交通运输系统工程与信息 ›› 2020, Vol. 20 ›› Issue (3): 54-60.

• 智能交通系统与信息技术 • 上一篇    下一篇

基于时延特性建模的多断面短时交通流预测

刘小明*1,田玉林1,唐少虎2,尚春琳1,魏路1   

  1. 1. 北方工业大学城市道路交通智能控制技术北京市重点实验室,北京 100043; 2. 北京联合大学城市轨道交通与物流学院,北京 100101
  • 收稿日期:2020-01-21 修回日期:2020-02-14 出版日期:2020-06-25 发布日期:2020-06-28
  • 作者简介:刘小明(1974-),男,河北唐山人,教授,博士.
  • 基金资助:

    国家重点研发计划/ National Key Research and Development Program of China(2018YFB1601003);北京市自然科学基金/ Natural Science Foundation of Beijing, China (8184070, 8172018).

Short-term Traffic Flow Prediction of Multi-sections Based on Time-delay Modeling

LIU Xiao-ming1, TIAN Yu-lin1, TANG Shao-hu2, SHANG Chun-lin1, WEI Lu1   

  1. 1. Beijing Key Lab of Urban Road Traffic Intelligent Technology, North China University of Technology, Beijing 100043, China; 2. College of Urban Rail Transit and Logistics, Beijing Union University, Beijing 100101, China
  • Received:2020-01-21 Revised:2020-02-14 Online:2020-06-25 Published:2020-06-28

摘要:

针对现有交通流预测方法未充分考虑多断面车流演变规律,提出基于时延特性建模的时空相关性计算方法. 该方法采用对不同断面、不同时刻交通流的分布相似性度量,对输入的车辆到达数据序列进行切割构建时空相似度矩阵,得到相邻断面之间的时延参数. 基于时延特性建模,将多断面之间的流量信息进行融合,使用长短时记忆(LSTM)网络进行流量预测. 通过对实际路段数据的预测和结果分析,验证所提方法的有效性和实用性.

关键词: 城市交通, 交通流预测, 时延建模, LSTM网络

Abstract:

In view of the existing traffic flow prediction methods failed to fully reveal the evolution rules of the traffic flow in multi- sections, this paper proposes a spatiotemporal correlation method based on traffic flow transmission delay modeling. This method introduces the similarity measure of traffic flow distribution in different traffic sections at different time, and constructs a spatial-temporal similarity matrix by segmenting the input data sequence of arrival vehicles. The delay parameters between adjacent sections are obtained. Based on the modeling of delay property, the traffic flow data between multi-sections is fused, and the traffic flow prediction is carried out by using the Long Short- Term Memory (LSTM) network. The validity and practicability of this method was verified by experimental analysis with actual traffic flow data.

Key words: urban traffic, traffic flow prediction, modeling of time delay, Long Short-Term Memory (LSTM) network

中图分类号: