Yilin Wang

Yilin Wang

Ph.D. Candidate @ Purdue | Cooperative Perception, Physical AI, Autonomous Driving

Developing a Cooperative Perception System

  • Role: Project Leader
  • Period: Jan 2023 – Jun 2025
  • Affiliation: CART Lab, Purdue University
  • Funding: Phase 1 by the NSF Cyber-Physical Systems (CPS) program; Phase 2 by the Center for Connected and Automated Transportation (CCAT), a USDOT University Transportation Center
  • Status: completed (Jun 2025)
  • Links: Phase 1 paper in Transportation Research Record

Phase 1: A Cooperative Perception System for CAV Navigation and Improving Safety

Funded by the NSF award “CPS: Medium: Collaborative Research: Transforming Connected and Automated Transportation with Smart Networking, Cooperative Sensing, and Edge Computing”

  • Objective: implement a cooperative perception system to improve road safety.
  • Contributions:
    1. Built a cooperative perception system that connects roadside LiDAR and the sensors of connected and automated vehicles (CAVs) through V2X.
    2. Developed a detection and navigation algorithm on the CAV with a sensor data fusion pipeline.
    3. Built and ran field tests demonstrating the safety benefits of detecting occluded vulnerable road users (VRUs) through the cooperative perception system.
  • Findings: the cooperative perception system helps CAVs plan safer trajectories (higher post-encroachment time) and smoother ones (lower deceleration rates).
  • Publication: Hanlin Chen, Vamsi K Bandaru, Yilin Wang, Mario A Romero, Andrew Tarko, Yiheng Feng (2024). “Cooperative Perception System for Aiding Connected and Automated Vehicle Navigation and Improving Safety.” Transportation Research Record, 2678(12), 1498-1510.

Phase 2: An Infrastructure-Vehicle Cooperative Perception System

Jul 2024 – Jun 2025. Funded by the CCAT grant “Cost Benefit Analysis of an Infrastructure-Vehicle Cooperative Perception System”

  • Objective: develop a cooperative perception system that includes multiple vehicles, infrastructure sensors, and drones.
  • Methodology: multi-sensor fusion of infrastructure sensors, drone-view cameras, and vehicle sensors.
  • Contribution: extracted and processed multi-sensor data (LiDAR, camera, and others) into vehicle trajectories; an open-source dataset is being prepared for public release.