Fixed-Time Neural Network-Based Dynamic Surface Control for Hypersonic Flight Vehicle With Historical Data Online Learning

Abstract

The motivation of this article is to solve the tracking control problem of hypersonic flight vehicle (HFV) with uncertainties. To this end, a fixed-time neural network (NN)-based dynamic surface control scheme is proposed. First, a historical data online learning NN is designed to handle the matched uncertainty. For the proposed NN, a fixed-time auxiliary system is utilized to introduce historical information into the update law of NN weights. This design not only improves the data utilization of the NN but also enhances the estimation performance. Then, based on the reconstructed information of NN, a fixed-time dynamic surface controller is proposed, in which a fixed-time filter is used to estimate the derivative of the virtual control input and unmatched uncertainty in the HFV system. Compared with existing results, the proposed method ensures that the tracking error converges within a fixed time while having a smaller computational burden. These properties are of great importance for the practical HFV application. Finally, the effectiveness of the proposed fixed-time control strategy is verified by a numerical simulation example.