VR-PAVIB: The Virtual Reality Pedestrian-Autonomous Vehicle Interaction Benchmark

Abstract

Autonomous vehicles (AVs) are an emerging theme for future transportation. However, research on pedestrian-AV interaction, which promotes pedestrian safety during autonomous driving, is not a well-explored domain. One challenge preventing the development of pedestrian-AV interaction research is that there is no publicly available and standardized benchmark to allow researchers to investigate how different interfaces could help pedestrians communicate with AVs. To resolve this challenge, we introduce the Virtual Reality Pedestrian-Autonomous Vehicle Interaction Benchmark (VR-PAVIB). VR-PAVIB is a standardized platform that can be used to reproduce interaction scenarios and compare results. Our benchmark provides state-of-the-art functionalities that can easily be implemented in any interaction scenario authored by a user. The VR-PAVIB can easily be used in a controlled lab space using low-cost virtual reality equipment. We have released our project code and include the automotive user interface community to extend VR-PAVIB.