OPTIMAL-DESIGN OF ROBUST PREDICTORS FOR LINEAR DISCRETE-TIME-SYSTEMS

The one-step-ahead prediction problem for systems subject to parameter uncertainty in the system dynamics and noise statistics is considered , The objective is the design of a robust predictor that minimizes an upper bound of the error covariance, Sufficient and necessary conditio ns for the existence of such an optimal robust estimator are given. Th e computation of the predictor is based on the stabilizing solution of a suitable H(i)nfinity-type Riccati equation, In the uncertainty-free case the robust predictor reduces to the standard Kalman predictor.