Presentation Information

[AI-02]Impact of Frozen Road Surfaces on Vehicle Safety: An Approach using CARSIM and Machine Learning

*Thuyet Quoc Dang1, Ha Vinh Tran 1, Akihiro Kuno1, Shuichi Kameyama2 (1. Taisei Group Future Technology Field/SATTE, Engineering Division, Taisei Rotec Corporation (Japan), 2. Department of Civil and Environmental Engineering, Faculty of Engineering, Hokkaido University (Japan))

Keywords:

Frozen road surface,Vehicle stability,Rut depth,Machine learning,CARSIM

Frozen road surface irregularities, such as rutting, significantly impact vehicle dynamics and safety. Conducting full-scale vehicle experiments in such conditions is hazardous, costly, and cannot easily reproduce the wide range of dangerous scenarios that may occur in practice. In this study, we used a vehicle dynamics simulation software, CARSIM, to reproduce vehicle motion under various road conditions. A wide range of simulation conditions, including different vehicle types, rut geometries, rut depths, friction coefficients, vehicle speeds, and tire positions within the ruts was analyzed using Double Lane Change test procedure based on ISO 3888-1. Instability was evaluated based on vehicle dynamic responses, including excessive lateral motion, loss of path-tracking, and abnormal yaw behavior. Machine learning algorithms were applied to identify which parameters most strongly govern hazardous maneuvers among the numerous vehicle motions demonstrated in CARSIM. The analysis showed that the fluctuation of lateral acceleration was the dominant indicator of instability, followed by the yaw angle within the lane-change region. Sudden changes in acceleration or deceleration caused by rutting can compromise stability, especially under high lateral acceleration during turning, increasing the likelihood of side slip and potential collision with adjacent vehicles. These findings suggest that the fluctuation of lateral acceleration is an effective surrogate measure for vehicle stability on frozen road surfaces. From a pavement management perspective, minimizing the fluctuation of lateral acceleration, maintaining adequate friction levels and managing rut depth can significantly reduce the occurrence of hazardous vehicle dynamics. The combining CARSIM and machine learning framework provides a practical tool for assessing surface properties and supporting data-driven maintenance and speed-management strategies in cold regions.