基于神经网络的车辆掉头对主线交通影响分析  

Analysis of the Influence of Vehicle U-turn on Main Line Traffic Based on Neural Network

作　　者：曹柯凡 杨震 CAO Kefan;YANG Zhen(College of Automobile and Traffic Engineering,Nanjing Forestry University,Nanjing 210037,China)

机构地区：[1]南京林业大学汽车与交通工程学院,南京210037

出　　处：《武汉理工大学学报（交通科学与工程版）》2023年第6期1010-1015,共6页Journal of Wuhan University of Technology（Transportation Science & Engineering）

基　　金：江苏省自然科学基金(BK20170932);南京林业大学青年科技创新基金(CX2017011);南京林业大学学科竞赛项目(162310168)。

摘　　要：文中分析了车辆掉头变道对主线车辆延误的影响因素,选取车均延误为评价指标,考虑掉头车数、掉头距离、主线流量,以及换道次数,运用单一变量原则分别设计四种仿真方案.结果表明:主线流量较小时,各因素对车流影响均有限,当流量达到一定阈值后,车均延误急剧增加;换道次数对道路车均延误影响较大.建立BP、DBN、GRNN三种神经网络模型,用于拟合仿真数据.DBN神经网络模型在各延误区间表现良好,其拟合优度(0.884)明显优于BP和GRNN神经网络(0.604和0.572),此外,在均方误差、均方根误差以及平均绝对误差上,DBN神经网络表现得更好.The influencing factors of vehicle delay caused by vehicle turning around and changing lanes were analyzed.Taking the average vehicle delay as the evaluation index,considering the number of U-turn vehicles,U-turn distance,main line flow and lane change times,four simulation schemes were designed by using the principle of single variable.The results show that when the main traffic flow is small,all factors have limited influence on the traffic flow,and when the traffic flow reaches a certain threshold,the average delay of vehicles increases sharply.The number of lane changes has a great influence on the average delay of road vehicles.Three neural network models,BP,DBN and GRNN,are established to fit the simulation data.DBN neural network model performs well in each delay interval,and its goodness of fit(0.884)is obviously better than BP and GRNN neural networks(0.604 and 0.572).In addition,DBN neural network performs better in mean square error,root mean square error and average absolute error.

关 键 词：掉头变道 神经网络 VISSIM仿真 车辆延误 

分 类 号：U495[交通运输工程—交通运输规划与管理]