The Motion Planning And Cooperation Mechanism Of Biped Robots In Robocup3D Simulation Environment

RoboCup3D robot soccer simulation competition has served as an excellent research domain forApollo3D—our team of Nanjing University of Posts and Telecommunications, and made anin-depth analysis on the individual action design and team coordination mechanism. In order tooptimize and improve the offensive and defensive performance of Apollo3D, this paper presented anew set of method on the robot walking, kicking and multi-agent cooperation. The main work andinnovations were as follows:First, this paper proposed a closed-loop control method of an omnidirectional walking model ofhumanoid robots which based on Cerebellar Mode Articulation Controller (CMAC). The cubicspline interpolation method were used to plan the foot space trajectory, and built the Double LinearPendulum (D-LIP). Subsequently, the trunk and foot poses were applied to compute the currentjoint angle value of the robot legs. Meanwhile, the sensor information according to the CMACmodel can be modified the angle value in the Inverse Kinematics to achieve a steady and fastomnidirectional walk.Secondly, aiming at the kicking design of humanoid robots, this paper presented a gradualaccumulation learning method which based on the Covariance Matrix Adaptation EvolutionStrategy (CMA-ES). By planning the best kicking point and the foot space motion trajectory, thefirst layer of learning optimization can be realized using the linear distance after kicking and thetime cost about kicking point as the target. Afterwards, the double balancing mechanism of therobot’s center of the mass and the gyroscope sensor feedback was employed to fulfill theoptimization of the next layer. The ultimate learning goal was the overall consideration of thefootball contact point selection, the weighted penalty of the ankle joint and the performance ofkicking. In general, Apollo3D adopted the above-mentioned kicking mechanism, not only keepingthe stability of the robot but also ensuring the kicking distance and angle.Finally, in order to perform the team collaboration strategy in the RoboCup3D multi-agentsystem, this paper described a hierarchical decision mechanism based on a role assignment function,a vote communication system and the ideal of behavior prediction model (IOBPM). The selectionof the Center Forward (CF) and the other roles successively, while establishing the IOBPM for theCF dribbling decision module, at last the dynamic planning algorithm was adopted to optimize thehigh dimensionality problem caused by the role assignment in order to ensure the fluency of the roles rotation with the continuous changing of the football positions.After the constant competition and experiment validation, the team Apollo3D demonstrated theabove methods have significant results whether in the action design and optimization of a singlerobot, or the continuity of the team collaboration.