The Design of Neural Network Controller of a Class of Nonlinear Systems with Actuator Saturation

Abstract

The problem of actuator saturation appears in many practical control systems. If the controller is designed only with conventional linearly techniques, the presence of saturation can debase the performance even lead the closed-loop system to an unstable behavior. In this paper, neural net-based actuator saturation compensation scheme with on-line weights tuning law for the nonlinear systems in Brnovsky form is presented to decrease the influence of saturation. In this scheme, RBF neural network is adopted to approximate the part exceeding the saturation limit of controller's output. Another most prominent feature of the scheme is which can ensure the system is uniformly ultimately bounded which is proved by Lyapunov theory, and considering the network reconstruction error and the system's external disturbance. The tracking error can be freely adjusted by known form. The simulation example is given to illustrate the effectiveness of this method.