Enhancing Traffic Flow Understanding by Two-Dimensional Microscopic Models - ETF2D
Final Report Abstract
The project focused on enhancing traffic flow understanding through comprehensive data collection, detailed driver behavior analysis, and the development of advanced traffic modeling and visualization tools. It was divided into three key work packages: data collection, driver behavior modeling, and traffic simulation and visualization. In Work Package 1, high-resolution vehicle trajectory data was collected over a 900-meter section of the A50 urban freeway in Milan, Italy, using six drones equipped with 4K cameras. The data, recorded for 135 minutes, captured diverse traffic states, including free-flow and congested conditions. Over 124 641 trajectories were extracted, providing rich insights into vehicle classifications, lane-changing behavior, and stop-and-go traffic dynamics. The dataset was rigorously validated, stored in the OPARA repository, and will be publicly available to support traffic research, autonomous vehicle development, and computer vision applications. Work Package 2 focused on analyzing driving behavior and developing a two-dimensional traffic model. A robust Leader-Follower (LF) identification methodology was introduced to address the complexities of non-lane-based traffic, leveraging high-resolution trajectory data from Chennai, India. This methodology applies to lane-based traffic as well. This approach improved LF identification accuracy and enabled detailed insights into vehicle interactions. Additionally, driving behavior heterogeneity was investigated using the Intelligent Driver Model (IDM), revealing significant variations across drivers and vehicle types. The findings motivated the development of the Intelligent-Agent Model (IAM), which integrates car-following and lane-changing behaviors with social force dynamics and can be used for lane-based and lane-free traffic. The IAM replicates realistic traffic dynamics by combining directional influences from surrounding vehicles, validated through extensive simulations. In Work Package 3, the developed IAM was integrated into an open-source traffic simulator. This simulator allows users to configure IAM parameters, implement traffic management measures such as signals, speed limits, and obstructions, and analyze their effects on traffic flow. Additionally, a 3D web-based traffic visualizer was developed using Three.js, enabling interactive visualization of real-world traffic data collected in Work Package 1. The visualizer allows users to analyze specific vehicle trajectories, integrate traffic flow models like IDM and IAM, and adjust model parameters to study their impacts. The visualizer will be made open-source and accessible via GitHub. Overall, the project successfully delivered a high-quality trajectory dataset, advanced insights into driving behavior, and a two-dimensional traffic model capable of simulating realistic vehicle interactions. The development of simulation and visualization tools provides powerful platforms for analyzing traffic dynamics, testing management strategies, and optimizing road operations. These outcomes contribute significantly to traffic flow research, benefiting academia and industry by enabling practical applications in traffic management, safety, and autonomous driving technologies.
Link to the final report
https://doi.org/10.25368/2025.122
Publications
-
Drone-Based Trajectory Data for an All-Traffic-State Inclusive Freeway with Ramps. Conference in Emerging Technologies in Transportation Systems (TRC-30).
Chaudhari, A. A., Treiber, M. & Okhrin O.
-
The Intelligent Agent Model—A Fully Two-dimensional Microscopic Traffic Flow Model. Lecture Notes in Civil Engineering, 319-326. Springer Nature Singapore.
Treiber, Martin & Chaudhari, Ankit Anil
-
Leader–follower identification with vehicle-following calibration for non-lane-based traffic. Transportation Research Part C: Emerging Technologies, 171, 104940.
Kulkarni, Mihir Mandar; Chaudhari, Ankit Anil; Srinivasan, Karthik K.; Chilukuri, Bhargava Rama; Treiber, Martin & Okhrin, Ostap
