DISC: Dataset for Analyzing Driving Styles In Simulated Crashes for Mixed Autonomy

TL;DR

DISC introduces a first-of-its-kind dataset for analyzing human driving behavior in pre-crash scenarios using an immersive VR environment. By coupling MDSI-based driving style labels with rich sensory data from 12 simulated accident scenarios collected from 110 participants, DISC enables correlations between personality traits and driving actions, and supports mixed-autonomy trajectory prediction. The dataset is designed for interoperability with ScenarioNet and UniTraj formats, facilitating integration with large-scale trajectory benchmarks and enabling driving-style conditioned forecasting. While VR realism imposes limitations, DISC offers a practical pathway to enhance autonomous vehicle safety and behavior analysis in mixed autonomy settings.

Abstract

Handling pre-crash scenarios is still a major challenge for self-driving cars due to limited practical data and human-driving behavior datasets. We introduce DISC (Driving Styles In Simulated Crashes), one of the first datasets designed to capture various driving styles and behaviors in pre-crash scenarios for mixed autonomy analysis. DISC includes over 8 classes of driving styles/behaviors from hundreds of drivers navigating a simulated vehicle through a virtual city, encountering rare-event traffic scenarios. This dataset enables the classification of pre-crash human driving behaviors in unsafe conditions, supporting individualized trajectory prediction based on observed driving patterns. By utilizing a custom-designed VR-based in-house driving simulator, TRAVERSE, data was collected through a driver-centric study involving human drivers encountering twelve simulated accident scenarios. This dataset fills a critical gap in human-centric driving data for rare events involving interactions with autonomous vehicles. It enables autonomous systems to better react to human drivers and optimize trajectory prediction in mixed autonomy environments involving both human-driven and self-driving cars. In addition, individual driving behaviors are classified through a set of standardized questionnaires, carefully designed to identify and categorize driving behavior traits. We correlate data features with driving behaviors, showing that the simulated environment reflects real-world driving styles. DISC is the first dataset to capture how various driving styles respond to accident scenarios, offering significant potential to enhance autonomous vehicle safety and driving behavior analysis in mixed autonomy environments.

Figures (2)

• Figure 1: Scenarios from the driver's point of view: The LEFT image depicts the deer-crossing scenario, where the participant encounters a running deer while driving on a country road. The MIDDLE image shows the running red lights scenario, where a participant encounters a car at an intersection that crosses the road after ignoring the red traffic signal. The RIGHT image displays a jaywalking pedestrian from the driver's point of view.
• Figure 2: Plots of sensory values: The figure displays plots of acceleration, jerk, speed, and steering angle, representing distinct driving styles amidst jaywalking pedestrian scenarios, arranged from the top right to the bottom left. Observations reveal diverse patterns in sensory dynamics, providing insights into the driving behavior of drivers encountering pedestrians and enabling driving personality-conditioned trajectory prediction.