Perceived safety and predictability of human/AV interactions : Deliverable 1.2 in the EC ITN project SHAPE-IT

Rapport, 2023

In the fast-evolving landscape of transportation, the integration of Automated Vehicles (AVs) holds immense promise for revolutionizing mobility and enhancing road safety. As we embark on this transformative journey, it is imperative that we investigate and understand the intricacies of the interactions between humans and AVs, because successful integration of AVs into the urban fabric hinges not only on their technical proficiency but also on human factors. Consequently, the fundamental premise of our exploration is that a thorough understanding of the perceived safety and predictability of AVs is of key importance.

Supporting the development of AVs that are perceived as safe requires the consideration and integration of different aspects and perspectives on predictability that are highly related to each other. On the one hand, the drivers and passengers inside the vehicle must, at least in principle, be able to understand and predict the behaviour of the AV, because any unexpected behaviour would reduce their acceptance and trust—and consequently their willingness to adopt this technology. To achieve this goal, the AV’s behaviour (and, ideally, its underlying goals and plans) must be made transparent to the humans inside the vehicle. On the other hand, the behaviour— specifically the movement—of an AV depends in part on the movements of the surrounding humans. In highly dynamic, social environments such as urban traffic, the ability of AVs to predict the movement of vulnerable road users (VRUs) is therefore paramount. The AVs’ proactive and anticipatory responses can enhance overall road safety and contribute to their efficient, harmonious integration into shared urban environments. Thus, to some degree, these prediction functions must be transparent to the humans inside the AV, to facilitate predictability of AV behaviour (so that the humans in the AVs know what to expect) and to ensure that they experience an appropriate level of perceived safety.

Although this deliverable is in SHAPE-IT work package 1 ("Safe and transparent interactions between AVs and humans inside the AVs"), we have included both in-AV and outside-AV perspectives. This decision was motivated by the interdependence of these two perspectives and the multi-disciplinary emphasis of SHAPE-IT. Essentially, we addressed the multifaceted challenges presented by the predictability of human movement behaviour from both inside and outside the AV. Our empirical results shed light on many nuances of the factors that

influence the subjective sense of safety experienced by individuals as they interact with AVs. Our approaches included subjective evaluations, neuroergonomics methodologies, and modelling approaches—as well as revelations about the degree to which AV-VRU interactions can be accurately predicted.

In the three chapters of this deliverable, we will explain our findings, which reveal the complex tapestry of human-AV interactions. Results from empirical and modelling research on the perceived safety of AVs and the predictability of human-AV interactions will be presented. By revealing the interwoven threads of perceived safety and predictability, we hope to contribute not only to academic discourse but also to the practical implementation of AV technologies. Our endeavour aligns seamlessly with the European Union's commitment to fostering innovation that prioritizes human-centric design, thereby ensuring that future mobility is not just automated but, more importantly, safe, and predictable for all.