The journey to safe autonomous driving: from interpretable design to rigorous verification and validation

Safety-critical systems such as autonomous driving need to operate safely in their real-world deployment. With the implementation of AI in these systems, new safety assurance paradigms are critical to mitigate the risks posed by AI as traditional rules-based controls are replaced by data-driven ones. In the talk, I will present an overview of existing methods for AI safety. Subsequently, I will introduce my contributions to this field by elaborating on the design of interpretable AI algorithms, explainable AI methods, and rigorous testing methodology. I will conclude my talk with future research directions towards safe and trustworthy AI.