An approach to reduction of parameter uncertainty for robust control system design

This paper considers system models with uncertain parameters belonging to a known region of tolerances and attacks the problem of how to reduce this region based on a knowledge of the noisy measurements. The main assumption is that bounds for the L/sup 2/ norms ("energies") of the disturbances are known. The "worst case" situation with respect to unknown state variables and disturbances is then considered in the sense that, for a given value of the uncertain parameter p, we try to explain the noisy measurement data using any unknown state variables and disturbances with minimum energy. If this minimum energy is greater than the assumed energy bound, the value of the parameter is excluded from the reduced uncertainty region. The determination of the reduced region of parameter uncertainty will facilitate the use of known robust control designs.<>