Low-order model identification of MIMO systems from noisy and incomplete data

In this paper, we provide preliminary results aimed at solving the following problem: Given a priori information on Multi-Input/Multi-Output (MIMO) plant, namely constraints on the pole location, and scattered input/output data, find the lowest order model that is compatible with both the a priori assumptions and the collected data. By combining concepts from signal sparsification and subspace identification, algorithms are developed that can determine a low order model from data that is both corrupted by measurement noise and has missing measurements. Effectiveness of the proposed approach is shown by an academic example.