Improving vehicle speed estimates using street network centrality

ABSTRACT

This paper describes a novel approach to improve prediction models which estimate vehicle speeds and their diurnal variation for road network links in urban street networks using only static map attributes. The presented approach takes into account previously neglected spatial information by integrating network centrality measures for closeness (indicating how central a link is) and betweenness (indicating how important a road link is) into the prediction model. The model is calibrated with a real-world dataset of 100 million individual speed measurements from a fleet of 3500 taxi probe vehicles in Vienna, Austria. Given that centrality can be derived directly from readily available street network data, the experimental results demonstrate that integrating centrality measures considerably improves the predictions without the need for adding a supplementary data source. Improvements for vehicle speed estimates are particularly prevalent on important street network links in the city center as well as rural streets in the periphery.

RÉSUMÉ

Ce papier décrit une nouvelle approche pour améliorer un modèle prédictif dont le but est d’estimer les vitesses des véhicules et leurs variations diurnes sur un réseau routier en relation avec un réseau de rues, en n’utilisant que des attributs cartographiques statiques. L’approche présentée prend en compte une information spatiale négligée précédemment en intégrant dans le modèle prédictif deux mesures de centralité : la centralité de proximité ’closeness’ (qui estime la centralité d’un arc) et la centralité intermédiaire ‘betweeness’ (qui estime l’importance d’un arc routier). Le modèle est calibré avec des données réelles à partir d’une base de 100 millions d’enregistrements de vitesses correspondant à une flotte de 3500 taxi à Vienne en Autriche. La centralité est directement dérivée des données de réseau de rues disponibles. Les résultats expérimentaux montrent que l’intégration des mesures de centralité améliore considérablement les prédictions sans avoir à ajouter des informations supplémentaires. Les améliorations de l’estimation de la vitesse des véhicules sont particulièrement significatives sur le réseau des rues importantes en centre-ville ainsi que sur les routes du réseau rural en périphérie.

KEYWORDS:

• Mobility
• network centrality
• traffic speed
• travel time
• prediction models

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Anita Graser http://orcid.org/0000-0001-5361-2885

M. Leodolter http://orcid.org/0000-0003-4531-7012

H. Koller http://orcid.org/0000-0002-4255-3530

N. Brändle http://orcid.org/0000-0002-2976-3138

Notes on Contributors

Anita Graser graduated in Computer Sciences focusing on Geomatics and received her MSc from the University of Applied Sciences in Wr. Neustadt, Austria in 2010. In 2006, she was a visiting student at the University of Technology Prague, Czech Republic. From 2007 to 2011, Anita worked as a contractor for Arsenal Research, later AIT. Since 2011 she has been part of the Dynamic Transportation Systems Team at AIT Mobility. Her work focuses on problems in the area of geographic information sciences, analysis and visualization of spatiotemporal data, especially concerning aspects of data quality, as well as research project management. Anita currently teaches at UNIGIS Salzburg and the Technical University in Vienna and is a member of the QGIS project steering committee and OSGeo board of directors.

Maximilian Leodolter studied Mathematics in Economics at the Vienna University of Technology. 2013 he finished his master studies with honours with his diploma thesis “Analysis of datasets with missing values; multiple imputation for traffic data”. Maximilian joined the Dynamic Transportation Systems team at AIT Mobility in 2013 and works on the statistical analysis and modelling of mobility data and behaviour.

Hannes Koller studied Applied Computer Science at Klagenfurt University where he received his Diplomingenieur degree in 2005 at the Institute for Intelligent Systems and Business Informatics. During his course he worked for a number of companies as a software developer. Since 2005 he has been working as an expert for Linux and software engineering at the Dynamic Transportation Systems team of AIT Mobility. Hannes is mainly focusing on traffic telematics, especially development of server software for data analysis and machine learning. Previous work includes the design and development of applications in the areas of knowledge management, artificial intelligence, computational linguistics and data mining.

Norbert Brändle graduated in Computer Sciences and received his PhD from the Vienna University of Technology, Austria in 2002. In 1995 he was visiting student at the Université Joseph Fourier, Grenoble, France. Norbert worked as a researcher at the Novartis Research Institute Vienna (1996–1997) and assistant professor at the Vienna University of Technology at the Pattern Recognition and Image Processing Group (1998–2003). After joining Arsenal Research in 2004, he was a key person to shape the path from the five-person business unit ‘Human Centered Mobility Technologies’ to AIT's 38-person business unit ‘Dynamic Transportation Systems’, both in terms of research topics and staffing. From 2009–2012 he was deputy head of the business unit. From 2006–2009 he was a lecturer for multimedia information systems at the University of Applied Sciences Technikum Wien. As a senior scientist at AIT for the field of mobility data analysis he leads and coordinates mobility-related research project acquisitions, their implementation and international publication of their results.

Additional information

Funding

This work was partially funded by the Joint Programming Initiative Urban Europe through the Austrian Ministry for Transport, Innovation and Technology (bmvit) [grant number 847350] (project e4share).