Abstract
Unstable zeros in discrete-time models and computation delays are key issues in digital controller design. To cope with these issues, we consider an application of the partial loop transfer recovery technique which has been proposed by Moore and Xia for continuoustime non-minimum phase systems. Taking account of computation delays, we construct a discrete-time LQG controller that depends only on the estimates of the minimum phase states of a plant model. The controller consists of a dynamical feedback gain matrix, a minimum phase state predictor and a Kalman filter estimating the minimum phase states. We consider feedback properties at the input of the plant. It is shown that, by increasing the variance of a fictitious disturbance injected into the minimum phase part of the plant model, we can recover feedback properties achieved by the minimum phase state feedback. To prove the recovery, we use simple decompositions of the sensitivity matrices. We give an explicit representation of the recovered sensitivity matrix. A numerical example is presented to illustrate the effectiveness of the proposed design.
