Adaptive control of a class of nonlinear systems using Support Vector Regression

Abstract

In this paper we demonstrate an improvement in the transient tracking performance when a Support Vector Regression (SVR) is used to identify a nonlinear ARMA plant. We use an on-line version of SVR, and at each instant, the identified model is used to determine the appropriate control law. A further improvement in the transient performance is shown with the methodology of multiple models, switching, and tuning.