Minimum Variance Prediction of MIMO Linear Time-varying Systems

In this paper we study the problem of minimum variance prediction of linear time-varying systems and develop a minimum variance predictor for a standard multi-input multi-output time-varying ARMA model. It is shown that the linear time-varying predictor allows time-varying degrees in the ARMA model and minimizes the output prediction error variance under colored noise and rapidly time-varying plant parameters. Assumptions made on the linear time-varying systems are natural extensions of those made on linear time-invariant plants by the standard linear time-invariant minimum variance predictor.