Abstract
An optimal reduced-order modeling problem in the presence of parameter uncertainty is considered. Necessary and sufficient conditions are derived for guaranteeing robust reduced-order model over a small range of parameter variations. A practical example is presented to illustrate the performance of the design method.