Generative Mobility Simulation

Generative Mobility Simulation creates realistic, scalable synthetic populations and mobility traces for locations where representative data are sparse, restricted, or unavailable.

Overview

This project advances patterns-of-life simulation from individual demonstrations to high-performance data-generation systems. Agents are assigned homes, activities, schedules, destinations, and social contexts, producing trajectories and visits that remain interpretable because each movement is linked to a modeled behavioral process. The resulting data support controlled experiments at scales that are difficult to achieve with proprietary mobility datasets.

Research Questions

• How can synthetic mobility preserve realistic spatial, temporal, and semantic patterns?
• How can generation scale to large populations and long simulation horizons?
• How should synthetic mobility be evaluated for realism, utility, privacy, and bias?

Methods

• High-performance and distributed simulation
• Activity schedules and destination choice models
• OpenStreetMap-based environment construction
• Statistical validation against aggregate mobility patterns

Applications

• Training and evaluating mobility models
• Scenario generation for transportation and public health
• Privacy-conscious data sharing
• Stress testing spatial analytics and infrastructure

Selected Outputs

• HD-GEN software system
• Large-scale patterns-of-life datasets
• Scalable multi-modal mobility simulation frameworks