Neural control for longitudinal dynamics of hypersonic aircraft

Abstract

This paper investigated the discrete adaptive controller with neural network for the longitudinal dynamics of a generic hypersonic flight vehicle. Based on functional decomposition, we design the controller for the altitude subsystem and the velocity subsystem separately. The altitude subsystem is transformed into the explicit 4-step ahead prediction model with four 1-step ahead prediction subsequences. The control design is based on the state feedback and neural approximation. For each subsystem only one neural network is employed to approximate the lumped system uncertainty. The controller is considerably simpler than the ones based on back-stepping scheme. The velocity subsystem is transformed into the output feedback form and the indirect discrete NN controller is applied. The semiglobal uniform ultimate boundedness stability and the output tracking error are made within a neighborhood of zero. The simulation is presented to show the effectiveness of the proposed control approach.