Neural Adaptive Finite-Time Sliding Mode Controller for Air-Breathing Hypersonic Vehicle

A neural adaptive sliding mode controller (NASMC) composed of an adaptive finite-time sliding mode controller (AFSMC) with the long short-term memory (LSTM)-based deep recurrent neural network is presented for air-breathing hypersonic vehicle (AHV) subject to difficulties of control system design including tight couplings between propulsion and aerodynamics, strong nonlinear, static instability, harsh flight conditions and parametric uncertainties. Firstly, the longitudinal nonlinear model of AHV is processed applying input/output feedback linearization method to transform the model into an affine nonlinear form. Secondly, the AFSMC is composed of a non-singular fast terminal sliding surface (NFTS) and a fast adaptive super-twisting reaching law (FAST). The NFTS is proposed in order to accelerate convergent speed. Meanwhile, the FAST is employed as a reaching law, alleviating chattering phenomenon. Strict proofs are given using Lyapunov theory for AFSMC, which demonstrates that the closed-loop system can reach stable state in finite time. Thirdly, the LSTM is utilized to approximate and adjust uncertain parameters online so as to enhance global robustness automatically and decrease tracking error. Finally, the simulation results of AHV illustrate the superiority and effectiveness of the proposed NASMC.