PI: Katherine Flanigan
Co-PI(s): Christopher McComb
University: Carnegie Mellon University
Industry partner: Robert Bosch LLC (Bosch Research and Technology Center)
A key challenge along Pennsylvania intersections and roadways (high-severity/fatal accident rate exceeds the national average), is gaining an understanding of the interactions between road users for improved safety operations. Recent years have seen the proliferation of cameras supporting computer vision (CV) aimed at classifying and tracking vulnerable travelers (e.g., pedestrians). However, work has focused on improving near-miss detection, mobility classification, and identification to capture aggregate statistics. CV-based approaches face numerous challenges—occlusion, poor anomalous activity detection, and poor re-identification, to name a few—preventing their ability to inform real-time prediction for operational management. This research overcomes these gaps by focusing not on what CV can do, but rather, on what it cannot do. That is, how can we help cameras “see” what they cannot see? With Bosch RTC, this research proposes the fusion of Wi-Fi and cameras on a standalone device for use in intersections and roadways to overcome important limitations of camera-based solutions, as well as to enable two-way communication between travelers and surrounding infrastructure, which supports real-time feedback. The research and technology development is widely generalizable and will benefit the City of Pittsburgh and other PA cities seeking to prioritize investments in user safety with an aim of more equitable outcomes.