Nonlinear system control using a self-organizing functional-linked neuro-fuzzy network

This study presents a self-organizing functional-linked neuro-fuzzy network (SFNN) for a nonlinear system controller design. An online learning algorithm, which consists of structure learning and parameter learning of a SFNN, is presented. The structure learning is designed to determine the number of fuzzy rules and the parameter learning is designed to adjust the parameters of membership function and corresponding weights. Thus, an adaptive self-organizing functional-linked neuro-fuzzy control (ASFNC) system, which is composed of a computation controller and a robust compensator, is proposed. In the computation controller, a SFNN observer is utilized to approximate the system dynamic and the robust compensator is designed to eliminate the effect of the approximation error introduced by the SFNN observer upon the system stability. Finally, to show the effectiveness of the proposed ASFNC system, it is applied to a chaotic system. The simulation results demonstrate that favorable control performance can be achieved by the proposed ASFNC scheme without any knowledge of the control plants and without requiring preliminary offline tuning of the SFNN observer.