Supermodular batch state estimation in optimal sensor scheduling

Prince Singh, Sze Yong, Emilio Frazzoli

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

This letter addresses the problem of activating, at each time-step in a finite time horizon problem, a subset of available sensors to generate a “high quality” estimate of the state of a discrete-time linear system operating under limited resources. We propose a sensor schedule that minimizes the mean square estimation error of the batch state vector of the system—the batch state estimation (BSE) problem. Due to the presence of limited resources, we address the cardinality-constrained BSE problem, which is inherently combinatorial and computationally intractable when working with large-scale systems. This NP-hard complexity is overcome by employing a greedy algorithm, which returns a near-optimal sensor schedule with performance guarantees when minimizing a supermodular objective over matroids. To this end, we prove (despite the existence of counter-examples in literature) that our objective function is supermodular when the batch prior information matrix is a strictly diagonally dominant M-matrix (with a constraint on its inverse and conditions on the measurement model). Hence, we obtain a near-optimal solution to the BSE problem via a greedy algorithm. Additionally, we provide its time complexity.

Original languageEnglish (US)
Pages (from-to)292-297
Number of pages6
JournalIEEE Control Systems Letters
Volume1
Issue number2
DOIs
StatePublished - Oct 1 2017

Fingerprint

State Estimation
State estimation
Batch
Scheduling
Sensor
Sensors
Greedy Algorithm
Schedule
Error analysis
Linear systems
Large scale systems
Discrete-time Linear Systems
Resources
Information Matrix
Performance Guarantee
M-matrix
Prior Information
Estimation Error
Large-scale Systems
Matroid

Keywords

  • Approximation algorithms
  • Batch state estimation
  • Large-scale systems
  • Optimal scheduling
  • Sensor networks

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Control and Optimization

Cite this

Supermodular batch state estimation in optimal sensor scheduling. / Singh, Prince; Yong, Sze; Frazzoli, Emilio.

In: IEEE Control Systems Letters, Vol. 1, No. 2, 01.10.2017, p. 292-297.

Research output: Contribution to journalArticle

Singh, Prince ; Yong, Sze ; Frazzoli, Emilio. / Supermodular batch state estimation in optimal sensor scheduling. In: IEEE Control Systems Letters. 2017 ; Vol. 1, No. 2. pp. 292-297.
@article{dee5bb0de592442b9ae049d7b7767a1f,
title = "Supermodular batch state estimation in optimal sensor scheduling",
abstract = "This letter addresses the problem of activating, at each time-step in a finite time horizon problem, a subset of available sensors to generate a “high quality” estimate of the state of a discrete-time linear system operating under limited resources. We propose a sensor schedule that minimizes the mean square estimation error of the batch state vector of the system—the batch state estimation (BSE) problem. Due to the presence of limited resources, we address the cardinality-constrained BSE problem, which is inherently combinatorial and computationally intractable when working with large-scale systems. This NP-hard complexity is overcome by employing a greedy algorithm, which returns a near-optimal sensor schedule with performance guarantees when minimizing a supermodular objective over matroids. To this end, we prove (despite the existence of counter-examples in literature) that our objective function is supermodular when the batch prior information matrix is a strictly diagonally dominant M-matrix (with a constraint on its inverse and conditions on the measurement model). Hence, we obtain a near-optimal solution to the BSE problem via a greedy algorithm. Additionally, we provide its time complexity.",
keywords = "Approximation algorithms, Batch state estimation, Large-scale systems, Optimal scheduling, Sensor networks",
author = "Prince Singh and Sze Yong and Emilio Frazzoli",
year = "2017",
month = "10",
day = "1",
doi = "10.1109/LCSYS.2017.2710051",
language = "English (US)",
volume = "1",
pages = "292--297",
journal = "IEEE Control Systems Letters",
issn = "2475-1456",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "2",

}

TY - JOUR

T1 - Supermodular batch state estimation in optimal sensor scheduling

AU - Singh, Prince

AU - Yong, Sze

AU - Frazzoli, Emilio

PY - 2017/10/1

Y1 - 2017/10/1

N2 - This letter addresses the problem of activating, at each time-step in a finite time horizon problem, a subset of available sensors to generate a “high quality” estimate of the state of a discrete-time linear system operating under limited resources. We propose a sensor schedule that minimizes the mean square estimation error of the batch state vector of the system—the batch state estimation (BSE) problem. Due to the presence of limited resources, we address the cardinality-constrained BSE problem, which is inherently combinatorial and computationally intractable when working with large-scale systems. This NP-hard complexity is overcome by employing a greedy algorithm, which returns a near-optimal sensor schedule with performance guarantees when minimizing a supermodular objective over matroids. To this end, we prove (despite the existence of counter-examples in literature) that our objective function is supermodular when the batch prior information matrix is a strictly diagonally dominant M-matrix (with a constraint on its inverse and conditions on the measurement model). Hence, we obtain a near-optimal solution to the BSE problem via a greedy algorithm. Additionally, we provide its time complexity.

AB - This letter addresses the problem of activating, at each time-step in a finite time horizon problem, a subset of available sensors to generate a “high quality” estimate of the state of a discrete-time linear system operating under limited resources. We propose a sensor schedule that minimizes the mean square estimation error of the batch state vector of the system—the batch state estimation (BSE) problem. Due to the presence of limited resources, we address the cardinality-constrained BSE problem, which is inherently combinatorial and computationally intractable when working with large-scale systems. This NP-hard complexity is overcome by employing a greedy algorithm, which returns a near-optimal sensor schedule with performance guarantees when minimizing a supermodular objective over matroids. To this end, we prove (despite the existence of counter-examples in literature) that our objective function is supermodular when the batch prior information matrix is a strictly diagonally dominant M-matrix (with a constraint on its inverse and conditions on the measurement model). Hence, we obtain a near-optimal solution to the BSE problem via a greedy algorithm. Additionally, we provide its time complexity.

KW - Approximation algorithms

KW - Batch state estimation

KW - Large-scale systems

KW - Optimal scheduling

KW - Sensor networks

UR - http://www.scopus.com/inward/record.url?scp=85052588693&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85052588693&partnerID=8YFLogxK

U2 - 10.1109/LCSYS.2017.2710051

DO - 10.1109/LCSYS.2017.2710051

M3 - Article

AN - SCOPUS:85052588693

VL - 1

SP - 292

EP - 297

JO - IEEE Control Systems Letters

JF - IEEE Control Systems Letters

SN - 2475-1456

IS - 2

ER -