Analysis of Behavioral Patterns of Lane Change in Weaving Areas

Based on vehicle trajectory data captured by unmanned aerial vehicle (UAV) video, this study investigates the differences in the behavior pattern distribution of different lane change (LC) maneuvers in the weaving area. The hierarchical Dirichlet process and hidden Markov model (HDP-HSMM) algorithm is first used to segment the LC sequences to generate behavior primitives with risk attributes, and then the Gaussian mixture model and latent Dirichlet allocation (GMM-LDA) algorithm is used to cluster the primitives to obtain the behavior patterns that make up the LC process. The results show that the composition of patterns differed significantly between consecutive LC and left/right LC, and that the proportion of riskier patterns was relatively high. This study can provide a basis for the development of LC models for autonomous vehicles and early warning of risky LC maneuvers through an in-depth analysis of the patterns of LC maneuvers occurring in the weaving area.