Feedback Control of an Inverted Pendulum by Use of Artificial Intelligence

Abstract

An inverted pendulum represents an unstable system which is excellent for demonstrating the use of feedback control with different kinds of control strategies. In this work feedback control of the inverted pendulum is examined from the point of view of a state space model of the system. First a pole placement algorithm is thoroughly explored. After this artificial intelligence (AI) methods are investigated to better cope with the nonlinearities of the physical model. The technique used is based on a hybrid system combining a Neural Network (NN) with a Genetic Algorithm (GA). The NN controller is trained by the GA against the behaviour of the physical model. The results of the training process show that the chromosome population tends to stabilize at a suboptimal level, and that changes in the environmental parameters have to take place to reach a new optimal level. The results, however, show that the NN controller gradually adapts to the behaviour of the pendulum.