Towards a Traffic Scenario Catalog for Collaborative Testing of Autonomous Vehicles

State-of-the-art safety assurance approaches for autonomous vehicles (AVs) rely on the existence of relevant, high-quality traffic scenarios as test cases. As a key drawback, traffic scenario synthesis approaches are often expensive to compute and hard to integrate in external AV testing workflows. While different approaches make varying assumptions about the AV-under-test, they rely on similar conceptual baselines for scenario representation, which yields an opportunity for unification and collaboration within the AV testing community. In this paper, we build up on these representation similarities and propose a traffic scenario catalog to support collaborative AV testing. Our proposed model-based architecture unifies common concepts in existing AV testing approaches, thus enabling a seamless integration of their derived scenarios and test execution results. Additionally, our catalog supports abstractions over scenarios through an integrated data aggregation methodology. This allows users to empirically evaluate and compare the behavior of AV controllers, thereby identifying potential anomalies (faults), e.g., through a domain-specific adaptation of metamorphic testing.