A simplified parametrization of the criterion for the prediction error dual controller

Abstract

This paper deals with analysis of suboptimal dual controller design for linear discrete time stochastic system with uncertain parameters. The design of the dual controller is based on augmentation of the criterion with suitable measure of the parameters uncertainty in order to explicitly ensure the probing property of the control law. As the uncertainty measure the weighted prediction error of the augmented state is chosen. This approach provides very good quality of the resulting control compared to other explicit dual control method. The drawback of the method can be, however, seen in the lack of methodology for selection of the supplemental weighting matrix. The aim of this article is thus to provide insight into a way how to chose this supplemental weighting matrix with minimum necessary optional parameters.