Tolerance-weighted l-optimal experiment design: A new approach to task-directed sensing

J. De Geeter; J. De Schutter; H. Bruyninckx; H. Van Brussel; M. Decréton

doi:10.1163/156855399X00252

Tolerance-weighted l-optimal experiment design: A new approach to task-directed sensing

J. De Geeter, J. De Schutter, H. Bruyninckx, H. Van Brussel, M. Decréton

Research output › peer-review

Abstract

The choice of 'where to look next' is a special case of an optimal experiment design. This paper proposes the tolerance-optimal experiment design, which is a special instance of the well-known L-optimal design, that minimizes the weighted trace of the covariance matrix of the estimatedstate under Gaussian assumptions. The weighting matrix is chosen such that the design is invariant to transformations with non-singular Jacobians, and such that the emerging sensing sequence reflects the information needs of the task. This tolerance-optimal design does not require more calculationsthan existing optimal experiment designs. Existing optimal experiment designs do not reflect the information needs of the task. In addition, some of them physically do not make sense if the estimated state has inconsistent units.

Original language	English
Pages (from-to)	401-416
Number of pages	16
Journal	Advanced Robotics
Volume	13
Issue number	4
DOIs	https://doi.org/10.1163/156855399X00252
State	Published - 1 Jan 1998

ASJC Scopus subject areas

Control and Systems Engineering
Software
Human-Computer Interaction
Hardware and Architecture
Computer Science Applications

Access to Document

10.1163/156855399X00252

Cite this

@article{d69c82e1ce4b47c68e1729c75f4f11bc,

title = "Tolerance-weighted l-optimal experiment design: A new approach to task-directed sensing",

abstract = "The choice of 'where to look next' is a special case of an optimal experiment design. This paper proposes the tolerance-optimal experiment design, which is a special instance of the well-known L-optimal design, that minimizes the weighted trace of the covariance matrix of the estimatedstate under Gaussian assumptions. The weighting matrix is chosen such that the design is invariant to transformations with non-singular Jacobians, and such that the emerging sensing sequence reflects the information needs of the task. This tolerance-optimal design does not require more calculationsthan existing optimal experiment designs. Existing optimal experiment designs do not reflect the information needs of the task. In addition, some of them physically do not make sense if the estimated state has inconsistent units.",

keywords = "End condition, Gaussian distribution, Task invariant, Task-directed sensing, Tolerance-optimal design",

author = "{De Geeter}, J. and {De Schutter}, J. and H. Bruyninckx and {Van Brussel}, H. and M. Decr{\'e}ton",

year = "1998",

month = jan,

day = "1",

doi = "10.1163/156855399X00252",

language = "English",

volume = "13",

pages = "401--416",

journal = "Advanced Robotics",

issn = "0169-1864",

publisher = "Taylor & Francis (CRC)",

number = "4",

}

TY - JOUR

T1 - Tolerance-weighted l-optimal experiment design

T2 - A new approach to task-directed sensing

AU - De Geeter, J.

AU - De Schutter, J.

AU - Bruyninckx, H.

AU - Van Brussel, H.

AU - Decréton, M.

PY - 1998/1/1

Y1 - 1998/1/1

N2 - The choice of 'where to look next' is a special case of an optimal experiment design. This paper proposes the tolerance-optimal experiment design, which is a special instance of the well-known L-optimal design, that minimizes the weighted trace of the covariance matrix of the estimatedstate under Gaussian assumptions. The weighting matrix is chosen such that the design is invariant to transformations with non-singular Jacobians, and such that the emerging sensing sequence reflects the information needs of the task. This tolerance-optimal design does not require more calculationsthan existing optimal experiment designs. Existing optimal experiment designs do not reflect the information needs of the task. In addition, some of them physically do not make sense if the estimated state has inconsistent units.

AB - The choice of 'where to look next' is a special case of an optimal experiment design. This paper proposes the tolerance-optimal experiment design, which is a special instance of the well-known L-optimal design, that minimizes the weighted trace of the covariance matrix of the estimatedstate under Gaussian assumptions. The weighting matrix is chosen such that the design is invariant to transformations with non-singular Jacobians, and such that the emerging sensing sequence reflects the information needs of the task. This tolerance-optimal design does not require more calculationsthan existing optimal experiment designs. Existing optimal experiment designs do not reflect the information needs of the task. In addition, some of them physically do not make sense if the estimated state has inconsistent units.

KW - End condition

KW - Gaussian distribution

KW - Task invariant

KW - Task-directed sensing

KW - Tolerance-optimal design

UR - http://www.scopus.com/inward/record.url?scp=0033332795&partnerID=8YFLogxK

U2 - 10.1163/156855399X00252

DO - 10.1163/156855399X00252

M3 - Article

AN - SCOPUS:0033332795

SN - 0169-1864

VL - 13

SP - 401

EP - 416

JO - Advanced Robotics

JF - Advanced Robotics

IS - 4

ER -