Investigation of Emerging Sensing and AI/ML Technologies to Enhance the Safety of Vulnerable Roadway Users at Signalized Intersection

Year: 2024-2026

Role: PI

Sponsor: US DOT – UTC TBD National Center

Description: Pedestrians, bicyclists, and other non-vehicle occupants are considered vulnerable roadway users (VRUs), accounting for around 19 percent of crash fatalities. However, it remains a great challenge to accurately identify and analyze VRUs due to their inclinations to deviate from well-defined routes, resulting in limited data sources available to inform the development of effective signal control strategies for enhancing intersection safety. This study aims to design an AI/ML (Artificial Intelligence/Machine Learning)-empowered data collection and analysis method by investigating and fusing two promising and complementary multimodal sensors, namely, Light Detection and Ranging (LiDAR) and Pan-tilt-zoom (PTZ) Network Cameras. Great efforts will be made to devise effective visual, temporal, and 3-D point cloud data (PCD) ML, especially deep learning (DL), approaches, to accurately identify and analyze the nature and behavioral properties of pedestrians, cyclists, and various vehicle types, providing high-quality information inputs for effective strategic planning and thus enhancing intersection safety.