交通运输系统工程与信息 ›› 2025, Vol. 25 ›› Issue (5): 91-102.DOI: 10.16097/j.cnki.1009-6744.2025.05.008

• 自动驾驶与智慧交通 • 上一篇    下一篇

基于粘滞性车辆组团识别的道路偶发拥堵预警研究

韩宝睿,纪宇轩,李根*,杨政,颜荣添,徐圣睿   

  1. 南京林业大学,汽车与交通工程学院,南京210037
  • 收稿日期:2025-06-05 修回日期:2025-07-11 接受日期:2025-07-18 出版日期:2025-10-25 发布日期:2025-10-25
  • 作者简介:韩宝睿(1973—),男,山东安丘人,副教授,博士。
  • 基金资助:
    江苏省自然科学基金(BK20240678);江苏省高校哲学社会科学项目(2024SJYB0142)。

Sporadic Road Congestion Early Warning Based on Viscous Vehicle Group Identification

HAN Baorui, JI Yuxuan, LI Gen*, YANG Zheng,YAN Rongtian, XU Shengrui   

  1. College of Automobile and Traffic Engineering, Nanjing Forestry University, Nanjing 210037, China
  • Received:2025-06-05 Revised:2025-07-11 Accepted:2025-07-18 Online:2025-10-25 Published:2025-10-25
  • Supported by:
    Natural Science Foundation of Jiangsu Province (BK20240678);Social Science Project of Colleges and Universities in Jiangsu Province (2024SJYB0142)。

摘要: 为研究车辆组团视角下的道路偶发拥堵预警方法,本文提出一种基于双重跟驰模式与多参数融合层次聚类的粘滞性车辆组团(Viscous Vehicle Group, VVG)识别框架。首先,基于普通跟驰(Normal Following, NF)与错位跟驰(Staggered Following, SF)两种微观跟驰模式,划分VVG的普通跟驰单元(Normal Following Unit, NFU)和错位跟驰单元(Staggered Following Unit, SFU);然后,从偶发拥堵特征、时空相似性和结构动态稳定性这3个维度出发,分别选取车速、时间间隔、空间间隔、速度差和加速度差这5个核心参数用以进行量化表征,结合Kolmogorov-Smirnov检验与统计分析进行参数筛选,并通过自底向上的层次聚类确定NFU与SFU的参数阈值范围,并识别VVG。以南京市道路实测交通流数据为例,实证结果表明:NFU与SFU参数的分布差异显著,支持将两种组团单元区分处理;在3088个车辆样本中成功识别出243辆参与VVG,有效验证了模型的识别能力;VVG多在偶发拥堵前出现,可作为拥堵前驱信号,预警成功率可达93.33%。本文为道路偶发拥堵预警提供理论和方法支持。

关键词: 交通工程, 拥堵预警方法, 层次聚类, 车辆组团, 跟驰行为

Abstract: To investigate an early-warning method for sporadic road congestion considering vehicle grouping, this paper proposes a framework to identify Viscous Vehicle Groups (VVG) based on dual car-following modes and multi-parameter fused hierarchical clustering. First, using two microscopic car-following patterns: Normal Following (NF) and Staggered Following (SF), the study defines the Normal Following Units (NFU) and Staggered Following Units (SFU) within a VVG. Then, with respect to the three dimensions: sporadic congestion characteristics, spatiotemporal similarity, and structural dynamic stability, five core parameters were selected for quantifying these features: vehicle speed, time gap, spatial gap, speed difference, and acceleration difference. Parameters are screened through the Kolmogorov-Smirnov tests and statistical analysis, and the bottom-up hierarchical clustering is applied to determine threshold ranges for NFU and SFU parameters and to identify VVGs. The case study is performed using real-world traffic flow data measured on roads in Nanjing city. The results show that: (1) NFU and SFU parameter distributions differ significantly, warranting separate treatment of the two unit types. (2) Among 3088 vehicle samples, 243 vehicles were successfully identified as participating in VVGs, validating the model's detection capability. (3) The VVGs predominantly emerge before sporadic congestion, serving as effective precursors, with an early warning success rate of 93.33%. This study provides theoretical and methodological support for the early warning of sporadic road congestion.

Key words: traffic engineering, congestion early warning method, hierarchical clustering, vehicle grouping, car-following behavior

中图分类号: