Massive Trajectory Data Based on Patterns of Life

Sumamry:

This study introduces a large-scale simulated human mobility dataset generated using an agent-based simulation referred as Patters of Life (POL) simulation. Addressing limitations of existing mobility datasets—like GeoLife—that are small, non-representative, and privacy-constrained, this dataset enables extensive research in human mobility patterns. The simulated data captures realistic trajectories, check-ins, and social interactions for agents in various U.S. cities, making it accessible for human mobility and social network studies without privacy concerns.

Methodology:

The simulation is based on “Patterns of Life,” where agents emulate realistic daily routines, driven by needs like socialization, work, and food. The simulation integrates real-world map data, generating location-based social network (LBSN) data that records social interactions, check-ins, and movement between points of interest. The data can be locally regenerated to avoid large file transfers.

Results:

The generated dataset encompasses hundreds of gigabytes, capturing millions of trajectories, social links, and check-ins across four cities. Each simulation instance provides dense, plausible human mobility data, enabling robust testing for applications such as traffic management, urban planning, and social network analysis. Comparisons show the data surpasses existing real-world datasets in scale and diversity.

Conclusion:

This simulated dataset provides a valuable resource for studying human mobility patterns at scale, free from privacy and sparsity issues. The study contributes reusable simulation code and parameters, enabling researchers to create region-specific data. Future improvements will focus on adding more nuanced agent movement and expanded social interaction dynamics.

Full Paper: