REINFORCEMENT FUZZY-RULE EMULATED NETWORKS AND ITS APPLICATION ON NONLINEAR DISCRETE-TIME SYSTEMS

Abstract

The adaptive control algorithm based on Multi-Input Fuzzy Rules Emulated Networks (MIFRENs) with the reinforcement learning algorithm is introduced for a class of nonlinear discrete-time systems. Because of the unknown future values of systems, the long term cost function is estimated by the first MIFREN through the human knowledge with the defined IF-THEN rules and the proposed learning algorithm. The main controller is constructed by another MIFREN and the parameters in side this network have been tuned to minimize the estimated cost function and the control system error. All designed parameters are given with the Lyapunov method and the proposed theorem. The numerical simulation results are demonstrated the system performance with the selected nonlinear discrete-time systems