Subspace identification by data orthogonalization and model decoupling

It has been observed that identification of state-space models with inputs may lead to unreliable results in certain experimental conditions even when the input signal excites well within the bandwidth of the system. This may be due to ill-conditioning of the identification problem, which occurs when the state space and the future input space are nearly parallel. We have in particular shown in the companion papers (Automatica 40(4) (2004) 575; Automatica 40(4) (2004) 677) that, under these circumstances, subspace methods operating on input-output data may be ill-conditioned, quite independently of the particular algorithm which is used. In this paper, we indicate that the cause of ill-conditioning can sometimes be cured by using orthogonalized data and by recasting the model into a certain natural block-decoupled form consisting of a ''deterministic'' and a ''stochastic'' subsystem. The natural subspace algorithm for the identification of the deterministic subsystem is then a weighted version of the PI-MOESP method of Verhaegen and Dewilde (Int. J. Control 56 (1993) 1187-1211). The analysis shows that, under certain conditions, methods based on the block-decoupled parametrization and orthogonal decomposition of the input-output data, perform better than traditional joint-model-based methods in the circumstance of nearly parallel regressors.

[1]  Alessandro Chiuso,et al.  Probing inputs for subspace identification , 2000, Proceedings of the 39th IEEE Conference on Decision and Control (Cat. No.00CH37187).

[2]  Alessandro Chiuso,et al.  Asymptotic variance of subspace methods by data orthogonalization and model decoupling: a comparative analysis , 2004, Autom..

[3]  P. Caines,et al.  Estimation, Identification and Feedback , 1976 .

[4]  A. Chiuso,et al.  The asymptotic variance of subspace estimates , 2004 .

[5]  Alessandro Chiuso,et al.  Numerical conditioning and asymptotic variance of subspace estimates , 2004, Autom..

[6]  Bo Wahlberg,et al.  On Consistency of Subspace Methods for System Identification , 1998, Autom..

[7]  B. Moor,et al.  Subspace identification for linear systems , 1996 .

[8]  E. J. Hannan,et al.  Unit canonical correlations between future and past , 1988 .

[9]  Alessandro Chiuso,et al.  Some Algorithmic aspects of Subspace Identification with Inputs , 2001 .

[10]  Alessandro Chiuso,et al.  On the ill-conditioning of subspace identification with inputs , 2004, Autom..

[11]  H. Akaike.  Markovian Representation of Stochastic Processes by Canonical Variables , 1975 .

[12]  Subspace identification by orthogonal decomposition , 1999 .

[13]  Giorgio Picci,et al.  Stochastic realization with exogenous inputs and 'subspace-methods' identification , 1996, Signal Process..

[14]  M. Verhaegen.  Subspace model identification Part 2. Analysis of the elementary output-error state-space model identification algorithm , 1992 .

[15]  M. Deistler,et al.  On the Impact of Weighting Matrices in Subspace Algorithms , 2000 .

[16]  Patrick Dewilde,et al.  Subspace model identification Part 1. The output-error state-space model identification class of algorithms , 1992 .

[17]  Anders Lindquist,et al.  Canonical correlation analysis, approximate covariance extension, and identification of stationary time series , 1996, Autom..

[18]  Alessandro Chiuso,et al.  Comparison of Two Subspace Identification Methods for Combined Deterministic-Stochastic System: Part 2 , 2000 .

[19]  U. Desai,et al.  A realization approach to stochastic model reduction , 1985 .

[20]  M. Gevers,et al.  Representations of jointly stationary stochastic feedback processes , 1981 .

[21]  Lennart Ljung,et al.  System Identification: Theory for the User , 1987 .