Abstract:

A novel nonlinear inversion dynamic neurocontrol method is proposed, which is successfully applied to a fighter nonlinear decoupling control problem. The method abolishes the conventional design way that directly gets the nonlinear inversion dynamic neurocontrol law by training the neurocontroller. The basic idea of the method is that the controlled system should be configured into a pseudo linear system with neural networks as nonlinear inversion dynamic compensators, and then, by means of modern control system synthesis methods such as linear quadratic regulator, the pseudo linear system should be configured into a loop closed system with the ideal system performance. The simulation results of a fighter nonlinear inversion dynamic decoupling neurocontrol system show the potential of this neural networks as a nonlinear inversion dynamic controller.

Key words: flight control, nonlinear, neurons-control