ARE LINEAR ALGORITHMS OPTIMAL IN GUARANTEED-ESTIMATION PROBLEMS

The question of whether linear (affine) in the measured signal algorit hms for guaranteed estimation are optimal is studied. Conditions under which the estimation results obtained in the classes of linear (affin e) estimates that are continuous in the measurements and of arbitrary linear estimates coincide under convex constraints on the indeterminat e parameters are obtained. A number of examples which demonstrate that such a coincidence does not occur, generally speaking, when the value s of a vector functional are estimated, is presented.