Unveiling driver workload dynamics and road safety risks in assisted driving systems

The progressive introduction of driving automation is profoundly transforming driver-vehicle interaction. Although automated driving systems are designed to improve vehicle stability and reduce driver workload, their influence on driver behavior, especially during transitions between manual control and assisted or automated control, remains insufficiently understood. The aim of this study is to investigate how different driving modes affect lateral vehicle control and visual behavior during curve negotiation, with specific attention to the transition from manual driving to Level 2 assisted driving and the subsequent resumption of manual control. A simulator-based experiment was conducted in which drivers sequentially experienced three road segments: manual driving, Level 2 assisted driving, and manual driving resumed after assisted driving. A two-stage analytical approach was adopted. First, linear mixed-effects models were fitted to assess the effect of driving mode on lateral trajectory (Lane Gap) and on two temporally defined visual indices related to curve negotiation. Subsequently, a fuzzy c-means clustering analysis was applied to the same variables to identify latent multivariate behavioral states, allowing the representation of uncertainty and gradual transitions in driver behavior. The results revealed significant effects of driving mode on both vehicle trajectory and visual engagement. Assisted driving was associated with reduced lateral variability and diminished anticipatory and guidance-related visual engagement. After assisted driving, manual driving behavior showed partial recovery but did not fully return to the initial manual-driving pattern. These findings provide insights relevant to road safety and the design of human-machine interfaces and demonstrate the value of studying driver behavior in assisted driving contexts.