Nuclear norm minimization methods for frequency domain subspace identification

Abstract

Frequency domain subspace identification is an effective means of obtaining a low-order model from frequency domain data. In the noisy data case using a singular value decomposition to determine the observable subspace has several problems: an incorrect weighting of the data in the singular values; difficulties in determining the appropriate rank; and a loss of the Hankel structure in the low-order approximation. A nuclear norm (sum of the singular values) minimization based method, using spectral constraints, is presented here and shown to be an effective technique for overcoming these problems.