ADAPTIVE ROBUST PREDICTIVE CONTROL FOR HYPERSONIC VEHICLES USING RECURRENT FUNCTIONAL LINK ARTIFICIAL NEURAL NETWORK

A novel adaptive predictive control methodology is presented for air-breathing hypersonic vehicles (AHVs) subjected to unknown dynamical disturbances and uncertainties. The control architecture is comprised of a continuous-time nonlinear generalized predictive controller, a recurrent functional link artificial neural network (RFLANN) control adjustment, and an adaptive robust control item. The RFLANN proposed by this paper, is a simple recurrent NN without hidden layers. Due to its strong ability of learning dynamic information and small computing cost, the RFLANN is utilized to approximate external disturbances and parameter uncertainties during hypersonic flight. The weights of the RFLANN are first online tuned by a derived adaptive law based on Lyapunov stability theorem, and the offline training is not necessary. By the stability analysis of the closed-loop control system, it is proven that all errors are uniformly ultimately bounded. Finally, simulation results show better performance of the controller for the AHV attitudes tracking than the methods compared, moreover, the robustness to disturbances and uncertainties are successfully accomplished.