Localization in Wireless Sensor Networks Using a Mobile Robot
Loading...
Journal Title
Journal ISSN
Volume Title
School of Electrical Engineering |
Doctoral thesis (monograph)
| Defence date: 2016-04-15
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.
Author
Date
2016
Major/Subject
Mcode
Degree programme
Language
en
Pages
272 + app. 34
Series
Aalto University publication series DOCTORAL DISSERTATIONS, 56/2016
Abstract
This thesis presents studies and methods relevant to the problem of localization in wireless sensor networks (WSN), with the ultimate goal of producing practical solutions that can be used in real adhoc deployments. The motivational sample application type is emergency and rescue operations, which are characterized by the lack of a pre-installed infrastructure and on-site training data. The base scenario is an unexplored environment in which the nodes of a WSN are distributed in random unknown positions. A robot capable of simultaneous localization and mapping is used as a mobile beacon, with the help of which the nodes' position can be estimated accurately using received signal strength (RSS) measurements. Using data collected in three different environments, we demonstrate sub-metre accuracy for some of the proposed methods, part of which are self-adaptive and can cope with changes in the environment. The localization algorithms are based on least squares (LS) and maximum likelihood (ML) estimation relying on parametric measurement models. In order to obtain a realistic confidence indicator on the estimates, special attention is paid to the calculation of their covariance. The work presented includes studies on the variability of the log-normal model parameters typically observed in WSNs, the sensitivity of ML position estimators due to this variability, the over-confidence of the estimates under the assumption of identically and independently distributed errors, the spatial autocorrelation of the RSS and the usage of concentrated log-likelihoods to jointly estimate position and model parameters while solving identifiability and convergence issues using regularization. In addition to ML, three families of linear position estimators are studied and evaluated, two of which are new. Unlike non-linear methods, they do not require initial estimates. Two of them have closed analytical forms, and therefore are computationally efficient. The third is iterative, and it has demonstrated an excellent performance comparable to ML in our experiments. All in all, besides contributing to the field of localization, this work represents a small step towards understanding and leveraging the potential benefits of using mobile robots as assistive localization devices.Description
Supervising professor
Kyrki, Ville, Prof., Aalto University, Department of Automation and Systems Technology, FinlandHalme, Aarne, Prof., Aalto University, Department of Automation and Systems Technology, Finland
Keywords
localization, wireless sensor networks, mobile robot, beacons, received signal strength, channel modelling, spatial autocorrelation, joint localization and model parameters estimation, regularization
