This research introduces a novel scenario-driven testing framework for Autonomous Driving Systems that transforms real accident data into realistic test scenarios.

• Addresses key limitations in current ADS testing by preserving critical map information
• Generates concrete, realistic test scenarios based on actual traffic accidents
• Improves trajectory accuracy and scene realism for more effective validation
• Enables scalable, automated testing while maintaining real-world relevance

For engineering teams, this approach significantly enhances safety validation by grounding tests in actual accident data rather than artificial scenarios, potentially accelerating ADS deployment while improving public safety.

From Accidents to Insights: Leveraging Multimodal Data for Scenario-Driven ADS Testing