A rear-end crash-causation model for virtual safety assessment
Safety assessment of crash and conflict avoidance systems is important for the automotive industry and other stakeholders. Although using computer simulation to assess safety systems is becoming increasingly common, it is not yet commonly used for systems that affect driver behavior, such as driver monitoring systems (DMS). Models that generate virtual crashes, considering crash-causation mechanisms, are needed to assess these systems, of which only a few exist. And those that do have not been thoroughly validated on real-world data...
This work highlights the need to; a) thoroughly understand the process of generating virtual scenarios, and b) have the tools to validate them. While more work to develop validation processes for scenario generation is needed across all levels of crash severity, the transform and other validation tools that were developed bring us one step closer to accurate validation methodologies.
This study demonstrates that to validate scenario generation for virtual safety assessment, there is a clear need to understand the details of scenario generation, as well as of the validation dataset used - especially for possible selection bias in the validation dataset. Research on methods for validating scenario generation for virtual assessment is rare; much more work is needed for virtual simulation to become the tool we all want it to be, providing accurate estimates of safety benefits across all levels of crash severity. This work also indicates that behavior-based crash causation models may be an efficient tool to assess systems that aim to influence behaviors, such as DMS.
Bärgman, J., Svärd, M., Lundell, S., & Hartelius, E. (2024). Methodological challenges of scenario generation validation: A rear-end crash-causation model for virtual safety assessment. Transportation Research Part F: Traffic Psychology and Behaviour, 104, 374-410. https://doi.org/10.1016/j.trf.2024.04.007