Disturbance attenuation controller design by Markov parameters

This paper proposes to design a dynamic controller, which minimizes a class of conflicting quadratic performance indices, by using only the unit-pulse response of the plant. The method does not employ any parametric models such as transfer functions or state-space equations and, hence, shows how to design a controller directly from the input-output data of the plant. We begin by showing how to estimate the unit-pulse response of the system and the statistical data of the noise from the measured unit-pulse responses corrupted by system and measurement noises. With these informations, we design a conflicting-objective compensator which minimizes a given quadratic cost function. A numerical result that helps validate the efficiency of the method is also included