Low Cost Fault Tolerant Flight Controls for UAV's Using Neural Networks, System Identification & Robust Control

Modern fligt control systems rely on hardware redundancy to overcome failures of sensors and actuators. This is an expensive approach which does not make use of analytical redundancy present in a flight control system. Analytical redundancy is achieved by comparison of sensor outputs over time and across sensors using physical and identified models. We propose here methods to detect failures using analytic redundancy and to compensate for these failures by reconfiguration of the control systems. The proposed innovation for UAV reconfigurable flight controls builds on the previous work by Scientific Systems and other companies under the sponsorship of AF on the Self-Repairing Control Systems for F-15 and integrates recent developments in the fields of Systems Identification, Robust Control and Neural Networks. Phase I will involve the following major tasks: (i) failure detection and identification using Neural Networks and on-line parameter identification (ii) reconfigurable controller design using robustcontrol theory and (iii) implementation and testing using Bell's Eagle Eye UAV simulation. Phase II will involve hardware-in-the-loop real time simulation or flight testing of the proposed system on a UAV. Bell Helicopter Textron, a manufacturer of Eagle Eye UAV will provide data and evaluation support during Phases I & II and commercialize the results during Phase III. Anticipated Military Benefits/Potential Commercial Applications of the Research or Development: The technologies of low cost fault detection, identification and reconfiguration have applications in Electric Power Systems, IVHS, Civil Aviation, Automated Manufacturing, telecommunications and Process Control.