analogs of pyrethrin 

from chrysanthemum flowers, are 

of particular interest because they are 

already approved for use as environmentally 

safe insecticides. These insecticides 

have low toxicity to mammals, 

do not persist, do not bioaccumulate 

and are available in industrial quantities 

(Feng et al., 2009). Tannin is present 

in terrestrial plants, mangroves and 

in some marine algae, primarily as an 

anti-herbivory chemical. However, 

some have found it to have antifouling 

properties as well (Brock et al., 2007; 

Lau & Qian, 1997; Perez et al., 2007; 

Wikstrom & Pavia, 2004). The longterm 

efficacy of tannin isolated from 

the quebracho tree was improved by 

precipitating it with aluminum forming 

a salt which increased the life 

span of the coating to 1 month in the 

field (Perez et al., 2007). 

Another avenue of research is isolating 

chemicals from microorganisms 

and using the microorganisms themselves. 

There are many benefits to 

this strategy including culturability, 

abundance and ability to trick or stress 

the organisms into producing large 

quantities of the necessary chemical 

(Dobretsov et al., 2006; Holmstrom 

& Kjelleberg, 1994). Holmstrom et al. 

(2000) were able to keep bacteria alive 

in a coating for 14 days in the laboratory. 

Microencapsulation is another 

strategy being investigated, not just 

for microorganisms but for all natural 

products, as a way to increase length 

of efficacy. Coatings with microencapsulated 

living bacteria were able 

to prevent fouling up to 7 weeks in 

field trials (Chambers et al., 2006; Yee 

et al., 2007). 

The use of enzymes and hormones 

that are commercially available is another 

strategy for chemical antifouling. 

Many patents have been awarded and 

there is an enzymatic coating available 

on the Danish yacht market, although 

little scientific evidence of effectiveness 

exists (Olsen et al., 2007). Enzymes 

may act directly by dissolving 

glues, lysing cells or decomposing 

exoskeletons of barnacles (Abarzua & 

Jakubowski, 1995; Evans & Clarkson, 

1993; Olsen et al., 2007). They may 

also act indirectly by increasing the 

effectiveness of an antifoulant or by 

acting on the coating to improve release 

or polishing rates (Olsen et al., 

2007). Hormones, such as noradrenaline, 

may also be used as a non-toxic 

deterrent in antifouling coatings 

(Gohad et al., 2010). However, both 

enzymes and hormones have several 

drawbacks to their widespread use 


including expense, instability, potentially 

specific response and need for 

environmental approval (Gohad et al., 

2010; Olsen et al., 2007; Rittschof, 

2000). 

There are several challenges in getting 

new chemicals approved for use in 

antifouling paints. The environmental 

problems associated with TBT have 

increased awareness of the potential 

risks associated with introducing new 

chemistries and attention must be 

given to proving environmental compliance. 

This greatly increases the 

time and the cost to go from the identification 

and isolation of a chemical 

to commercialization, which can cost 

millions of dollars and take over 

10 years to get approval. Furthermore, 

many natural products are structurally 

complex and only available in small 

amounts in the organism so there are 

issues with obtaining or synthesizing 

the active chemicals (Fusetani, 2004; 

Rittschof, 2000; Rittschof, 2001). 

Natural products tend to have a short 

life span as they cannot be toxic to the 

organism.Thisleadstoissueswhen 

incorporated into a coating as coatings 

need to maintain efficacy for 

3-12 years of service life (de Nys & 

Steinberg, 2002; Fusetani, 2004; 

Ingle, 2007; Marechal & Hellio, 

2009; Rittschof, 2000; Rittschof, 

2001). Conversely, the short life span 

is beneficial for environmental compliance 

as one of the characteristics of an 

ideal chemical antifoulant is short half 

life (Clare, 1996). Other factors for an 

ideal chemical antifoulant include 

non-toxicity, activity against a wide 

variety of fouling organisms, easy incorporation 

into a controlled release 

coating and should come from a culturable 

organism or have an active 

chemistry that can be industrially synthesized 

(Clare, 1996; Hellio et al., 

2002; Marechal & Hellio, 2009; 

Raveendran & Mol, 2009; Rittschof, 

2000; Yebra et al., 2004). 

Physical 

Slippery coatings are not new 

technology. Commercially available 

fouling release coatings have been 

available since the mid-1970s and are 

proving to be an increasingly successful 

method for fouling control. These 

coatings combine polydimethylsiloxane 

silicone with low surface energy, 

oils and compliance to reduce the 

adhesion strength of organisms to a 

surface. Fouling is removed by hydrodynamic 

shear forces or with gentle 

cleaning (Anderson et al., 2003). 

There are several lanolin based waxes 

on the market for use on ship hulls 

and propellers. While the wax may 

provide short-term antifouling, the 

main purpose is to lessen attachment 

strength and make cleaning easier. 

These products are often used over a 

tough epoxy, however the duration of 

effect is unknown and we were unable 

to find any published data in a peer reviewed 

journal to back the claims of 

manufacturers. Results obtained from 

these coatings may vary depending 

on the fouling communities. Effects 

of surface energy and wettability on 

surface colonization are species specific 

with some responding favorably to hydrophobic 

surfaces and some to hydrophilic 

or intermediate surfaces (Callow 

et al., 2002, Dahlstrom et al., 2004). 

Additionally primary colonizers often 

change the surface energy of surfaces 

which will change the effect on subsequent 

settlers (Scardino & de Nys, 

2011). 

Mimicking the surface texture of 

marine organisms has been investigated 

as an environmentally friendly 

antifoulant. “Sealcoat” is a commercially 

available antifouling coating 

that is flocked with fibers mimicking 

the fur of a seal. According to the company’s 

website, fouling is prevented for 

up to 5 years. However, no scientific 

data exists to back this claim. Other 

studies looking at flocked or furred 

coatings found mixed responses with 

green and brown algae, encrusting 

bryozoans and barnacles deterred, red 

algae and hydroids unaffected and 

solitary tunicates and tube worms 

increased by these coatings (Phillippi 

et al., 2001). It must also be remembered 

that seals do not only depend 

on their fur to keep them fouling free 

but also groom and spend large amounts 

of time out of the water. 

Mimics of topographies from other 

organisms such as crustose coralline 

algae, molluscs, crabs, brittle stars, 

soft corals and dogfish egg cases, have 

been investigated for antifouling activity 

and have shown short-term efficacy 

in laboratory assays. Additionally, 

topographies from pilot whale and 

shark skins have been characterized 

and had an antifouling activity attributed 

to the microstructures. These active 

topographies range in scale from 1 

to 300 μm with multiple length scales 

occurring on natural surfaces (Baum 

et al., 2002; Bers & Wahl, 2004; 

Scardino & de Nys, 2004; Scardino 

& de Nys, 2011). The “Sharklet” is 

an example of a biomimetic texture 

used as an engineered surface to prevent 

fouling. It has performed well in 

laboratory assays against Ulva spores 

and Balanus amphitrite cyprids (Carman 

et al., 2006; Schumacher et al., 2007). 

In order to improve the effect of this 

and other topographies, a mathematical 

model was created called 

the “Engineered Roughness Index”; 

this index can also be used to predict 

settlement of marine organisms on 

the engineered topographies (Long 

et al., 2010). Engineered surfaces 

with hierarchically wrinkled surfaces 


have shown promising results in field 

trials, especially against barnacles 

(Efimenko et al., 2009; Scardino & de 

Nys, 2011). 

Sound has been suggested as an 

antifouling method. However, there 

are no published field test data that scientifically 

prove that it can provide 

long-term antifouling. This is not a 

truly biomimetic method as it is not reported 

as a natural antifouling mechanism. 

Sound is used by competent 

larvae of fish and invertebrates like 

crabs to navigate to appropriate settlement 

sites (Radford et al., 2010; 

Simpson et al., 2008; Stanley et al., 

2010). Specific habitats have different 

auditory signatures and larvae can use 

these to differentiate and pilot to their 

adult habitats (Radford et al., 2010). 

Both high- and low-frequency sound 

waves have been shown to be effective 

at inhibiting settlement of barnacles 

and mussels (Branscomb & Rittschof, 

1984; Donskoy & Ludyanskiy, 1995; 

Guo et al., 2011). Additionally, ultrasound 

waves have been used to destroy 

barnacle larvae via cavitation for ballast 

water treatment (Seth et al., 2010). 

