Investigation of Discretionary Lane-Change Characteristics Using Next-Generation Simulation Data Sets

In this article, we investigate discretionary lane change (DLC) characteristics using Next Generation SIMulation (NGSIM) data sets. We first develop a set of heuristic rules to automatically filter out abnormal samples from a massive trajectory data set and identify DLC trajectories from a mixture of mandatory lane change (MLC) and DLC trajectories. Then, we investigate a variety of DLC characteristics. We demonstrate that the kernel part of every normal DLC trajectory can be approximately depicted by a certain fifth-order polynomial. Moreover, we discuss the definition of begin/end points of DLC actions and show that the duration time of DLC actions follows a log-normal distribution with respect to navigation speed. All the findings can help promote temporal and spatial accuracy of lane changing models.