ABSTRACT
At present, many researches use vehicle trajectory data to identify rapid deceleration driving behaviors, but the existing method use fixed threshold to judge such behaviors, which cannot distinguish different driving scenes and lack of modeling analysis. Based on the car following model, we propose a method to judging rapid deceleration driving behavior, which includes multiple driving scenes. The method considers some parameters, such as lighting conditions, weather and road speed, and solves the problem of lacking scene classification. Clustering algorithm is used to distinguish the rapid deceleration driving behavior in the historical data, actual threshold is extracted and compared with the model results to verify the accuracy of the method. The comparison shows that the model has good adaptability and high accuracy for different driving scenes. Compared with the existing method it is proved that the method can better identify the rapid deceleration behavior of vehicles to lay a foundation for driving safety research.
