Translating Crowdsourced Scenarios for use in Simulation Testing for Autonomous Driving Systems

Scenario-based testing is an industry-standard method for building safety cases for Autonomous Vehicles (AV) and Advanced Driver Assistance Systems (ADAS). Safety Pool™ Studio is a crowd sourcing platform that enables users to create and submit highly abstract, functional scenarios. Users compose scenarios using a tile-based editor, whereby each tile represents a road segment. Road users can be placed on these tiles and assigned behaviours via a simplified and intuitive simple interface. Prioritising accessibility is key to ensuring broad public participation on the platform. Crowdsourcing scenarios is a novel way of scenario identification, as it focuses on linking public experience into the engineering process and can help identify scenarios which may have been overlooked by engineers. This approach has the potential to increase public trust in the autonomous technology, and ensure that real-world, relevant cases are captured in the testing and training datasets. This paper explores a pipeline for translating abstract, crowdsourced scenarios into concrete, simulation-ready scenarios for use by engineers in simulation-based testing. Thistesting. This step is crucial in ensuring that the output of the application can be useful, while still offering a simple and intuitive interface which is accessible to the public. The process focuses on maintaining the user’s intent when creating the scenario by maintaining the timeline of events through translation. We output the scenarios as ASAM OpenDRIVE® and ASAM OpenSCENARIO® XML, which can then be bundled into a scenario library. , ready for use by experts in simulation-based testing.