Public value
Turn “when can this feature be used, when should it exit, and what is not publicly stated” into a verifiable list, reducing the risk of confusing L2 assistance with automation.
GB/T 45312—2025 · Evidence-first registry
A public, machine-readable registry for AD operating boundaries.
OpenODC does not rank AD systems. It asks a narrower question: how much of a system's operational boundary is backed by public manuals, official pages, government rules, clearly marked community extraction, boundary combinations, or public gaps?
All current samples are community extractions from public sources, not vendor-confirmed ODC declarations; there are no vendor-confirmed records yet.
PUBLIC SAMPLE LIBRARY
Turn “when can this feature be used, when should it exit, and what is not publicly stated” into a verifiable list, reducing the risk of confusing L2 assistance with automation.
Use one 144-element national-standard vocabulary to compare public boundaries across OEMs, automation levels, and robotaxi domains, while highlighting typical boundary combinations that single elements cannot express.
Provide JSON Schema, an editor, compare views, a matrix view, and a vendor intake workbench so manuals, configuration tables, and operating rules become reviewable ODC tables and trigger-condition candidates.
WORKFLOW
OPEN STEWARDSHIP
OpenODC now has a standard schema, sample gallery, comparison matrix, and intake workbench. To become real-time, credible, shared, and open, it cannot depend on one person’s spare time. We are looking for contributors with standards, ADAS / ADS development, safety, testing, data engineering, or open-source operations experience.
OPEN AD SAFETY STACK
Boundary combinations are not meant to turn OpenODC into a full SOTIF platform. They provide the shared interface: OpenODC proposes public-source-supported combined boundaries, DRIVEResearch estimates how often those combinations appear in real traffic and with what parameter ranges, and ROAM records what happens when boundaries are exceeded or operational interventions occur.
Turns manuals, feature pages, and operating rules into traceable single-element ODC entries and combined-boundary candidates.
Uses naturalistic driving data to estimate how frequent these combinations are and how their parameters are distributed on real roads.
Uses robotaxi remote-operation anomalies and incident records to see which combinations lead to takeover, degradation, or intervention.