Intersection Safety for the Vulnerable

Vulnerable road users are considered people that are not in a vehicle and are, consequently, at a higher risk for serious injury because they have less crash protection than a vehicle occupant. Pedestrians, bicyclists, motorcyclists, and road workers are common vulnerable road users. Vulnerable road users can be further categorized by their degree of mobility, perception, and cognition. Shown in Figure 1 are numerous examples for each of the vulnerability categories. Vulnerabilities also have spatial and temporal dependencies, which can be quantified by a vulnerability index that might range from very low risk to very high risk (Figure 2). For example, running on a sidewalk in the middle of the day has an associated very low index. However, jaywalking across the road during rush hour might have a high or very high vulnerability index. The goal of the proposed work is to enhance the safety of the vulnerable at intersections because these are locations of planned conflict and thus have an inherent risk. To accomplish this goal, the research team proposes a cyber-physical system (Figure 3) that detects vulnerable road users, calculates a vulnerability index, then takes an appropriate action or actions to minimize the opportunity for injury. For example, if a person falls out of their wheelchair in the middle of a signalized intersection, all of the traffic signals would stay red, emergency medical vehicles would be dispatched, and audiovisual warnings would be broadcast. Warnings could also be sent via wireless communication to personal devices and even connected autonomous vehicles. The core of the system is based on detecting the vulnerable in visual data captured from cameras. Accomplishing this task requires an annotated dataset of people with vulnerabilities, e.g., walking cane, bicycle, etc. There are some public datasets available with annotated wheelchair users for example, but not nearly enough vulnerable road users are available. To fill the dataset gap, the research team will deploy cameras in areas with expected high vulnerable activity. If the team is unable to capture enough image examples, they will augment the dataset with synthetic images (e.g., project a 3D model of a man using crutches into an image) and/or perform enactments. Then the research team will train models for detecting the vulnerable, develop a method for calculating a vulnerability index, and develop a warning system. The research team have a longstanding deployment partnership with the City of Pittsburgh Department of Mobility and Infrastructure (DOMI). DOMI has already approved a camera deployment at the intersection of Forbes and Morewood. The research team also has a relationship with Easterseals Massachusetts in an advisory capacity for issues related to people with vulnerabilities.


  • English


  • Status: Active
  • Funding: $97194
  • Contract Numbers:


  • Sponsor Organizations:

    Carnegie Mellon University

    Safety21 National UTC for Promoting Safety
    Pittsburgh, PA  United States  15213

    Office of the Assistant Secretary for Research and Technology

    University Transportation Center Program
  • Managing Organizations:

    Carnegie Mellon University

    Safety21 National UTC for Promoting Safety
    Pittsburgh, PA  United States  15213
  • Project Managers:

    Stearns, Amy

  • Performing Organizations:

    Carnegie Mellon Universit

  • Principal Investigators:

    Narasimhan, Srinivasa

  • Start Date: 20230701
  • Expected Completion Date: 20240630
  • Actual Completion Date: 0
  • USDOT Program: University Transportation Centers

Subject/Index Terms

Filing Info

  • Accession Number: 01900368
  • Record Type: Research project
  • Source Agency: Safety21
  • Contract Numbers: 69A3552344811
  • Files: UTC, RIP
  • Created Date: Nov 21 2023 6:52PM