中文摘要：

针对任务约束下的航天器姿态控制系统（ACS）在轨重构问题,提出了一种基于自适应动态规划（ADP）的在轨重构算法。首先,综合考虑航天器在轨任务约束条件设计效用函数和性能指标函数,获得离散Hamilton-Jacobian-Bellman（HJB）方程形式的最优重构策略。其次,采用执行依赖启发式动态规划（ADHDP）方法近似求解HJB方程,避免了直接求解HJB方程的＂维数灾难＂问题。通过简化设计ADHDP执行网络,提高了迭代训练速度;同时在训练中引入ε-greedy因子,避免了训练算法过早陷入局部最优解。仿真结果验证了所提方法的有效性。

英文摘要：

An on-orbit reconfiguration algorithm based on adaptive dynamic programming（ ADP） is proposed to reconfigurate the controller of a spacecraft attitude control system（ ACS） in case of fault with the consideration of the ongoing missions. Firstly,a utility function and a performance index function regarding the scheduled spacecraft mission are designed,and the optimal reconfiguration policy is obtained as a solution of the Hamilton-Jacobian-Bellman（ HJB）equation. Secondly,the action-dependent heuristic dynamic programming（ ADHDP） approach is used to approximately solve the HJB equation,avoiding the curse of the dimensionality of directly solving HJB. Finally,simplifying the design of the ADHDP action network is conducted to make the iterative training process faster,and the ε-greedy factor is introduced in the training process to avoid getting in local minimum too early. Simulation results show the effectiveness of the proposed method.