ITSC 2025 Paper Abstract

Paper TH-LA-T20.2

Takedomi, Shogo (Bridgestone corporation), ISHII, KEITA (Bridgestone Corporation), Mori, Teppei (Bridgestone Corporation), Saida, Akira (Civil Engineering Research Institute for Cold Region), OKUMURA, Kota (Traffic Engineering Research Team Civil Engineering Research Ins), Nakamura, Yuki (Civil Engineering Research Institute for Cold Region), matsushima, tetsuro (CIVIL ENGINEERING RESEARCH INSTITUTE FOR COLD REGION), Ono, Shunsuke (Institute of Science Tokyo)

High-Resolution Forecasting of Road Surface Temperature Using Graph Neural Networks

Scheduled for presentation during the Invited Session "S20c-Foundation Model-Enabled Scene Understanding, Reasoning, and Decision-Making for Autonomous Driving and ITS" (TH-LA-T20), Thursday, November 20, 2025, 16:20−16:40, Surfers Paradise 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 18, 2025

Keywords Data Analytics and Real-time Decision Making for Autonomous Traffic Management, AI, Machine Learning and Predictive Analytics for Traffic Incident Detection and Management, Testing and Validation of ITS Data for Accuracy and Reliability

Abstract

Effective winter road maintenance (WRM) is critical for the safety and mobility of road users in cold regions. High spatial resolution predictions of road surface conditions greatly improve WRM efficiency. Previous studies have explored spatial modeling to achieve this; however, traditional statistical models, such as Kriging, face limitations in capturing complex spatial dependency. In this study, we utilized graph neural networks (GNNs) for spatial modeling to predict road surface temperature (RST), which an essential road weather parameter, at high spatial resolution. Furthermore, by integrating this approach with time-series prediction model at road weather observatories and weather mesh forecast data, we propose a framework that enables high spatial-resolution RST predictions over future horizons. Experiments using real-world data from Sapporo, Japan demonstrated that the proposed method could predict RST at approximately 100-meter intervals along road networks with an root mean square error below 2 °C for future horizons up to 24 h. In conclusion, the study findings represent a significant improvement in accuracy over current operational methods, offering valuable support for decision-making in WRM planning to enhance road safety and mobility.