Crossing the Road in Front of Autonomous Vehicles: An Investigation of Pedestrian Behavior Using Survival Analysis

Abstract

As cities strive to create more livable environments and adapt to an emerging reality in which autonomous vehicles (AVs) share the road with vulnerable road users, questions remain about the safety of these technologies and how pedestrians will respond to them. This study examined pedestrian waiting times in front of both human-driven vehicles (HDVs) and AVs, and assessed the effects of vehicle type, local covariates, and regional heterogeneity on pedestrian behavior. Utilizing a large-scale multicity real-world AV data set from cities in North America and Asia, a random intercepts accelerated failure time (RI-AFT) survival model was applied to capture regional variations in crossing behavior. Results indicate that pedestrians tend to wait longer when interacting with HDVs than with AVs, suggesting an increased level of comfort with AVs. In addition, higher vehicle and pedestrian volumes are associated with shorter waiting times, which is consistent with the safety-in-numbers effect. In addition, severer interactions (i.e., lower postencroachment times) are linked to longer waiting durations, which reflects increased caution. Finally, the RI-AFT model demonstrates the importance of accounting for city-level heterogeneity, revealing that pedestrians in Singapore behave more cautiously than those in US cities (Boston, Las Vegas, and Pittsburgh). These findings provide insights that are useful for designing pedestrian-friendly infrastructure and developing targeted strategies to improve safety and efficiency in environments with growing AV adoption.