交通运输系统工程与信息

交通运输系统工程与信息 ›› 2016, Vol. 16 ›› Issue (1): 58-63.

• 智能交通系统与信息技术 • 上一篇    下一篇

基于驾驶人决策机制的换道意图识别模型

倪捷，刘志强*   

  1. 江苏大学汽车与交通工程学院，江苏镇江212013
  • 收稿日期:2015-05-21 修回日期:2015-09-22 出版日期:2016-02-25 发布日期:2016-02-25
  • 作者简介:倪捷（1982-），女，江苏启东人，讲师，博士生
  • 基金资助:

    国家自然科学基金/National Natural Science Foundation of China（51108209）；教育部博士点基金/The Ph.D.
    Programs Foundation of Ministry of Education of China（20113227110014）；江苏省普通高校研究生科研创新计划项目/The
    Research Innovation Program for College Graduates of Jiangsu Province（CXLX12_0657）.

A Recognition Model of Lane Change Intention Based on Driver's Decision Mechanism

NI Jie, LIU Zhi-qiang   

  1. School of Automobile and Traffic Engineering, Jiangsu University, Zhenjiang 212013, Jiangsu, China
  • Received:2015-05-21 Revised:2015-09-22 Online:2016-02-25 Published:2016-02-25

PDF

被引次数

3

摘要：

依据驾驶人换道决策的产生机制，提出速度期望满足度、危险感知系数和换道 可行性系数作为换道决策的识别指标并确定其量化方法.通过实车试验数据的分析表明： 量化指标与换道决策存在不同程度的相关性；同时在换道初期、车道保持及过渡状态阶 段存在显著差异.以速度期望满足度、危险感知系数和换道可行性系数为特征输入参数， 建立基于模糊神经网络的驾驶人换道意图识别模型，进行驾驶人换道意图的识别.结果表 明，该模型在换道初期的预测准确率达到89.93%，虚警率为9.52%，优于以碰撞时间TTC 为输入指标的BP神经网络模型，以及以RV、RP、RS为变量的Logistic 模型，说明模型具 有较好的预测准确性.

关键词: 智能交通, 换道意图, 决策机制, 换道辅助系统, 模糊神经网络

Abstract:

According to the producing mechanism of driver's lane change decision, desired speed satisfaction, risk perception coefficient and change feasibility coefficient are put forward and quantified as the identification parameters of lane change decision. The results of analyzing real vehicle test data indicate that quantitative indicators have different correlation with lane change decision, and there is a significant difference among the beginning of lane changing, lane keeping and transition state stage. Fuzzy neural network model is established to identify driver's lane change intention by using desired speed satisfaction, risk perception coefficient and feasibility coefficient of lane change as the input feature index. The research results show that the model accuracy in the early stage of lane change is 89.93%, and the false alarm rate is 9.52%, which both are better than BP neural network model by taking the collision time TTC as input vectors and the Logistic model by using RV, RP and RS as variables. It shows that the model has a good predictive accuracy.

Key words: intelligent transportation, lane change intention, decision mechanism, lane assistant change, fuzzy neural network

中图分类号: