People Flow Estimation in Urban Environments Using BLE Advertising Packets

Abstract

In recent years, the analysis of people flow has become increasingly important for urban management and commercial strategies. While cameras and 3D sensors have been used for people flow estimation, they involve high costs and raise privacy concerns. In contrast, Wi-Fi and Bluetooth signals enable low-cost, non-intrusive data collection. However, widespread MAC address randomization in modern devices limits the effectiveness of conventional people tracking methods that rely on static MAC addresses. To overcome this limitation, we propose a method to estimate movement trajectories by associating dynamically randomized MAC addresses. Our approach utilizes the appearance/disappearance times and advertising (Adv) data of BLE packets collected by multiple scanners. Unlike prior work that focused on single-sensor setups or simulations, we conducted a citywide field experiment spanning 2.5 km in central Saga City, Japan. Thirty BLE scanners were deployed over a 60-day period, and the collected data were evaluated against GPS logs as the ground truth. The proposed method achieved an average macro coverage of 70.9% for walking, 46.2% for cycling, and 66.0% overall. This represents improvements of 9.5%, 3.4%, and 8.3% over baseline methods, respectively.

Award

Links

code:references.bib

@inproceedings{bib:kotaro_BReID-Walk_IJABC2026,

author={Hayashi, Kotaro and Yoshimura, Taito and Kobayashi, Hiroyuki and Imamura, Ryo and Itou, Mina and Ushijima, Kiyohide and Suwa, Hirohiko and Matsuda, Yuki},

title={{People Flow Estimation in Urban Environments Using BLE Advertising Packets}},

journal={International Journal of Activity and Behavior Computing},

volume={2026},

number={1},

year={2026},

pages={1--17},

doi={10.60401/ijabc.132}

}

https://scrapbox.io/files/698fb8663583dfb074ce4540.png

https://scrapbox.io/files/69b95412b9e150750ae75a93.png

Category

Conference

Project

Keywords

Collaborating Organization