The use of low-frequency sound is limited 

because it is audible to humans 

and other organisms and therefore 

noise pollution is an issue. Several 

companies worldwide (Ultrasonic Antifouling, 

ASM, Sonihull and others) 

offer ultrasonic units that can be installed 

that are purported to prevent 

fouling or conversely to kill settling 

fouling organisms thereby making 

them easy to remove. However, this 

method is variable in effect, with settlement 

rates ranging from 1% up to 

55% for low and high frequency, respectively 

(Branscomb & Rittschof, 

1984; Guo et al., 2011). Guo and colleagues 

(2011) reported a settlement 

rate for barnacle cyprids in the laboratory 

of about 20% for their best ultrasonic 

treatment compared to a rate of 

around 70% for the control so the 

method is not perfect. Additionally, 

Sonihull reports changes in fish behavior 

when their ultrasonic units are 

in use so there are noise pollution concerns 

with high-frequency sound as 

well. 

Physical methods of antifouling are 

often inferred but seldom proved due 

to challenges associated with testing 

living materials. Surface topography 

effects are scale dependent (Scardino 

et al., 2006; Schumacher et al., 2007) 

and effectiveness in fouling prevention 

may vary geographically (Bers et al., 

2010). Finding a universal physical 

antifoulant may be difficult and the 

results are often short lived, lasting a 

monthorlessinfield testing (Holm 

et al., 1997). 

Mechanical 

Mechanical cleaning is performed 

on ships, aquaculture nets, instruments 

and other marine structures 

when they become fouled, either because 

an antifouling coating was not 

used or if that coating becomes fouled. 

The U.S. Navy cleans their vessels 

when a set level of fouling is reached 

as set out in the Naval Ships’ Technical 

Manual (Cologer, 1984; NSTM, 

2006). Cleaning is reactive and has 

been shown to speed the rate of recolonization 

and increase the risk of 

transport of nonindigenous species 

(Floerl et al., 2005). Additionally, 

commercially available brush cleaning 

devices (i.e., SCAMP, Mini-Pamper, 

etc.) are harsh and may damage the 

antifouling coating. A new direction 

for mechanical antifouling is to mimic 

natural grooming. This is the idea behind 

the HullBUG (Hull Bioinspired 

Underwater Grooming), an autonomous 

robot that will proactively pass 

over a hull while a ship is in port. Its 

mode of action is a gentle wiping or 

brushing of the surface on a frequent 

schedule sufficient to remove fouling 

at its earliest stages before it can become 

established (Borchardt, 2010; 

Tribou & Swain, 2010). Results 

from field testing of fouling release 

and copper-coated panels subjected 

to grooming are so promising that further 

experiments and scale up on this 

method are being investigated. 

Ecospeed is a commercially available 

hull coating system. It is a tough 

glass flake reinforced vinyl ester 

coating. When combined with hull 

cleaning, this non-toxic coating is 

purported to maintain a fouling free 

surface with no repainting for up to 

25 years. Additionally, the coating 

smooths during cleaning, decreasing 

drag. Again, no scientifically published 

data exists to back the claims made 

by the manufacturer. 

Surface renewal has been attempted 

as an antifoulant for ship 

hull coatings. Polymers were developed 

that hydrolyze in seawater leaving 

a clean surface as they dissolve 

(Candries et al., 2000). To date, however, 

these coatings have only been 

successful when combined with biocides 

as the rate of dissolution and 

thickness of the coating required 

would be too great without the help 

of toxic chemicals. 

Mechanical antifouling has proven 

to be an effective but imperfect method 

of keeping submerged surfaces clean. 

Cleaning requires the deployment of 

equipment and usually divers, which 

increases both the expense and human 

risk factor of this antifouling 

mechanism. When applied to toxic 

coatings, cleaning may increase the release 

of biocides, at least in the short 

term (Schiff et al., 2004). Cleaning 

may cause damage to coatings which 

increases the rate of re-colonization 


and may increase the risk of transport 

of invasive species (Floerl et al., 2005; 

Piola & Johnston, 2008). Mechanical 

antifouling works better when combined 

with another antifouling method 

such as a biocidal coating or a fouling 

release surface. Grooming, however, 

is proactive and more closely matches 

many of the behavioral activities 

found in nature. Many organisms 

benefit from self or mutual grooming 

to maintain their surfaces free of 

fouling. 

Behavioral 

Behavioral methods include removing 

a vessel from the water when 

not in use or moving between fresh 

and salt water. The former is commonly 

practiced by recreational boat 

owners; however, removing a large 

vessel from the water is impractical, especially 

if that ship is frequently used. 

Moving vessels between fresh and 

salt water, by traversing through the 

Panama Canal, for instance, is performed 

occasionally and has been 

credited with preventing the unobstructed 

movement of Caribbean and 

Pacific species between the two bodies 

of water. Brock and colleagues (1999) 

found that moving a ship into fresh 

waterfor9dayswassufficient to remove 

90% of fouling from the hull. 

However, tolerant fouling organisms 

will not be affected by this antifouling 

method as shown by the survival and 

subsequent introduction of the mussel 

Mytilus galloprovincialis to Oahu, 

Hawaii, from Washington. The mussel 

was one of the 10% of fouling organisms 

remaining on the USS Missouri 

after its Pacific transit and was seen 

spawning shortly after arrival in Pearl 

Harbor and later found colonizing 

the ballast tanks of a submarine (Apte 

et al., 2000). 

Combined 

It is unlikely that any one antifouling 

mechanism will be sufficient to 

prevent all fouling in all situations 

that may be encountered by submerged 

structures. Indeed, every 

organism that is well studied with 

regards to natural antifouling uses a 

combination of strategies to maintain 

a clean surface. The most effective 

coatings in use today also use more 

than one antifouling mechanism; antifouling 

coatings use a biocide combined 

with self-polishing or ablative 

mechanism to keep an active layer at 

the surface. Fouling release coatings 

combine low surface energy, oils and 

compliance to maximize self-cleaning. 

To date, most researchers investigating 

a biomimetic solution to antifouling 

have focused on only one method. 

However, that is beginning to change 

with the recent publication of reviews 

focusing on combined antifouling 

mechanisms (Ralston & Swain, 2009; 

Scardino & de Nys, 2011). 

Bioinspired Approaches 

The examples highlighted above 

represent biomimetic solutions for 

biofouling control. Very little research 

exists that takes lessons from nature 

and adapts or alters them for a true 

bioinspired solution. It has only been 

recently that novel uses have been proposed 

from biological models. For example, 

the dopamine based adhesive 

system in mussels, a common fouling 

organism, has been investigated as a way 

to obtain better adhesion of non-stick 

coatings to a substrate. The dopamine 

adhesive allows testing on polyethylene 

glycol (PEG) and other slippery 

polymers where before it was not possible 

because the polymers would not 

sticktoanything.Thosecoatings 

using the bioinspired PEG-DOPA system 

outperformed traditional silicone 

fouling release coatings in laboratory 

assays, comparing both the settlement 

and adhesion of a common fouling 

diatom and alga (Statz et al., 2006). 

Conclusions 

Marine organisms can achieve 

long-term protection from fouling 

using short-lived renewable mechanisms. 

This is attributed to using a 

combination of chemical, physical, 

mechanical and behavioral mechanisms. 

Much research has been published 

investigating the specific ways 

that organisms maintain a clean surface 

but frequently focus on only one 

mechanism without considering the efficacy 

of a holistic combined method. 

The challenge, for us, is to identify and 

select the best natural systems to solve 

the problem of biofouling. Through 

improved knowledge of natural systems,wewillbebetterabletoboth 

mimic and innovate using biological 

models to find engineering solutions. 

Results so far have been promising 

but better interactions between biologists, 

ecologists, engineers, chemists 

and materials scientists are needed 

and publishing of results is vitally important. 

Despite some issues, biomimetics 

and bioinspiration hold great 

promise for new antifouling solutions. 

For example, the HullBUG grooming 

method now being developed by the 

Office of Naval Research demonstrates 

how a proactive grooming method will 

enhance the long-term effectiveness 

of the presently available commercial 

antifouling or fouling release surfaces. 



the Office of Naval Research (grants 

N000140210217, N000140810034 

and N000140910843), who has 


funded much of this work and continues 

to support research directed 

towards discovering improved antifouling 

technology and the participants 

in the ONR Coatings Research 

Group. 

