Changing Regularity of the Interaction Effects of Multi-Scale Factors on Drivers’ Merging Behaviors in the Highway Work Zone

This study aimed to explore the impacting mechanism of macro- and micro-factors and multi-scale synergy on drivers’ merging behaviors in the highway work zone, and then to facilitate future merging prediction research. The merging behavior, requiring drivers to detect adjacent vehicles and real-time traffic conditions simultaneously, is a complex cognitive process. Previous studies have mainly focused on the stable impacts of limited factors on merging probability, but ignored the varying states and drivers’ performance. Status-changing positions in the whole merging process were first identified by constructing and analyzing the relationship between running speed and distance from the construction area. Subsequently, the interaction analysis was conducted among the multi-scale traffic factors, utilizing optimized logistic models and interaction estimates, thus establishing macro- and micro-factor connections. Besides, the marginal effect was calculated to analyze the fluctuation degree of these connections. Finally, a multilevel identification framework was proposed, whose effectiveness and practicality were validated using 744 naturalistic vehicular trajectories from a real highway work area. At different positions, drivers are affected by various factors to varying degrees. While approaching the construction area, drivers become more passive, thus trying to avoid rear-ending the lead vehicle but ignoring the lag vehicle. Besides, traffic volume also affects drivers’ merging decisions by confusing their cognition toward the time headway. This research indicates that dynamic interaction effects between multi-scale factors could provide far-reaching benefits for lane-changing prediction. The findings provide a basis for formulating traffic management policies and constructing driving assistance systems.