Research On Lane-changing Risk Discrimination And Personalized Forewarning Of Expressway

The rapid growth of China’s highway construction mileage has eased the pressure of economic development on the transportation industry,but at the same time also brought great challenges to road traffic safety.In recent years,the proportion of automobile traffic accidents in China accounted for more than 65%of the total traffic accidents,up to 70%.Research indicates that human factors are the main cause of traffic accidents,and 95%of traffic accidents are related to the driver’s behavior.As a common driving behavior,lane-changing behavior requires drivers not only to pay attention to the longitudinal motion characteristics of the vehicle and the running conditions of the preceeding and following vehicles in the lane,but also to pay attention to the lateral motion characteristics of the vehicle and its interaction with vehicles in adjacent lane.Collision accidents are caused by insufficient driver experience and subjective judgment errors.Therefore,it is the expectation of drivers to analyze the dangerous driving behavior of lane-changing and evaluate the driving risk to avoid the potential collision risk.Firstly,lane-changing intention recognition model based on support vector machine(SVM)is established in this paper.The feature index after dimensionality reduction using principal component analysis(PCA)is used as the model input,and the two driving states of lane keeping and lane change is used as the model output.According to the classification accuracy and the AUC value of the ROC curve verifies the accuracy of the model,and analyzes the"intention revocation phenomenon"existing in the lane change process.Secondly,by analyzing the dangerous driving behavior that affects the lateral stability of the self-vehicle and the inter-vehicle safety during the lane change of the highway,a lane change risk measurement index system is constructed.The lateral instability risk of the lane change vehicle is simulated by Car Sim software,and adopting the K-means clustering algorithm discriminates the interaction risk between vehicles to complete the classification of lane change risk levels.Finally,on the basis of determining the type of driver based on the fuzzy mathematical theory,a safe distance model of vehicle lane-changing considering the characteristics of the driver is established to predict the potential risk of lane change.The main research conclusions of this paper are as follows:(1)Principal component analysis is used to reduce the dimension of the feature space,and a driver lane change intention recognition model based on support vector machine is established.Its classification accuracy and AUC value are 84.99%and 0.8924 respectively,which proves that the model has good recognition performance.Aiming at the“intention revocation phenomenon”existing in the lane change process,through independent index t test,it is clear that the mean lateral speed and mean steering wheel angle of the vehicle can be used as indicators for the second judgment,which can more accurately determine the beginning time of lane change intention.(2)Consider the lateral stability of the vehicle and the safety of vehicle interaction,the risk level of lane change is divided into three categories:comfort,safety and danger.By comparing the discriminant consistency of the two types of indicators,it is concluded that the risk of lateral instability and the risk of interaction between vehicles may occur simultaneously when the vehicle changes lanes.(3)The long axis of the elliptical vehicle geometry model is used to reflect the longitudinal potential danger of the vehicle,and the characteristics of conservative,conventional and aggressive drivers are characterized by different T_d values.Based on the minimum safety distance model,a lane change safety warning model considering driving characteristics is established.The false alarm rate,false negative rate and accuracy rate of this early warning model are 9.6%,5.9%and 91.3%respectively,which can basically meet the driver’s demand for lane-changing warning under the highway driving environment.