Motion Planning for Three-Branch Space Robotic System

Energy-efficient motion planning for a three-branch space robot under obstacle avoidance constraints holds significant application value in on-orbit services. This article proposes an innovative method that utilizes a single-layer neural network to approximate the signed distance field of the robot's links. By integrating this with the robot's kinematic chain, an obstacle avoidance and self-avoidance model for the three-branch robot is constructed. This model provides efficient and precise distance queries while ensuring the smoothness of the displacement derivatives. Subsequently, an optimal control-based motion planner is proposed to minimize energy consumption, with obstacle avoidance as a hard constraint. To validate the effectiveness of the proposed method, a three-branch robot ground experimental platform was developed, and experimental studies were conducted on it. Two sets of motion planning experiments are conducted: first, point-to-point motion planning for the three-branch robot, and second, on-orbit crawling motion planning for the three-branch robot. The experimental results show that the proposed method reduces energy consumption by 9.87% and 5.71%, respectively, compared to the benchmark method, confirming its superiority in terms of both obstacle avoidance and energy optimization.