Autonomous driving technology has seen significant advancements in recent years, yet safety and reliability remain major barriers to large-scale deployment. Traditional testing methods rely heavily on mileage accumulation and synthetic simulation, which fail to capture the rare but critical human errors that lead to real-world accidents. This paper proposes a testing and validation framework driven by the analysis of human-related accidents and conflicts. By integrating accident causality modeling, near-miss scenario mining, and behavior-driven simulation, the framework aims to construct high-risk, high-value testing scenarios that accelerate the safety verification process of autonomous vehicles (AVs).
Arif Hossain (Tue,) studied this question.