城市交通流聚类仿真与特征识别研究  

Research on Clustering Simulation and Feature Recognition of Urban Traffic Flow

作　　者：洪莹 丁飞[1,2] 崔峻 李永军 HONG Ying;DING Fei;CUI Jun;LI Yong-jun(Nanjing University of Posts and Telecommunications,Nanjing Jiangsu 210003,China;Urban Management and Big Data Joint Laboratory,Nanjing University of Posts and Telecommunications,Nanjing Jiangsu 210003,China)

机构地区：[1]南京邮电大学,江苏南京210003 [2]南京邮电大学城市管理与大数据联合实验室,江苏南京210003

出　　处：《计算机仿真》2024年第5期133-139,146,共8页Computer Simulation

基　　金：教育部-中国移动科研基金(MCM20170205);江苏省“六大人才高峰”高层次人才培养资助项目(.DZXX-008);中国博士后科学基金面上资助项目(2019M661900);江苏省博士后科研资助计划(.2019K026);江苏省研究生科研与实践创新计划(KYCX20_0770);南京邮电大学大学生创新训练项目(SYB2020035)。

摘　　要：城市交通流特征识别是交通管理和优化的关键,据此提出一种利用不同聚类算法提取动态交通流特征并性能评估的方法。首先,定义路网多交叉路口交通流的波动特征、信号灯配时和车辆分布场景,基于VISSIM仿真环境输出车流量、平均车速和交叉口排队时长;然后,利用聚类算法对生成的交通流场景进行特征提取,并研究不同信号配时对应的交通流仿真及聚类特征对比,通过对交通流聚类特征分析提取道路车辆期望车速分布优化目标,优化道路车速以提高道路通行效率;最后依据轮廓系数法对不同交叉路口的交通流场景进行聚类性能评价,对比K-means、AGNES和DBSCAN算法对交通流特征识别的差异性并分析各自场景适应性。所提交通流仿真与特征识别研究结果,为评估区域交通的通行效率并分析其影响因素提供理论依据,从而提高城市交通服务水平。Urban traffic flow feature recognition is the key to traffic management and optimization.Based on this,a method of extracting dynamic traffic flow features and performance evaluation using different clustering algorithms is proposed.Firstly,the signal timing and vehicle distribution scene at the multi-intersection of the road network are defined,and the traffic flow and queuing time are output based on the VISSIM simulation environment.Then,the clustering algorithm was used to extract the features of the traffic flow scene,and study the traffic flow clustering feature comparison corresponding to different signal timings,through the analysis of traffic flow clustering features,the optimization target of road vehicle expected speed distribution was extracted to improve road traffic efficiency.Finally,the clustering performance of traffic flow scenes at different intersections was evaluated according to the contour coefficient method,the differences in traffic flow feature recognition by K-means,Agnes and DBSCAN algorithms were compared,and their scene adaptability was analyzed.The research results of traffic flow feature recognition provide a theoretical basis for evaluating the traffic efficiency of regional traffic and analyzing its influencing factors,thereby improving the urban traffic service level.

关 键 词：交通流 仿真 聚类 交叉口 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]