Trajectory Characteristic Analysis for Abnormal Lane-Changing Behaviors near Congested Expressway Merge Bottleneck

Lane-changing is a common phenomenon in the driving process. It has a key impact on traffic flow and traffic safety due to its complex interaction with surrounding vehicles. Thus, this study aims to explore the influence of abnormal lane-changing behaviors on traffic oscillation by analyzing vehicular trajectory data. The video captured by Unmanned Aerial Vehicle (UAV) is used to extract vehicular trajectory data by deep learning target detection and tracking algorithm. In the whole switching-process from free flow to congestion, lane-changing behaviors increased significantly. Among the 621 vehicles observed, 448 vehicles had lane-changing behaviors, 69 vehicles were greater than one lane-change, in which most are abnormal. The study found that abnormal lane-changing behaviors worsened traffic oscillation.