Design and Implementation of a Robot Soccer System - ResearchGate

Proceedings of 2005 CACS Automatic Control ConferenceTainan, Taiwan, Nov 18-19, 20052. Omni-directional Movement ControlMechanismIn the robot mechanical design, as shown in Fig. 3, anomni-directional movement mechanism with fouromni-directional wheels is constructed. The systemarchitecture is shown in Fig. 4, four omni-directionalwheels in four directions perpendicularly of sides offront, back, left and right. In this way, the controlmethod can be more easily designed and the moreacceleration can be obtained when it moves forward orbackward. Two shock absorbers, as shown in Fig. 5,are installed on each wheel so that the robot can movesmoothly and avoid idle running caused of ruggedground.Let the mobile robot be rigid moving on the playfield. As shown in Fig. 6, it is assumed that themoving coordinate system O:xmymis fixed on themobile robot. The omni-directional wheels aremounted on robot symmetrically such that the anglebetween each wheel and the xm- and ym-coordinates is . The dynamic equations of4these four wheels [2-5] can be described byv 1 R sin1 v 2 R sin2 v 3 R sin3 v R4 sin4wherei coscoscoscos Lxm L (1)ymL Lv and iare respectively the driving forceand the rotational velocity of the i-th wheel;xmandymare the velocity of the center of the robot in themoving coordinate system; is the vehiclerotational velocity; L is the distance from each to thecenter of the omni-directional mobile robot O ; R isthe radius of the omni-directional wheel. Theanti-clockwise rotational direction of the robot ispositive. The clockwise rotational direction is negative.The velocity of the omni-directional wheel is obtainedfrom Equation (1). The velocity of the wheels isprovided from the move vector of robot and the anglebetween the absolute coordinate and movingcoordinate to provide. The DC motor with the velocityfeedback is equipped on the robot. The feedbacksignal of the motors is able to be used to adjust thevelocity of the wheel.3. Omni-directional Vision SystemIn order to let the robot capture objects quickly, anomni-directional vision system [6-7] is constructed inthe soccer robot so that two goals and four cornercylinders can be recognized to estimate the location ofthe robot. A camera and a hyperbolic mirror are usedto the environmental image around the robot. Itsperformance is 640 pixels in width and 480 pixels inheight per frame and 30 frames per second. A highspeed notebook (ASUS M300N/NP) is used to be theimage processing unit. The robot can shareinformation to other team members by UDPcomponent.In the image processing, we compare the imageand the color model to find out the objects in the gamefield. A cluster index method is proposed to model theassigned colors of objects (ball, goals, cornercylinders) so that these objects can be find out fromthe captured image. Three methods: (a) color modeladjust automatically method, (b) object shape analysismethod, and (c) object position estimation method, areproposed to enhance the image program to adapt thedynamic environment in the competition. The colormodel adjusted automatically method is proposed tofix the inaccuracy image data making from brightnessand the distance between the robot and objects. Theobject shape analysis method is proposed to recognizethe object if it is real advanced. The object positionestimation method is proposed to use the objectposition in the last state to estimate its position in thenext state when the robot or the ball move too fast orthe objects are covered or lose in the visionIn the self-localization, the (absolute or relative)location of the robot on the field can be determined bythe information of the ball, two goals, and fourcylinders obtained in the image processing. Each robotcan share its own information with the other teammember to build a virtual field by collecting all sharedinformation. Then the best strategy for the offensive ordefensive behavior can be decided by the analysis ofthe constructed virtual field.In order to obtain the location of the robot on thefield, the relative positions of objects must bedetermined in the vision. In the omni-directionalvision system, all information has problems by thehyperbolic mirror. The object information of thedistance between the robot and the objects need to bemodified. Owing to the object information of theangle is more exact than that of the distance, theinformation of the angle in the vision is used to findout the location of robot on the field. In Fig. 7, therobot on the field can recognize some objects like twogoals and two cylinders near itself. Four angles can bemade from these four objects and robot. Based on theangles and rules of triangle, we can obtain the locationof the robot on the field.4. Experiment ResultsSome experiment results are presented in this section.

Proceedings of 2005 CACS Automatic Control ConferenceTainan, Taiwan, Nov 18-19, 2005The speed control results are described in Fig. 8,where the speed command, the feedback velocities ofwheel 1, wheel 2, wheel 3, and wheel 4 are described.We can see that the rise time that from velocity 0 to100 about 1 second. An omni-directional image isshown in Fig. 9, where the ball, two goals, and fourcorner cylinders are recognized and their positioncoordinates are described on the left of the picture.Furthermore, the angle relative to the robot can beobtained.5. ConclusionsTable 1. Description of main equipments installed inthe implemented robot.NameMain ProcessorOmni-directionalVision SystemControllerProductASUS S300NVstone VS-C42N-TRAltera NIOS 1S10DC Motor Maxon 110406Motor DriverTA8429H ICIn 2004, we first developed a vision-basedautonomous soccer robot for the middle-size leagueof RoboCup. Until now, three types of middle-sizerobots had been developed. In 2005, some systemstructures and algorithms are applied andimplemented in the new robot soccer system toenhance the ability of robots in the competition. Ateam with five soccer robots will be constructed toattend the middle-size robot soccer game ofRoboCup-2006.6. References[1] URL:http://www.robocup.org[2] L. Husng, Y.S. Lim, D. Li, and C.E.L. Teoh,“Design and analysis of a four wheelomni-directional mobile robot,” InternationalConference on Autonomous Robots and Agents,Dec. 2004.[3] K.S. Byun, S.J. Kim, and J.B. Song, “Design of afour-wheeled omnidirectional mobile robot withvariable wheel arangement mechanism,”International Conference on Robot & Automation,2002.[4] M. Wada and H.H. Asada, “Design and control ofa variable footprint mechanism for holonomicomnidirectional vehicles and its application towheelchairs,” IEEE Transactions on Robotics andAutomation, vol. 15, no. 6, pp.978-989, 1999.[5] R. Holmberg and O. Khatib, “Development andcontrol of a holonomic mobile robot for mobilemanipulation tasks,” International Journal ofRobotics Research, vol.19, no.11, pp.1066-1074,2000.[6] A. Clerentin, L. Delahoche, and E. Brassart,“Cooperation between two omnidirectionalperception systems for mobile robot localization,”IEEE/RSJ International Conference on IntelligentRobots and Systems, vol.2, pp. 1499-1504, 2000.[7] B.Krose, R. Bunschoten, S. Hagen, B. Terwijn,and N. Vlasis, “Household robots look and learn:environment modeling and localization from anomnidirectional vision system,”IEEE Robotics &Automation Magazine, vol. 11, pp.45-52, 2004.Fig. 1. Field description of middle-size league ofRoboCup.Fig. 2. Photograph of the constructed two soccerrobots.Fig. 3. Photograph of the omni-directional wheel.

Proceedings of 2005 CACS Automatic Control ConferenceTainan, Taiwan, Nov 18-19, 2005Fig. 7. Description of the self-localization of a robot inthe field.Fig 4. System architecture of the implementedvision-based soccer robot.Fig. 8. Speed control results of four wheels.Fig. 5. Description of two shock absorbers installed onthe omni-directional wheel.Fig. 9. Description of the object recognition resultsin an omni-directional image.AcknowledgementFig. 6. Coordinate description of theomni-directional mobile robot.This research was supported in part by the NationalScience Council of the Republic of China undercontract NSC 93-2213-E-032-004.