Discrete Choice Models for Gap Acceptance at Urban Expressway Merge Sections Considering Safety, Road Geometry, and Traffic Conditions

Abstract

A number of discrete choice models, including a multinomial logit model (MNL), a nested logit model (NL), and a latent choice set model (LCS), are applied to the representation of gap acceptance by merging vehicles (MVs) on urban expressways and their results are compared. The results show that all models give consistent signs to the estimated parameters. It is evident that the LCS model is capable of representing gap acceptance and allows exploration of the latent choice made by MVs. Comparison of the models shows that LCS is superior to the traditional MNL and NL models in terms of goodness of fit and provides a more realistic simulation of MV behavior. The effects of geometry and traffic conditions are fully considered in this study, and they are found to significantly influence the choices made by MVs. A longer acceleration lane may provide a relatively larger gap for MVs and, therefore, motivate them to accept the gap. With respect to traffic conditions, an MV tends to reject a gap under low (<20 vehicles/km/lane) or high density (>40 vehicles/km/lane) conditions. However, gap acceptance is more likely when the density is from 20 to 40 vehicles/km/lane. With regard to safety, the concept of time to collision (TTC) is adopted to capture the reactions of MVs to mainline vehicles (MLVs). Both leading and following TTCs (between the MV and the leading/following MLV, respectively) are implemented in the models. Analyses indicate that the TTC thresholds for an MV to reject or accept a gap are 5 and 6 s for the leading and the following TTC, respectively.