SF-LIFE: A Large-Scale Simulated Movement Dataset for the San Francisco Bay Area
This paper introduces SF-LIFE, a large-scale, privacy-preserving simulated dataset containing 3.024 trillion noise-free, multi-modal location records for 500,000 agents in the San Francisco Bay Area, designed to overcome the limitations of real-world tracking data and accelerate research in transportation, mobility, and machine learning.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine you want to study how people move around a giant city like San Francisco. In the real world, this is like trying to film a movie with a shaky camera, missing scenes, and actors who refuse to be on screen because of privacy rules. You get a messy, incomplete picture.
The paper introduces SF-LIFE, which is like a perfect, crystal-clear, 70-day movie of 500,000 imaginary people living in the San Francisco Bay Area. It's a massive, computer-generated dataset designed to help researchers understand transportation and human movement without the headaches of real-world data.
Here is a simple breakdown of how this "movie" was made and what it contains:
1. The Cast: 500,000 Digital People
The dataset features 500,000 simulated agents. These aren't real people; they are digital characters created by a computer.
- Who are they? They have realistic lives. Some are workers, some are students, and some are homemakers. They have ages, genders, and even different types of homes and jobs, all based on real census data.
- How do they think? Instead of just wandering randomly, they are driven by needs, much like Maslow's hierarchy of needs.
- Hungry? They go to a restaurant.
- Bored? They go to a park or a bar with friends.
- Need to work? They go to their office.
- Need to sleep? They go home.
- Social? They meet friends.
The computer simulates these needs ticking away until the "need" is high enough to trigger a trip.
2. The Stage: A Perfect Map of the Bay Area
The simulation takes place on a digital version of the San Francisco Bay Area built from OpenStreetMap (a free, open-source map).
- The Set: It includes every building (homes, schools, offices, restaurants) and every road.
- The Bus System: It connects to real public transit schedules from over 40 different transit agencies (like BART, buses, and ferries) across 9 counties.
- The Rules: The agents follow the rules of the road. If a bus leaves at 8:00 AM, the agent waits for it. If they have a car, they drive. If they are walking, they walk.
3. The Plot: A 70-Day Story
The dataset covers 70 days of continuous movement.
- The Script: First, the computer decides what each agent plans to do each day (their "agenda"). For example: "Wake up, go to work, eat lunch, go to a bar, go home."
- The Action: Then, a routing engine (called Valhalla) figures out exactly how they get there. It calculates the fastest path, whether that's walking, biking, driving, or taking a train.
- The Result: The output is a 1Hz frequency recording. This means the dataset captures the exact location of every single agent once every second for 70 days. That adds up to 3 trillion location records.
4. Why This "Movie" is Special
Real-world data (like GPS from phones) is often like a broken film reel:
- Missing Scenes: Phones lose signal in tunnels or tall buildings.
- Static Noise: Data is often "noisy" or inaccurate.
- Privacy: You can't track real people without their permission.
SF-LIFE fixes all of this:
- No Missing Scenes: Every second is recorded perfectly.
- No Noise: The data is clean and precise.
- Privacy Safe: Since the people are imaginary, there are no privacy concerns.
- Full Picture: It includes every mode of transport (walking, biking, bus, train, car) in one big network.
5. What's in the Box?
The researchers are giving this dataset away for free. It comes with:
- The Trajectories: The 3 trillion location points showing where everyone was, every second.
- The Agendas: The "scripts" showing what people planned to do.
- The Cast List: Demographic info (age, job, gender) for every agent.
- The Map: The underlying road and building data used to build the world.
The Bottom Line
Think of SF-LIFE as a massive, high-definition sandbox for scientists. Instead of trying to figure out how traffic works by watching messy, real-world cameras, researchers can now run experiments on this perfect, simulated city. They can test new traffic lights, plan better bus routes, or train AI to predict movement, all without worrying about privacy or missing data. It's a "what-if" machine for the San Francisco Bay Area.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.