Composite learning control of hypersonic flight dynamics without back-stepping

Abstract

In this paper, composite neural control is proposed for hypersonic flight control in presence of unknown dynamics. Using high gain observer (HGO), the controller of attitude subsystem is designed without back-stepping. This strategy simplifies the process of controller design and reduces the computation burden of parameter updating. To construct the composite neural controller, the filtered modeling error is further considered in the weight updating of RBF NN. Moreover, the composite neural controller can achieve the fast learning of system uncertainty. Simulation is presented to demonstrate the effectiveness of the design.