Neural Networks-based Multiple Model Control of a Class of Nonlinear Systems with Unknown Parameters

Abstract

A novel nonlinear system adaptive control method based on neural networks is proposed for a class of nonlinear discrete-time systems with unknown parameters. The nonlinear dynamics are first represented by the linear part and the nonlinear part. For the linear part, several fixed models are established by the localization method. At the same time, in order to improve the control quality and accelerate the convergence of the system parameters, two adaptive models are introduced. For the nonlinear part, its model can be set up by a neural network. Then, robust adaptive controllers are designed based on the fixed model, adaptive model and nonlinear model. In practice, the sub-model which is most suitable for the system is selected according to the switching rule, and the corresponding control law is implemented. Finally, the simulation results show that the proposed method can effectively improve the transient performance of the system.