摘要
变道模型是微观交通仿真中重要的模块,对其研究具有十分重要的意义。目前提出的变道模型大多采用基于驾驶员思维规则和基于机器学习的建模方法。然而,这些模型都没有考虑紧急程度,并且模型精度较低。引入紧急程度,提出了一种新的基于机器学习的变道模型。通过聚类算法将数据集根据紧急程度进行划分,使用梯度提升决策树在不同紧急程度上的数据集进行学习,得到不同紧急程度下的变道模型。通过实验验证,所提出的基于机器学习的变道模型相较于其他机器学习变道模型有更高的预测精度。最后,基于梯度提升决策树的特征重要度分析表明紧急程度在变道决策过程中具有十分重要的作用。
The lane-changing model is an important module in microscopic traffic simulation,and it is of great significance for its research.Most of the proposed lane change models are based on drivers’thinking and machine learning based modeling methods.However,none of these models consider urgency and the model accuracy is low.This paper introduces urgency and proposes a new machine learning based lane change model.The data set is divided according to the urgency by the clustering algorithm,and the gradient boosting decision tree is used to learn the data sets of different urgency levels,and the lane change model with different urgency degrees is obtained.The experimental results show that the proposed machine learning-based lane change model has higher prediction accuracy than other machine learning lane change models.Finally,based on the analysis of the feature importance of the gradient decision tree,it is shown that the urgency plays an important role in the decision-making process.
作者
李捷
李韬
徐大林
LI Jie;LI Tao;XU Da-lin(Jiangsu Automation Research Institute, Lianyungang 222061, China)
出处
《指挥控制与仿真》
2020年第4期88-92,共5页
Command Control & Simulation
关键词
机器学习
变道模型
紧急程度
GBDT
聚类
交通仿真
machine learning
lane-changing model
urgency
GBDT
clustering
traffic simulation
