计算物理 ›› 2022, Vol. 39 ›› Issue (1): 83-95.DOI: 10.19596/j.cnki.1001-246x.8339

• 研究论文 • 上一篇    下一篇

基于守恒高阶模型和支持向量机的多车道车辆换道模型

张立灿1(), 郭明旻2, 林志阳3, 张鹏4, 段雅丽1,*()   

  1. 1. 中国科学技术大学数学科学学院, 安徽 合肥 230026
    2. 复旦大学航空航天系, 上海 200433
    3. 同济大学经济与管理学院, 上海 200092
    4. 上海市应用数学和力学研究所, 上海大学力学与工程学院, 上海 200072
  • 收稿日期:2021-02-02 出版日期:2022-01-25 发布日期:2022-09-03
  • 通讯作者: 段雅丽
  • 作者简介:

    张立灿(1995-), 男, 汉族, 硕士, 主要研究方向为计算数学中的交通流问题, E-mail:

  • 基金资助:
    国家重点研发计划(2018YFB1600900); 国家自然科学基金(11972121); 云南省交通运输厅科技创新项目(2019303); 陆地交通气象灾害防治技术国家工程实验室开放研究基金(NEL-2019-02)

A Lane Changing Model Based on High Order Conservation Model and Support Vector Machine

Lican ZHANG1(), Mingmin GUO2, Zhiyang LIN3, Peng ZHANG4, Yali DUAN1,*()   

  1. 1. School of Mathematical Science, University of Science and Technology of China, Hefei, Anhui 230026, China
    2. Department of Aeronautics and Astronautics, Fudan University, Shanghai 200433, China
    3. School of Economics and Management, Tongji University, Shanghai 200092, China
    4. Shanghai Institute of Applied Mathematics and Mechanics, School Mechanics and Engineering Science, Shanghai University, Shanghai 200072, China
  • Received:2021-02-02 Online:2022-01-25 Published:2022-09-03
  • Contact: Yali DUAN

摘要:

针对高速公路车辆换道问题, 提出一个多车道车辆换道模型。利用支持向量机(SVM)在多维特征下二分类问题的优势, 将SVM和Lagrange坐标下的高阶守恒模型(CHO)结合, 通过全离散跟车模型生成原始数据, 采用SMOTE(Synthetic Minority Oversampling Technique)算法对数据进行预处理, 采用双指标评估度SVM进行训练, 建立多车道车辆换道仿真模型。仿真结果表明: 基于支持向量机和CHO模型的换道模型, 驾驶车能够就当前的驾驶环境, 准确地作出决策, 有效地模拟高速公路上真实的多车道驾驶情况。

关键词: 支持向量机, SOMTE算法, 特征选择, CHO模型, 多车道交通流

Abstract:

A lane changing model for multi-lane traffic flow is proposed.It makes use of advantages of Support Vector Machine (SVM) in a binary classification problem with multi-dimensional features and combines with Conserved Higher-Order traffic flow model (CHO) in Lagrange coordinates.The original data is generated with a fully discrete car following model and preprocessed by Synthetic Minority Oversampling Technique (SMOTE) algorithm.The SVM is trained with two indexes evaluation.It shows that the lane changing model based on SVM and CHO simulates effectively real multi-lane driving behavior based on current driving environment on expressway.

Key words: SVM model, SMOTE algorithm, feature selection, CHO model, multi-lane traffic flow