Authors: 

Emily Ralston and Geoffrey Swain 


150 W. University Blvd., 

Melbourne, FL 32901 

Emails: eralston@fit.edu; 

swain@fit.edu 

References 

Abarzua, S., & Jakubowski, S. 1995. Biotechnological 

investigation for the prevention 

of biofouling. I. Biological and biochemical 

principles for the prevention of biofouling. 

Mar Ecol Prog Ser. 123:301-12. doi: 10.3354/ 

meps123301. 

Acosta, C., & Poirrier, M. 1992. Grooming 

behavior and associated structures of the 

mysid Mysidopsis bahia. J Crustacean Biol. 

12:383-91. doi: 10.2307/1549032. 

Anderson, C., Atlar, M., Callow, M., Candries, 

M., Milne, A., & Townsin, R. 2003. The 

development of foul-release coatings for 

seagoing vessels. J Mar Des Operations. 

B4:11-23. 

Apte, S., Holland, B., Godwin, L., & Gardner, 

J. 2000. Jumping ship: A stepping stone 

event mediated transfer of a non-indigenous 

species via a potentially unsuitable environment. 

Biol Invasions. 2:75-79. doi: 10.1023/ 

A:1010024818644. 

Armstrong, E., Boyd, K., & Burgess, J. 2000. 

Prevention of marine biofouling using natural 

compounds from marine organisms. Biotech 

Ann Rev. 6:221-41. doi: 10.1016/S1387- 

2656(00)06024-5. 

Baier, R. 1972. Influence of the initial surface 

condition of materials on bioadhesion. 

In: Proc. 3rd Int. Congr. Mar. Corr. Foul. 

Maryland: Northwestern University Press. 

Baier, R., & Meyer, A. 1986. Biosurface 

chemistry for fun and profit. Chemtech. 

16:178-85. 

Bakus, G., Targett, N., & Schulte, B. 1986. 

Chemical ecology of marine organisms. 

J Chem Ecol. 12:951-87. doi: 10.1007/ 

BF01638991. 

Batang, Z., & Suzuki, H. 2003. Gill-cleaning 

mechanisms of the burrowing thalassinidean 

shrimps Nihonotrypaea japonica and Upogebia 

major (Crustacea: Decapoda). J Zool. 261: 

69-77. doi: 10.1017/S0952836903003959. 

Bauer, R. 1981. Grooming behavior and 

morphology in the decapod crustacea. 

J Crustacean Biol. 1:153-73. doi: 10.2307/ 

1548154. 

Baum, C., Meyer, W., Stelzer, R., & 

Fleischer, L-G. 2002. Average nanorough 

skin surface of the pilot whale (Globicephala 

melas, Delphinidae): Considerations on the 

self-cleaning abilities based on nanoroughness. 

Mar Biol. 140:653-7. doi: 10.1007/s00227- 

001-0710-8. 

Baum, C., Simon, F., Meyer, W., Fleischer, 

L-G., Siebers, D., Kacza, J., & Seeger, J. 2003. 

Surface properties of the skin of the pilot whale 

Globicephala melas. Biofouling. 19:181-6. 

doi: 10.1080/0892701031000061769. 

Bavington, C., Lever, R., Mulloy, B., 

Grundy, M., Page, C., Richardson, N., & 

McKenzie, J. 2004. Anti-adhesive glycoproteins 

in echinoderm mucus secretions. 

Comp Biochem Physiol B. 139:607-17. 

doi: 10.1016/j.cbpc.2004.07.008. 

Becker, K., & Wahl, M. 1996. Behaviour 

patterns as natural antifouling mechanisms 

of tropical marine crabs. J Exp Mar Biol 

Ecol. 203:245-58. doi: 10.1016/0022- 

0981(96)02575-0. 

Bers, A., Diaz, E., Da Gama, B., Vieira-Silva, 

F., Dobretsov, S., Valdivia, N., … Wahl, M. 

2010. Relevance of mytilid shell microtopographies 

for fouling defence—A global 

comparison. Biofouling. 26:367-77. 

doi: 10.1080/08927011003605888. 

Bers, A., & Wahl, M. 2004. The influence of 

natural surface microtopographies on fouling. 

Biofouling. 20:43-51. doi: 10.1080/ 

08927010410001655533. 

Borchardt, J. 2010. Grooming the fleet. 

Mech Eng. 132:33-35. 

Branscomb, E., & Rittschof, D. 1984. An 

investigation of low frequency sound waves 

as a means of inhibiting barnacle settlement. 

J Exp Mar Biol Ecol. 79:149-54. doi: 10.1016/ 

0022-0981(84)90215-6. 

Brock, E., Nylund, G., & Pavia, H. 2007. 

Chemical inhibition of barnacle larval settlement 

by the brown alga Fucus vesiculosus. 

Mar Ecol Prog Ser. 337:165-74. doi: 10.3354/ 

meps337165. 

Brock, R., Bailey-Brock, J., & Goody, J. 

1999. A case study of the efficacy of freshwater 

immersion in controlling introduction of 

alien marine fouling communities: The USS 

Missouri. Pac Sci. 53:223-31. 

Brown, B., & Bythell, J. 2005. Perspectives on 

mucus secretion in reef corals. Mar Ecol Prog 

Ser. 296:291-309. doi: 10.3354/meps296291. 

Brown, B., Creed, R., & Dobson, W. 2002. 

Branchiobdellid annelids and their crayfish 

hosts: Are they engaged in a cleaning symbiosis 

Oecologia. 132:250-5. doi: 10.1007/s00442- 

002-0961-1. 

Bryan, P., Rittschof, D., & McClintock, J. 

1996. Bioactivity of echinoderm ethanolic 

body-wall extracts: An assessment of marine 

bacterial attachment and macroinvertebrate 

larval settlement. J Exp Mar Biol Ecol. 196: 

79-96. doi: 10.1016/0022-0981(95)00124-7. 

Callow, M., Jennings, A., Brennan, A., 

Seegert, C., Gibson, A., Wilson, L., … 

Callow, J. 2002. Microtopographic cues for 

settlement of zoospores of the green fouling 

alga Enteromorpha. Biofouling. 18:237-45. 

doi: 10.1080/08927010290014908. 

Campbell, A., & Rainbow, P. 1977. The 

role of pedicellariae in preventing barnacle 

settlement on the sea-urchin test. Mar Behav 

Physiol. 4:253-60. doi: 10.1080/ 

10236247709386957. 

Candries, M., Atlar, M., & Anderson, C. 

2000. Considering the use of alternative 


antifoulings: The advantages of foul-release 

systems. In: Conference Proceedings, ENSUS 

2000. pp. 88-95. Newcastle, UK: University 

of Newcastle-upon-Tyne. 

Carbery, K., Owen, R., Frickers, T., Otero, 

E., & Readman, J. 2006. Contamination of 

Caribbean coastal waters by the antifouling 

herbicide Irgarol 1051. Mar Pollut Bull. 

52:635-44. doi: 10.1016/j.marpolbul.2005. 

10.013. 

Carman, M., Estes, T., Feinberg, A., 

Schumacher, J., Wilkerson, W., Wilson, L., … 

Brennan, A. 2006. Engineered antifouling 

microtopographies—correlating wettability 

with cell attachment. Biofouling. 22:11-21. 

doi: 10.1080/08927010500484854. 

Carson, R., Damon, M., Johnson, L., & 

Gonzalez, J. 2009. Conceptual issues in 

designing a policy to phase out metal-based 

antifouling paints on recreational boats in 

San Diego Bay. J Environ Manag. 90:2460-8. 

doi: 10.1016/j.jenvman.2008.12.016. 

Chambers, L., Stokes, K., Walsh, F., & 

Wood, R. 2006. Modern approaches to 

marine antifouling coatings. Surf Coat Tech. 

201:3642-52. doi: 10.1016/j.surfcoat.2006. 

08.129. 

Clare, A. 1996. Marine natural product 

antifoulants: Status and potential. Biofouling. 

9:211-29. doi: 10.1080/08927019609378304. 

Cologer, C. 1984. Six year interaction of 

underwater cleaning with copper based antifouling 

paints on Navy surface ships. Nav Eng J. 

96:200-8. doi: 10.1111/j.1559-3584.1984. 

tb01829.x. 

Dafforn, K., Glasby, T., & Johnston, E. 

