Synthetic People Flow Generation

Human Mobility
Privacy
Simulation

Synthetic people flow generation models people's movement and stay tendencies from large-scale mobility logs and generates city-scale synthetic people-flow data. By generating data that does not contain real users' trajectories, it aims to build people-flow data that is easy to use for urban planning, traffic analysis, epidemic simulation, and other applications.

Journal Paper / Conference Paper / CiNii / Award

Visualization of generated synthetic people flow (Nagoya, 240k agents)

Overview

Urban planning, traffic simulation, and epidemic risk assessment all rely on people-flow data that describes how people move and stay across an entire city. However, real mobility logs can contain personal information such as home and workplace locations, so they are not easy to share or use directly.

This research learns users' movement and stay tendencies from large-scale GPS-based mobility logs and generates synthetic people-flow data that does not directly contain real users' trajectories. Specifically, it represents each user's behavioral tendencies through Agent2Vec-based mobility modeling, and generates synthetic agents that move through the city from the trained model. By probabilistically determining each agent's destinations and stays, it builds synthetic trajectory data that approximates the city-wide people flow.

This approach makes it possible to generate data usable for mobility analysis and simulation without directly sharing real users' mobility logs. Because it models user movement tendencies through unsupervised learning, it requires no special labeling and can generate city-scale synthetic people flow from GPS-based mobility logs.

Example of one synthetic agent's daily trajectory (visualized in OrGIS, a web app under development at Ornis)

Demo / Web App

Coming soon.

Papers

Journal Article

田村直樹, 浦野健太, 青木俊介, 米澤拓郎, 河口信夫.
都市を対象とした大規模移動履歴に基づく疑似人流データ生成手法.
情報処理学会論文誌, Vol.64, No.1, pp.123–133, 2023.
DOI: 10.20729/00223417

International Conference

Naoki Tamura, Kenta Urano, Shunsuke Aoki, Takuro Yonezawa, and Nobuo Kawaguchi.
Synthetic People Flow: Privacy-Preserving Mobility Modeling from Large-Scale Location Data in Urban Areas.
MobiQuitous 2021, pp.553–567, 2022.
DOI: 10.1007/978-3-030-94822-1_36

Domestic Conference

田村直樹, 浦野健太, 青木俊介, 米澤拓郎, 河口信夫.
都市を対象とした大規模移動履歴に基づく疑似人流データ生成手法.
Multimedia, Distributed, Cooperative, and Mobile Symposium (DICOMO2021), 2021.07. Best Paper Award & Outstanding Presentation Award

@article{tamura2023syntheticHumanFlow,
  title={都市を対象とした大規模移動履歴に基づく疑似人流データ生成手法},
  author={田村 直樹 and 浦野 健太 and 青木 俊介 and 米澤 拓郎 and 河口 信夫},
  journal={情報処理学会論文誌},
  volume={64},
  number={1},
  pages={123--133},
  year={2023},
  doi={10.20729/00223417}
}

@incollection{tamura2022syntheticPeopleFlow,
  title={Synthetic People Flow: Privacy-Preserving Mobility Modeling from Large-Scale Location Data in Urban Areas},
  author={Tamura, Naoki and Urano, Kenta and Aoki, Shunsuke and Yonezawa, Takuro and Kawaguchi, Nobuo},
  booktitle={Mobile and Ubiquitous Systems: Computing, Networking and Services},
  pages={553--567},
  year={2022},
  publisher={Springer},
  doi={10.1007/978-3-030-94822-1_36}
}

@inproceedings{tamura2021syntheticPeopleFlowDicomo,
  title={都市を対象とした大規模移動履歴に基づく疑似人流データ生成手法},
  author={田村 直樹 and 浦野 健太 and 青木 俊介 and 米澤 拓郎 and 河口 信夫},
  booktitle={マルチメディア，分散，協調とモバイルシンポジウム（DICOMO2021）},
  year={2021}
}