Spacecraft Attitude Maneuver Planning Based on Deep Reinforcement Learning under Complex Constraints

During the execution process of the spacecraft mission, a large number of attitude maneuvers are required. When the external environment and the spacecraft's dynamic constraints are coupled, it will cause certain difficulties in solving the spacecraft's attitude maneuver path planning. Therefore, the spacecraft attitude maneuver planning must be improved. However, traditional methods often have the problems of low solution efficiency and strong model dependence. We propose the Deep Q-network (DQN) method to solve these problems and directly perform end-to-end inference. Finally, by comparing with the Rapid-exploration Random Tree (RRT) algorithm, the results show that the planning results meet all constraints and energy consumption of the DQN method is reduced by about 62% compared to the RRT method, and the planning time is also reduced by about 81 %.