Robust estimation and filtering in uncertain linear systems under unknown covariations

In the problems of estimation and filtering under uncertainty in the regressors, parameters, and covariances of random noise and perturbations, the best possible upper boundary of the proportionality coefficient between the root-mean-square error of estimate or filter and the sum of variances of all random factors was determined. This boundary which was named the level of suppression of random perturbations is characterized in terms of the linear matrix inequalities. The minimax estimate and minimax filter optimizing this index were established. The optimal robust estimate and filter were obtained using additional information about the membership of the covariance matrix in the given convex polyhedron.