The Study On The Formation Mechanism Of Dangerous Driving Behavior At Urban Expressway Entrance And Exit

The traffic flow at the entrances and exits of urban expressways achieves the purpose of inter-vehicle flow by changing the driving direction.This phenomenon is the main reason for the large number of traffic conflicts,the deterioration of the stability of the traffic flow,and the increase in traffic accidents year by year.Traffic accidents at the entrances and exits of urban expressways are largely caused by the bad driving habits of drivers.Typical dangerous driving behaviors are "too close to the car","violation of markings","mandatory lane change","line change lane change" and "parking waiting lane change",etc.The research on the formation mechanism of these dangerous driving behaviors can provide a theoretical basis for improving the traffic safety of entrances and exits.The research on dangerous driving behavior needs to extract the driving behavior characteristic index.In order to avoid the instability of the driving behavior characteristic caused by the multiple effects of the driver,vehicle and environment,the "driver-vehicle" is regarded as a "black box" system based on the NGSIM database.Historical data,extract the key indicators that can reflect the driving behavior characteristics of the "black box" system,draw on the Newell car-following theory,and determine the three parameters of reaction time,minimum safety distance,and driving behavior stability coefficient as the key indicators for the study of driving behavior characteristics.The thesis divides the driving behavior characteristics into longitudinal driving behavior characteristics and lateral driving behavior characteristics.The longitudinal driving behavior characteristics are studied by a variety of clustering analysis methods,and the driving behaviors are divided into six categories.Among them,the minimum safety distance is small and the reaction Long time and large driving behavior stability coefficient(the clustering centers are:d=7.88m,τ=2.29s,η=0.34)This type is classified as dangerous driving behavior that is too close to the car;the lateral driving behavior characteristics study will change lanes according to the operation of the car behind the target lane The state is divided into three stages,using Pearson correlation coefficient to divide the stage,study the change of driving behavior characteristics of the vehicle behind the target lane before and after the lane change,determine whether the following car is safe to follow the car(d<8.5m),and distinguish between forced lane change and free lane change.Fitting two types of lane-changing behavior trajectories separately,comparing the differences of the trajectories to find out the difference between the driving behavior characteristics of forced lane change and free lane change,and found that the lane change trajectory of the forced lane change vehicle is relatively irregular,and the lateral displacement change of lane change Big.Finally,the paper uses linear kernel,quadratic kernel and tertiary kernel support vector machine model(SVM)to identify the longitudinal driving behavior characteristic index and lateral driving behavior characteristic index.It is found that the three-core support vector machine model recognizes the accuracy of dangerous driving behavior Higher,can be used as a recognition or early warning model of dangerous driving behavior.