Combining Scenario Elements for Scenario-based Testing

Ensuring the safety and functionality of automated vehicles (AVs) is critical in their validation and deployment. Traditional real-world testing is costly, time-consuming, and often lacks reproducibility. To address these challenges, scenario-based testing has emerged as an effective approach, allowing controlled and repeatable evaluations of AV performance under diverse conditions. This thesis presents a novel methodology for combining scenario elements to generate complex, realistic testing environments. The study proposes a structured framework for merging driving scenarios by analyzing scenario similarity and integrating elements from multiple real-world datasets. Using real-world traffic recordings from inD Maps and OMEGA files, the research identifies similar driving scenarios and replaces ego vehicle trajectories to create new, dynamic test cases. The methodology is implemented within the CARLA simulation environment, leveraging automated controllers to evaluate AV behavior under modified conditions. Key findings highlight the effectiveness of scenario-merging techniques in producing realistic test cases that accurately reflect urban traffic interactions. The analysis of trajectory-based interactions, crash occurrences, and time-to-collision metrics demonstrates the robustness of the generated scenarios. Furthermore, by applying synchronous and asynchronous simulation modes, the study assesses the impact of different testing environments on AV response accuracy. The results confirm that merging scenarios enhances AV validation by introducing diverse, scalable, and repeatable test conditions, addressing limitations in existing scenario-generation techniques. This research contributes to the ongoing development of efficient and safe validation frameworks for AV technology, paving the way for improved automation reliability in real-world driving environments.