On the Class of Dynamic Multicriteria Problems in the Design of Experiments

Abstract

The optimal design of experiments constitutes an important part in data processing. In the theory of the inverse problems in control system dynamics two basic procedures of the design of experiments are usually considered: the measurement allocation for estimation of the state of the system and the design of optimal inputs for systems parameters identification. Both these problems were studied carefully in control literature (Aoki and Staley, 1970; Athans, 1972; Herring and Melsa, 1974; Mehra, 1974 and others) for the systems with statistical description of unknowns. At the same time there exist applied problems concerned with the design of experiments for dynamic systems where either statistical description of disturbances is not available or disturbances are not probabilistic in their character. The models with setmembership description of unknowns, considered in the theory of control and observation under uncertainty conditions (Krasovski, 1968; Kurzhanski, 1977) appear to be more suitable in the last case.