Methods for Finding a Regularized Solution When Identifying Linear Multivariable Multiconnected Discrete Systems

This article considers the problem of structural and parametric identification of a complex multi-input multi-output (MIMO) discrete system in a class of state-space models. It is assumed that only the input and output coordinates of the system on some time interval and the range of measurement erro...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Cybernetics and systems analysis 2019-11, Vol.55 (6), p.881-893
Hauptverfasser: Gubarev, V. F., Romanenko, V. D., Miliavskyi, Yu. L.
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:This article considers the problem of structural and parametric identification of a complex multi-input multi-output (MIMO) discrete system in a class of state-space models. It is assumed that only the input and output coordinates of the system on some time interval and the range of measurement errors are known. The subspace method (4SID) underlies this approach, which presumes that the dimension of the system (state vector) is known, which is not always fulfilled in practice. Moreover, the dependence on the noise level makes it impossible to correctly identify a high-dimensional system. Therefore, it is proposed to consider dimension as a regularizing parameter. Three methods are developed for choosing an approximate model dimension depending on the duration of the observation interval and the possibility of designing an active experiment. The proposed methods are approved by an example of the problem of identifying a cognitive map of a commercial bank in an impulse process.
ISSN:1060-0396
1573-8337
DOI:10.1007/s10559-019-00198-5