Citation

Choundhury C, Ben-Akiva M, Rao A, Lee G, Toledo T. Transp. Traffic Theory 2007; 17: 711-733.

Copyright

(Copyright © 2007, Publisher varies)

DOI

unavailable

PMID

unavailable

Abstract

This paper demonstrates a methodology to model state dependency in lane changing behavior by applying it to model the merging behavior of drivers on a congested freeway. The model has explicit normal, courtesy and forced merging components sequenced in a single detection framework. The decision to initiate a merge and the acceptance of gaps to complete the merge are affected by the decision state of the driver as well as neighborhood variables and driver characteristics (agent effect). The model parameters for state-transition are estimated simultaneously with the parameters of the gap acceptance models with detailed vehicle trajectory data using maximum likelihood estimation technique. The statistical model section criteria using the estimation results showed that the proposed state dependent merging model is superior to a single level instantaneous model estimated with the same data ignoring state dependency. This result was further strengthened by a validating case study, which compared the results obtained from simulation runs from each of the model implementation in the microscopic traffic simulator MITSIMLab. It may be noted that the methodology presented in this paper to model state dependency in merging behavior can be extended to other driving behavior models as well and this will be explored in future research.