Vehicle-pedestrian interaction analysis for evaluating pedestrian crossing safety at uncontrolled crosswalks − a geospatial approach using multimodal all-traffic trajectories

Abstract

Introduction: Pedestrian crossing safety has gained increased attention due to the high rate of pedestrian fatalities and injuries, especially at uncontrolled crosswalks. Method: In this study, we proposed a novel GIS-based method for detecting motorist yield behaviors using multi-modal trajectory data collected from LiDAR (Light Detection and Ranging) sensors at uncontrolled crosswalks. The approach classifies diverse types of motorist-pedestrian interactions and calculates motorist compliance rates, enabling us to assess the safety performance of different geometric crossing treatments. The method was applied to four uncontrolled crosswalks in midtown Reno, NV to analyze the impact of different crossing treatments, including curb extensions, pedestrian refuge islands, and Danish Offset, on motorist yield rates. Results: The findings indicated that refuge islands significantly improve driver yield rates, with further improvement observed when implementing Danish Offset designs. Among the four sites, the highest motorist yield rate (78.0%) was observed at Taylor (Danish Offset), followed by St. Lawrence (refuge island) with 71.9%. Martin and LaRue (curb extension only) exhibited lower yield rates of 57.9% and 61.3%, respectively. Practical applications: This study emphasized the importance of considering different directions when evaluating pedestrian safety at crosswalks, an aspect currently not considered in the latest Highway Capacity Manual (HCM). This research also provides valuable insights into applying multimodal all-road-user geospatial trajectory data for initiative-taking traffic safety performance evaluation of pedestrian crossing facilities at uncontrolled crosswalks and can guide future efforts in improving pedestrian safety.