From closed-loop identification to dynamic networks: Generalization of the direct method

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

1 Citation (Scopus)
25 Downloads (Pure)

Abstract

Identification methods for identifying (modules in) dynamic cyclic networks, are typically based on the standard methods that are available for identification of dynamic systems in closed-loop. The commonly used direct method for closed-loop prediction error identification is one of the available tools. In this paper we are going to show the consequences when the direct method is used under conditions that are more general than the classical closed-loop case. We will do so by focusing on a simple two-node (feedback) network where we add additional disturbances, excitation signals and sensor noise. The direct method loses consistency when correlated disturbances are present on node signals, or when sensor noises are present. A generalization of the direct method, the joint-direct method, is explored, that is based on a vector predictor and includes a conditioning on external excitation signals. It is shown to be able to cope with the above situations, and to retain consistency of the module estimates.

Original languageEnglish
Title of host publication2017 IEEE 56th Annual Conference on Decision and Control, CDC 2017
Place of PublicationPiscataway
PublisherInstitute of Electrical and Electronics Engineers
Pages5845-5850
Number of pages6
ISBN (Electronic)978-1-5090-2873-3
DOIs
Publication statusPublished - 18 Jan 2018
Event56th IEEE Conference on Decision and Control (CDC 2017) - Melbourne, Australia
Duration: 12 Dec 201715 Dec 2017
Conference number: 56
http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8253407

Conference

Conference56th IEEE Conference on Decision and Control (CDC 2017)
Abbreviated titleCDC 2017
CountryAustralia
CityMelbourne
Period12/12/1715/12/17
Internet address

Fingerprint

Dynamic Networks
Direct Method
Closed-loop
Sensors
Dynamical systems
Feedback
Excitation
Disturbance
Module
Sensor
Prediction Error
Vertex of a graph
Conditioning
Dynamic Systems
Predictors
Generalization
Network dynamics
Estimate

Cite this

Van den Hof, P. M. J., Dankers, A. G., & Weerts, H. H. M. (2018). From closed-loop identification to dynamic networks: Generalization of the direct method. In 2017 IEEE 56th Annual Conference on Decision and Control, CDC 2017 (pp. 5845-5850). Piscataway: Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/CDC.2017.8264543
Van den Hof, Paul M.J. ; Dankers, Arne G. ; Weerts, Harm H.M. / From closed-loop identification to dynamic networks : Generalization of the direct method. 2017 IEEE 56th Annual Conference on Decision and Control, CDC 2017. Piscataway : Institute of Electrical and Electronics Engineers, 2018. pp. 5845-5850
@inproceedings{97084d2505814f419827a1745bee2887,
title = "From closed-loop identification to dynamic networks: Generalization of the direct method",
abstract = "Identification methods for identifying (modules in) dynamic cyclic networks, are typically based on the standard methods that are available for identification of dynamic systems in closed-loop. The commonly used direct method for closed-loop prediction error identification is one of the available tools. In this paper we are going to show the consequences when the direct method is used under conditions that are more general than the classical closed-loop case. We will do so by focusing on a simple two-node (feedback) network where we add additional disturbances, excitation signals and sensor noise. The direct method loses consistency when correlated disturbances are present on node signals, or when sensor noises are present. A generalization of the direct method, the joint-direct method, is explored, that is based on a vector predictor and includes a conditioning on external excitation signals. It is shown to be able to cope with the above situations, and to retain consistency of the module estimates.",
author = "{Van den Hof}, {Paul M.J.} and Dankers, {Arne G.} and Weerts, {Harm H.M.}",
year = "2018",
month = "1",
day = "18",
doi = "10.1109/CDC.2017.8264543",
language = "English",
pages = "5845--5850",
booktitle = "2017 IEEE 56th Annual Conference on Decision and Control, CDC 2017",
publisher = "Institute of Electrical and Electronics Engineers",
address = "United States",

}

Van den Hof, PMJ, Dankers, AG & Weerts, HHM 2018, From closed-loop identification to dynamic networks: Generalization of the direct method. in 2017 IEEE 56th Annual Conference on Decision and Control, CDC 2017. Institute of Electrical and Electronics Engineers, Piscataway, pp. 5845-5850, 56th IEEE Conference on Decision and Control (CDC 2017), Melbourne, Australia, 12/12/17. https://doi.org/10.1109/CDC.2017.8264543

From closed-loop identification to dynamic networks : Generalization of the direct method. / Van den Hof, Paul M.J.; Dankers, Arne G.; Weerts, Harm H.M.

2017 IEEE 56th Annual Conference on Decision and Control, CDC 2017. Piscataway : Institute of Electrical and Electronics Engineers, 2018. p. 5845-5850.

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

TY - GEN

T1 - From closed-loop identification to dynamic networks

T2 - Generalization of the direct method

AU - Van den Hof, Paul M.J.

AU - Dankers, Arne G.

AU - Weerts, Harm H.M.

PY - 2018/1/18

Y1 - 2018/1/18

N2 - Identification methods for identifying (modules in) dynamic cyclic networks, are typically based on the standard methods that are available for identification of dynamic systems in closed-loop. The commonly used direct method for closed-loop prediction error identification is one of the available tools. In this paper we are going to show the consequences when the direct method is used under conditions that are more general than the classical closed-loop case. We will do so by focusing on a simple two-node (feedback) network where we add additional disturbances, excitation signals and sensor noise. The direct method loses consistency when correlated disturbances are present on node signals, or when sensor noises are present. A generalization of the direct method, the joint-direct method, is explored, that is based on a vector predictor and includes a conditioning on external excitation signals. It is shown to be able to cope with the above situations, and to retain consistency of the module estimates.

AB - Identification methods for identifying (modules in) dynamic cyclic networks, are typically based on the standard methods that are available for identification of dynamic systems in closed-loop. The commonly used direct method for closed-loop prediction error identification is one of the available tools. In this paper we are going to show the consequences when the direct method is used under conditions that are more general than the classical closed-loop case. We will do so by focusing on a simple two-node (feedback) network where we add additional disturbances, excitation signals and sensor noise. The direct method loses consistency when correlated disturbances are present on node signals, or when sensor noises are present. A generalization of the direct method, the joint-direct method, is explored, that is based on a vector predictor and includes a conditioning on external excitation signals. It is shown to be able to cope with the above situations, and to retain consistency of the module estimates.

UR - http://www.scopus.com/inward/record.url?scp=85046276899&partnerID=8YFLogxK

U2 - 10.1109/CDC.2017.8264543

DO - 10.1109/CDC.2017.8264543

M3 - Conference contribution

AN - SCOPUS:85046276899

SP - 5845

EP - 5850

BT - 2017 IEEE 56th Annual Conference on Decision and Control, CDC 2017

PB - Institute of Electrical and Electronics Engineers

CY - Piscataway

ER -

Van den Hof PMJ, Dankers AG, Weerts HHM. From closed-loop identification to dynamic networks: Generalization of the direct method. In 2017 IEEE 56th Annual Conference on Decision and Control, CDC 2017. Piscataway: Institute of Electrical and Electronics Engineers. 2018. p. 5845-5850 https://doi.org/10.1109/CDC.2017.8264543