Approximation of Subnetwork Models using Frequency-Domain Data

In this paper, we approximate models of interconnected systems that are to be used for decentralized control design. The suggested approach is based on approximation of so-called subnetwork models. A subnetwork model is a model of the interconnected system, as seen from one specific position in the network. The simplification is done by using weighted model reduction, and several approximation criteria are given. A new method for weighted model reduction is used. The method is based on a combination of known techniques that use semidefinite programming and frequency-data samples of transfer functions. The method is guaranteed to preserve stability and does not depend strongly on the order of the original model. This is particularly important for large interconnected systems. Two examples are given to illustrate the technique.