Abstract: An approach to neural network-based robust and adaptive dynamic surface control is proposed for supermaneuverable flight with violent changes of aero-dynamics parameters. To achieve high precision, a radical basis function neural network (RBFNN) is used to compensate for parameter uncertainties and unknown disturbance. Control laws are obtained from dynamic surface control which reduces the complexity of backstepping controller. In addition, adaptive tuning rules of neural network weights are improved to achieve good performance of transient processes. Stability analysis using the Lyapunov stability theorem shows that all closed-loop signals are bounded. Output tracking error and approximate error of neural network weights exponentially converge to small compacts. Finally, results of the Herbst maneuver simulation show that the designed control systems have good performance under stall conditions.

Key words: flight control, dynamic surface control, RBF neural network (RBFNN), supermaneuver