Investigating driver responses to automated vehicles in a bottleneck scenario: The impact of lateral offset and eHMI

Abstract

This driving simulator study investigated drivers’ responses to an approaching automated or manual vehicle in a bottleneck scenario. Participants were asked to decide whether to pass through the bottleneck, or yield for the approaching vehicle, across numerous trials. Prior to each trial, they were informed whether the approaching vehicle was an automated vehicle (AV) or a manually driven vehicle (MV). Although participants were told that the MV was controlled by the experimenter using a distributed simulator, both vehicles were actually controlled by the system, and behaved in the same way. The kinematics of the approaching vehicle, such as its yielding behaviour (with or without lateral offset), and the presence of external Human Machine Interfaces (eHMIs, AV only) were manipulated. 40 participants took part in this study. Results indicated that participants’ subjective responses and behaviours did not differ between the AVs and MVs. The approaching vehicle’s lateral offset was seen to be the most influential source of information for participants, followed by information from the eHMI. Participants were more likely to pass through the bottleneck first, and had a shorter decision time, when encountering yielding vehicles with “away offsets”, which involved the vehicle moving away from the road centre line. This condition also led to higher perceived safety, comprehension, and trust ratings. Conversely, drivers were more likely to yield and had a shorter decision time when encountering non-yielding vehicles without any lateral offset. The lateral offset of non-yielding vehicles did not have an impact on drivers’ perceived safety and trust. However, non-yielding with “towards offsets” (towards the centre line) led to a higher comprehension score. Participants also passed through the bottleneck significantly more often and provided higher ratings for perceived safety and trust when the yielding vehicles presented an eHMI. This was regardless of lateral deviation. However, the eHMI only led to a higher rating of comprehension when the AV yielded without an offset. This study shows the value of using lateral offsets to communicate vehicles’ intentions in bottleneck scenarios. While the eHMI could enhance the driver’s understanding of the yielding AV, some participants also noted that it introduced uncertainty. Therefore, the need for eHMI should be further discussed.