A least squares approach to reduce stable discrete linear systems preserving their stability
A new stability preserving model reduction algorithm for discrete linear SISO-systems based on a least squares approach is proposed. Similar to the Padé approximation, an equation system for the Markov parameters involving a high dimensional Hankel matrix is considered. It is proved that approximate solutions, computed via the Moore-Penrose pseudo-inverse, give rise to a stability preserving reduction scheme. Furthermore, the proposed algorithm is compared to the balanced truncation method, showing comparable performance of the reduced systems.