Traffic speed modelling to improve travel time estimation in OpenRouteService

• Verfasst von: Ludwig, Christina [VerfasserIn]
• Verfasst von: Psotta, Julian [VerfasserIn]
• Verfasst von: Buch, Anna [VerfasserIn]
• Verfasst von: Kolaxidis, Nikolaos [VerfasserIn]
• Verfasst von: Fendrich, Sascha [VerfasserIn]
• Verfasst von: Zia, Mohammed [VerfasserIn]
• Verfasst von: Fürle, Johannes [VerfasserIn]
• Verfasst von: Rousell, Adam [VerfasserIn]
• Verfasst von: Zipf, Alexander [VerfasserIn]
• Titel: Traffic speed modelling to improve travel time estimation in OpenRouteService
• Verf.angabe: C. Ludwig, J. Psotta, A. Buch, N. Kolaxidis, S. Fendrich, M. Zia, J. Fürle, A. Rousell, A. Zipf
• E-Jahr: 2023
• Jahr: 22 Jun 2023
• Umfang: 18 S.
• Fussnoten: Gesehen am 01.08.2024
• Titel Quelle: Enthalten in: FOSS4G (2023 : Prizren)Free and Open Source Software for Geospatial (FOSS4G) 2023 - Academic Track
• Ort Quelle: Hannover : ISPRS, 2023
• Jahr Quelle: 2023
• Band/Heft Quelle: (2023), Seite 109-116
• DOI: doi:10.5194/isprs-archives-XLVIII-4-W7-2023-109-2023
• Datenträger: Online-Ressource
• Sprache: eng
• Sach-SW: Centrality
• Sach-SW: OSM
• Sach-SW: Routing
• Sach-SW: Traffic speed
• Sach-SW: Twitter
• K10plus-PPN: 1897360738

Abstract

Time-dependent traffic speed information at a street level is important for routing services to estimate accurate travel times and to recommend routes which avoid traffic congestion. Still, most open-source routing machines that use OpenStreetMap (OSM) as the primary data source rely on static driving speeds derived from OSM tags, since comprehensive traffic speed data is not openly available. In this study, a method was developed to model traffic speed by hour of day at a street level using open data from OpenStreetMap, Twitter and population data. The modelled traffic speed data was subsequently integrated into the open-source routing engine openrouteservice to improve travel time estimation in route planning. Machine learning models were trained for ten cities worldwide using traffic speed data from Uber Movement as reference data. Different indicators based on geolocation and timestamp of Twitter data as well as a geographically adapted betweeness centrality indicator were evaluated for their potential to improve prediction accuracy. In all cities, the Twitter indicators improved the model, although this effect was only visible for certain road types. The centrality indicator improved the model as well but to a lesser extent. The Google Routing API was used as reference to evaluate the accuracy in travel time estimation. Deviations in travel times were regionally different and were partly alleviated by including the raw traffic data by Uber or the modelled traffic speed data in openrouteservice.