基于贝叶斯网络模型的车辆碰撞概率预测  

Prediction of vehicle collision probability based on bayesian networks

作　　者：张俊友 李鹏飞 王树凤 廖亚萍 ZHANG Jun-you;LI Peng-fei;WANG Shu-feng;LIAO Ya-ping(College of Transportation,Shandong University of Science and Technology,Qingdao 266590,China;Operating Branch of Qingdao Metro Group Co., Ltd., Qingdao 266000, China)

机构地区：[1]山东科技大学交通学院,山东青岛266590 [2]青岛地铁集团有限公司运营分公司,山东青岛266000

出　　处：《广西大学学报（自然科学版）》2018年第6期2332-2340,共9页Journal of Guangxi University（Natural Science Edition）

基　　金：国家自然科学基金资助项目(51178231);山东科技大学人才引进科研启动基金项目资助(2015RCJJ035)

摘　　要：为准确研判道路交通环境中车辆发生碰撞的可能性,辅助驾驶人及时采取有效的避障决策,文中结合驾驶人、车辆、道路、环境四方面影响因素构建贝叶斯网络,提出了基于贝叶斯网络模型的车辆碰撞概率预测方法。利用Uc-road仿真软件对青岛市黄岛区前湾港路周边交通场景建立仿真模型,采用仿真数据与交通调查实际数据相结合的形式对模型进行验证。研究表明,利用贝叶斯网络模型预测的车辆碰撞概率与实际计算概率的最大有效绝对误差为0. 016,模型能够准确地预测车辆碰撞的概率,可为驾驶人在异常工况下避障提供决策依据。In order to accurately determine the possibility of vehicle collisions in traffic environment,and to help drivers to adopt effective obstacle avoidance decisions in a timely manner,a Bayesian network was constructed by combining the influence factors of driver,vehicle,road,and environment,and a vehicle collision probability prediction method based on a Bayesian network model was proposed.A simulation model was designed for the traffic scene around Qianwangang Road in Huangdao District of Qingdao City by means of Uc-road simulation software.The model was tested and verified by combining the simulation data with the actual traffic survey data.The results show that the maximum effective absolute error between the predicted vehicle collision probability value and the actual calculated probability value is 0.016.This prediction model can accurately predict the probability of vehicle collision and provide decision-making basis for the driver to avoid obstacles under abnormal conditions,and provide technical and theoretical support for traffic safety.

关 键 词：贝叶斯网络 车辆碰撞 概率预测 交通安全 

分 类 号：U471.15[机械工程—车辆工程]