Abstract: Autonomous Vehicle (AV) technology has the potential to revolutionize the future transportation landscape. Safety validation for AVs, however, is a significant challenge because of the rarity of safety-critical events in complex, dynamic driving environments. Simulation-based testing faces two major hurdles: constructing high-fidelity simulations that accurately replicate real-world driving environments, and developing efficient testing methods that accelerate the validation process. To ensure the accuracy of test results, the simulation system must generate a naturalistic driving environment (NDE) with distribution-level accuracy to capture rare, long-tail events. To tackle this, we developed a deep-learning-based framework to model multi-agent interaction behavior with statistical realism. However, due to the high dimensionality of the environment and the rarity of safety-critical events, directly using the NDE for testing is highly inefficient. We discovered that making sparse but adversarial adjustments to the NDE, resulting in the naturalistic and adversarial driving environment (NADE), can significantly reduce the required test miles without compromising evaluation accuracy. The proposed NDE and NADE environments form a powerful AV testing tool, successfully deployed in real-world testing facilities including Mcity, the American Center for Mobility, and the Transportation Research Center.

Short Bio: Xintao Yan received the bachelor's degree in automotive engineering from Tsinghua University, China, in 2018, and the Ph.D. degree in civil engineering and scientific computing from the University of Michigan, Ann Arbor in 2023. He is currently a Postdoc in the Department of Civil and Environmental Engineering at the University of Michigan, Ann Arbor. His research interests lie in enhancing the safety performance of connected and automated vehicles, including naturalistic driving environment modeling and automated driving system evaluation. He has served as a member of the SAE On-Road Automated Driving (ORAD) Verification & Validation task force. He was the recipient of the Exceptional Paper Award from the Transportation Research Board (TRB) Annual Meeting in 2019 and the Intelligent Transportation Systems Best Paper Award from the INFORMS Transportation Science and Logistics Society in 2021.