Bidirectional parallel Monte Carlo tree search gait planning for hexapod robot
To solve the gait planning problem of the hexapod robot in sparse foothold terrain,and improve the planning time efficiency,passing ability,arrival accuracy and motion speed,a bidirectional parallel Monte Carlo tree search algorithm(BPMCTS)is proposed.The gait planning problem is transformed into a Markov sequence optimization process.A bidirectional parallel extended Monte Carlo tree structure is constructed to search for the best base position and form gait sequences.In the simulation phase,the deep-root parallelization simulation method is adopted to improve the convergence speed of the algorithm.The encounter evaluation index is introduced in the reward evaluation mechanism to enhance the orientation of the algorithm.The results of simulation experiments show that the planning time efficiency increases by 46.9％of the proposed algorithm,the passing ability increases by 7.7％,the arrival accuracy increases by 32.6％and the motion speed increases by 16.8％of the robot,which verifies the feasibility and superiority of the proposed algorithm.
hexapod robotgait planningreinforcement learningMonte Carlo tree search
国家科技期刊平台
NSTL
万方数据
