小天体柔性着陆器姿轨耦合智能控制

A method for attitude-orbit coupling intelligent control of flexible lander based on maximum entropy reinforcement learning is proposed in this paper，aiming at solve the adverse effects of the complex perturbation environment and the inaccurate flexible deformation force. Firstly，the orbital dynamics model of the equivalent agent is established by introducing the internal flexible force of the lander. The datum plane method is used to characterize the attitude of the flexible lander with complex deformation. The attitude-orbit coupling dynamic environment of the lander is constructed to train the intelligent controller. Then，an intelligent controller with deep neural network architecture is designed according to the soft actor-critic （SAC） algorithm of maximum entropy reinforcement learning theory. Each thruster can keep the lander attitude stable and track the navigation trajectory with high precision by self-adapting the output thrust. Finally，the landing process with the controller deployed is simulated. The simulation results show that compared with the classic PD control method，the intelligent control method proposed in this paper has stronger robustness.