Control of a sliding inverted pendulum using a neural network

Abstract

It is well known that neural networks can be used to control nonlinear unstable systems. In this paper a rule–based neural controller is proposed to balance an inverted pendulum, which is a classic example of an inherently nonlinear unstable system. The control objective is to swing up the pendulum from the stable position to the unstable position, and bring its slider back to the origin of the track. The overall on–line control algorithm is decomposed into three separate neural control areas based upon the angular and velocity values of the pendulum. The training process is performed using an on–line back propagation learning algorithm without the off–line learning process. The experimental results show that this neural controller is able to swing up and balance the inverted pendulum and guide its slider to the centre of the track. It also has the robustness to balance the inverted pendulum in suffering an external impact acting on the pendulum.