Decoding Road-User Intent: Advancing Surrogate Safety Measures through Behavior-Aware Generative World Models

This project investigates how roadside LiDAR sensing and generative scene prediction models can support more informative and uncertainty-aware traffic safety assessment at urban intersections. Conventional surrogate safety measures, such as time-to-collision and post-encroachment time, are widely used to identify potential traffic conflicts. However, these measures are primarily based on observed kinematic relationships and may not fully capture road-user behavior, interaction context, or uncertainty caused by sensing noise, occlusion, and tracking errors. The project will develop and evaluate a pilot infrastructure-centric modeling framework that uses roadside LiDAR data to forecast short-term movements of vehicles, pedestrians, cyclists, and other road users from a fixed roadside perspective. Rather than treating observed trajectories as deterministic, the framework will explore probabilistic scene prediction methods that represent multiple plausible future interactions and provide uncertainty estimates. Building on these forecasts, the project will develop prototype intention-aware and uncertainty-aware surrogate safety measures. These measures may include probabilistic conflict indicators, uncertainty-calibrated risk scores, and other metrics that supplement conventional kinematic safety measures. The proposed metrics will be evaluated using roadside LiDAR data and manually reviewed conflict events, with attention to prediction accuracy, uncertainty calibration, interpretability, and practical relevance for proactive safety analysis. The project will also produce implementation-oriented resources, including a prototype Roadside Safety Forecasting Toolkit, example workflows, evaluation scripts, documentation, and a practitioner-facing brief.

Language

• English

Project

• Status: Active
• Funding: $200,000.00
• Contract Numbers:

  69A3552344815

  69A3552348320

• Sponsor Organizations:

  Office of the Assistant Secretary for Research and Technology

  University Transportation Centers Program
  Department of Transportation
  Washington, DC  United States  20590
• Managing Organizations:

  University of Texas at Austin

  Austin, TX  United States  78712
• Performing Organizations:

  City College of New York

  Civil Engineering, Steinman T-127
  140th Street and Convent Avenue
  New York, NY  United States  10031
• Principal Investigators:

  Li, Yiqiao

• Start Date: 20260601
• Expected Completion Date: 20270531
• Actual Completion Date: 0
• USDOT Program: University Transportation Centers

Subject/Index Terms

• TRT Terms: Behavior; Intersections; Laser radar; Predictive models; Safety analysis; Traffic safety; Uncertainty; Urban areas
• Subject Areas: Data and Information Technology; Highways; Operations and Traffic Management; Pedestrians and Bicyclists; Planning and Forecasting; Safety and Human Factors;

Filing Info

• Accession Number: 02001769
• Record Type: Research project
• Source Agency: Center for Understanding Future of Travel Behavior and Demand
• Contract Numbers: 69A3552344815, 69A3552348320
• Files: UTC, RIP
• Created Date: Sep 4 2026 3:14PM