ON-LINE ADAPTIVE NETWORKS APPLIED TO AIRCRAFT CONTROL

RECONFIGURABLE, FLIGHT CONTROL SYSTEMS OFFER SIGNIFICANT IMPROVEMENTS IN SURVIVABILITY, RELIABILITY, AND MAINTAINABILITY OF MILITARY AIRCRAFT AND IN THE SAFETY OF FLIGHT OPERATIONS OF BOTH MILITARY AND COMMERCIAL AIRCRAFT. HOWEVER, ON-LINE LEARNING MUST BE INCLUDED TO REALIZE THE FULL POTENTIAL OF THESE SYSTEMS BECAUSE IT IS IMPRACTICAL TO DESIGN IN ADVANCE FOR THE VARIETY AND COMBINATORIAL COMPLEXITY OF THE UNFORESEEN CONTROL EFFECTOR AND AIRFRAME IMPAIRMENTS THAT CAN OCCUR. BARRON ASSOCIATES, INC. (BAI) PROPOSES TO EXPAND UPON CURRENT DEVELOPMENTS IN ALGORITHMS FOR ON-LINE ADAPTATION OF POLYNOMIAL NETWORK CONTROLLERS TO SOLVE THE PROBLEM OF COMPENSATION FOR THESE IMPAIRMENTS. UNDER CONTRACT F33615-88-C-3615, BAI HAS ESTABLISHED THE ADVANTGES IN ACCURACY, DECISION SPEED, AND REDUCED COMPUTATION OF FIXED POLYNOMIAL NETWORKS SYNTHESIZED OFF-LINE FOR GLOBAL FAILURE DETECTION, ISOLATION, AND ESTIMATION. BAI IS ALSO APPLYING FIXED NETWORKS TO SUPERVISORY CONTROLLER CONFIGURATIONS IN WHICH CONTROL GAINS ARE ADJUSTED ON-LINE. IN THE PROPOSED WORK, BAI WILL REFINE ALGORITHMS CURRENTLY BEING DEVELOPED FOR ON-LINE ADAPTATION OF POLYNOMIAL NETWORKS, USE THESE ALGORITHMS IN PREDICTIVE MODELS OF AIRCRAFT RESPONSE AND IN SUPERVISORY CONTROLLERS, INTEGRATE THE ON-LINE ADAPTATION METHODS INTO EXISTING SYNTHESIS ALGORITHMS, AND DEMONSTRATE VIA SIMULATIONS (WITH A NONLINEAR AIRCRAFT MODEL) THE USE OF ON-LINE ADAPTIVE NETWORKS IN THE CRCA OR F-15.