Adaptive Neurocontroller for a Nonlinear Combat Aircraft Model

This paper introduces an adaptive controller, based on neural networks use, for a nonlinear six-degrees-of-freedom combat aircraft model. This controller is based on the determination of the inverse dynamics of aircraft through a state feedback, taking advantage of the neural network online learning ability in dealing with any changes of the aircraft dynamics during the flight. By comparing the online and offline training, how effective the neural controller is in adaptation is investigated and highlighted in situations involving highly demanding maneuvers as well as sudden environmental disturbances. The neural controller is designed according to the reference model adaptive direct inverse scheme. The behavior of this controller is compared with that of a conventional linear stability and control augmentation system (normal acceleration limiter), implemented under military handling qualities and high maneuverability requirements. The online training of the nonlinear neural controller is based on a recursive prediction error algorithm, whose performance results from a proportional derivative performance index formulation. The stability analysis demonstrates how the extra degree of freedom, provided by the derivative term, makes the algorithm more robust than the standard recursive least-squares method. Performance is verified through numerical simulations.