How Do Pedestrians Perceive the Risk Against an Oncoming Autonomous Vehicle? Visualization of the Pedestrian’s Risk Field

Abstract

User acceptance is important for the successful deployment of shared mobility services using autonomous vehicles (AVs). However, current engineers who design the control logic of AVs does not consider the perceived risk of other road users, especially pedestrians, surrounding the AV. The aim of this research was to develop a model of the “perceived risk field” of a pedestrian in relation to an AV, through an experiment in a virtual reality environment, LargeSpace. In the experiment, a pedestrian (a real participant) and an AV (a virtual object) approached each other under various conditions. For each trial, the pedestrians responded with assessments of their perceived risk (i.e., subjective responses they felt when faced with the approaching AV) and tolerance level. The results showed that the pedestrian’s risk field could be modeled by fitting the perceived risk values for each participant using a Gaussian function. However, the perceived risk and tolerance level differed participant to participant, making it difficult to derive a common control logic of AV control for pedestrians. On the other hand, we found the possibility of inferring perceived risk from pedestrian behavioral data. Although the modeling of the pedestrians’ perceived risk needs improvement, our approach is promising for establishing a clear model of pedestrians’ perceived risk of an oncoming AV. This paper contributes to the future deployment of shared AV mobility services by providing results and insights that can be applied to AV control.