Simulated Urban Dynamics

SF-LIFE

A full simulated, noise-free, open-access movement dataset for the San Francisco Bay Area.

3.02T Location Records
500K Simulated Agents
70 Days @ 1Hz Freq.
40+ Transit Agencies

Needs-Driven Agendas

Agents don't just move randomly. Built on a Maslowian behavioral framework, daily schedules emerge dynamically based on needs for food, recreation, social interaction, and mandatory obligations like work or school.

Multi-Modal Kinematics

Agendas are translated into high-fidelity trajectories using the Valhalla routing engine. Agents seamlessly navigate OpenStreetMap infrastructure across 9 counties utilizing walking, biking, driving, bus, and rail modes based on GTFS data.

Synthesized Population Demographics

The simulated population is procedurally generated to reflect realistic urban distributions, accurately categorized by behavior, occupation, and vehicle access.

Occupation

  • Workers ~53%
  • Students ~24%
  • Homemakers ~23%

Vehicle Access

  • Car Owners 48.3%
  • No Vehicle 40.6%
  • Bicycle Only 11.1%

Core Data Included

  • ✓ 1Hz Trajectory Records
  • ✓ Semantic Activity Agendas
  • ✓ OSM Road & Building Graphs

Scalable Sub-Datasets

Available in agent-centric and bucketed Parquet formats to support environments of all sizes.

Sampling Rate 15 Agents 100 Agents 1K Agents 10K Agents 500K Agents
1 Second (1Hz) 0.24 GB - - - 443.0 GB
5 Seconds - 0.36 GB 3.60 GB 37.8 GB -
1 Minute - 0.06 GB 0.67 GB 2.70 GB 6.20 GB
10 Minutes - - 0.07 GB 0.08 GB 1.70 GB