2008. Differential effects of tributyltin and 

copper antifoulants on recruitment of nonindigenous 

species. Biofouling. 24:23-33. 

doi: 10.1080/08927010701730329. 

Dahlstrom, M., Jonsson, H., Jonsson, P., 

& Elwing, H. 2004. Surface wettability as a 

determinant in the settlement of the barnacle 

Balanus improvisus (Darwin). J Exp Mar Biol 

Ecol. 305: 223-32. doi: 10.1016/j.jembe. 

2003.12.013. 

Dahms, H., Ying, X., & Pfeiffer, C. 2006. 

Antifouling potential of cyanobacteria: 

A mini-review. Biofouling. 22:317-27. 

doi: 10.1080/08927010600967261. 

Davis, A., & Wright, A. 1989. Interspecific 

differences in fouling of two congeneric ascidians 

(Eudistoma olivaceum and E. capsulatum): Is 

surface acidity an effective defense Mar Biol. 

102:491-7. doi: 10.1007/BF00438350. 

Davies, M., & Hawkins, S. 1998. Mucus 

from marine molluscs. In: Advances in Marine 

Biology, eds. Blaxter, J., Southward, A., & 

Tyler, P., 1-71. San Diego: Academic Press. 

Denny, M. 1989. Invertebrate mucous secretions: 

Functional alternatives to vertebrate 

paradigms. Symp Soc Exp Biol. 43:337-66. 

de Nys, R., & Steinberg, P. 2002. Linking 

marine biology and biotechnology. Curr Opin 

Biotech. 13:244-8. doi: 10.1016/S0958- 

1669(02)00311-7. 

Dexter, S. 1979. Influence of substratum 

critical surface tension on bacterial adhesion— 

in situ studies. J Colloid Interf Sci. 70: 

346-54. doi: 10.1016/0021-9797(79)90038-9. 

Dixon, D., Sole-Cava, A., Pascoe, P., & 

Holland, P. 1995. Periostracal adventitious 

hairs on spat of the mussel Mytilus edulis. 

J Mar Biol Assoc UK. 75:363-72. 

doi: 10.1017/S0025315400018233. 

Dobretsov, S., Dahms, H., & Qian, P. 2006. 

Inhibition of biofouling by marine microorganisms 

and their metabolites. Biofouling. 

22:43-54. doi: 10.1080/08927010500504784. 

Dobretsov, S., Dahms, H-U., Tsoi, M., & 

Qian, P-Y. 2005. Chemical control of epibiosis 

by Hong Kong sponges: The effect of sponge 

extracts on micro- and macrofouling communities. 

Mar Ecol Prog Ser. 297:119-29. 

doi: 10.3354/meps297119. 

Donskoy, D., & Ludyanskiy, M. 1995. Low 

frequency sound as a control measure for zebra 

mussel fouling. In: Proceedings of the 5th 

International Zebra Mussel and Other Aquatic 

Nuisance Organisms Conference. pp. 103-112. 

Pembroke, Ontario: Professional Edge. 

Dyrynda, P. 1986. Defensive strategies of 

modular organisms. Philos T Roy Soc Lon B. 

313:227-43. doi: 10.1098/rstb.1986.0035. 

Ebran, N., Julien, S., Orange, N., Auperin, 

B., & Molle, G. 2000. Isolation and characterization 

of novel glycoproteins from fish 

epidermal mucus: Correlation between their 

pore-forming properties and their antibacterial 

activities. Biochim Biophys Acta. 1467:271-80. 

doi: 10.1016/S0005-2736(00)00225-X. 

Efimenko, K., Finlay, J., Callow, M., Callow, 

J., & Genzer, J. 2009. Development and 

testing of hierarchically wrinkled coatings 

for marine antifouling. ACS Appl Mat Int. 

1:1031-40. doi: 10.1021/am9000562. 

Evans, L., & Clarkson, N. 1993. Antifouling 

strategies in the marine environment. J Appl 

Bacteriol Symp Suppl. 74:119S-24S. 

Faimali, M., Garaventa, F., Terlizzi, A., 

Chiantore, M., & Cattaneo-Vietti, R. 2004. 

The interplay of substrate nature and biofilm 

formation in regulating Balanus amphitrite 

Darwin, 1854 larval settlement. J Exp Mar 

Biol Ecol. 306:37-50. doi: 10.1016/j.jembe. 

2003.12.019. 

Feng, D., Ke, C., Li, S., Lu, C., & Guo, F. 

2009. Pyrethroids as promising marine antifoulants: 

Laboratory and field studies. Mar 

Biotechnol. 11:153-60. doi: 10.1007/s10126- 

008-9130-9. 

Finlay, J., Callow, M., Ista, L., Lopez, G., 

& Callow, J. 2002. The influence of surface 

wettability on the adhesion strength and 

settled spores of the green algae Enteromorpha 

and the diatom Amphora. Integr Comp Biol. 

42:1116-22. doi: 10.1093/icb/42.6.1116. 

Fish, F., & Rohr, J. 1999. Review of 

dolphin hydrodynamics and swimming 

performance. Technical Report no. 1801. SSC. 

136 pp. 

Floerl, O., Inglis, G., & Marsh, H. 2005. 

Selectivity in vector management and investigation 

of the effectiveness of measures used to 

prevent transport of non-indigenous species. 

Biol Invasions. 7:459-75. doi: 10.1007/ 

s10530-004-4863-5. 


Fusetani, N. 2004. Biofouling and antifouling. 

Nat Prod Res. 21:94-104. doi: 10.1039/ 

b302231p. 

Glantz, P., Baier, R., Attstrom, R., Meyer, A., 

& Gucinski, H. 1991. Comparative clinical 

wettability of teeth and intraoral mucosa. 

J Adhes Sci Technol. 5:401-8. doi: 10.1163/ 

156856191X00459. 

Gohad, N., Shah, N., Metters, A., & Mount, 

A. 2010. Noradrenaline deters marine invertebrate 

biofouling when covalently bound 

in polymeric coatings. J Exp Mar Biol Ecol. 

394:63-73. doi: 10.1016/j.jembe.2010.07.014. 

Guenther, J., Heimann, K., & de Nys, R. 

2007. Pedicellariae of the crown-of-thorns sea 

star Acanthaster planci are not an effective 

defence against fouling. Mar Ecol Prog Ser. 

340:101-108. doi: 10.3354/meps340101. 

Guo, S., Lee, H., Chaw, K., Miklas, J., Teo, 

S., Dickinson, G., … Khoo, B. 2011. Effect 

of ultrasound on cyprids and juvenile barnacles. 


08927014.2010.551535. 

Hall, J., Giddings, J., Solomon, K., & Balcomb, 

R. 1994. An ecological risk assessment for the 

use of Irgarol 1051 as an algaecide for antifouling 

paints. Crit Rev Toxicol. 29:367-437. 

Hay, M. 1996. Marine chemical ecology: 

What’s known and what’s next J Exp Mar 

Biol Ecol. 200:103-34. doi: 10.1016/S0022- 

0981(96)02659-7. 

Hellio, C., Berge, J., Beaupoil, C., Le Gal, Y., 

& Bourgougnon, N. 2002. Screening of 

marine algal extracts for anti-settlement 

activities against microalgae and macroalgae. 


08927010290010137. 

Hirose, E., Yamashiro, H., & Mori, Y. 2001. 

Properties of tunic acid in the ascidian Phallusia 

nigra (Ascidiidae, Phlebobranchia). Zool Sci. 

18:309-14. doi: 10.2108/zsj.18.309. 

Holm, E., Cannon, G., Roberts, D., Schmidt, 

A., Sutherland, J., & Rittschof, D. 1997. 

The influence of initial surface chemistry on 

development of the fouling community at 

Beaufort, North Carolina. J Exp Mar Biol 

Ecol. 215:189-203. doi: 10.1016/S0022- 

0981(97)00040-3. 

Holmstrom, C., & Kjelleberg, S. 1994. The 

effect of external biological factors on settlement 

of marine invertebrate and new antifouling 

technology. Biofouling. 8:147-60. 

doi: 10.1080/08927019409378269. 

Holmstrom, C., Rittschof, D., & Kjelleberg, 

S. 1992. Inhibition of settlement by larvae 

of Balanus amphitrite and Ciona intestinalis 

by a surface-colonizing marine bacterium. 

Appl Environ Microb. 58:2111-5. 

Holmstrom, C., Steinberg, P., Christov, V., 

Christie, G., & Kjelleberg, S. 2000. Bacteria 

immobilised in gels: Improved methodologies 

for antifouling and biocontrol applications. 


08927010009386302. 

Ingle, M. 2007. Naval Sea Systems Command 

Antifouling Program. Presented at the National 

Paint & Coatings Association International 

Marine and Offshore Coatings Expo. Virginia 

Beach, VA. 

Keats, D., Knight, M., & Pueschel, C. 1997. 

Antifouling effects of epithallial shedding in 

three crustose coralline algae (Rhodophyta, 

Coralinales) on a coral reef. J Exp Mar Biol 

Ecol. 213:281-93. doi: 10.1016/S0022- 

0981(96)02771-2. 

Kohler, J., Hansen, P., & Wahl, M. 1999. 

Colonization patterns at the substratum-water 

interface: How does surface microtopography 

influence recruitment patterns of sessile 

organisms Biofouling. 14:237-48. doi: 10.1080/ 

08927019909378415. 

Krishnan, S., Wang, N., Ober, C., Finlay, J., 

Callow, M., Callow, J., … Fischer, D. 2006. 

Comparison of the fouling release properties 

of hydrophobic fluorinated and hydrophilic 

PEGylated block copolymer surfaces: Attachment 

strength of the diatom Navicula and 

the green algae Ulva. Biomacromolecules. 

7:1449-62. doi: 10.1021/bm0509826. 

Krug, P. 2006. Defense of benthic invertebrates 

against surface colonization by larvae: A 

chemical arms race. In: Progress in Molecular 

and Subcellular Biology Subseries Marine 

Molecular Biotechnology, eds. Fusetani, N., 

& Clare, A., 1-53. Berlin Heidelberg: Springer- 

Verlag. 

Lau, S., & Qian, P. 1997. Phlorotannins 

and related compounds as larval settlement 

inhibitors of the tube-building polychaete 

Hydroides elegans. Mar Ecol Prog Ser. 

159:219-27. doi: 10.3354/meps159219. 

Long, C., Schumacher, J., Robinson, P., III, 

Finlay, J., Callow, M., Callow, J., & Brennan, 

A. 2010. A model that predicts the attachment 

behavior of Ulva linza zoospores on surface 

topography. Biofouling. 26:411-9. doi: 10.1080/ 

08927011003628849. 

Marechal, J-P., & Hellio, C. 2009. Challenges 

for the development of new non-toxic antifouling 

solutions. Int J Mol Sci. 10:4623-37. 

doi: 10.3390/ijms10114623. 

Marquis, C., Baird, A., de Nys, R., Holmstrom, 

C., & Koziumi, N. 2005. An evaluation 

of the antimicrobial properties of the eggs 

of 11 species of scleractinian corals. Coral 

Reefs. 24:248-53. doi: 10.1007/s00338-005- 

0473-7. 

McKenzie, J., & Grigolava, I. 1996. The 

echinoderm surface and its role in preventing 

microfouling. Biofouling. 10:261-72. 

doi: 10.1080/08927019609386285. 

Meyer, A., Baier, R., & King, R. 1988. Initial 

fouling of nontoxic coatings in fresh, brackish, 

and sea water. Can J Chem Eng. 66:55-62. 

doi: 10.1002/cjce.5450660108. 

Meyer, W., & Seegers, U. 2004. A preliminary 

approach to epidermal antimicrobial 

defense in the Delphinidae. Mar Biol. 

144:841-4. doi: 10.1007/s00227-003- 

1256-8. 

Molino, P., & Weatherbee, R. 2008. The 

biology of biofouling diatoms and their role 

in the development of microbial slimes. 


08927010802254583. 

Naval Ships’ Technical Manual. 2006. 

Chapter 081—Waterborne underwater hull 

cleaning of Navy ships. Publication # S9086- 

CQ-STM-010/CH-081 Revision 5. Washington, 

DC: Naval Sea Systems Command. 


Nehring, S. 2001. After the TBT era: Alternative 

anti-fouling paints and their ecological 

risks. Senck Marit. 31:341-51. 

Nylund, G., & Pavia, H. 2005. Chemical 

versus mechanical inhibition of fouling in the 

red alga Dilsea carnosa. Mar Ecol Prog Ser. 

299:111-21. doi: 10.3354/meps299111. 

Olsen, S., Pedersen, L., Laursen, M., Kiil, S., 

& Dam-Johansen, K. 2007. Enzyme-based 

antifouling coatings: A review. Biofouling. 

23:369-83. doi: 10.1080/08927010701566384. 

Omae, I. 2003. Organotin antifouling paints 

and their alternatives. Appl Organometal 

Chem. 17:81-105. doi: 10.1002/aoc.396. 

Omae, I. 2006. General aspects of natural 

products antifoulants in the environment. 

Handbook Env Chem. 5:227-62. doi: 10.1007/ 

698_5_057. 

Owen, R., Knap, A., Toaspern, M., & Carbery, 

K. 2002. Inhibition of coral photosynthesis 

by the antifouling herbicide Irgarol 1051. Mar 

Pollut Bull. 44:623-32. doi: 10.1016/S0025- 

326X(01)00303-4. 

Pawlik, J. 1992. Chemical ecology of the 

settlement of benthic marine invertebrates. 

Oceanogr Mar Biol Annu Rev. 30:273-335. 

Perez, M., Garcia, M., Blustein, G., & Stupak, 

M. 2007. Tannin and tannate from the 

quebracho tree: An eco-friendly alternative 

for controlling marine biofouling. Biofouling. 

23:151-9. doi: 10.1080/08927010701189484. 

Phillippi, A., O’Connor, N., Lewis, A., & 

Kim, Y. 2001. Surface flocking as a possible 

anti-biofoulant. Aquaculture. 195:225-38. 

doi: 10.1016/S0044-8486(00)00556-1. 

Piola, R., & Johnston, E. 2008. The potential 

for translocation of marine species via small 

scale disruptions to antifouling surfaces. 


08927010801930480. 

Poulin, R., & Grutter, A. 1996. Cleaning 

symbioses: Proximate and adaptive explanations. 

Bioscience. 46:512-7. doi: 10.2307/1312929. 

Qian, P., Xu, Y., & Fusetani, N. 2010. 

Natural products as antifouling compounds: 

Recent progress and future perspectives. 


08927010903470815. 

Radford, C., Stanley, J., Tindle, C., 

Montgomery, J., & Jeffs, A. 2010. Localised 

coastal habitats have distinct underwater 

sound signatures. Mar Ecol Prog Ser. 

401:21-9. doi: 10.3354/meps08451. 

Ralston, E., & Swain, G. 2009. Bioinspiration—the 

solution for biofouling control 

Bioinspir Biomim. 4:015007. doi: 10.1088/ 

1748-3182/4/1/015007. 

Ramasamy, M., & Murugan, A. 2007. Fouling 

deterrent chemical defense in three muricid 

gastropod egg masses from the Southeast 

coast of India. Biofouling. 23:259-65. 

doi: 10.1080/08927010701278857. 

Raveendran, T., & Mol, V. 2009. Natural 

products antifoulants. Curr Sci. 97:508-20. 

Ritchie, K. 2006. Regulation of microbial 

populations by coral surface mucus and 

mucus-associated bacteria. Mar Ecol Prog Ser. 

322:1-14. doi: 10.3354/meps322001. 

Rittschof, D. 2000. Natural product antifoulants: 

One perspective on the challenges 

related to coating development. Biofouling. 

15:119-27. doi: 10.1080/08927010009386303. 

Rittschof, D. 2001. Natural product antifoulants 

and coatings development. In: 

Marine Chemical Ecology, eds. McClintock, 

J., & Baker, B., 543-66. Boca Raton, Florida: 

CRC Press. 

Rittschof, D., & Costlow, J., Jr. 1989. 

Bryozoan and barnacle settlement in relation 

to initial surface wettability: A comparison 

of laboratory and field studies. Topics Mar 

Biol. 53:411-6. 

Scardino, A., & de Nys, R. 2004. Fouling 

deterrence on the bivalve shell Mytilus galloprovincialis: 

A physical phenomenon. Biofouling. 

20:249-57. doi: 10.1080/08927010400016608. 

