On H-infinity model reduction of MIMO systems

In this paper we study multi-input-multi-output (MIMO) extensions of a recently proposed model reduction algorithm for single-input-multi-output (SIMO) linear time-invariant (LTI) systems. We discuss three versions, including a trivial modification of the SIMO method. Reduced models are found by solving a convex problem with Linear Matrix Inequality (LMI) constraints given a state space model or a frequency-sampled version. We construct examples that illustrate the properties of methods, compare algorithms and apply them on an industrial bench-mark.