GRVA-23-25
Guidance proposal for AI Dataset Assessment for Automated Driving Systems
Downloads
UNECE server
Related documents
Excerpts from session reports
GRVA | Session 23 | 22-25
Sep 2025
8. The representative of the University of Warwick presented informal documents GRVA-23-25 and GRVA-23-34, which proposed guidance for AI Dataset Assessment for Automated Driving Systems (ADS). The proposal introduces the ODD-based AI Safety In autonomouS Systems (OASISS) methodology, a framework designed to assess whether AI training and testing datasets are complete and representative of the system’s Operational Design Domain (ODD), in a technology-agnostic, measurable approach aligned with document ISO PAS 8800. GRVA exchanged views on these proposals and recommended that the presentation be transmitted to the IWG on AI for further consideration.