An Autonomus Obstacle Avoidance Method for Finite-thrust Spacecraft

This paper outlines an online autonomous decision-making method for obstacle avoidance planning with finite-thrust spacecraft. Based on the analysis of rapid orbit maneuver method in small range for spacecraft, this paper focus on decision-making and control of maneuvering opportunity which is in the process of on-orbit autonomous obstacle avoidance for a finite-thrust spacecraft. Reinforcement learning theory is applied to find the change rules of maneuvering opportunity and motion state during obstacle avoidance process. An autonomous obstacle avoidance decision-making training model for space vehicle is established, which is based on "offline learning and online decision-making" frame. Study on the typical parameters that affect orbital maneuver, and a reinforcement learning evaluation mechanism is constructed with time as reward function parameter. The method performs energy optimal small-scale orbital maneuver planning. As compared to the finite thrust trajectory planning with traditional Gauss pseudo spectral method, this approach is better in solving speed and operation performance with simulation case studies.