Data-driven Minimum Variance Control Using Regulatory Closed-loop Data Based on the FRIT Method

Abstract

The paper proposes data-driven minimum variance regulatory control using a Fictitious Reference Iterative Tuning (FRIT) method. Our previous work has extended the original FRIT for reference signal tracking to the case of a stochastic disturbance attenuation problem. However, the method parametrizes the controller by using plant parameters based on a specific controller structure for minimum variance control. The present work uses a linearly parametrized controller for the inverse of the controller. Thus, a general parametrization is realized, and the integrator is straightforwardly introduced. An analytical result for the data-driven cost function is also given. A numerical example shows effectiveness of the proposed method.