Research On Causes Of Rear-end Collision And Behavior Of Avoiding Collision Based On Deep Data Analysis

With the improvement of China’s road network construction and the increasing number of motor vehicles,the number of road traffic accidents also increases,and rear end accidents account for a large proportion of road traffic accidents.It is important for theoretical value and practical significance to prevent rear end collision and improve road traffic safety by studying the temporal and spatial distribution of rear-end collisions,exploring the causes of rear-end collisions,simulating and reproducing of different types of rear-end collision accident dynamic parameter changes,and studying the driver’s braking and steering collision avoidance behavior under the risk of rear-end collision accidents.First,based on the data of more than 8000 traffic accidents in Harbin City from 2015 to 2019,the traffic tailgating accident data are screened.1127 data are finally identified as the research object of this paper by applying data preprocessing and geocoding methods,and the ArcGIS software is used to analyze the temporal and spatial distribution characteristics of rear-end collision accidents.From the spatial dimension,the mean center method and standard deviation ellipse method are used to obtain the spatial diffusion direction and aggregation degree of tailgating accidents.Two spatial data mining techniques,density analysis and cluster analysis are applied to identify the accident-prone areas and areas with high severity of accidents in two cases.From the temporal dimension,density analysis is used to compare the frequency of tailgating accidents in different seasons and different time periods.In terms of time dimension,density analysis is applied to compare the frequency distribution of rear-end accidents in different seasons and time periods.Second,based on Harbin traffic accident database and weather database.Nine influencing factors were selected:time of occurrence,week of occurrence,season of occurrence,accident pattern,wind speed,climate,location of occurrence,type of vehicle and type of accident.Building LightGBM,Random Forest and Support Vector Machine models.By comparing and analyzing the accuracy of each model,the LightGBM algorithm model was finally determined as the tailgating accident severity prediction model.Subsequently,the causal factors of rear-end accidents are deeply explored.The ranking of the importance of the features affecting the severity of the accident was calculated.Third,based on the theory of rigid body kinematics and dynamics,according to the brake traces,scattered objects,the final position of the vehicle and the design parameters of the vehicle at the scene of the accident,the motion state of the vehicle before the accident is inferred.The simulation and reproduction of four typical rear-end collision accident types:car rear-end cars,car rear-end trucks,truck rear-end cars and truck rear-end trucks are realized using PC-Crash.The dynamics of the four rear-end accident vehicles is analyzed using the control variable method.Under the influence of different offset degrees,different relative collision speeds,different braking behaviors and different steering behaviors,the law of changes in the maximum acceleration of the vehicle,the duration of the collision,the deflection angle and the angular velocity of the vehicle of the four types of rear-end collision accidents mentioned above is explored.Finally,based on a high-simulation driving simulation platform,the trafic experiment scene under the risk of rear end collision is built,and driving simulation experiments are carried out using an expressway in Harbin as a prototype.In the experiment,under the speed levels of 60km/h,80km/h and 100km/h,the leading vehicle simulates the emergency stop with the deceleration of 0.6g.The driver follows the car in front with a certain space headway,and adopts different behaviors to avoid rear end collision.Research has shown that drivers mainly use two collision avoidance behaviors combining only braking with steering and braking to avoid rear-end collisions.Further statistical analysis and variance analysis are carried out on the braking and steering data under the two behaviors to obtain the characteristics and rules of collision avoidance driving behavior data of different experience experimenters,which provides a certain reference for the design or optimization of driving assistance system or collision avoidance system.