Predefined-Time Fault-Tolerant Control for Spacecraft Attitude Systems with Learning Observer

Abstract

This paper focus on a novel observer-based fault-tolerant control method engineered for spacecraft attitude control systems subject to actuator faults, external disturbances, and model uncertainties. Firstly, by combining a traditional iterative learning observer with a PD control strategy, a PD-type learning observer is proposed to reconstruct the composite disturbance without requiring prior knowledge of disturbance boundaries. Subsequently, an RBF neural network is employed for model uncertainty approximation. Leveraging these derived estimations, a predefined-time fault-tolerant controller is developed, guaranteeing that the attitude tracks the command state within a specified time. The stability properties of the closed-loop faulty system are rigorously validated through Lyapunov analysis, and the superiority of the formulated control architecture is demonstrated via numerical simulations.