Modeling the Pedestrian Crossing Decision Behavior Based on Vehicle Deceleration Patterns Using Virtual Reality Environment

Abstract

Since a considerable number of traffic fatalities are crossing pedestrians all over the world, understanding the characteristics of the pedestrian’s decision to cross unsignalized crosswalks is important for applying proper countermeasures. This study aims to clarify the quantitative impact of approaching vehicle maneuvers and road geometry on pedestrian crossing decision probability and its timing. A virtual reality experiment was designed to collect pedestrian crossing behaviors at mid-block crosswalks under various conditions, including yielding and non-yielding vehicles to pedestrians. A binary logistic regression model was developed to describe the crossing decision probability at each time. This model can be helpful in designing the motion planning of AVs to improve their reliability on the road.