Interaction between Pedestrians and Highly Automated Vehicles: Effects of External Human-Machine Interfaces and Vehicle Size

Abstract

Pedestrians will interact with automated vehicles of higher automation levels on the road in forthcoming urban traffic. In highly automated vehicles, vehicle automation takes over the driving task, and the former human driver becomes a passenger. The challenge of introducing highly automated vehicles is to guarantee a safe interaction with their surroundings, particularly pedestrians. Highly automated vehicles should have proficient communication abilities to ensure the safety of pedestrians. External human-machine interfaces (eHMIs) and vehicle kinematics are possible means of communication. An eHMI can communicate the vehicle’s automation status or the vehicle’s intention to the surrounding traffic environment, e.g., via light signals. Previous results revealed that light-based eHMIs, in the form of light-bands, could positively affect pedestrian behavior, perceived safety, and trust when interacting with highly automated vehicles in low-speed traffic areas. Moreover, vehicle kinematics support pedestrians to understand the vehicle’s intention. Nevertheless, the precise interplay of eHMI and vehicle kinematics regarding pedestrians’ interactions with highly automated vehicles remains unclear. In addition, most studies applied eHMIs to only one vehicle size, even though pedestrian evaluations could differ depending on the vehicle size, e.g., larger vehicles were perceived as a greater threat by pedestrians than smaller vehicles. Considering the relevance of vehicle size, more research is required to explore the applicability of eHMIs to other vehicle sizes. Therefore, the focus of this dissertation is on the effects of eHMIs, vehicle size, and the interplay of eHMI and vehicle kinematics on pedestrians’ interactions with differently sized, highly automated vehicles in shared space settings. Four studies were conducted in three experimental settings (online setting, virtual reality, and test track). Study 1 was an experimental online study. The focus of this study was on the transferability of light-band eHMI communication strategies, previously designed and evaluated for a smaller, highly automated vehicle, to a larger, highly automated vehicle in a shared space. The findings revealed that light-band eHMIs enhanced the perceived safety of pedestrians when interacting with small and large highly automated vehicles. The use of a dynamic eHMI, which informed about the vehicle automation status and the yielding intent, increased the perceived safety and led to more positive affective reactions than no eHMI or a static eHMI, which only informed about the vehicle automation status. Additionally, pedestrians perceived a smaller, highly automated vehicle as safer than a larger one. Study 2 was built upon Study 1 and was also conducted online. This study aimed to investigate the interplay of eHMI and vehicle kinematics for pedestrians’ interactions with differently sized, highly automated vehicles. According to the findings, the participants relied on explicit communication via dynamic eHMI rather than on vehicle kinematics to indicate their perceived safety, i.e., they felt safe even when the eHMI falsely communicated a yielding intent. Moreover, the pedestrians were more willing to cross the road and reported feeling safer when the highly automated vehicle was equipped with dynamic eHMI than with static eHMI or no eHMI. Furthermore, the pedestrians’ ratings on willingness to cross were higher for a smaller, highly automated vehicle than for a larger one. Study 3 was conducted in a virtual reality setting, considering the crossing behavior of pedestrians and their subjective assessments. Focusing on pedestrian behavior, pedestrians initiated their crossing earlier with a dynamic eHMI compared to a static eHMI or no eHMI. In Study 3, the potential of light-band eHMIs was underlined when they were well-coordinated with vehicle kinematics. Pedestrians felt safest when the yielding intent of the vehicle was communicated explicitly via dynamic eHMI and implicitly via early yielding. Additionally, pedestrians felt more aroused by a larger, highly automated vehicle than a smaller vehicle. Study 4 focused on the effects of eHMI communication and the interplay of eHMI and vehicle kinematics in a real-world pedestrian crossing. The test vehicle was instructed to be highly automated, and a light-band eHMI prototype was installed on the vehicle. Pedestrians initiated their crossing significantly earlier with dynamic eHMI than with static eHMI or no eHMI. Furthermore, they evaluated a dynamic eHMI combined with an early yielding as safety-enhancing and with a more positive affective valence than with a late yielding. In conclusion, the findings of this thesis suggest the transferability of light-band eHMI communication strategies, previously evaluated for a highly automated car, to a highly automated bus. The use of lightband eHMIs positively affected pedestrians’ interactions with both vehicle sizes. Nevertheless, pedestrians relied on the dynamic eHMI, even in cases where the explicit signals conflicted with the vehicle kinematics. The size of the highly automated vehicles had an impact on pedestrians' perceived safety, their willingness to cross, and their affective arousal in this thesis. The results revealed that the pedestrians felt less safe and were less willing to cross for the larger, highly automated vehicle vs. the smaller one. Moreover, the participants reported to feel more aroused when interacting with the larger, highly automated vehicle than with the smaller one. In summary, eHMIs and vehicle kinematics should be considered as means of communication for differently sized, highly automated vehicles in a coordinated manner for their interactions with pedestrians in the future.