An Integrated Scenario-Based Testing and Explanation Framework for Autonomous Vehicles

The validation and verification approaches for autonomous vehicles have commonly employed scenario-based testing in various environments. Simulation platforms play a key role in the execution and evaluation of test scenarios, particularly for validating perception, decision-making, and actions in various traffic scenarios. It is imperative to establish a systematic validation methodology and use a formal specification for these scenarios, which significantly contributes to standardization, repeatability, and verification processes. In this paper, we propose an integrated scenario-based testing and explanation framework for autonomous vehicles that is modular, reusable, and available. Our proposed framework compartmentalizes testing processes into interchangeable and extendable components. In addition, we identify bottlenecks and suggest improvements to the testing process while also implementing solutions, followed by examples of experiments. This approach also provides the ability to explain autonomous vehicle scenarios by adapting state-of-the-art explainable artificial intelligence to the framework.