3D Humanoid Robot Multi-gait Switching and Optimization

Abstract

In RoboCup3D humanoid robot competition, players need to have a better offensive and defensive ability, efficient walking is the basis of collaboration and confrontation. RoboCup3D is a highly simulated complex environment, for different real-time forms of robots need to switch different walking modes, and the speed characteristics of different walking modes and the stability of switching will play an important role in the final competition results. For different interference degree of environment, adopt the method of reinforcement learning to optimize different gait parameters, to ensure the fastest under different disturbance degree walking speed, and use the reduction strategy based on PID control to realize smooth switching of gait can further enhance the efficiency of the walking robot, the intelligence that has more initiative in complex environment. Experimental results demonstrate the effectiveness of this method.