Construction of road anomaly event detection method for occurrences of disasters via state-space model that utilizes weather and probe data

In this research we propose a method to detect road abnormality events (standstill) in the event of natural disasters, using weather data and probe data. The proposed method compares the predicted value by the state space model with the filtered estimator and detects abnormality when the deviation is large. The proposed method was applied to cases of standstill at heavy snow occurring in Fukui and Tokyo, and showed that abnormality can be detected before standstill happening. In addition, we compare the accuracy of abnormality detection between the proposed method and the conventional time series model, and the proposed method has higher prediction accuracy.