Optimization - based approach for motion planning of a robot walking on rough terrain

The paper presents a motion planning algorithm for a robot walking on rough terrain. The motion-planer is based on the improved RRT (Rapidly Exploring Random Tree)-Connect algorithm. The Particle Swarm Optimizati on (PSO) algorithm is proposed to solve a posture optimization problem. The steepest descent method is used to determine the postion of the robot's feet during the swing phase. The gradient descent method is used for smoothing the final path. The properties of the motion planning algorithm are presented in four cases: motion planning over a bump, concavity, step and rough terrain mocup. The maximal sizes of particular obstacle types traversable by the Messor robot with the new, optimized motion plan are given.