Impact Analysis of Maneuver and Avoidance Performance of Autonomous Vehicles on Pedestrians' Crossing Decisions

Abstract

In order for autonomous vehicles (AVs) to get public acceptance by pedestrians, it is important not just to technically improve the avoidance performance of AVs to pedestrians, but also for pedestrians to behave properly in front of AVs without overconfidence of their performance. This study aims at clarifying the impact of deceleration maneuver and avoidance performance reliability of AVs on pedestrians’ crossing behavior at unsignalized crosswalks. Virtual reality experiments were conducted so that participants as crossing pedestrians at unsignalized crosswalks experience the approaching AVs with various maneuver and reliability settings. Crossing decisions and their timings, subjective impressions at each trial were collected. A time-dependent pedestrian crossing decision model was developed from the experimental data. The proposed model revealed that not only the expected arrival time of AVs but also the pedestrians’ belief of avoidance reliability and the frequency with which the participants usually give way as drivers affect their crossing decisions.