Abstract

With the development of Automated Vehicles (AVs), interaction between cars and pedestrians will soon become a challenging issue for road safety. In today’s world, human drivers and pedestrians interact using non-verbal communication based on eye contacts to manage their interactions and to clarify their intentions. Several studies have shown the importance of this informal communication to avoid accident. This is particularly true in the case of road crossing decisions when priority rules are not clear (e.g., absence of zebra crossing). Introduction of automated vehicles on public road will radically change the situation: car occupant could be totally inattentive to the traffic scene (when sleeping or being visually distracted by a secondary task, for instance), and it is also possible that there are no human on board. In this context, pedestrians will have to discover new behaviours of vehicles, to learn new rules to make their crossing decision, and to train themselves in new practices to interact safely with automated cars. Because automated cars are not currently traveling on public roads, one way to study this issue is to use immersive simulation to allow real humans to « practically experience » (even if in a virtual way) future interactions with such automated vehicles. In order to study how interactions between pedestrians and automated vehicles may look like in the future, and how road safety may be potentially affected by the AV behaviour, an immersive experiment was implemented in the SUaaVE european project (SUpporting acceptance of automated VEhicle, H2020). Forty participants were involved in this experiment: 20 young mean-aged of 25.1 years old, and 20 elderly people mean-aged of 67.5 years old (50% women for each group). They all have to experience a set 8 driving scenarios (in a randomized order), from two different points of view: As an AV passenger (to experience on a driving simulator how the AV reacts when facing pedestrians’ road crossing) and as a pedestrian (equipped with a virtual reality headset and located on a pavement, facing a continuous flow of a randomized number of approaching vehicles). For each scenario, they have to assess the safety of the different braking strategies implemented by the AV. Results show that AV braking strategies stopping the car at 5 meters of the pedestrians are assessed as more safe than braking stopping closer to them (i.e. 2 meters). In addition, a common hierarchy of the safety of the different AV yielding manoeuvres is identified for all the participants. Nevertheless, significant differences are found between the two groups: compared to young people, who assessed the safety of the AV manoeuvres similarly whether as a pedestrian or as an AV passenger, older participants have a tendency - as pedestrians - to misperceived and/or misestimate the “velocity/brutality” of some AV braking manoeuvers (typically for AEB emergency braking strategies).