Switching control of a rotary inverted pendulum

In this paper we consider the problem of swinging-up and stabilization of a rotary inverted pendulum. We design switching controllers and analyse whether the performance can be improved using more controllers. We optimize both the control gains and the switching points. All the controllers are tested experimentally on a rotary inverted pendulum. Our results indicate that having a large number of switching controllers does not necessarily lead to a performance increase.

I. Introduction

For linear models there are many methods to design con-trollers. Therefore we usually linearize the nonlinear model about an operating point and then we design the controller for the linear model. The problem with this approach is that when we apply the controller on the nonlinear model it will work only in a small domain. A possible solution for this problem is to linearize the model in several points, design several controllers and then to combine them. These type of controllers are called switching controllers. In this paper we analyze whether having more switching points and controllers leads to better performance, or there is a maximum number of switching points and controllers over which we can not obtain better performance.