Robust LQG control system under parameter uncertainty

Abstract

This paper presents a method of designing a robust linear quadratic Gaussian (LQG) controller for systems with unknown parameters when an accurate model is not available. The proposed robust LQG control scheme results in the prevention of the system performance deterioration caused by the parameter uncertainty without estimating unknown parameters explicitly. First, an approximate model is introduced to describe the uncertain parts of the plant and a target model which the controlled system should track is constructed based on the known nominal system and the approximate signal generated by this approximate model. Next, this tracking problem is changed to the LQG control one for an augmented system. This target trajectory is determined automatically in control operation for the augmented system. Further, both the controller and filter gain are determined as a solution of a deterministic suboptimal control problem.