A homotopy approach to improving PEM identification of ARMAX models

Takayuki Furuzuki, Kotaro Hirasawa, Kousuke Kumamaru

Research output: Contribution to journalArticle

19 Citations (Scopus)

Abstract

This paper presents a homotopy approach to improving PEM identification of ARMAX model. PEM estimates of ARMAX model parameters are determined as the global minimum of criterion function, which is however not always unimodal because of the MA noise model part. An optimization-based PEM identification algorithm has a potential risk to be stuck at a local minimum that results in a poorly identified model. A homotopy continuation method is introduced to solve this problem. The idea is to start the estimation with the criterion function for PEM identification of the ARX model, which is gradually deformed into the actual one for PEM identification of the ARMAX model as the algorithm iterates. By building the deformation into the usual recursive procedure for the ARMAX identification and introducing a scheme to control the solution continuously staying in the global minima of the deformed criterion functions, the homotopy-based PEM identification algorithm is implemented in such a way that it has very good convergence performance, with only little increase in computation load compared to the usual PEM algorithm.

Original languageEnglish
Pages (from-to)1323-1334
Number of pages12
JournalAutomatica
Volume37
Issue number9
DOIs
Publication statusPublished - 2001 Sep
Externally publishedYes

Fingerprint

Identification (control systems)

Keywords

  • ARMAX model
  • Homotopy continuation method
  • Multimodality problem
  • Parameter estimation
  • Prediction error method

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Electrical and Electronic Engineering

Cite this

A homotopy approach to improving PEM identification of ARMAX models. / Furuzuki, Takayuki; Hirasawa, Kotaro; Kumamaru, Kousuke.

In: Automatica, Vol. 37, No. 9, 09.2001, p. 1323-1334.

Research output: Contribution to journalArticle

Furuzuki, Takayuki ; Hirasawa, Kotaro ; Kumamaru, Kousuke. / A homotopy approach to improving PEM identification of ARMAX models. In: Automatica. 2001 ; Vol. 37, No. 9. pp. 1323-1334.
@article{4c181a1e902f496a9c6f2bfbdb972ba1,
title = "A homotopy approach to improving PEM identification of ARMAX models",
abstract = "This paper presents a homotopy approach to improving PEM identification of ARMAX model. PEM estimates of ARMAX model parameters are determined as the global minimum of criterion function, which is however not always unimodal because of the MA noise model part. An optimization-based PEM identification algorithm has a potential risk to be stuck at a local minimum that results in a poorly identified model. A homotopy continuation method is introduced to solve this problem. The idea is to start the estimation with the criterion function for PEM identification of the ARX model, which is gradually deformed into the actual one for PEM identification of the ARMAX model as the algorithm iterates. By building the deformation into the usual recursive procedure for the ARMAX identification and introducing a scheme to control the solution continuously staying in the global minima of the deformed criterion functions, the homotopy-based PEM identification algorithm is implemented in such a way that it has very good convergence performance, with only little increase in computation load compared to the usual PEM algorithm.",
keywords = "ARMAX model, Homotopy continuation method, Multimodality problem, Parameter estimation, Prediction error method",
author = "Takayuki Furuzuki and Kotaro Hirasawa and Kousuke Kumamaru",
year = "2001",
month = "9",
doi = "10.1016/S0005-1098(01)00081-4",
language = "English",
volume = "37",
pages = "1323--1334",
journal = "Automatica",
issn = "0005-1098",
publisher = "Elsevier Limited",
number = "9",

}

TY - JOUR

T1 - A homotopy approach to improving PEM identification of ARMAX models

AU - Furuzuki, Takayuki

AU - Hirasawa, Kotaro

AU - Kumamaru, Kousuke

PY - 2001/9

Y1 - 2001/9

N2 - This paper presents a homotopy approach to improving PEM identification of ARMAX model. PEM estimates of ARMAX model parameters are determined as the global minimum of criterion function, which is however not always unimodal because of the MA noise model part. An optimization-based PEM identification algorithm has a potential risk to be stuck at a local minimum that results in a poorly identified model. A homotopy continuation method is introduced to solve this problem. The idea is to start the estimation with the criterion function for PEM identification of the ARX model, which is gradually deformed into the actual one for PEM identification of the ARMAX model as the algorithm iterates. By building the deformation into the usual recursive procedure for the ARMAX identification and introducing a scheme to control the solution continuously staying in the global minima of the deformed criterion functions, the homotopy-based PEM identification algorithm is implemented in such a way that it has very good convergence performance, with only little increase in computation load compared to the usual PEM algorithm.

AB - This paper presents a homotopy approach to improving PEM identification of ARMAX model. PEM estimates of ARMAX model parameters are determined as the global minimum of criterion function, which is however not always unimodal because of the MA noise model part. An optimization-based PEM identification algorithm has a potential risk to be stuck at a local minimum that results in a poorly identified model. A homotopy continuation method is introduced to solve this problem. The idea is to start the estimation with the criterion function for PEM identification of the ARX model, which is gradually deformed into the actual one for PEM identification of the ARMAX model as the algorithm iterates. By building the deformation into the usual recursive procedure for the ARMAX identification and introducing a scheme to control the solution continuously staying in the global minima of the deformed criterion functions, the homotopy-based PEM identification algorithm is implemented in such a way that it has very good convergence performance, with only little increase in computation load compared to the usual PEM algorithm.

KW - ARMAX model

KW - Homotopy continuation method

KW - Multimodality problem

KW - Parameter estimation

KW - Prediction error method

UR - http://www.scopus.com/inward/record.url?scp=0035452451&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0035452451&partnerID=8YFLogxK

U2 - 10.1016/S0005-1098(01)00081-4

DO - 10.1016/S0005-1098(01)00081-4

M3 - Article

VL - 37

SP - 1323

EP - 1334

JO - Automatica

JF - Automatica

SN - 0005-1098

IS - 9

ER -