FlowSense Traffic Flows - estimated from vehicle trajectories based on sparse mobile phone geolocation data

This repository contains data of estimated traffic flows for every street segment in Stockholm and Gothenburg, Sweden, from sparse mobile phone geolocation data, as well as road network and ground truth data. Geolocation points and trajectories cannot be made public due to privacy restrictions. The estimated flows yield significant and (very) strong correlations with the various ground truth datasets. This research is part of the FlowSense research project. With this research project, and making our code and data openly available, we intend to open up new possibilities for data-driven transport, environment and planning research as well as practice.

More information can be found in the accompanying paper "Estimating traffic flows from vehicle trajectories based on sparse mobile phone geolocation data", to be presented in October 2025 at the NetMob 2025 conference.

Using the FlowSense Traffic Flows dataset?

If you use this repository in your work, please cite the accompanying conference paper:

Citation info: [accepted for publication] Teeuwen, R., & Gil, J. (2025). Estimating traffic flows from vehicle trajectories based on sparse mobile phone geolocation data. NetMob 2025 conference. Paris, France: Conservatoire national des arts et métiers.

Code: Associated code to process this data and to estimate traffic flows are available at https://github.com/rflteeuwen/flowsense_trafficflows

License: This repository is licensed under the GNU General Public License v3.0 (GPL-3.0).

Abstract: Large-scale empirical traffic flow data are instrumental in spatial planning, transport research, and data-driven decision making. However, existing data, collected using traffic sensors or counts, lack spatio-temporal coverage and granularity. While mobile phone geolocation data are deemed promising for capturing traffic at scale given their size, granularity, and coverage, their potential for such cases remains unexplored. In this study, we investigate how traffic flows with high granularity and extensive coverage can be estimated from vehicle trajectories based on sparse mobile phone gelocation data. We introduce our data-processing methodology, implement it on two major cities in Sweden, and compare our outcomes to ground truth data. The flows resulting from our methodology yielded significant and strong correlations with the ground truth data. At highway locations in Gothenburg, very strong correlations were observed when selecting trajectories of at least 20 km/h on average as input. Future work remains to understand where and why deviations remain, and how correlations can be strengthened further, opening up new possibilities for data-driven transport, environmental and planning research as well as practice.

Acknowledgements: This research is funded by Chalmers University of Technology Transport Area of Advance under grant agreement no. 2024-0299.