Comparative analysis of continuous and single lane-changing behaviors in freeway merging areas

Lane-changing behavior in freeway merging areas comprises two heterogeneous types, single lane-changing (SLC) and continuous lane-changing (CLC). This study investigates these differences in maneuver characteristics, safety, and interaction patterns using high-resolution trajectory data. Hierarchical Dirichlet process hidden semi-Markov models (HDP-HSMMs) and Gaussian mixture hidden Markov models (GMM-HMMs) are applied to capture temporal evolution, with representative patterns identified through dynamic time warping (DTW)-based K-means clustering. Results show that CLC is more aggressive: it begins 22.22 m earlier and completes the same lateral displacement 1.12 s faster than SLC. When both mainline preceding and following vehicles are present, TTC ≤ 5.5 s events account for 18.84% of CLC cases, versus 6.15% for SLC. CLC and SLC exhibit distinct interaction patterns and evolutionary dynamics, with CLC more frequently involving speed adjustments by surrounding vehicles, whereas SLC is predominantly characterized by speed-maintaining interactions. These findings inform merging behavior modeling and automated-driving risk assessment.