Efficient Appearance-Based Topological Mapping and Navigation with Omnidirectional Vision

Christopher Burbridge, Joan Condell

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Because of a mobile robot’s ability to move in itsenvironment, one of the most important and common tasks formobile robots is, arguably, the task of navigation. This has tobe carried out in practically every mobile robot, and it hasto be carried out quickly (real-time constraints) and cheaply(computational constraints). SLAM [7] is a reliable and veryaccurate method which is widely used now in mobile robotics.However, these algorithms are not computationally cheap. Inthis paper we therefore investigate a computationally efficientalgorithm for simultaneous mapping and navigation that issuitable for application in simple mobile robots. We apply selforganisingmaps in a novel way to image data compression andindexing of topological map nodes at the same time.Our proposed system builds a dense topological map using onlythe visual appearance of the environment, with no need for anyfeature extraction or matching. This is made possible throughthe novel use of a self-organising map and a relaxed attitudetowards loop closure and metric consistency. The inconsistenciesand uncertainties within the map are not considered duringmapping, but rather only during navigation at which point aBayesian approach is taken to allow accurate navigation. Thepaper concludes by presenting mobile robot experiments
LanguageEnglish
Title of host publicationUnknown Host Publication
Pages26-33
Number of pages8
Publication statusPublished - 1 Sep 2010
EventTAROS 2010 - Towards Autonomous Robotic Systems 2010 - University of Plymouth, Devon, UK
Duration: 1 Sep 2010 → …

Conference

ConferenceTAROS 2010 - Towards Autonomous Robotic Systems 2010
Period1/09/10 → …

Fingerprint

Mobile robots
Navigation
Self organizing maps
Data compression
Robotics
Robots
Experiments

Cite this

@inproceedings{fed435c2b4264e8fa6d4a487af25621b,
title = "Efficient Appearance-Based Topological Mapping and Navigation with Omnidirectional Vision",
abstract = "Because of a mobile robot’s ability to move in itsenvironment, one of the most important and common tasks formobile robots is, arguably, the task of navigation. This has tobe carried out in practically every mobile robot, and it hasto be carried out quickly (real-time constraints) and cheaply(computational constraints). SLAM [7] is a reliable and veryaccurate method which is widely used now in mobile robotics.However, these algorithms are not computationally cheap. Inthis paper we therefore investigate a computationally efficientalgorithm for simultaneous mapping and navigation that issuitable for application in simple mobile robots. We apply selforganisingmaps in a novel way to image data compression andindexing of topological map nodes at the same time.Our proposed system builds a dense topological map using onlythe visual appearance of the environment, with no need for anyfeature extraction or matching. This is made possible throughthe novel use of a self-organising map and a relaxed attitudetowards loop closure and metric consistency. The inconsistenciesand uncertainties within the map are not considered duringmapping, but rather only during navigation at which point aBayesian approach is taken to allow accurate navigation. Thepaper concludes by presenting mobile robot experiments",
author = "Christopher Burbridge and Joan Condell",
year = "2010",
month = "9",
day = "1",
language = "English",
pages = "26--33",
booktitle = "Unknown Host Publication",

}

Burbridge, C & Condell, J 2010, Efficient Appearance-Based Topological Mapping and Navigation with Omnidirectional Vision. in Unknown Host Publication. pp. 26-33, TAROS 2010 - Towards Autonomous Robotic Systems 2010, 1/09/10.

Efficient Appearance-Based Topological Mapping and Navigation with Omnidirectional Vision. / Burbridge, Christopher; Condell, Joan.

Unknown Host Publication. 2010. p. 26-33.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Efficient Appearance-Based Topological Mapping and Navigation with Omnidirectional Vision

AU - Burbridge, Christopher

AU - Condell, Joan

PY - 2010/9/1

Y1 - 2010/9/1

N2 - Because of a mobile robot’s ability to move in itsenvironment, one of the most important and common tasks formobile robots is, arguably, the task of navigation. This has tobe carried out in practically every mobile robot, and it hasto be carried out quickly (real-time constraints) and cheaply(computational constraints). SLAM [7] is a reliable and veryaccurate method which is widely used now in mobile robotics.However, these algorithms are not computationally cheap. Inthis paper we therefore investigate a computationally efficientalgorithm for simultaneous mapping and navigation that issuitable for application in simple mobile robots. We apply selforganisingmaps in a novel way to image data compression andindexing of topological map nodes at the same time.Our proposed system builds a dense topological map using onlythe visual appearance of the environment, with no need for anyfeature extraction or matching. This is made possible throughthe novel use of a self-organising map and a relaxed attitudetowards loop closure and metric consistency. The inconsistenciesand uncertainties within the map are not considered duringmapping, but rather only during navigation at which point aBayesian approach is taken to allow accurate navigation. Thepaper concludes by presenting mobile robot experiments

AB - Because of a mobile robot’s ability to move in itsenvironment, one of the most important and common tasks formobile robots is, arguably, the task of navigation. This has tobe carried out in practically every mobile robot, and it hasto be carried out quickly (real-time constraints) and cheaply(computational constraints). SLAM [7] is a reliable and veryaccurate method which is widely used now in mobile robotics.However, these algorithms are not computationally cheap. Inthis paper we therefore investigate a computationally efficientalgorithm for simultaneous mapping and navigation that issuitable for application in simple mobile robots. We apply selforganisingmaps in a novel way to image data compression andindexing of topological map nodes at the same time.Our proposed system builds a dense topological map using onlythe visual appearance of the environment, with no need for anyfeature extraction or matching. This is made possible throughthe novel use of a self-organising map and a relaxed attitudetowards loop closure and metric consistency. The inconsistenciesand uncertainties within the map are not considered duringmapping, but rather only during navigation at which point aBayesian approach is taken to allow accurate navigation. Thepaper concludes by presenting mobile robot experiments

M3 - Conference contribution

SP - 26

EP - 33

BT - Unknown Host Publication

ER -