Scardino, A., & de Nys, R. 2011. Minireview: 

Biomimetic models and bioinspired 

surfaces for fouling control. Biofouling. 27: 

73-86. doi: 10.1080/08927014.2010.536837. 

Scardino, A., Guenther, J., & de Nys, R. 

2008. Attachment point theory revisited: The 

fouling response to a microtextured matrix. 


08927010701784391. 

Scardino, A., Harvey, E., & de Nys, R. 2006. 

Testing attachment point theory: Diatom 

attachment on microtextured polyamide biomimetics. 


08927010500506094. 

Scarlett, A., Donkin, P., Fileman, T., & 

Morris, R. 1999. Occurrence of the antifouling 

herbicide, Irgarol 1051, within coastalwater 

seagrasses from Queensland, Australia. 

Mar Pollut Bull. 38:687-91. doi: 10.1016/ 

S0025-326X(99)00003-X. 

Schiff, K., Diehl, D., & Valkirs, A. 2004. 

Copper emissions from antifouling paint 

on recreational vessels. Mar Pollut Bull. 

48:371-7. doi: 10.1016/j.marpolbul.2003. 

08.016. 

Schumacher, J., Aldred, N., Callow, M., 

Finlay, J., Callow, J., Clare, A., & Brennan, A. 

2007. Species-specific engineered antifouling 

topographies: Correlations between the settlement 

of algal zoospores and barnacle cyprids. 


08927010701393276. 

Seth, N., Chakravarty, P., Khandeparker, L., 

Anil, A., & Pandit, A. 2010. Quantification 

of the energy required for the 

destruction of Balanus Amphitrite larva 

by ultrasonic treatment. J Mar Biol Assoc 

UK. 90:1475-82. doi: 10.1017/ 

S0025315409991548. 

Shephard, K. 1994. Functions for fish 

mucus. Rev Fish Biol Fisher. 4:401-29. 

doi: 10.1007/BF00042888. 

Simpson, S., Meekan, M., Jeffs, A., 

Montgomery, J., & McCauley, R. 2008. 

Settlement-stage coral reef fish prefer the 

higher-frequency invertebrate-generated 

audible component of reef noise. Anim 

Behav. 75:1861-8. doi: 10.1016/j.anbehav. 

2007.11.004. 

Stanley, J., Radford, C., & Jeffs, A. 2010. 

Induction of settlement in crab megalopae by 

ambient underwater reef sound. Behav Ecol. 

21:113-20. doi: 10.1093/beheco/arp159. 


Statz, A., Finlay, J., Dalsin, J., Callow, M., 

Callow, J., & Messersmith, P. 2006. Algal 

antifouling and fouling release properties of 

metal surfaces coated with a polymer inspired 

by marine mussels. Biofouling. 22:391-9. 

doi: 10.1080/08927010601004890. 

Stoeker, D. 1980. Relationships between 

chemical defense and ecology in benthic 

ascidians. Mar Ecol Prog Ser. 3:257-65. 

doi: 10.3354/meps003257. 

Swain, G. 1999. Redefining antifouling 

coatings. J Prot Coat Linings. 16:26-35. 

Swain, G. 2010. The importance of ship 

hull coatings and maintenance as drivers for 

environmental sustainability. In: Ship Design 

and Operation for Environmental Sustainability. 

pp 8. London: Royal Institute of 

Naval Architects. 

Swain, G., Kovach, B., Touzot, A., Casse, F., 

& Kavanagh, C. 2007. Measuring the 

performance of today’s antifouling coatings. 

J Ship Prot. 23:164-170. 

Targett, N., Bishop, S., O’Connell, O., & 

Yoder, J. 1983. Antifouling agents against the 

benthic marine diatom, Navicula salinicola 

Homarine from the gorgonians Leptogorgia 

virgulata and L. setacea and analogs. J Chem 

Ecol. 9:817-29. doi: 10.1007/BF00987807. 

Thomas, K., & Brooks, S. 2010. The environmental 

fate and effects of antifouling paint 

biocides. Biofouling. 26:73-88. doi: 10.1080/ 

08927010903216564. 

Thomas, K., Fileman, T., Readman, J., & 

Waldock, M. 2001. Antifouling paint booster 

biocides in the UK coastal environment and 

potential risks of biological effects. Mar 

Pollut Bull. 42:677-88. doi: 10.1016/S0025- 

326X(00)00216-2. 

Tribou, M., & Swain, G. 2010. The use of 

proactive in-water grooming to improve the 

performance of ship hull antifouling coatings. 


08927010903290973. 

Videler, H., Geertjes, G., & Videler, J. 1999. 

Biochemical characteristics and antibiotic 

properties of the mucus envelope of the 

queen parrotfish. J Fish Biol. 54:1124-7. 

doi: 10.1111/j.1095-8649.1999.tb00864.x. 

Vrolijk, N., Targett, N., Baier, R., & Meyer, 

A. 1990. Surface characterisation of two 

gorgonian coral species: Implications for a 

natural antifouling defence. Biofouling. 

2:39-54. doi: 10.1080/08927019009378128. 

Wahl, M. 1989. Marine epibiosis. I. Fouling 

and antifouling: Some basic aspects. Mar 

Ecol Prog Ser. 58:175-89. doi: 10.3354/ 

meps058175. 

Wahl, M., Kroger, K., & Lenz, M. 1998. 

Non-toxic protection against epibiosis. 


08927019809378355. 

Wahl, M., & Sonnichsen, H. 1992. Marine 

epibiosis. IV. The periwinkle Littorina littorea 

lacks typical antifouling defences—why are 

some populations so little fouled Mar Ecol 

Prog Ser. 88:225-35. doi: 10.3354/meps088225. 

Walters, L., Smith, C., & Hadfield, M. 2003. 

Recruitment of sessile marine invertebrates 

on Hawaiian macrophytes: Do pre-settlement 

or post-settlement processes keep plants free 

from fouling Bull Mar Sci. 72:813-39. 

Walters, L., & Wethey, D. 1996. Settlement 

and early post-settlement survival of sessile 

marine invertebrates on topographically 

complex surfaces: The importance of refuge 

dimensions and adult morphology. Mar 

Ecol Prog Ser. 137:161-71. doi: 10.3354/ 

meps137161. 

Wikstrom, S., & Pavia, H. 2004. Chemical 

settlement inhibition versus post-settlement 

mortality as an explanation for differential 

fouling of two congeneric seaweeds. Oecologia. 

138:223-30. doi: 10.1007/s00442-003-1427-9. 

Xu, Q., Barrios, C., Cutright, T., & Newby, 

B. 2005. Assessment of antifouling effectiveness 

of two natural product antifoulants 

by attachment study with freshwater bacteria. 

Environ Sci Pollut Res. 12:278-84. 

doi: 10.1065/espr2005.04.244. 

Yebra, D., Kiil, S., & Dam-Johansen, K. 2004. 

Antifouling technology—past, present and 

future steps towards efficient and environmentally 

friendly antifouling coatings. Prog Org 

Coat. 50: 75-104. doi: 10.1016/j.porgcoat. 

2003.06.001. 

Yee, L., Holmstrom, C., Fuary, E., Lewin, N., 

Kjelleberg, S., & Steinberg, P. 2007. Inhibition 

of fouling by marine bacteria immobilised 

in K-carrageenan beads. Biofouling. 23: 

287-94. doi: 10.1080/08927010701366280. 


BOOK REVIEW 

Sex, Drugs, and Sea Slime: The Oceans’ 

Oddest Creatures and Why They Matter 

By Ellen Prager 

University of Chicago Press, April 15, 2011 (International Publication: May 15, 2011) 

184 pp., $26.00 (Hardcover) 

Reviewed by Jason Goldberg 

U.S. Fish and Wildlife Service 

When it comes to reviewing this book, 

please let me be candid about a personal 

bias: as a volunteer at the Smithsonian’s 

National Museum of Natural History, 

I’ve been known to give tours focused 

on the Sant Ocean Hall’s oddities, such 

as the two-horned narwhal skull, giant 

squid, and the seadevil. There’s a method 

to such madness: if you can capture a lay 

audience’s attention with the exciting and 

unusual, they may be more receptive to 

discussing more serious messages about 

the importance of our oceans. With Sex, 

Drugs, and Sea Slime, Dr. Ellen Prager has 

unquestionably written one of the more 

bizarre and fascinating books in ocean literature. 

