ANN Control of Non-Linear and Unstable System and its Implementation on Inverted Pendulum

Abstract

The inverted pendulum is a classical problem for control problem, therefore it is used to benchmark new control techniques This process is best suited for control engineers working on developing new controllers as it is highly nonlinear and unstable. The inability of normal controllers like PID to adapt to changes in the process forces our hand in developing controller with intelligence. Controllers with intelligence means the ability of the controller to react or adapt to uncertainty in the unknown process. Neural Network Controller has the ability to learn and also ability to tolerate incorrect or noisy data. Moreover developing a controller using Adaptive Linear Element (ADALINE) and Radial Basis Function based (RBF) based controllers doesn’t require mathematical modelling of the system. The main aim of the project is to keep the pendulum erect. The use of Neural Network based controller reduces the error. The Adaptive Neural Network toolbox is used to do a comparative study of the two types of controllers.