Design of minimax linear filters by a local and an integral criteria

The article is conterned with linear filtering of basic coordinates of linear dynamic processes in continuous time with incompletely defined perturbing processes whose variance is constrained. Game statements are given of minimax linear filtering problems with the payoff function being a local criterion of generalized filtering error variance and an integral square criterion. The solution is found by a local criterion as a system of differential equations with boundary conditions which act only early in the observation. The design methods are compared by the local and integral criteria. A matrix function is found of the integral criterion with which the solution obtained by the local criterion is minimax by the integral criterion as well. An example is provided.