基于贝叶斯网络的车辆换道模型

发布时间：2018-01-19 14:32

本文关键词： 公路运输贝叶斯网络机器学习车辆换道特征离散　出处：《交通运输系统工程与信息》2015年05期 　论文类型：期刊论文

【摘要】：车辆换道行为是微观交通流中最基本的驾驶行为之一,研究车辆换道行为可以提高车辆换道模型的仿真精度和减少由不合适的车辆换道行为引发的交通事故.当前车辆换道模型大多是基于驾驶员的决策思维方式建立的决策模型,这类模型的缺点是很难捕捉到驾驶员在决策过程中一些潜在决策模式和考虑的影响因素.鉴于此,本文引入了一种典型的人工智能方法——贝叶斯网络,建立了一个全新的车辆换道模型,试图通过机器学习的途径来提高车辆换道模型的精度.采用了分段离散化的方法对数据进行预处理,然后使用处理后的数据对贝叶斯网络的结构和参数进行学习,并分别建立了与两种贝叶斯网络结构相对应的车辆换道模型,最后对建立的模型分别进行验证.模型的验证结果表明,建立的基于贝叶斯网络的车辆换道模型对换道行为的识别率可以达到88%以上.此模型还可进一步应用到驾驶员辅助系统的开发中.
[Abstract]:Vehicle change behavior is one of the most basic driving behaviors in microscopic traffic flow. The research on the vehicle changing behavior can improve the simulation accuracy of the vehicle change model and reduce the traffic accident caused by the inappropriate vehicle changing behavior. At present, the vehicle changing model is mostly based on the driver's decision-making thinking mode. Decision model. The disadvantage of this kind of model is that it is difficult to capture some potential decision patterns and factors considered by drivers in the decision-making process. In view of this, a typical artificial intelligence method, Bayesian network, is introduced in this paper. A new vehicle change model is established to improve the accuracy of the vehicle change model by means of machine learning. The segmentation discretization method is used to preprocess the data. Then, the structure and parameters of Bayesian network are studied by using the processed data, and the vehicle changing models corresponding to two Bayesian network structures are established respectively. Finally, the established model is verified, and the verification results show that. The model based on Bayesian network can recognize the change behavior of vehicle more than 88%. The model can be further applied to the development of driver-assisted system.
【作者单位】：西南交通大学交通运输与物流学院;综合交通运输智能化国家地方联合工程实验室;综合运输四川省重点实验室;
【基金】：国家自然科学基金(51278429,51408509) 四川省科技厅项目(2013GZX0167,2014ZR0091) 中央高校基本业务经费(SWJTU11CX080) 成都市科技局项目(2014-RK00-00056-ZF,2014-RK00-00072-ZF)
【分类号】：U491
【正文快照】： 1引言车辆换道是实际中最为常见的驾驶行为之一,也是微观交通流仿真软件中的基本模块之一.1985年,美国为了研究微观交通流,利用航拍数据建立了车辆运行的微观信息数据库,促进了车辆换道模型的进一步发展[1].当前研究中,根据驾驶员追求获益的动机不同,可以将换道行为分为两类:

【参考文献】