State/driving-variable representations and model reduction for l2 2D systems

We present a reduction procedure for separable state/driving-variable (SDV) representations of l₂ 2D systems. These representations can describe a broader class of systems than the classical 2D input/state/output (ISO) models; moreover, similarly to what happens for separable Roesser models, they have the advantage of being decomposable into two ID state representations (one of which is a SDV representation and the other is regarded as being ISO). Following the ideas developed in [2] for separable Roesser models, our reduction procedure is based on the separate balancing of these two 1D representations.