An Application of Gaussian Process Models for Control Design

This paper describes a method of modelling nonlinear dynamical systems from measurement data for control design purpose. The method merges linear local model blending approach with Bayesian Gaussian process modelling. FixedStructure Gaussian Process model can be interpreted as linear model structure with varying and probabilistic parameters, which are represented with Gaussian process models. It can be applied for extended local linear equivalence class of nonlinear systems. The obtained nonlinear system model can be used for control system design based on parametric process model. The modelling and control design will be illustrated with a simple example. Copyright c © 2006 USTARTH