eHMI for Automated Vehicles - Benefits of Augmented Reality for Pedestrian-Vehicle Interaction

Abstract

Future urban road traffic may be characterized by a mix of automated vehicles (AVs) and nonautomated road users, including pedestrians. AVs may have different communication capabilities from manually operated vehicles, which can currently pose as interaction partners for pedestrians. External human–machine interfaces (eHMIs) support pedestrians when encountering AVs in road-crossing scenarios to facilitate safe and efficient interactions. The present thesis addresses three questions that need answers regarding eHMIs. The effort of this thesis includes adding evidence to the questions regarding what type of information to display to pedestrians and how to transmit these messages to those needing it. Furthermore, communication in complex scenarios in which multiple AVs approach a pedestrian has been evaluated. This study also tested augmented reality (AR) as a possible modality for eHMIs because this technique allows AVs to send clear, specific, and contextually situated information to the user. An experimental, monitor-based online study was conducted using simulated videos of crossing scenarios with either one or two approaching AVs. A 360° light band and an AR interface presenting either the vehicle automation status (VAS) alone or the AV’s intention to yield in addition to the VAS were used, as well as scenarios without any communicated information. Participants’ willingness to cross, their perception of traffic safety when crossing, and their perceived mental demands during decision-making were assessed in the within-design study. Among 94 participants, communicating yielding intention in addition to the VAS led to higher willingness to cross, greater feelings of safety, and less mental demand than VAS-based interfaces or no eHMI conditions. An intention-based eHMI using AR compared to the 360° light band resulted in even better ratings for all three measures. Communicating the VAS alone did not produce advantages for pedestrians in crossing situations. Further, the study failed to show differences for any measure when comparing simple to complex interaction scenarios. Despite methodical limitations regarding the lack of realism of the scenarios and the monitor-based setup, this thesis demonstrates that communicating the yielding intention of AVs appears to be essential and that AR can be used to support pedestrians, with potential benefits over light-based eHMIs. Further studies should investigate this modality in more realistic settings and in combination with other eHMIs that pedestrians can access even if they do not own an AR device.