Yousefzadeh, Nooshin (University of Florida), Sengupta, Rahul (University of Florida), Dilmore, Jeremy (Florida Department of Transportation), Ranka, Sanjay (University of Florida)

TGDT: A Temporal Graph-Based Digital Twin for Urban Traffic Corridors

Scheduled for presentation during the Regular Session "S08c-Intelligent Modeling and Prediction of Traffic Dynamics" (WE-LA-T8), Wednesday, November 19, 2025, 16:00−16:20, Coolangata 2

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on October 19, 2025

Abstract

Traffic congestion along urban corridors formed by sequences of signalized intersections results in delays, economic losses, and increased emissions. Existing deep learning models for arterial traffic control lack spatial generalizability, rely on complex architectures, or struggle with real-time deployment. We propose the Temporal Graph-based Digital Twin (TGDT), a scalable framework that combines Temporal Convolutional Networks and Attentional Graph Neural Networks to dynamically model and assess corridor-wide traffic. TGDT estimates key Measures of Effectiveness (MOEs) at both the intersection level (e.g., queue length, waiting time) and the corridor level (e.g., volume, travel time). Its modular design supports any number of intersections and MOEs, delivering accurate, concurrent multi-output predictions. TGDT outperforms state-of-the-art baselines, remains robust under diverse traffic conditions, and requires only minimal input features. Fully parallelized, it can simulate thousands of scenarios in seconds, providing a cost-effective, interpretable, and real-time tool for optimizing urban traffic flow.