Robust neural adaptive control of a hypersonic aircraft

Abstract

Design of a robust Multi Input Multi Output (MIMO) neural adaptive controller for the longitudinal dynamics of a generic hypersonic aircraft is presented. This aircraft model is highly nonlinear, multivariable, unstable and is assumed to have unknown dynamics. The unknown dynamics are approximated by single layer neural networks. A main feature of the proposed adaptive controller is that it bypasses the stabilizability problem frequently encounted with an adaptive controller based on inverse dynamics, where the estimated model can become uncontrollable. The tracking error at steady state can be guaranteed to converge to inside of a small residue set whose size can be determined by following a design procedure. Simulation studies were conducted for trimmed cruise conditions of 110,000 ft and Mach 15 where the responses of the vehicle to a step change in altitude and airspeed were evaluated. The commands were 80ft/s step-velocity and 100ft step-altitude. The simulation studies demonstrated that the performance requirements were met. © 2003 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.