(November 20) How Can an AI Dashcam Help Improve Mobility and Safety?

Today’s driver-assistance systems have impressive “physical” intelligence — they perceive the road and act on it — but almost no understanding of the human driver they share control with. This seminar presents a research program that moves from physical autonomy toward human-centric shared autonomy, built on some of the largest open naturalistic driving datasets in the world (OpenLKA, ADAS-TO, BATON, DriveMotion, and DriveDNA), collected with low-cost AI dashcams across hundreds of drivers and vehicle models. We will show where production ADAS fails in the real world, how drivers hand over and take back control, and how advanced AI models can both warn drivers seconds before automation fails and anticipate the driver’s intent, motion, and individual style. Together, these results point toward vehicles and drivers that genuinely understand each other, with practical implications for traffic safety, infrastructure readiness, and the design of next-generation driver assistance.

Hao Zhou, PhD
Assistant Professor, Intelligent Transportation Systems, Traffic Operations, and Safety
Center for Urban Transportation Research

Dr. Hao Zhou is an Assistant Professor in the Department of Civil & Environmental Engineering and the Center for Urban Transportation Research (CUTR) at the University of South Florida, where he directs the MOTIF Lab (Mobility Optimized by Traditional Ideas and Frontier technologies). His research spans traffic flow theory, connected and automated vehicles, and AI for transportation safety, with a focus on real-world evaluation of driving automation and driver–automation interaction. His lab has created several of the largest open naturalistic datasets on production driver-assistance systems — including OpenLKA, ADAS-TO, BATON, and DriveDNA — and its AI-dashcam work has been featured by USF News and FOX 13 Tampa Bay.

Yuhang Wang
Graduate Research Assistant
Center for Urban Transportation Research

Yuhang Wang is a Ph.D. student in Civil & Environmental Engineering at the University of South Florida and a graduate research assistant in the MOTIF Lab at CUTR, advised by Dr. Hao Zhou. His research uses AI dashcams, large-scale naturalistic driving data, and vision–language models to understand how human drivers and vehicle automation share control. He is the lead author of several open datasets and benchmarks in this area — including OpenLKA, ADAS-TO, BATON, DriveMotion, and DriveDNA — as well as VLM-based early-warning systems that anticipate ADAS failures before they occur.