Research On Evaluation Method Of Driving Risk Based On Natural Driving Data

In recent years,as my country's industrial strength continues to increase,the automobile manufacturing industry has also been upgraded.The advanced automobile manufacturing industry has made the number of motor vehicles increase year by year,but the rapid growth in the number of motor vehicles is showing China's growing industrial strength.At the same time,it also brings huge traffic safety hazards and other social problems.As the economic and property losses caused by traffic safety accidents have been increasing year by year,this issue has gradually become the focus of public opinion in the country and society.The "Statistical Annual Report on Road Traffic Accidents of the People's Republic of China" pointed out that among the causes of various accidents,the driver causing the accident is still the subject of the accident.In the driver's accident behavior,the driver's misjudgment and improper handling measures are the main behaviors.This shows that the driver's dangerous driving behavior has become one of the leading factors in traffic accidents.Except for speeding,lane changing and other violations clearly stipulated by the law in the Road Traffic Safety Law of the People's Republic of China,some potential safety hazards that may cause traffic accidents have not received sufficient attention,such as: Misjudgments of other drivers' behaviors,misjudgments of vehicle performance and deviations in the estimation of driving scenes,etc.;and there is no reliable evaluation method.Therefore,effectively discovering driving behaviors with potential safety hazards,improving driving safety,and protecting people's lives and property is one of the important topics of current traffic safety research.Aiming at the misjudgment of driver behavior and the estimated deviation of driving scenes,this paper proposes a driving risk assessment model for drivers with different driving styles in different scenarios,and studies the driving risk based on different judgments of drivers.,To build a driving risk assessment model.And based on the NGSIM natural driving data set for experimental verification.The main research contents are as follows:(1)A vehicle driving state recognition model based on the NGSIM natural driving data set is proposed.This article first starts from the analysis of driving style classification,analyzes the driving style trends of drivers in different scenarios,builds a driving scene recognition model based on the vehicle motion data in NGSIM,and extracts vehicle driving state models in different scenarios from the data set,Laid a good foundation for the subsequent clustering research on driving styles in different scenarios.(2)After extracting vehicle driving data in different scenarios,analyze the relationship between each driving data parameter and driving style,and extract feature parameters that can better describe the driving style to form a feature vector.After obtaining different feature parameters in different scenarios,the fuzzy C-means clustering method is used to use the degree of membership as the similarity rule to cluster the feature vectors that characterize the driving style of the driver in each scenario,and finally obtain the driving in each scenario Probability of driver's driving style trend.(3)In the driving risk assessment model,considering that the driver's driving style trend probability and the discrimination probability of the scene are not static,it is necessary to introduce a feedback mechanism to adjust the driver's judgment on the driving scene in real time,and The driver's judgment of driving style,because for the driver,in the uncertain scene,the probability of appearance of each scene is not much different.In this case,the judgment result of the scene tends to the driver's judgment of the scene.It has a great impact,and this feature is highly similar to that of the information entropy theory.Therefore,the information entropy theory is introduced to construct a feedback model for real-time adjustment of the occurrence probability of each scene,so that the next construction is based on the Bayesian network Can be converted to a dynamic Bayesian network.So as to better evaluate the driving risk.(4)Based on a large number of previous studies,a dynamic Bayesian network can be constructed to evaluate driving risk.After establishing a relationship between driving styles in different scenarios through a directed acyclic graph,it is necessary to check each terminal in the graph.The node is to analyze the relationship between driving style and driving risk in different scenarios,and establish a driving risk-driving style-driving scene correspondence scale,so that the model's judgment of driving style can be transformed into a judgment of driving risk.Complete the assessment of driving risk according to driving style.Research shows that the model can accurately distinguish driving scenes.In the NGSIM data set,the accuracy of the discrimination reaches 92.5%.