Speaker

Dr. Henry Liu, Director of the University of Michigan Transportation Research Institute (UMTRI) and Bruce D. Greenshields Collegiate Professor of Engineering at the University of Michigan, Ann Arbor

Abstract

Autonomous vehicles (AVs) have experienced increased real-world deployment in recent years. However, safety remains a major barrier to large-scale adoption. Existing AV testing primarily focuses on functional safety, a vehicle-centric approach aimed at verifying the reliability, robustness, and sufficiency of an AV’s hardware and software systems. While essential, this approach is insufficient for large-scale deployment, as it fails to capture the safety impact of the AV's behavior on traffic environment. To address this limitation, we propose a paradigm shift towards behavioral safety, to evaluate AV’s behavioral responses and interactions with its traffic environment. To systematically evaluate behavioral safety, we introduce a third-party AV safety assessment framework consisting of two complementary components: the Driver Licensing Test, which evaluates an AV’s reactive behavior in controlled scenarios to ensure basic behavioral competency, and the Driving Intelligence Test, which assesses its interactive behavior in naturalistic traffic environments, measuring the occurrence rate of safety-critical events, thus providing statistically meaningful safety insights when before large-scale deployment. To validate the framework, we applied it to Autoware.Universe, an open-source Level 4 Automated Driving System (ADS), in both the simulated environment and the physical test track of the Mcity testing facility at the University of Michigan. Results show that Autoware.Universe passes 7 out of 14 scenarios and exhibits a crash rate of 4.16 × 10^-3 crash per mile, approximately 2,000 times higher than that of an average human driver.

Bio

Dr. Henry Liu is the Director of the University of Michigan Transportation Research Institute (UMTRI) and Bruce D. Greenshields Collegiate Professor of Engineering at the University of Michigan, Ann Arbor. He is a Professor in Civil and Environmental Engineering, and a Professor in Mechanical Engineering. He also directs the Center for Connected and Automated Transportation, a USDOT funded Region 5 university transportation center. Dr. Liu conducts interdisciplinary research at the interface of transportation engineering, automotive engineering, and artificial intelligence. He is recognized for his foundational work in cyber-physical transportation systems, particularly on the development of smart traffic signal systems and testing/evaluation of autonomous vehicles. His work on safety validation of autonomous vehicles has been published in Nature and featured as the cover story. Prof. Liu is the managing editor of Journal of Intelligent Transportation Systems and a board member for the ITS America. He is also the VP for Educational Activities for IEEE ITS Society.