Input Matrix Factorizations for Constrained Control Allocation

Research output: Contribution to journalArticleResearchpeer-review

Abstract

Control allocation (CA) is part of a hierarchical control architecture and distributes the desired control effort of a controller among a set of redundant actuators as it is used, for example, to meet high reliability requirements. Before applying CA, a factorization of a linear plant's input matrix must be carried out. This paper investigates the influence of this factorization on two common algorithms for constrained CA: direct allocation and redistributed pseudoinverse (RPINV). It is shown why the factorization does not affect the first method, whereas it influences the latter one as soon as the number of actuator saturations exceeds a certain threshold. On this basis a modification of RPINV is proposed, which allows the prioritization of virtual control vector components. If the actuator saturations prevent an exact solution, the errors of those components with high priority are preferentially forced to zero. Finally simulations demonstrate the main results.
Original languageEnglish
Pages (from-to)1163-1170
Number of pages8
JournalIEEE Transactions on Automatic Control
Volume63
Issue number4
Early online date14 Aug 2017
DOIs
Publication statusPublished - 4 Apr 2018

Fingerprint

Factorization
Actuators
Controllers

Cite this

Input Matrix Factorizations for Constrained Control Allocation. / Kirchengast, Martin; Steinberger, Martin; Horn, Martin.

In: IEEE Transactions on Automatic Control, Vol. 63, No. 4, 04.04.2018, p. 1163-1170.

Research output: Contribution to journalArticleResearchpeer-review

@article{d76a00027c644438b058b570704762fa,
title = "Input Matrix Factorizations for Constrained Control Allocation",
abstract = "Control allocation (CA) is part of a hierarchical control architecture and distributes the desired control effort of a controller among a set of redundant actuators as it is used, for example, to meet high reliability requirements. Before applying CA, a factorization of a linear plant's input matrix must be carried out. This paper investigates the influence of this factorization on two common algorithms for constrained CA: direct allocation and redistributed pseudoinverse (RPINV). It is shown why the factorization does not affect the first method, whereas it influences the latter one as soon as the number of actuator saturations exceeds a certain threshold. On this basis a modification of RPINV is proposed, which allows the prioritization of virtual control vector components. If the actuator saturations prevent an exact solution, the errors of those components with high priority are preferentially forced to zero. Finally simulations demonstrate the main results.",
author = "Martin Kirchengast and Martin Steinberger and Martin Horn",
year = "2018",
month = "4",
day = "4",
doi = "10.1109/TAC.2017.2739648",
language = "English",
volume = "63",
pages = "1163--1170",
journal = "IEEE Transactions on Automatic Control",
issn = "0018-9286",
publisher = "IEEE Control Systems Society",
number = "4",

}

TY - JOUR

T1 - Input Matrix Factorizations for Constrained Control Allocation

AU - Kirchengast, Martin

AU - Steinberger, Martin

AU - Horn, Martin

PY - 2018/4/4

Y1 - 2018/4/4

N2 - Control allocation (CA) is part of a hierarchical control architecture and distributes the desired control effort of a controller among a set of redundant actuators as it is used, for example, to meet high reliability requirements. Before applying CA, a factorization of a linear plant's input matrix must be carried out. This paper investigates the influence of this factorization on two common algorithms for constrained CA: direct allocation and redistributed pseudoinverse (RPINV). It is shown why the factorization does not affect the first method, whereas it influences the latter one as soon as the number of actuator saturations exceeds a certain threshold. On this basis a modification of RPINV is proposed, which allows the prioritization of virtual control vector components. If the actuator saturations prevent an exact solution, the errors of those components with high priority are preferentially forced to zero. Finally simulations demonstrate the main results.

AB - Control allocation (CA) is part of a hierarchical control architecture and distributes the desired control effort of a controller among a set of redundant actuators as it is used, for example, to meet high reliability requirements. Before applying CA, a factorization of a linear plant's input matrix must be carried out. This paper investigates the influence of this factorization on two common algorithms for constrained CA: direct allocation and redistributed pseudoinverse (RPINV). It is shown why the factorization does not affect the first method, whereas it influences the latter one as soon as the number of actuator saturations exceeds a certain threshold. On this basis a modification of RPINV is proposed, which allows the prioritization of virtual control vector components. If the actuator saturations prevent an exact solution, the errors of those components with high priority are preferentially forced to zero. Finally simulations demonstrate the main results.

U2 - 10.1109/TAC.2017.2739648

DO - 10.1109/TAC.2017.2739648

M3 - Article

VL - 63

SP - 1163

EP - 1170

JO - IEEE Transactions on Automatic Control

JF - IEEE Transactions on Automatic Control

SN - 0018-9286

IS - 4

ER -