Search for: All records

The design of self-driving vehicles requires an understanding of the social interactions between drivers in resolving vague encounters, such as at un-signalized intersections. In this paper, we make the case for social situation awareness as a model for understanding everyday driving interaction. Using a dual-participant VR driving simulator, we collected data from driving encounter scenarios to understand how (N=170) participant drivers behave with respect to one another. Using a social situation awareness questionnaire we developed, we assessed the participants’ social awareness of other driver’s direction of approach to the intersection, and also logged signaling, speed and speed change, and heading of the vehi- cle. Drawing upon the statistically significant relationships in the variables in the study data, we propose a Social Situation Awareness model based on the approach, speed, change of speed, heading and explicit signaling from drivers. 

The characterization of driver interactions is im- portant for a variety of problems associated with the design of autonomy for vehicles. We consider the role of cultural context in driver interactions, by evaluating the differences in driving interactions in simulated driving experiments conducted in New York City, New York, USA, and in Haifa, Israel. The same experiment was conducted in both locations, and focused on naturalistic driving interactions at unsigned intersections, in which interaction with another vehicle was required for safe navigation through the intersection. We employ conditional dis- tribution embeddings, a nonparametric machine learning tech- nique, to empirically characterize differences in the distribution of trajectories that characterize driver interactions, across both locations. We show that cultural variability outweighs individual variability in intersections that require turning ma- neuvers, and that clear distinctions amongst driving strategies are evident between populations. Our approach facilities a data-driven analysis that is amenable to rigorous statistical testing, in a manner that minimizes filtering, pre-processing, and other manipulations that could inadvertently bias the data and obscure important findings. 

Rerun is a software system to support post-facto analysis in sim- ulation research. In this submission, we show it working inside a multiplayer driving simulator. Rerun is built in Unity 3D and captures the virtual behavior of participants and their interactions with virtual objects. These recorded behaviors can then be played back from any perspective in the virtual space. This is useful in multi-agent interaction studies because researchers can sift through scenarios carefully from each participant’s perspective or even from an outside observer’s perspective. This enables a fine-grained un- derstanding of implicit and explicit signaling between participants and other human or AI-controlled agents.