基于SVM的快速路合流区车辆间隙选择模型

Gap choice model at urban expressway merging sections based on SVM

通过2种典型快速路合流区车辆行为观测,发现匝道车辆在选择间隙时存在多次超车行为,表明其汇入过程是一个多次决策的动态过程.根据合流过程中车辆速度的变化特性,判定车辆在选择间隙时的决策点并采集决策点处的微观交通流参数.在此基础上比较了2种不同渠化设计下入口匝道车辆汇入行为的差异.考虑到合流车辆不同行为判别所需的关键参数不同,使用2个支持向量机模型(SVM)进行分类,建立了合流区车辆多次决策的间隙选择模型.通过对采集的交通流参数进行训练,SVM模型的预测精度能够达到91%以上,实现预测车辆间隙选择的目的.最后与Logistic回归模型进行比较,结果证明所提出的模型能够获得较高精度.

Through the observation of the vehicle behaviors at the merging sections of two typical expressways,multiple overtaking behaviors are found to exist when the vehicle chooses the gap.It shows that the merging process is a dynamic process of multiple decisions.According to the variation of vehicle speed during the merging process,the decision point of the vehicle is determined.Then the video processing software is used to collect the microscopic traffic flow parameters at the decision point.Then a comparison analysis is conducted to discuss merging behaviors on different channelized on-ramps based on the survey data.Considering different key parameters required for different behaviors of merging vehicles,two support vector machine(SVM)models are used for behavior classification.Finally,a gap choice model for vehicle multiple decision-making in a merging section is established.The model can predict gap choice maneuver through the traffic parameters for training and the accuracy of the SVM model is over 91%.The proposed model is proved to obtain higher accuracy compared with the logistic regression model.