Modeling and analysis of mandatory lane-changing behavior considering heterogeneity in means and variances

The lane-changing drivers are typical heterogeneous groups, and risky mandatory merging behavior greatly affects road traffic safety and disturbs the normal flow of traffic. This study focuses on the mandatory merging decision behavior at the on-ramp of the interweaving area of a freeway, aiming to build a merging decision assist model which can consider the potential heterogeneity, has strong interpretability, and good model performance. The merging decision data are extracted from the Federal Highway Administration’s Next Generation Simulation dataset. Spearman rank correlation test and step regression method are used to eliminate the multicollinearity among candidate explanatory variables and select the best combination of explanatory variables. A grouped random parameter logit model with heterogeneity in means and variances (GRPMV) and its baseline models are established in this study. The model estimation results show that these selected explanatory variables significantly affect the merging decision behavior, and compared with the baseline model, the GRMPV model has the best model performance and can better capture the unobserved heterogeneity. In the GRMPV model, two random parameters were found. The speed difference between the putative leading vehicle and the merging vehicle significantly affected the random parameter’s mean of the space headway between vehicle M and vehicle PL and the random parameter’s variance of the lateral position of vehicle PF. The research results of this paper can be applied to the autonomous driving assistance system and traffic flow simulation software and can shed light on the mechanism of mandatory merging behaviors.