Fixed-Time Fault-Tolerant Optimal Attitude Control of Spacecraft With Performance Constraint via Reinforcement Learning

The attitude stabilization control problem of spacecraft with actuator fault, external disturbance, and performance constraint is studied via reinforcement learning (RL). The attitude stabilization error constrained by prescribed performance is first transformed into an unconstrained variable. Unlike the existing optimal controllers ensuring uniformly ultimately bounded stability, an RL-based fixed-time optimal control framework is then proposed. In this control framework, a neural network (NN) weight updating law with the persistent excitation condition eliminated is designed. Moreover, a fixed-time estimator is developed and added into the classical RL-based optimal controller to synthesize a fixed-time fault-tolerant controller. The closed-loop system and the estimation errors of the NN weights are stabilized within fixed time. The control cost is also significantly reduced. The effectiveness of the control policy is finally examined through numerical simulation.