A Language Modeling Framework for Generating and Adapting Human Mobility Trajectories

Takayuki Mizuno; Taizo Horikomi; Shouji Fujimoto; Atushi Ishikawa

Abstract

The ability to generate realistic and adaptive synthetic human mobility data is vital for applications ranging from urban planning and epidemiology to disaster response. However, most existing generative models are static, limiting their usefulness in dynamic, real-world scenarios. We introduce a paradigm shift by treating human mobility as a language. We represent sequences of locations and inter-event times as discrete tokens and train a Transformer (GPT-2) model from scratch to learn the underlying grammar of movement. Our approach offers two key capabilities. (1) Conditional generation: by prepending special tokens that encode personal attributes (e.g., gender, age) and environmental context (e.g., weekday/weekend, weather), the model produces trajectories consistent with subgroup-specific mobility patterns. (2) Rapid adaptation: a pre-trained mobility model can be fine-tuned to new, anomalous conditions (e.g., post-disaster mobility) using a small amount of data, achieving faster convergence and higher final accuracy than training from scratch. Across large-scale datasets, our Transformer outperforms Markov-chain and autoregressive baselines in long-horizon location prediction and inter-event time modeling, while closely matching real-world distributional statistics. These findings establish mobility-as-language as a powerful, flexible paradigm for controllable and adaptive trajectory simulation in social physics.

