Research on the Spatiotemporal Evolution of Disturbed Through Vehicles at Signalized Intersections Based on Trajectory Data

Abstract

Disturbances in vehicle trajectories are noticeable in the front section of weaving areas at signalized intersections, primarily due to merging conflicts. These disturbances often pose heightened safety risks. To investigate the mechanisms of risk formation, this paper proposes a method for analyzing the dynamic evolution of vehicle disturbances, integrating time series modeling with panel data analysis. Grid-based trajectory data were used to quantify localized disturbances in vehicle movements, while a fuzzy logic algorithm was employed to comprehensively output disturbance values. Based on this, the spatial Markov model that considers spatial lag variations across different regions was constructed to reproduce the transition trends of disturbances in through vehicles affected by merging right-turn vehicles. The spatial dynamic Durbin panel models (SDDPM) were constructed to identify key factors influencing these disturbances. The experiments were conducted at selected signalized intersections in Xi’an, China. Results indicate that the spatial Markov model outperforms the traditional Markov model in describing vehicle disturbance transition trends under various spatiotemporal lag conditions. The SDDPM provides a better fit than other panel data models that consider only spatial or temporal lag effects. Key factors influencing the disturbance level of a unit include the speed standard deviation and maximum yaw angle within the unit, as well as the maximum yaw angle within neighboring units. This analysis of vehicle disturbance patterns enables more precise identification of safety transitions and their underlying triggers, offering a foundation for risk identification and prediction of disturbed trajectories.