Minimum variance prediction for linear time-varying systems (Q1361315)

A minimum variance predictor for linear time-varying systems, described by an autoregressive moving average model, is developed. The plant parameters are assumed to be known and uniformly bounded. The development is based on a pseudocommunication technique for dealing with noncommutativity of linear time-varying operators in a transfer function framework. The performance of a designed input-output predictor and the connections between the proposed and the known Kalman predictor are demonstrated by simulation examples.