A map-matching algorithm dealing with sparse cellular fingerprint observations

The widespread availability of mobile communication makes mobile devices a resource for the collection of data about mobile infrastructures and user mobility. In these contexts, the problem of reconstructing the most likely trajectory of a device on the road network on the basis of the sequence of o...

Full description

Saved in:
Bibliographic Details
Published inGeo-spatial information science Vol. 22; no. 2; pp. 89 - 106
Main Authors Dalla Torre, Andrea, Gallo, Paolo, Gubiani, Donatella, Marshall, Chris, Montanari, Angelo, Pittino, Federico, Viel, Andrea
Format Journal Article
LanguageEnglish
Published Wuhan Taylor & Francis 03.04.2019
Taylor & Francis Ltd
Taylor & Francis Group
Subjects
Accuracy
Algorithms
Availability
cellular fingerprint
Cellular telephones
Communication
Energy conservation
Energy consumption
Fingerprints
Global positioning systems
GPS
Hidden Markov Model
Map-matching algorithm
Markov chains
Matching
Roads
Satellite observation
Spatial data
trajectory
Urban environments
Online AccessGet full text

Cover

More Information
Summary:The widespread availability of mobile communication makes mobile devices a resource for the collection of data about mobile infrastructures and user mobility. In these contexts, the problem of reconstructing the most likely trajectory of a device on the road network on the basis of the sequence of observed locations (map-matching problem) turns out to be particularly relevant. Different contributions have demonstrated that the reconstruction of the trajectory of a device with good accuracy is technically feasible even when only a sparse set of GNSS positions is available. In this paper, we face the problem of coping with sparse sequences of cellular fingerprints. Compared to GNSS positions, cellular fingerprints provide coarser spatial information, but they work even when a device is missing GNSS positions or is operating in an energy saving mode. We devise a new map-matching algorithm, that exploits the well-known Hidden Markov Model and Random Forests to successfully deal with noisy and sparse cellular observations. The performance of the proposed solution has been tested over a medium-sized Italian city urban environment by varying both the sampling of the observations and the density of the fingerprint map as well as by including some GPS positions into the sequence of fingerprint observations.
ISSN:1009-5020
1993-5153
DOI:10.1080/10095020.2019.1616933