事故条件下高速公路分级限速策略研究  

Research on the Grade-Based Speed Limit Strategy on Highways Under Accident Conditions

作　　者：卢胜聪 王芳 柳稳强 戴红良 于颖舟 LU Shengcong;WANG Fang;LIU Wenqiang;DAI Hongliang;YU Yingzhou(Zhejiang Zhijiang Intelligent Transportation Technology Co.,Ltd.,Hangzhou 311101,China;School of Information and Electrical Engineering,Hangzhou City University,Hangzhou 310015,China;Hangzhou City University Binjiang Innovation Center,Hangzhou 310056,China;Zhejiang Scientific Research Institute of Transport,Hangzhou 310039,China;Zhejiang Expressway Co.,Ltd.,Hangzhou 310020,China)

机构地区：[1]浙江之江智能交通科技有限公司,杭州311101 [2]浙大城市学院信息与电气工程学院,杭州310015 [3]浙大城市学院滨江创新中心,杭州310056 [4]浙江省交通运输科学研究院,杭州310039 [5]浙江沪杭甬高速公路股份有限公司,杭州310020

出　　处：《交通工程》2025年第3期7-15,共9页Journal of Transportation Engineering

基　　金：国家重点研发计划(2023YFB3209803);浙江省教育厅科研项目(Y202454294)。

摘　　要：针对高速公路交通事故易造成道路瓶颈的问题,研究一种基于深度学习的事故瓶颈区分级限速策略。在仿真软件SUMO中搭建事故模型,基于LSTM架构建立SumoNet模块用于拟合仿真输出的评价指标,基于CNN架构建立PolicyNet模块用于输出包含限速位置、限速值、情报板间距以及缓冲距离的分级限速策略。以事故占用两车道为例进行仿真实验表明:与无控制策略相比,在所提方法作用下,事故点前道路流量在90~95 veh/min,车速在27~29 m/s时,单车平均行程时间降幅均值达3.72%;事故点前流量在90~115 veh/min,车速在25~29 m/s时,平均车速提升8.09%。进一步地,对事故占用单车道和三车道场景所有评价指标进行宏观评价,限定流量和车速对平均速度进行微观分析,均证明了提出的分级限速策略的有效性。Based on the problem of traffic accidents causing road bottlenecks on expressways,a deep learning-based accident bottleneck distinction and graded speed limit strategy was studied.A traffic accident model was built in the simulation software SUMO,and the SumoNet module was constructed to fit the evaluation indicators of the simulation output.The PolicyNet module was developed to comprehensively consider the traffic flow and vehicle speed before the accident,and to output a graded speed limit strategy including the speed limit position,value,distance between information boards,and buffer distance.Simulation experiments were conducted using a two-lane accident scenario,showing that compared to no control strategy,the proposed method resulted in a 3.72%reduction in average travel time when the road flow before the accident was 90~95 veh/min and the vehicle speed was 27~29 m/s.Additionally,when the road flow before the accident was 90~115 veh/min and the vehicle speed was 25~29 m/s,the average vehicle speed increased by 8.09%.Furthermore,a macro evaluation of all evaluation indicators for single-lane and three-lane accident scenarios,and a microscopic analysis of the average speed with limited flow and vehicle speed,all demonstrated the effectiveness of the proposed graded speed limit strategy.

关 键 词：智能交通 分级限速 深度学习 事故瓶颈区 交通仿真 

分 类 号：U491.4[交通运输工程—交通运输规划与管理]