Parametrized End-to-End Scenario Generation Architecture for Autonomous Vehicles

Validation of highly automated or autonomous vehicles is still a major challenge. One main reason is the need to identify the uncountable potential situations in which the vehicle should be evaluated. Even when they are completely identified, the full-coverage of all these situations is not possible in real-world tests. Virtual validation however can be used to generate infinitesimally changing cases for testing. Nevertheless, the main challenge then becomes the collection and description of realistic test scenarios. This paper proposes an end-to-end approach for scenario generation from various sources. A flexible database structure enables fast and efficient querying that is used to generate an extensive set of test cases. The proposed schema is evaluated end-to-end—from scenario to test case generation to automated simulation and data gathering—populated by a data source based on the experience gained through real world tests. The proposed approach is a further step towards more efficient testing of autonomous functions.