Combine Tina Fey with Carl 

Sagan or perhaps Mel Brooks and Rachel 

Carson and you have this book. Prager’s 

writing is reminiscent of the works of 

Mary Roach, Anthony Aveni, and 

Michael Shermer, all of whom also have 

a talent for writing about science’s odder 

side. To the best of my knowledge, 

however, none has written as eloquently 

about, as Prager calls it, the lobster’s 

“Super Soaker Pee Blaster.” As she writes, 

the purpose of her book is to be “a brief 

and entertaining look at some of the 

oceans’ most fascinating creatures, their 

unusual tactics for survival, and their invaluable 

links to humankind. The end 

goal is to showcase the importance of 

the great diversity of life in the sea, why 

it is at risk, and why we should all care.” 

With this book, she has succeeded exceptionally 

well in achieving her goals. 

Everyone in the Marine Technology 

Society likely has some favorite story of a 

bizarre creature or other factoid about 

the oceans they learned while in school 

or on the job. Prager has thoroughly researched 

and captured the best of these 

tales. More importantly, what she has 

really done is write entertainingly about 

why the ocean is relevant regardless of 

where you might live. She has an ability 

to wax poetic about the ocean’s strangest 

denizens, whether it is the unassuming 

dinoflagellate or the majestic humpback 

whale. The range of material she covers 

in such a slender volume is really quite astonishing. 

I often found myself wondering 

as I read the book whether she would cover 

this species or that, and inevitably she did. 

Plankton, hagfishes, corals, eels, parrotfish, 

conesnails, cephalopods, kelp, and more— 

they’re all in here, along with all the 

(copious) mucus and other excretions 

they produce. As the book’s title indicates, 

there’s also plenty of sex, as well as a few sex 

changes. For those of you who might be 

wondering, yes, she included the pearlfish, 

a species that provides clear proof that 

evolution has a sense of humor. 

Some chapters focus on specific species, 

such as those on plankton or the 

denizens of a coral reef, while others target 

specialized functions, such as species that 

might compete in the “X-Games” or live 

in extreme environments. The descriptions 

of the species and their unusual habits are 

always entertaining and sometimes laughout-loud 

funny. She then turns serious at 

the end of each chapter when she covers 

why these species matter. I was very 

pleased to see that Prager doesn’t just 

cover the usual reasons, such as food, 

drugs, and recreation, but that she highlights 

things many people don’t realize, 

such as the value of corals in mitigating 

storm damages. She concludes the book 

onamoreseriousbutoptimisticnote, 

highlighting the dangers that still lurk in 

ocean conservation and suggesting actions 

we can all take so we can continue to enjoy 

the ocean for generations to come. 

Overall, Prager’s work makes for fascinating 

reading for anyone interested in 

marine science. While it does sometimes 

get a little technical, it is certainly appropriate 

for lay audiences. It’s also an invaluable 

resource for anyone who talks 

about the oceans and needs some good references 

to spice up their talk, although the 

inclusion of an index would have been 

beneficial for such purposes. If you want 

to get another sense of her book, you can 

download a free National Public Radio 

podcast from http://www.npr.org/ 

2011/04/07/135043954/under-the-seasex-is-slimy-business. 

Thesamecombination 

of titillating humor and practical 

discussion of the ocean’s valueisdemonstrated 

in Prager’s interview. 

Our livelihoods depend on the ocean, 

and the talent that has helped develop and 

effectively use technology is extraordinary. 

It is therefore incumbent upon each of us 

to be able to talk in some way with the 

public and decision-makers about how 


the well-being of everyone living on Terra 

Firma relies on blue, brown, or white 

water. Books such as Prager’s offer an outline 

that many of us can use to foster a discussion 

about our own respective fields. 

Yes, the book is about weird science. The 

giggle factor is unmistakable. Even so, each 

time it gets weird, it casts light on the wonders 

of the ocean and makes that weirdness 

important and meaningful. Perhaps for 

that reason, more than any other, the 

book is highly recommended. 


UPCOMING MTS JOURNAL ISSUES 

September/October 2011 

General Issue 

November/December 2011 

Legacy Underwater Munitions: Assessment, 

Evaluation of Impacts, and Potential 

Response Technologies 

Guest Editors: Geoffrey Carton and Terrance Long 

CALL FOR PAPERS 

The MTS Journal 

includes technical 

papers, notes and 

commentaries 

of general interest in 

most issues. 

Contact Managing Editor 

Amy Morgante 

(morganteeditorial@verizon.net) 

if you have material you 

would like considered. 

Specifications for 

submitting a manuscript 

are on the MTS website 

under the Publications menu. 

www.mtsociety.org 

E-mail: 

morganteeditorial@verizon.net 

Or visit our homepage at 

Check the Society website for future Journal topics. 

www.mtsociety.org 

www.mtsociety.org

Open Positions 

The MTS Career Center is an 

online resource to help employers 

and job seekers make career 

connections in the marine 

technology industry. Career 

professionals bring fresh ideas 

and new energy, and can keep 

your company current with modern 

trends and tactics. 

MTS Career Center 

Visit www.mtsociety.org/careers 

The MTS Career Center Provides: 

• Unmatched, Targeted Exposure for Job Postings 

MTS represents the largest audience of qualied marine technology 

industry professionals. 

• Easy Online Job & Applicant Management 

You can enter job descriptions, check the status of postings, renew or 

discontinue postings and even make payments online. Plus, you can 

easily manage your applicants through automatic email noti cations 

and applicant history access. 

• Resume Database Access 

With a paid job posting, you can search the resume database and 

use an automatic notication system to receive email notications 

when new resumes match your criteria. 

• Employment Branding Opportunities 

Along with each job posting, you can include information about your 

company and a link to your website. 

International Marine Forensics Symposium 

April 3 – 5, 2012 • National Harbor, MD 

The Symposium is sponsored by the 

Marine Technology Society, 

American Society of Naval Engineers, 

Society of Naval Architects & Marine Engineers. 

This symposium will bring together marine professionals 

and historians to exchange information on historic marine 

losses, on marine forensic investigation processes and 

tools, and on case studies where causes of failures and 

losses have been determined or are under continued 

study. The Symposium will report on the latest research 

and understanding of the Titanic, Lusitania, Edmund 

Fitzgerald, the Monitor and Passaic, HMS Prince of 

Wales, Bismarck, HMS Hood, and the Andrea Doria. 

Keynote speaker James 

Cameron (conrmed). 

His undersea documentaries 

include Expedition Bismarck, 

Ghosts of the Abyss, 

Volcanoes of the Deep Sea, 

Aliens of the Deep and 

Last Mysteries of the Titanic. 

The International Marine Forensics Symposium will be held at the Gaylord Hotel, National Harbor, MD. 

Exhibition space is available at $650 for a 10 x 10-foot space. Freeman Exhibit Services will handle the 

show decorating. A number of Sponsorship Opportunities are also available to maximize your impact 

with Symposium attendees. Contact Mary Beth Loutinsky at mbloutinsky@gmail.com for details.

Oceans of Opportunity: 

International Cooperation and Partnerships Across the Pacic 

September 19–22, 2011 

Kona, Hawaii 

Register Today for Outstanding Topics in Oceanology, 

the Latest Technology in the Exhibition Hall, and 

Excellent Networking Opportunities! 

go to: 

http://www.oceans11mtsieeekona.org/main.cfm/CID/16/Registration/ 

Of cial Notice of Marine Technology Society Annual Meeting 

2011 Ofcer Elections and Notice of Annual Membership Meeting 

The terms of the individuals holding the following 

Marine Technology Society (“MTS”) ofces 

end on December 31, 2011: Vice President of 

Publications, Vice President of Industry and 

Technology, Vice President of Education and 

Research, and Vice President of Government and 

Public Affairs. Elections for these ofces will 

be conducted at the MTS Annual Membership 

Meeting and Awards Luncheon held in conjunction 

with the OCEANS’11 MTS/IEEE Kona 

Conference, September 21, 2011, at the Hilton 

Waikoloa Village, Kona, Hawaii. The meeting will 

begin at noon PST. 

Article IV, Section 1.6 of the MTS Bylaws denes 

the nominations and election procedures for MTS 

ofcer elections. On June 20, 2011, the MTS Nominating 

Committee presented MTS’s President- 

Elect with at least two (2) nominees for each of the 

ofces to be elected at this year’s Annual Meeting. 

