基于强化学习的燃料最优卫星编队重构控制方法

In the context of orbital rendezvous problems with environmental perturbations and thrust limitations，a fuel-optimal orbital rendezvous control method based on reinforcement learning is used to address issues of high computational complexity and sensitivity to initial values in traditional algorithms. This method can be employed for the initialization and reconfiguration of low Earth orbit satellite formations. The method describes the relative motion between satellites using relative orbital elements and establishes multi-impulse and continuous-thrust control models considering J₂perturbation. By discretizing the constrained optimization problem into a sequential decision process，a soft actor-critic （SAC）reinforcement learning algorithm is used to train the agent for generating fuel-optimal rendezvous trajectories，and an analytical method is employed to correct terminal orbit deviations. Meanwhile，a Monte Carlo tree search （MCTS） algorithm is integrated to optimize the formation reconfiguration strategy，selecting the target orbital positions of member satellites to reduce total fuel consumption. Simulation results show that the proposed method can accurately and efficiently generate fuel-optimal rendezvous and formation reconfiguration schemes，with computational time and performance metrics superior to traditional optimization methods. It also demonstrates good generality and robustness，providing an effective solution for large-scale satellite formation mission analysis.