Safety and Human Factors

13:30 - 17:30, November 18, 2025.

Southport 1, Star Grand, Broadbeach, Gold Coast, Australia.

Schedule

Invited talks

From Micro to Macro: Multi-Scale Traffic Flow Modeling and Analysis Based on Trajectory Data
Spearker: Prof. Zhiyuan Liu, Professor in the School of Transportation at Southeast University and Zelin Wang from Southeast University

Highlight: Prof. Liu is a pioneer in data and AI in transportation. He has been honored as the world’s top 2% of scientists and has won more than 20 championships and runner-up prizes in international top-tier artificial intelligence competitions.

How Continuous Video Analytics Delivers Unique Insights for Transport Network Safety and Operations with Particular Focus on Road User Trajectories?
Spearker: Dr. Simon Washington, Vice President, Transoft Solutions Inc.

Highlight: Dr. Washington is a long time leader in road safety research, with a distinguished career spanning academia, education, and industry. He has authored a widely used textbook on Transportation Data Analysis. In the industry, Dr. Washington co-founded and led the Advanced Mobility Analytics Group (AMAG) and now serves as VP for Transoft Solution Inc, pioneering AI-driven predictive analytics for road safety.

High-Altitude Drone Sensing for Large-Scale Urban Traffic Monitoring: The Songdo Open Dataset Initiative
Spearker: Robnert Fonod (EPFL), Haechan Cho (KAIST), Hwasoo Yeo (KAIST, Presenter), Nikolas Geroliminis (EPFL)

Highlight: We introduce a novel drone-based framework for large-scale urban traffic monitoring, developed to overcome the spatial and operational limitations of traditional ground-based sensors. Our system combines high-altitude object detection, trajectory stabilization via exclusion-masked image registration, and a robust georeferencing method using orthophotos and master frames. Deployed in the Songdo International Business District, South Korea, the experiment covered 20 intersections and yielded over 12TB of ultra-high-definition video data. From this, we constructed two open datasets: the Songdo Traffic Dataset with 700,000 georeferenced vehicle trajectories, and the Songdo Vision Dataset with 300,000 annotated vehicle instances across 5,000 images. Both datasets, along with the complete extraction pipeline, are publicly released to advance reproducibility and innovation in traffic research. This open data initiative sets a new benchmark for scalable, precise, and cost-effective traffic monitoring using drone and vision technologies—providing valuable resources for the intelligent transportation systems and smart city communities.

Bucheon Open Traffic Dataset: City-Scale Lane-Level Traffic Data from a Dense Urban CCTV Network
Spearker: Byungha Jeon (Team Leader of Traffic Policy Division, Bucheon City), Dongsoon Kye (Director of Smart City Business Department, Bucheon City Urban Corporation) (Presenter), Hwasoo Yeo (Professor, KAIST)

Highlight: Bucheon City, located 25 kilometers west of Seoul, South Korea, is a national leader in intelligent transportation system (ITS) deployment and innovation. With a population of approximately 790,000 and an urban area of 53.44 km², the city has established a comprehensive traffic monitoring network comprising 1,453 surveillance cameras—including 1,191 at 286 intersections and 262 at 262 midblock segments. Leveraging this infrastructure, Bucheon has developed a fine-grained urban traffic dataset using CCTV and YOLOv7-based vehicle detection, operating at 7 frames per second. The system captures lane-level traffic volume, approach speeds, and queue lengths at intervals of 5 to 60 minutes. To ensure data reliability, an extensive validation was conducted using manual counts at 40 intersections during peak hours, yielding high-performance metrics (Precision: 0.967, Recall: 0.884, F1 Score: 0.913). The dataset includes detailed movement-level and signal-phase-aligned information, covering up to 7 lanes per approach and both intersection and midblock conditions. The Bucheon Traffic Dataset is now publicly available through a website and is actively used for real-world applications such as signal optimization and travel speed improvements. As an open data initiative, this release aims to support cutting-edge research in AI-based traffic prediction, autonomous driving, and next-generation traffic control systems. By sharing high-quality, city-scale traffic data, Bucheon City hopes to global efforts in developing intelligent, data-driven transportation solutions.

Effect of the Moving Light Guidance System on Car-Following Model
Spearker: Ms. Mariko Nakai, PhD candidate at Ritsumeikan University, a visiting student at TU Delft, and an employee of Hanshin Expressway Company.

Highlight: Unconscious speed reductions in bottleneck sections induce chronic traffic congestion on freeways.To address this issue, Japanese highway authorities have implemented the Moving Light Guidance System (MLGS) for traffic flow management. Previous studies have shown that the MLGS helps mitigate congestion by reducing the number of traffic jams and their duration. However, the underlying mechanism, i.e., how these lights impact driver behavior, remains unclear due to limitations in conventional data collection methods. This study investigates the impact of MLGS on car-following behavior through parameter estimation of car-following models, examining differences in parameters between operations with and without MLGS, using trajectory data of all vehicles(ZTD). We expect our findings to clarify the mechanism by which MLGS reduce congestion. With this finding, road operators can set more effective MLGS operation strategies.

Capturing Freeway Congestion: What 100 Days of Data from 294 Cameras tell us?
Spearker: I-24 MOTION team

Highlight: The Tennessee Department of Transportation’s I-24 Mobility Technology Interstate Observation Network (MOTION) is a four-mile section of I-24 in the Nashville-Davidson County Metropolitan area with 294 ultra-high definition cameras. Those images are converted into a digital model of how every vehicle behaves with unparalleled detail. This is all done anonymously using trajectory processing algorithms developed by Vanderbilt University. By unlocking a new understanding of how these vehicles influence traffic, vehicle and infrastructure design can be optimized to reduce traffic concerns in the future to improve safety, air quality, and fuel efficiency.

Organizers

• Junyi Ji, Vanderbilt University
• Kai Li Lim, University of Queensland
• Xingmin Wang, University of Michigan Ann Arbor
• David Kan, Texas State University
• Gergely Zachár, Vanderbilt University
• William Barbour, Vanderbilt University

Advisory committee

• Henry Liu, University of Michigan Ann Arbor
• Dan Work, Vanderbilt University
• Jonathan Sprinkle, Vanderbilt University
• Zuduo Zheng, University of Queensland