Adaptive Neural Network Inverse Controller for General Aviation Safety

Abstract

An advanced flight control system developed for and demonstrated to compensate for unanticipated failures in military aircraft is investigated for use in general aviation. This method uses inverse control to decouple the flight controls and modify the handling qualities of the aircraft. The purpose of the system is to render a general aviation aircraft easier to fly by decoupling its flight control system and making the aircraft handling more natural to a nonpilot. Artificial neural networks are used to counteract the modeling errors in the inverse controller but, more important, to adapt to unanticipated failures during flight, thus, allowing the pilot to continue to control the aircraft safely. Because a decoupled flight control system is software based, it is a fly-by-wire system. For such a system, it is difficult from a cost standpoint for general aviation to incorporate the level of redundancy required in such flight control systems; therefore, the demonstration of this system's capability to handle control system failures is critical to future certification efforts. The system is verified with MATLAB® simulations for longitudinal flight. In simulations, the control system is shown to be able to track pilot velocity and pitch angle and flight-path angle commands. Simulations of changing configurations, payload, and partial control system failures have shown that the controller does rapidly adapt to these changes without a need for a pilot response. A pilot-in-the-loop flight simulator has verified the MATLAB simulations and ongoing work to flight test the control system on a Raytheon Bonanza F33C fly-by-wire testbed is discussed.