After reviewing these nominations, the President- 

Elect directed MTS staff to prepare the ballot for 

this year’s election and Member Proxy, both of 

which contain the names of the nominees for each 

ofce to be elected at this year’s election. The 

MTS Bylaws establish August 1, 2011 as the record 

date for determining the MTS members entitled to 

notice of the Annual Meeting and eligible to vote in 

this year’s election. If you are an active member of 

MTS as of August 1, 2011 (membership dues paid), 

you are a member in good standing and therefore 

entitled to vote in this year’s ofcer elections. 

Pursuant to MTS’s Bylaws, a member entitled to 

vote may do so in person at the Annual Meeting 

or by appointing a proxy to vote on the member’s 

behalf. If you wish to appoint a proxy to vote on 

your behalf at the MTS Annual Membership 

Meeting, please complete the online Member 

Proxy (http://www.votenet.com). 

This Proxy form must be completed no later than 

5:00 p.m. (Eastern Standard Time) on September 

11, 2011 (“Proxy Deadline”). Proxies that are 

incomplete or received after the Proxy Deadline 

will not be accepted or counted. A hardcopy proxy 

form may be requested by calling the MTS ofce at 

410.884.5330.

Marine Technology Society Member Organizations 

CORPORATE MEMBERS 

ABCO Subsea 

Houston, Texas 

AMETEK Sea Connect Products, Inc. 

Westerly, Rhode Island 

C & C Technologies, Inc. 

Lafayette, Louisiana 

C-MAR Group 


Compass Publications, Inc. 

Arlington, Virginia 

Converteam 


Delcor USA 


DOF Subsea USA 


Dynacon, Inc. 

Bryan, Texas 

E.H. Wachs Company 


Fluor Offshore Solutions 

Sugar Land, Texas 

Fugro Chance, Inc. 


Fugro Geoservices, Inc. 


Fugro-McClelland Marine Geosciences 


Fugro Pelagos, Inc. 

San Diego, California 

Geospace Offshore Cables 


GS-Hydro US 



Melbourne, Florida 

Hydroid, LLC 

Pocasset, Massachusetts 

Innerspace Corporation 

Covina, California 

INTECSEA 


InterMoor, Inc. 


iRobot Corporation 

Durham, North Carolina 

J P Kenny, Inc. 


Kongsberg Maritime, Inc. 


L-3 Communications 


L-3 MariPro 

Goleta, California 

Lockheed Martin Sippican 

Marion, Massachusetts 

Marine Cybernetics AS 

Trondheim, Norway 

Mitsui Engineering and Shipbuilding Co. Ltd. 

Tokyo, Japan 

Mohr Engineering & Testing 


Oceaneering Advanced Technologies 

Hanover, Maryland 

Oceaneering International, Inc. 


Odyssey Marine Exploration 

Tampa, Florida 

Oil States Industries, Inc. 

Arlington, Texas 

Perry Slingsby Systems, Inc. 


Phoenix International Holdings, Inc. 

Largo, Maryland 

QinetiQ North America – Technology Solutions 

Group 

Slidel, Louisiana 

Raytheon Technical Service Company, LLC 

Saipem America, Inc. 


Schilling Robotics, LLC 

Davis, California 

SEA CON Brantner and Associates, Inc. 

El Cajon, California 

SonTek/YSI, Inc. 


South Bay Cable Corp. 

Idyllwild, California 

Subconn, Inc. 

Burwell, Nebraska 

Subsea 7 (US), LLC 


Technip 


Teledyne Geophysical Instruments 


Teledyne Oil & Gas 

Daytona Beach, Florida 

Teledyne RD Instruments, Inc. 

Poway, California 

Tyco Electronics Subsea Communications 

Morristown, New Jersey 

UniversalPegasus International, Inc. 


BUSINESS MEMBERS 

Aanderaa Data Instruments, Inc. 

Attleboro, Massachusetts 

Ashtead Technology, Inc. 


Bastion Technologies, Inc. 


Bennex Subsea, Houston 


BioSonics, Inc. 

Seattle, Washington 

BIRNS, Inc. 

Oxnard, California 

C.A. Richards and Associates, Inc. 


Cochrane Technologies, Inc. 


Compass Personnel Services, Inc. 

Katy, Texas 

Contros Systems & Solutions GmbH 

Kiel, Germany 

DeepSea Power and Light 


Deepwater Rental and Sypply 

New Iberia, Louisiana 

DOER Marine 

Alameda, California 

DPS Offshore Inc. 


DTC International 


Energy Sales, Inc. 

Redmond, Washington 

Environ-Tech Diving 

Stanwood, Washington 

Falmat, Inc. 

San Marcos, California 

Fugro Atlantic 

Norfolk, Virginia 

Fugro GeoSurveys, Inc. 

St. John’s, Newfoundland and Labrador, 

Canada 

Global Industries Offshore, LLC 


Horizon Marine, Inc. 

Marion, Massachusetts 

ICAN 

Mt. Pearl, Newfoundland and Labrador, Canada 

Intrepid Global, Inc. 


IPOZ Systems, LLC 

Katy, Texas 

IVS 3D 

Portsmouth, New Hampshire 

iXBlue, Inc. 

Cambridge, Massachusetts 

KDU Worldwide Technical Services 

Sarjah, United Arab Emirates 

KnightHawk Engineering 


Liquid Robotics, Inc. 

Palo Alto, California 

Makai Ocean Engineering, Inc. 

Kailua, Hawaii 

Matthews-Daniel Company 


Oceanic Imaging Consultants, Inc. 

Honolulu, Hawaii 

The Marine Technology Society gratefully acknowledges the critical support of the Corporate, Business, and Institutional members listed. 

Member organizations have aided the Society substantially in attaining its objectives since its inception in 1963. 

OceanWorks International 


Poseidon Offshore Mining 

Oslo, Norway 

Quest Offshore Resources 

Sugar Land, Texas 

Remote Ocean Systems, Inc. 


RRC Robotica Submarina 

Macaé, Brazil 

SeaBotix 


SeaLandAire Technologies, Inc. 

Jackson, Mississippi 

SeaView Systems, Inc. 

Dexter, Michigan 

Sonardyne, Inc. 


SIMCorp Marine Environmental, Inc. 

St. Stephens, Canada 

Sound Ocean Systems, Inc. 

Redmond, Washington 

Stress Subsea, Inc. 


Subsea Riser Products, Inc. 


SURF Subsea, Inc. 

Magnolia, Texas 

Team Trident 

Cypress, Texas 

Technology Systems Corporation 

Palm City, Florida 

Teledyne Impulse 


Tension Member Technology 

Huntington Beach, California 

VideoRay, LLC 

Phoenixville, Pennsylvania 

WET Labs, Inc. 

Philomath, Oregon 

Xodus Group 


INSTITUTIONAL MEMBERS 

Associacio Institut Ictineu Centre, Catala De 

Recerca Submarina 

Barcelona, Spain 

Canadian Coast Guard 

St. John’s, Newfoundland and Labrador, Canada 

City of St. John’s 

Newfoundland and Labrador, Canada 

CLS America, Inc. 

Largo, Maryland 

Consortium for Ocean Leadership 

Washington, DC 

Department of Innovation, Trade and Rural 

Development 

St. John’s, Newfoundland and Labrador, Canada 

Fundação Homem do Mar 

Rio de Janeiro, Brazil 

Harbor Branch Oceanographic Institute 

Fort Pierce, Florida 

International Seabed Authority 

Kingston, Jamaica 

Marine Applied Research & Exploration 

Richmond, California 

Marine Institute 

Newfoundland and Labrador, Canada 

Monterey Bay Aquarium Research Institute 

Moss Landing, California 

National Research Council Institute for Ocean 

Technology 

St. John’s, Newfoundland and Labrador, 

Canada 

Naval Facilities Engineering Service Center 

Port Hueneme, California 

NOAA/PMEL 

Seattle, Washington 

Noblis 

Falls Church, Virginia 

OceanGate, LLC 

Everett, Washington 

Oregon State University College of Oceanic and 

Atmospheric Sciences 

Corvalis, Oregon 

Society of Ieodo Research 

Jeju-City, South Korea

5565 Sterrett Place, Suite 108 

Columbia, Maryland 21044 

Postage for periodicals 

is paid at Columbia, MD, 

and additional mailing offices.

Biomimetics and Marine Technology - Marine Technology Society

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?