The invention provides a spacecraft attitude planning method under complex constraints based on deep reinforcement learning in order to solve the problem of spacecraft attitude maneuver planning under prohibition pointing constraints. According to the method, the shortest rotation path, the optimal energy, the forbidden pointing constraint, the dynamic constraint and the like are comprehensively considered, a simple multi-constraint model is obtained through deep reinforcement learning training, the approximate optimal path of attitude maneuver can be determined, and the method is suitable for control over micro spacecrafts. The trained model can directly carry out end-to-end reasoning on initial and target attitudes, and compared with a traditional method which is low in solving efficiency and high in complexity and precision model degree, the method has the advantages of being relatively efficient and simple in solving process and high in generalization performance. A simulation result shows that a planning result meets all constraints, including four prohibition pointing constraints, dynamic constraints and the like.
本发明针对禁止指向约束下的航天器姿态机动规划问题,提出了一种基于深度强化学习的复杂约束下航天器姿态规划方法。本发明方法综合考虑旋转路径最短、能量最优、禁止指向约束、动力学约束等,通过深度强化学习训练得出了一个简洁的多重约束模型,可以确定姿态机动的近似最优路径,适用于微型航天器的控制。训练好的模型可以直接对起始、目标姿态进行端到端的推理,相较于传统方法的低求解效率,以及高复杂精确模型度,本发明具有相对高效且简单的求解过程以及较强的泛化性能。仿真结果表明规划结果满足所有约束,包括4个禁止指向约束、动力学约束等等。
Spacecraft attitude planning method under complex constraint based on deep reinforcement learning
基于深度强化学习的复杂约束下航天器姿态规划方法
2023-08-15
Patent
Elektronische Ressource
Chinesisch
IPC: | B64G Raumfahrt , COSMONAUTICS |
Spacecraft attitude redirection control method based on potential function under complex constraint
Europäisches Patentamt | 2023
|REINFORCEMENT LEARNING FOR SPACECRAFT ATTITUDE CONTROL
TIBKAT | 2020
|Failure spacecraft attitude adjustment method based on reinforcement learning
Europäisches Patentamt | 2023
|