Novel adaptive neuro-fuzzy control scheme for eliminating rule explosion in dynamic systems

This article presents a neuro-fuzzy adaptive control scheme for avoiding the problem of rule explosion inherently associated with fuzzy control systems. The study proposes a novel neuro-fuzzy model designed using only two membership functions and four if-then fuzzy rules for control of highly nonlinear two-stage cart and pendulum system. The neuro-fuzzy controller has been designed with the minimum possible number of rules thus solving the problem of rule explosion completely. The study further compares proposed ANFIS controller with conventional proportional-integral-derivative (PID) and artificial neural network controllers in terms of settling time, overshoot ranges and steady state error. The results show better performance of ANFIS controller compared to PID and neural controllers.