An empirical investigation of Pedestrians’ crossing behaviors and trust in human-driven versus autonomous vehicles on high-speed arterials: Insights from a virtual reality study

With the advancement of autonomous vehicle (AV) technology, the road environment is expected to be mixed between AVs, human-driven vehicles (HDVs), and other road users (e.g., pedestrians). The purpose of this study was to thoroughly investigate pedestrians’ crossing behaviors and trust towards AVs at high-speed arterial roads in a virtual reality (VR) environment and examine how such behaviors are comparable to their interactions with HDVs. It also sought to examine significant factors such as demographics, past behaviors/experiences along with visibility conditions, affecting various pedestrian behavior and safety measures.

Fifty-six adults participated in six different pedestrian crossing VR scenarios (336 crossings analyzed). Participants also responded to a survey asking about their confidence, predictability, dependability, responsibility, reliability, and faith in both HDVs and AVs. The survey results indicated higher confidence in HDVs, while AVs were rated higher in predictability, dependability, responsibility, and reliability. Several Generalized Linear Mixed Models were developed.

The results revealed significant differences in pedestrian behaviors towards AVs and HDVs. Overall, pedestrians experienced longer post-encroachment time (PET) with AVs compared to HDVs. Pedestrians also showed shorter waiting times at the curb and median and faster overall crossing times when dealing with AVs. Furthermore, AVs were associated with fewer pedestrian crashes compared to HDVs. Age and gender significantly influenced crossing behavior, in that pedestrians aged ≥ 25 took longer to cross, and males waited longer at the median. Familiarity with AVs reduced waiting times at curbs. Finally, low visibility resulted in longer waiting times and decreased safety, reflected by shorter PET.