Factors Affecting Lane-changing Duration Based on Time-dependent Proportional-hazards Regression Model

Abstract:

Based on the time-dependent proportional-hazards regression model， factors affecting lane-changing duration（LCD） are studied. Five hundred and sixty complete lane-changing （LC） trajectories are extracted from the newly opened HighD dataset， and the overall distribution of lane-changing duration is estimated using the non-parametric estimator. Then， the time-dependent proportional-hazards regression model is introduced to analyze the factors affecting LCD. It is found that most drivers finish LC in 3?8 s. There are six variables， of which the most important one is current vehicle’s speed. Additionally， the preceding vehicle in current lane and the following vehicle in target lane have significant effects on LCD.