Backstepping adaptive high maneuvers flight control based on neural network

A backstepping adaptive control method based on fully tuned neural network is proposed in the presence of model nonlinearity and parameters uncertainty for high maneuvers flight. Parameter uncertainties are compensated for online by the fully tuned radical basis function (RBF) neural network. The control law and the adaptive law of neural network are recursively achieved through a backstepping method. The fixed parameters optimization is done using a chaotic particle swarm optimization algorithm with adaptive parameter strategy for achieving a good transient performance. The final control surface deflections are derived by a weighted pseudo-inverse control allocation method. Simulation results show that precise high maneuvers can be performed with fast convergence and good robustness properties in spite of large aerodynamic parameters uncertainty and unknown control gain matrix. Moreover, the estimation errors of neural networks' parameters are remained in compact sets.