Lane-changing models are important components of microscopic traffic simulation tools. State-of-the-art models adopt a two-level decision framework comprising of lane selection followed by gap acceptance to explain lane changing behavior. However, lane changes have been modeled as an instantaneous process that is initiated and completed at the same time step following gap acceptance. But in reality, lane-changing is a continuous process with an average duration of 3-7 seconds, as reported in past works. This practical maneuvering requirement might seriously undermine the modeling assumptions of the standard lane changing models, especially in low density traffic conditions where lane change duration, owing to ample gap sizes, can take longer than average. This paper presents a framework for modeling the duration of lane changes. An additional level is added in the decision framework of the driver that governs the exact instance of completion of the lane change execution post gap acceptance. The model is estimated using detailed vehicle trajectory data collected from a stretch of Lankershim Boulevard, an urban arterial located in California. The execution decision is found to be affected by the speed of the driver and the difference between the actual and acceptable gaps. The goodness-of-fit achieved by the proposed model is found to be significantly better as compared to a base model that does not capture the duration of lane changes. The necessity to model the duration of lane changes is further strengthened by a validation study within MITSIMLab where the extended model consistently outperforms the base model in terms of its ability to accurately simulate the observed traffic characteristics.
Abstract