Compensator design for dynamic optimization†

This paper investigates the problem of designing a multiple-input, single-output linear compensator for a single-input, multiple-output linear plant in such a manner that a quadratic performance index is minimized. It is shown how optimal control may be achieved without measuring or estimating the state variables of the plant. A design procedure is suggested which depends for its success on the controllability and observability properties of the plant. It is pointed out how these observability properties are related to Luenberger's observer theory.