Abstract: In order to study the safe lane changing behavior in highway weaving areas and reveal the influence of potential heterogeneous interaction characteristics in the merging decision-making behavior of vehicles on entrance ramps on the merging behavior, the grouped mixed parameter logit model that took into account the mixed parameter correlation and heterogeneity in means was proposed. The proposed model was calibrated by the variables extracted from the NGSIM dataset, and the influence of each variable on the merging decision was quantitatively analyzed by combining with marginal effect. The research shows that the grouped mixed parameter logit model, which combines mixed parameter correlation and mean heterogeneity, exhibits significantly better data fitting performance than the conventional grouped mixed parameter logit model, the grouped mixed parameter logit model, which only deals with mixed parameter correlation, and the grouped mixed parameter logit model, which only explores the phenomenon of mean heterogeneity. Two significant mixing parameters captured by the proposed model and the mean value of the mixed parameter of the longitudinal distance difference between the front and rear vehicles in the target lane increase with the increase of the speed difference between the converging vehicle and the vehicle in the front of the target lane, and decrease with the increase of the lateral distance difference between the converging vehicle and the vehicle in the front of the target lane. There is a significant positive correlation between the above two mixed parameters, whose interaction will increase the probability of the merging vehicle driver changing lanes.

Key words: traffic and transportation engineering; traffic system; merging behavior; interweaving area; mixed parameter correlation; heterogeneity in means

摘要： 为研究高速公路交织区安全换道行为，揭示入口匝道车辆汇合决策行为中潜在的异质交互特性对汇合行为的影响，提出一种考虑混合参数相关性和均值异质性的分组混合参数logit模型，采用NGSIM数据集提取变量进行模型标定，结合边际效应量化分析了各变量对汇合决策的影响。研究表明：相较于常规分组混合参数logit模型、仅对混合参数相关性加以处理的分组混合参数logit模型以及仅探究均值异质现象的分组混合参数logit模型，综合考量混合参数相关性与均值异质性的分组混合参数logit模型在对数据的拟合性能上展现出显著更优的效果。提出的模型捕捉到了2个显著的混合参数，且目标车道中的前车和后车之间的纵向距离差的混合参数均值会随着汇入车辆和目标车道前车的速度差的增加而增加，随着汇入车辆和目标车道前车的横向距离差的增加而减小，上述2个混合参数之间存在显著的正相关性，其交互作用会提高汇入车辆驾驶员换道的概率。

关键词: 交通运输工程 ；交通系统；汇合行为；交织区；混合参数相关性；均值异质性