Motion planning of space robot obstacle avoidance based on DDPG algorithm

In order to solve the problem of unstructured environment and complex operation task of space robot, this paper use DDPG algorithm which is data-driven and model free in the obstacle avoidance motion planning of 7-DOF space robot of free-floating base. Aiming at the difficulty and the low speed of the convergence of DDPG algorithm, artificial guidance term is added on the basis of traditional sparse reward function. This paper proposes an improved method of reward function based on task decomposition architecture, in which the weight coefficients of obstacle avoidance guidance term and target guidance term are dynamically adjusted. Simulation results show that the proposed method effectively improves the convergence performance of the algorithm, and enables the space robot to better complete the task of avoiding collision and approaching the target point.