Soft Computing Based Adaptive Error Optimisation for Control of Nonlinear System

Abstract

This paper elaborates a novel hybrid learning approach for training error optimisation and control of highly dynamic triple-link inverted pendulum on cart. The study demonstrates a relationship between shape and number of membership functions (MFs) of both linear and constant type to determine training error tolerance of ANFIS controller. The results are plotted which clearly highlighted supremacy of constant type three triangular shape MFs. Mathematical model and simulink of proposed system has also been analysed. The learning ability and designing methodology of adaptive networks and robustness of PID controllers are briefly described. Finally, the study illustrates an offline mode comparison of PID based ANFIS and Neural controllers in terms of settling time, steady state error and overshoot.