Meta-learning-based fault-tolerant controller for aircraft attitude control system

In this study, for aircraft system with actuator fault, a meta-learning-based fault-tolerant control scheme is designed by combining integral reinforcement learning algorithm and fault compensator. For a fault-free system, a neural networks-based integral reinforcement learning algorithm is first developed to solve the Hamilton-Jacobi-Bellman equation, which could directly derive an approximate optimal control law. Then, a fault compensator synthesized into the optimal control law achieve fault-tolerant control of an aircraft system. The meta-learning concept is also adopted to address the unknown and sudden faults of the aircraft system, which can improve the convergence speed of the system state and the practical applicability of the designed fault-tolerant controller. It is proven that the closed-loop system is ultimately uniformly bounded by Lyapunov method. Finally, a simulation example is conducted to verify the proposed scheme's effectiveness.