Meta-learning-based Fault-Tolerant Attitude Control of Hypersonic Flight Vehicle with Input Constraints

Xiaoxiang Hu,Kejun Dong,Changhua Hu,Bing Xiao,Xiaosheng Si
DOI: https://doi.org/10.1007/s11071-024-10209-6
IF: 5.741
2024-01-01
Nonlinear Dynamics
Abstract:In this study, a meta-learning-based fault-tolerant control (MLFTC) strategy is proposed for the accurate attitude tracking control of hypersonic flight vehicle (HFV) with actuator fault and input constraints. The proposed MLFTC combines the advantage of integral reinforcement learning (IRL) algorithm and meta-learning, can greatly reduce the calculation amount of IRL. By recalling the control purpose of HFV’s attitude system, a tracking error system is derived and the control objective is obtained. Then a neural networks-based IRL algorithm is developed to solve the Hamilton-Jacobi-Bellman equation of the tracking error system, and an approximate optimal control law is directly derived. With considering the actuator fault of HFV, a fault compensator is adopted to achieve optimal fault tolerant control law. The meta-learning ideas are also adopted to address the unknown and sudden faults of HFV, and the convergence speed of the proposed IRL-based FTC law can improved by meta-learning. The ultimately uniformly bounded of the proposed MLFTC is proven by Lyapunov method. Finally, simulation results show that the proposed MLFTC can achieve accurate attitude tracking control for HFV with different actuator faults.
What problem does this paper attempt to address?