Overview of datasets for challenging automated driving scenarios

The current Autonomous Vehicle (AV) technology research frontier is defined by operational scenarios characterized by low probability but high impact for the widespread deployment of Level 4 and Level 5 systems.

Among the most challenging of this edge cases are construction zones, traffic incidents, tunnels, and urban environments. The EC CCAM project iEXODDUS has released a report providing an exhaustive analysis of the public dataset landscape relevant to these three critical domains.

The report highlights current dataset gaps and explores how cooperative perception, simulation, and multi-modal sensing can help address them. In particular, the report underlines that ego-only datasets are no longer sufficient for the most difficult safety-critical use cases, and that better integration of infrastructure data and synthetic scenario generation will be essential for robust deployment. This document serves as a valuable resource for researchers and engineers working on safety-critical scenarios and ODD extension.

Access the report here