An Improved Approach of Neural Network to Non-Linear Inverted Pendulum System

Abstract

This paper introduced a best approach techniques for control the non linearity of the inverted pendulum (IP) by using Artificial Neural Network (ANN). In A Artificial Neural network type is Feed-Forward Network (FFN).Training is done by “Trainlm” network function that is use to updates weight and bias states according to Levenberg Marquardt (LM) back-propagation algorithm. This technique is focused on how to solve non linear behaviour the of “parameter modification” such as angle and position of inverted pendulum. We use controllers(like conventional controller) to compare their response with Neural network controller to find their linear behaviour with improve accuracy and precious time by using MATALAB simulation. The method is applied to the Nonlinear IP model with the application of some precariousness, and the experimental results show that the system responds very well to handle those precariousness.