Safe Proximity to Space Target Based on DDPG Algorithm

Safe proximity to space target has always been the forefront problem in utilizing space. Deep reinforcement learning algorithms have achieved substantial developments in decision making. Integrating deep reinforcement learning algorithms and this typical space task scenario can make the spacecraft smarter and more suitable to the uncertain environment. In this paper, DDPG algorithm is used to instruct the service spacecraft to approach the expected region around the target spacecraft safely. Assistant reward functions are designed to deal with the sparse reward problem. Experiment has been simulated to verify the feasibility of the algorithm.