Vision-based cooperative pose estimation for localization in multi-robot systems equipped with RGB-D cameras

Xiaoqin Wang, Yasar Ahmet Sekercioglu, Thomas William Drummond

Research output: Contribution to journalArticleResearchpeer-review

14 Citations (Scopus)

Abstract

We present a new vision based cooperative pose estimation scheme for systems of mobile robots equipped with RGB-D cameras. We first model a multi-robot system as an edge-weighted graph. Then, based on this model, and by using the real-time color and depth data, the robots with shared field-of-views estimate their relative poses in pairwise. The system does not need the existence of a single common view shared by all robots, and it works in 3D scenes without any specific calibration pattern or landmark. The proposed scheme distributes working loads evenly in the system, hence it is scalable and the computing power of the participating robots is efficiently used. The performance and robustness were analyzed both on synthetic and experimental data in different environments over a range of system configurations with varying number of robots and poses.
Original languageEnglish
Pages (from-to)1-22
Number of pages22
JournalRobotics
Volume4
Issue number1
DOIs
Publication statusPublished - Mar 2015

Cite this

@article{96ce5a1665e641408644e54c59fc1574,
title = "Vision-based cooperative pose estimation for localization in multi-robot systems equipped with RGB-D cameras",
abstract = "We present a new vision based cooperative pose estimation scheme for systems of mobile robots equipped with RGB-D cameras. We first model a multi-robot system as an edge-weighted graph. Then, based on this model, and by using the real-time color and depth data, the robots with shared field-of-views estimate their relative poses in pairwise. The system does not need the existence of a single common view shared by all robots, and it works in 3D scenes without any specific calibration pattern or landmark. The proposed scheme distributes working loads evenly in the system, hence it is scalable and the computing power of the participating robots is efficiently used. The performance and robustness were analyzed both on synthetic and experimental data in different environments over a range of system configurations with varying number of robots and poses.",
author = "Xiaoqin Wang and Sekercioglu, {Yasar Ahmet} and Drummond, {Thomas William}",
year = "2015",
month = "3",
doi = "10.3390/robotics4010001",
language = "English",
volume = "4",
pages = "1--22",
journal = "Robotics",
issn = "2218-6581",
publisher = "MDPI AG",
number = "1",

}

Vision-based cooperative pose estimation for localization in multi-robot systems equipped with RGB-D cameras. / Wang, Xiaoqin; Sekercioglu, Yasar Ahmet; Drummond, Thomas William.

In: Robotics, Vol. 4, No. 1, 03.2015, p. 1-22.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Vision-based cooperative pose estimation for localization in multi-robot systems equipped with RGB-D cameras

AU - Wang, Xiaoqin

AU - Sekercioglu, Yasar Ahmet

AU - Drummond, Thomas William

PY - 2015/3

Y1 - 2015/3

N2 - We present a new vision based cooperative pose estimation scheme for systems of mobile robots equipped with RGB-D cameras. We first model a multi-robot system as an edge-weighted graph. Then, based on this model, and by using the real-time color and depth data, the robots with shared field-of-views estimate their relative poses in pairwise. The system does not need the existence of a single common view shared by all robots, and it works in 3D scenes without any specific calibration pattern or landmark. The proposed scheme distributes working loads evenly in the system, hence it is scalable and the computing power of the participating robots is efficiently used. The performance and robustness were analyzed both on synthetic and experimental data in different environments over a range of system configurations with varying number of robots and poses.

AB - We present a new vision based cooperative pose estimation scheme for systems of mobile robots equipped with RGB-D cameras. We first model a multi-robot system as an edge-weighted graph. Then, based on this model, and by using the real-time color and depth data, the robots with shared field-of-views estimate their relative poses in pairwise. The system does not need the existence of a single common view shared by all robots, and it works in 3D scenes without any specific calibration pattern or landmark. The proposed scheme distributes working loads evenly in the system, hence it is scalable and the computing power of the participating robots is efficiently used. The performance and robustness were analyzed both on synthetic and experimental data in different environments over a range of system configurations with varying number of robots and poses.

U2 - 10.3390/robotics4010001

DO - 10.3390/robotics4010001

M3 - Article

VL - 4

SP - 1

EP - 22

JO - Robotics

JF - Robotics

SN - 2218-6581

IS - 1

ER -