Adaptive Control Research On Nonlinear System Based On Neural Network

Neural network adaptive control is proposed combining adaptive control's advantages with neural network's characters and provides a new method for nonlinear control.Based on Lyapunov stability theorem and neural network direct adaptive control idea, the control problem of a class of uncertain nonlinear system with disturbance is researched. The controller is designed arming at two kinds of uncertainties existing in nonlinear system--the unknown functions and the uncertain disturbance. In controller, the radial basis function neural network is used as approximation model for the unknown functions, and nonlinear damping term is used to counteract the disturbances, so neural network's better approximation capabilities can be utilized richly and the modeling difficulties can be avoided. Meanwhile, the controlled system's stability and convergence can be guaranteed under some assumptions. At last, the program is designed to verify the effectiveness of the controller. In presented programs, Guassian function is used as basis function, Simulation results show that the boundness effects of weights and control input are better.The rough framework of this thesis is as following: the artificial neural network and neural network control are introduced in detail from the first chapter to the third chapter; the radial basis function neural network is described, and its approximation performance is simulated in the fourth chapter; the researched system's controller is designed and tested in the fifth chapter; the development of neural network control is expected and the further research prospect is proposed in the end words.