Abstract: A design method of flight control systems with neural networks is presented,by which the designed flight neurocontrol system possesses a strong robustness and can achieve the ideal system performance. The controlled flight system has a nonlinear state feedback configuration and an optimal system performance specification. A self learning optimal control algorithm is proposed to solve the optimal control problem and get training samples of the optimal control law, and an improved backpropagation algorithm is employed to train the neurocontroller in the whole flight envelope. The method is implemented into a fighter nonlinear simulation. By self learning and being trained, the fighter controlled by the neurocontroller has achieved the optimal system performance in the whole flight envelope. The simulation results show that neural networks used as flight controllers are full of promise.

Key words: flight control, optimal control, neural nets