Using eHMI, acoustic signal, and pitch motion to communicate the intention of automated vehicles to pedestrians: A Wizard of Oz study

Communication between automated vehicles (AVs) and pedestrians is crucial to coordinate traffic maneuvers and hence maintain road safety. Automated driving changes the driver's role to that of a mere passenger, who can perform non-driving related tasks and therefore may not interact with other road users. In this novel traffic environment, both explicit and implicit cues might contribute to an effective communication between AVs and pedestrians by providing relevant information.So far, physical prototypes of AV communication have rarely been evaluated under real world conditions. The few studies managed to do so used a Wizard of Oz technique, where a human operator simulates the behavior of an AV. To measure the communication effectiveness, these studies instructed participants to explicitly state their crossing decision, but not to cross. While safety is assured, the generalizability of the results to pedestrians' actual crossing behavior is unclear. We established a more realistic experimental setup for a Wizard of Oz study that allows pedestrians to actually start crossing while a of simulated AV is approaching.In a Wizard of Oz study, participants (N = 35) interacted with a simulated AV and had to infer its intention. In a fully factorial design, the effects of three different communication cues were explored: an external human-machine interface (eHMI, absent vs. present), an acoustic signal (absent vs. present), and pitch motion (active, normal, or none). The eHMI and acoustic signal were considered as explicit and the pitch motion as implicit cues. Participants were asked to exhibit natural crossing behavior which was captured by two optical tracking systems.Consistently to previous work, explicit AV communication showed a positive effect on pedestrians' crossing behavior and safety feeling. Pedestrians initiated their crossings earlier and reported to feel safer when approaching an AV with an eHMI or an acoustic signal, compared to only having only a pitch motion. There was no main effect of pitch motion on crossing behavior and safety feeling. Nevertheless, there was an interaction between eHMI and active pitch motion. The effect of eHMI on crossing initiation time (CIT) was smaller in combination with an active pitch motion compared to normal pitch conditions. These results underline the importance of implementing appropriate AV communication.