Analyzing differences of highway lane-changing behavior using vehicle trajectory data

Abstract

Analyzing fine-resolution lane-changing behavior is of great significance in the traffic flow modeling and safety assessment. The fact that lane-changing behavior varies over different environment and operation factors poses great challenges for lane-changing behavior modeling and analysis. To explore the differences of highway lane-changing behavior in complicated scenarios, the analysis of this study stems from a large volume of lane-changing trajectories (i.e., 6506 lane-changing trajectories) collected from the Highway Drone (HighD) dataset. In real-world scenarios, to identify and extract lane-changing start and end points of lane-changing behavior from trajectories, we present a fusion method by using frame interval and lateral displacement variation in trajectory data, which can be calibrated by the visualization. The empirical results show that: (a) the relative error between the extracted lane-changing time and real one is less than 4%; (b) the lane-changing time is ranged from 2.88 s to 7.32 s, while the lane-changing speed is ranged from 18.83 m/s to 43.30 m/s. For the difference analysis, we use the tests of statistical significance and heatmaps to analyze the differences of lane-changing behavior under complicated scenarios that involve different lane numbers, lane-changing directions, vehicle types, and speed limit condition. The results demonstrate that the lane-changing behavior varies over different scenarios significantly. Overall, the systematical analysis of lane-changing behavior in this study provides insight into the microscopic traffic flow modeling.