Model reduction for controllable systems

In the papers [1], [7] a new scheme for passivity-preserving model reduction has been proposed. We have shown in [2] that the approach can also be interpreted from a dissipativity theory point of view, and we put forward two procedures in order to compute a driving variable or output nulling representation of a reduced order model for a given behavior. In this paper we illustrate improved versions of both algorithms, which produce a controllable reduced-order model. The new algorithms are based on several original results of independent interest.