Integrating and Learning on Spatial Data via Multi-Agent Simulation
PI : Jeff M. Phillips
co-PI : Xiaoyue Cathy Liu
NSF III-2311954 (June 2023 - May 2026)
This project builds foundation models for human mobility patterns on the scale of a city. This needs to capture many aspects of human mobility including movement and demographics, be statistically valid, and should be applicable to policy tasks. Towards this larger objective, there are three major technical goals:

1. We aim to formulate and build a general model that can incorporate available and demographic data, movement data from both global and local sources. This should be trainable through modern machine learning methods, but also be useful to parameterize simulations so it can be used to generate simulated data which allows for asking, and then answering, a variety of pressing questions.

2. We plan to design scalable and statistically meaningful mechanisms to query and interact with both the large real data used as input, and the simulated data generated from our model. These should be able to discover interesting anomalous patterns in the spatial mobility data.

3. We set out to leverage the foundation model and query mechanisms towards intervention and prediction tasks. These may involve challenges which require a blend of demographic and mobility information such as assessing if profiling occurs in traffic stops, or optimizing the placement of charging stations.

Moreover, we aim to test out the use of these models, methods, and analysis to help influence policy, and further pressing educational development in related areas of data science and deep learning for training next-generation spatial scientists.

Publications:
  • A data-driven framework for agent-based modeling of vehicular travel using publicly available data.
    Yirong Zhou, Xiaoyue Cathy Liu, Bingkun Chen, Tony Grubesic, Ran Wei, and Danielle Wallace. Computers, Environment and Urban Systems, v.110. 2024

  • AI-enabled airport runway pavement distress detection using dashcam imagery.
    Arman Malekloo, Xiaoyue Cathy Liu, and David Sacharny. Computer-Aided Civil and Infrastructure Engineering. 2024

  • Trajectory Minimum Touching Ball.
         Jens Kristian Refsgaard Schou and Jeff M. Phillips. arXiv:2505.02472. May 2025.

    Educational Development and Outreach:
    We organized two Utah data science events that helped promote spatial data science.
  • 2024: Data Science Day
  • 2025: Data Science & AI Day, where co-PI Liu gave a Research Highlights talk on this project.

  • The PIs are leading the development of AI emphasis and minors, with aim to develop training programs to benefit the next generation of spatial data scientists. More details to come.

    Personnel:
  • PI: Jeff Phillips
  • co-PI : Xiaoyue Cathy Liu
  • GRA: Arman Malekloo
  • GRA: Hamid Shafieasl
  • GRA: Shouzheng Pan