Creating Swing-Up Patterns of an Acrobot Using Evolutionary Computation

Abstract

In recent years, the various controller design schemes have been examining for nonlinear systems with uncertainties. As one of such mechanical systems, an acrobot attract attention in last decade, which belongs to nonholonomic systems. Several controller design schemes have been proposed for the acrobot.
In this paper, a new approach to control the acrobot is considered, in which a suitable set of swing up pattern is created using the real-coded genetic algorithm(GA). The fitness function includes some terms, i.e., the time, the extra motion and the height in the inverted state. The effectiveness of the newly proposed scheme is numerically verifyed on some simulation results.