Characteristic Analysis And Identification Method Of Dangerous Driving Behavior Of Urban Expressway

As an vital link between freeways and city roads,urban expressways play an important role in relieving urban traffic congestion and improving residents’ travel experience.With a complex traffic environment,vehicle merging,diverging and intertwining behaviors happen frequently,urban expressways have become accident-prone among urban roads,and its’ traffic safety issues are receiving increasing attention.Driver’s dangerous driving behavior is directly related to the occurrence of traffic accidents.Therefore,focusing on urban expressway,this paper analyzes the vehicle driving characteristics,studies the driver’s dangerous driving behavior identification method,and proposes the application of dangerous driving behavior identification technology on traffic management measures.By summarizing the current situation of domestic and foreign research,the paper analyzes the traffic operation characteristics of urban expressway weaving area including merging,diverging and intertwining characteristics;based on the close relationship between driving behavior and traffic conflicts,proposes the definition of dangerous driving behavior and analyzes its’ influencing factors;with the help of UAV aerial photography technology and in-depth investigation of driving behavior,five typical dangerous driving behaviors of urban expressway are summarized.Based on the on-site traffic investigation,the vehicle driving characteristics data and traffic conflict data of the urban expressway were obtained.The vehicles were divided into incoming vehicles,outgoing vehicles and straight vehicles according to the trajectories,and the driving characteristics data such as speed,acceleration,distance from the centerline of the deviated road,speed difference with the previous vehicle,spacing,time headway,and modified time to collision of different types of vehicles were analyzed.The results show that there are significant differences(pvalue <0.05)in most of the characteristics among different vehicles.Traffic conflict data and distance from the centerline of the deviated road were used as indicators to obtain the dangerous driving behavior labels through cluster analysis,and driving behaviors with smaller time headway,larger distance from the centerline of the deviated road and smaller modified time to collision were labeled as dangerous driving behaviors;the random forest algorithm was adopted to rank the importance of the feature parameters to obtain the input parameters of the recognition model;the genetic-particle swarm algorithm optimization of BP neural network model were established to recognize dangerous driving behavior.The paper proposes the application of dangerous driving behavior recognition method in urban expressway traffic management measures,based on three application scenarios: two-lanes main road/ single-lane side road,three-lanes main road / single-lane side road,and three-lanes main road / twolanes side road,the application conditions of implementing road "virtual solid line" in the interweaving area of expressway were proposed.