References

1.
Zhu, L.; Yu, F.R.; Wang, Y.; et al. Big data analytics in intelligent transportation systems: A survey. IEEE Trans. Intell. Transp. Syst. 2019, 20, 383–398.
2.
Chang, S.; Pierson, E.; Koh, P.W.; et al. Mobility network models of COVID-19 explain inequities and inform reopening. Nature 2021, 589, 82–87.
13
3.
Deb, P.; Furceri, D.; Ostry, J.D.; et al. The economic effects of COVID-19 containment measures. Open Econ. Rev. 2022, 33, 1–32.
4.
Mizuno, T.; Ohnishi, T.; Watanabe, T. Visualizing social and behavior change due to the outbreak of Covid-19 using mobile phone location data. Open Econ. Rev. 2021, 39, 453–468.
5.
Sudo, A.; Kashiyama, T.; Yabe, T.; et al. Particle filter for real-time human mobility prediction following unprecedented disaster. In Proceedings of the SIGSPACIAL’16: 24th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, Burlingame, CA, USA, 31 October–3 November 2016.
6.
Rotman, A.; Shalev, M. Using location data from mobile phones to study participation in mass protests. Sociol. Methods Res. 2020, 51, 1357–1412.
7.
Cutter, S.L.; Ahearn, J.A.; Amadei, B.; et al. Disaster resilience: A national imperative. Environ. Sci. Policy Sustain. Dev. 2013, 55, 25–29.
8.
WMO-UNISDR. Disaster Risk and Resilience; Thematic Think Piece; UN System Task Force on the Post-2015 UN Development Agenda: Geneva, Switzerland, 2012.
9.
Yabe, T.; Tsubouchi, K.; Sekimoto, Y. Cityflowfragility: Measuring the fragility of people flow in cities to disasters using GPS data collected from smartphones. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 2017, 1, 1–17.
10.
Yabe, T.; Tsubouchi, K.; Sudo, A.; et al. A framework for evacuation hotspot detection after large scale disasters using location data from smartphones: Case study of Kumamoto earthquake. In Proceedings of the SIGSPACIAL’16: 24th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, Burlingame, CA, USA, 31 October–3 November 2016; Volume 44.
11.
Fiore, M.; Katsikouli, P.; Zavou, E.; et al. Privacy in trajectory micro-data publishing: A survey. Trans. Data Priv. 2020, 13, 91–149.
12.
Mir, D.J.; Isaacman, S.; Caceres, R.; et al. Dp-where: Differentially private modeling of human mobility. In Proceedings of the 2013 IEEE International Conference on Big Data, Santa Clara, CA, USA, 6–9 January 2013; pp. 580–588.
13.
Pellungrini, R.; Pappalardo, L.; Simini, F.; et al. Modeling adversarial behavior against mobility data privacy. IEEE Trans. Intell. Transp. Syst. 2020, 23, 1145–1158.
14.
Schlapfer, M.; Dong, L.; O’Keeffe, K.; et al. The universal visitation law of human mobility. Nature 2021, 593, 522–527.
15.
Song, C.; Koren, T.; Wang, P.; et al. Modelling the scaling properties of human mobility. Nat. Phys. 2010, 6, 818–823. https://doi.org/10.1038/nphys1760.
16.
Luca, M.; Barlacchi, G.; Lepri, B.; et al. A survey on deep learning for human mobility. ACM Comput. Surv. 2023, 55, 1–44. https://doi.org/10.1145/3485125.
17.
Toch, E.; Lerner, B.; Ben-Zion, E.; et al. Analyzing large-scale human mobility data: A survey of machine learning methods and applications. Knowl. Inf. Syst. 2019, 58, 501–523.
18.
Jiang, W.; Zhao, W.X.; Wang, J.; et al. Continuous trajectory generation based on two-stage GAN. In Proceedings of the AAAI Conference on Artificial Intelligence, Washington, DC, USA, 7–14 February 2023; pp. 4374–4382.
19.
Chu, C.; Zhang, H.; Wang, P.; et al. Simulating human mobility with a trajectory generation framework based on diffusion model. Int. J. Geogr. Inf. Sci. 2024, 38, 847–878.
20.
Rao, J.; Gao, S.; Kang, Y.; et al. LSTM-TrajGAN: A Deep Learning Approach to Trajectory Privacy Protection. In the Proceedings of the 11th International Conference on Geographic Information Science, Pozna´n, Poland, 27–30 September 2021; Volume 12.
21.
Kong, X.; Chen, Q.; Hou, M.; et al. Mobility trajectory generation: A survey. Artif. Intell. Rev. 2023, 56, 3057–3098.
22.
Hong, Y.; Zhang, Y.; Schindler, K.; et al. Deep Generative Model for Human Mobility Behavior. arXiv 2025, arXiv:2510.06473. 2025.
23.
Abbar, H.; Kassan, S.; Bidet, F.; et al. TrajDD-GAN: A Synthetic Mobility Trajectory Generation Solution Based on Diffusion Models. IEEE Access 2025, 13, 158018–158032.
24.
Demetriou, A.; Alfsv˚ag, H.; Rahrovani, S.; et al. A Deep Learning Framework for Generation and Analysis of Driving Scenario Trajectories. SN Comput. Sci. 2023, 4, 251.
25.
Graser, A.; Jalali, A.; Lampert, J.; et al. MobilityDL: A review of deep learning from trajectory data. GeoInformatica 2024, 29, 115–147.
26.
Vaswani, A.; Shazeer, N.; Parmar, N.; et al. Attention is all you need. arXiv 2017, arXiv:1706.03762.
27.
Zhang, Q.; Gao, Y.; Zhang, Y.; ; et al. TrajGen: Generating Realistic and Diverse Trajectories With Reactive and Feasible Agent Behaviors for Autonomous Driving. IEEE Trans. Intell. Transp. Syst. 2022, 23, 24474–24487.
28.
Mizuno, T.; Fujimoto, S.; Ishikawa, A. Generation of individual daily trajectories by GPT-2. Front. Phys. 2022, 10, 1021176.
29.
Horikomi, T.; Mizuno, T. Generating in-store customer journeys from scratch with GPT architectures. Eur. Phys. J. B 2024, 97, 144.
30.
Barbi, T.; Nishida, T. Trajectory Prediction using Conditional Generative Adversarial Network. Proceedings of the 2017 International Seminar on Artificial Intelligence, Networking and Information Technology, Bangkok, Thailand, 2–3 December 2017; pp. 193–197.
31.
Fujimoto, S.; Ishikawa, A.; Mizuno, T. RoBERTa Trained from Scratch on GPS Trajectory Data. In Proceedings of IEEE/WIC International Conference on Web Intelligence and Intelligent Agent Technology, Venice, Italy, 26–29 October 2023; pp. 636–639.
32.
Li, D.; Li, C.; Wang, Y.; et al. Learning Personalized Driving Styles via Reinforcement Learning from Human Feedback. arXiv 2025, arXiv:2503.10434.
33.
Schwartz, J. Bing Maps Tile System (2024). Microsoft Corporation. Available online: https://learn.microsoft.com/enus/bingmaps/articles/bing-maps-tile-system (accessed on 28 December 2025).
34.
Brodsky, I. H3: Uber’s Hexagonal Hierarchical Spatial Index. Uber Engineering Blog. 2018. Available online: https://www.uber.com/blog/h3/ (accessed on 28 December 2025).
35.
Radford, A.; Wu, J.; Child, R.; et al. Language models are unsupervised multitask learners. OpenAI Blog 2019, 1, 9.
36.
Agoop-Corp. Dynamic Population Data (2025). Available online: https://www.agoop.co.jp/ (accessed on 1 October 2025).

Scilight Press

Author Information

Abstract

Keywords

References

About Scilight

Journals

Publishing Policies

Contact Us