Intelligent Road Icing Early Warning System Based On Machine Learning

the winter seasons, road icing is amongst the most significant threats towards road safety, which is considered as a super dangerous weather condition. This study aims to optimize road deicing predictions using machine learning (ML) techniques. By collecting data from sensors including pavement temperature (PT), pavement friction coefficient (PFC), pavement condition (PC), thickness of water film (TWF), freezing temperature, ice content (IC) and ice warning value (IWV), we analyzed crucial parameters affecting road deicing - road surface temperature. Predicting pavement icing (PI) is critical in the transportation field. To achieve this, we utilized the Long Short -Term Memory (LSTM) ML model to estimate icing conditions on the 2nd Ring South Road in Jinan, China. By considering relevant parameters of the road surface within a specific timeframe, we attempted to forecast road temperature, providing a novel approach for predicting road icing. Experimental results demonstrated the model's ability to accurately predict icing conditions. Furthermore, by utilizing observed data from the current road condition, we were able to precisely predict road temperature and thereby forecast road icing occurrences.