Robust Study Of Uncertain Discrete System Based On Neural Network Control

With the progress of science and technology,there are more control systems which show strong uncertainty and discreteness.The classical control theory often cannot make a good performance when deal with these uncertain discrete-time systems.Intelligent control technology which is springing up in recent years may solve the control problems of complex, uncertain,multi-tasking systems,and it has become a hot area of research.Neural network(NN) control is an important method of intelligent control with a highly adaptive capacity.NN has a capability for nonlinear function approximation,and the NN control method establishes the model or controller of an uncertain system by amending connection values between neurons in the NN.Compared to the traditional control method,NN control can easily be applied to non-linear system identification and control,and has very good stability and security.Although NN control can solve many problems which are difficult for traditional control methods,but the theory study of NN control is still relatively small.In the control area,stability is very important of a control system.But current researches are often based on the continuous-time control systems,and study of stability of discrete control systems is also relatively small.In this paper,a control model is designed based on NN for a class of uncertain discrete-time systems which doesn't have a clear system structure.In the standard BP algorithm, instability occurs in the iterative process of NN coefficients,so a robust iterative algorithm is proposed to solve this problem.According to Lyapunov stability theory,the robust iterative algorithm introduces a "dead zone" to the standard BP algorithm.It avoids a result of local minimum in the iterative process which could make the control error of the entire system misconvergence.Two sets of simulation results show that compared with the standard BP algorithm,the control quality and tracking error convergence rate has been significantly improved by using robust iterative algorithm.