山区双车道公路超车持续时间影响因素分析及预测模型  

Analysis of Influencing Factors and Prediction Model for Overtaking Duration on Mountainous Two-Lane Roads

作　　者：覃文文[1,2] 刘云贵 臧津延 李武 韩春阳[1,2] 戢晓峰 QIN Wenwen;LIU Yungui;ZANG Jinyan;LI Wu;HAN Chunyang;JI Xiaofeng(Faculty of Transportation Engineering,Kunming University of Science and Technology,Kunming 650504,China;Yunnan Modern Logistics Engineering Research Center,Kunming 650504,China;YCIC Investment Co.,Ltd.,Kunming 650028,China)

机构地区：[1]昆明理工大学交通工程学院,云南昆明650504 [2]云南省现代物流工程研究中心,云南昆明650504 [3]云南交投集团投资有限公司,云南昆明650028

出　　处：《昆明理工大学学报(自然科学版)》2025年第2期147-159,共13页Journal of Kunming University of Science and Technology(Natural Science)

基　　金：国家自然科学基金项目(72461015,52062024);云南省基础研究计划面上项目(202401AT070309);云南省“兴滇英才支持计划”青年人才专项(KKXX202402046);教育部人文社科基金青年项目(24YJC630065);云南产业发展研究项目(2024Z01);云南省研究生导师团队建设项目(2024);昆明理工大学哲学社会科学创新团队项目(CXTD2024007).

摘　　要：为探究山区双车道公路行驶车辆的超车行为特性,基于无人机高空拍摄的超车行为视频数据,提出一种基于随机生存森林的超车持续时间预测模型.首先,利用Tracker软件提取超车行为相关车辆的行驶轨迹,并对超车行为特征进行分析;其次,采用非参数Kaplan-Meier模型和全参数加速失效时间(Accelerated Failure Time, AFT)模型确定影响超车持续时间的关键协变量;最后,构建随机生存森林模型预测山区双车道公路的超车持续时间.结果表明:山区双车道公路平均超车持续时间为12.3 s,考虑超车类型的超车持续时间在无对向来车时表现出显著差异;相较于固定效应AFT模型,全参数AFT模型具有更好的拟合优度,超车距离、超车车辆最终速度、被超车辆类型是影响超车持续时间的关键变量;对比随机森林模型、生存支持向量模型、XG-Boost模型,随机生存森林模型在一致性指数和整合布里尔分数等方面均优于前三者,且考虑超车类型对山区双车道公路超车持续时间的预测效果影响较小;根据变量重要性排名,超车距离和两车的初始速度差对超车持续时间的影响较大.研究结果可为提高山区公路行车安全提供参考.To investigate the characteristics of overtaking behavior on mountainous two-lane roads,this study proposes a prediction model for overtaking duration based on video data captured by drones.First,the Tracker software was used to extract the driving trajectories of vehicles involved in overtaking maneuvers,and the characteristics of overtaking behavior were analyzed.Second,the non-parametric Kaplan-Meier model and the fully parametric Accelerated Failure Time(AFT)model were employed to identify key covariates affecting overtaking duration.Finally,a Random Survival Forest(RSF)model was constructed to predict overtaking duration on mountainous two-lane roads.The results show that the average overtaking duration on mountainous two-lane roads is 12.3 seconds.The overtaking duration,when considering the type of overtaking,exhibits significant differences in the absence of oncoming traffic.Compared to the fixed-effect AFT model,the fully parametric AFT model demonstrates better goodness of fit.The key variables influencing overtaking duration are overtaking distance,the final speed of the overtaking vehicle,and the type of overtaken vehicle.The RSF model outperforms the Random Forest model,Survival Support Vector Machine model,and XG-Boost model in terms of the concordance index and integrated Brier score.Additionally,the consideration of overtaking type has a minimal impact on the prediction accuracy of overtaking duration on mountainous two-lane roads.According to the variable importance ranking,overtaking distance and the initial speed difference between the two vehicles have the most significant impact on overtaking duration.The findings of this study can provide valuable insights for improving driving safety on mountainous roads.

关 键 词：交通工程 超车持续时间 随机生存森林 山区双车道公路 预测模型 

分 类 号：U491[交通运输工程—交通运输规划与管理]