Investigating Drivers’ Heterogeneity and Its Impacts on Interactive Lane-Changing Maneuvers

Abstract

This paper investigates the impacts of driving-style heterogeneity on safety, interaction efficiency, and comfort in discretionary and mandatory lane-changing interaction scenarios. Leveraging two naturalistic driving datasets, highD and exiD, complete lane-changing interactions are extracted through data processing, and critical longitudinal and lateral features influencing these interactions are identified and quantified through principal component analysis. Based on these features, vehicle pairs consisting of a lane-changing vehicle (LCV) and its following vehicle (FV) in the receiving lane are clustered using the k-means method, resulting in four representative LCV–FV compositions: aggressive–aggressive, aggressive–mild, mild–aggressive, and mild–mild. Analysis of variance and post hoc paired tests revealed three main findings: (i) aggressive–aggressive compositions were associated with elevated collision risks and reduced comfort levels; (ii) aggressive–mild compositions tended to prioritize interaction efficiency; and (iii) FVs were more likely to refrain from yielding to LCVs in discretionary maneuvers compared to mandatory cases, indicating a stronger emphasis on efficiency. These findings provide insights into how driving-style heterogeneity shapes lane-changing interactions and may inform the design of advanced driver assistance systems and autonomous vehicles operating in highly interactive scenarios.