Mitigating cognitive workload in automated driving takeovers: Effects of non-driving-related tasks, takeover request modalities, and lead times

In Level 3 conditional automated driving, non-driving-related tasks (NDRTs) may reduce drivers’ readiness to resume control, while behavioral indicators alone may not fully capture concurrent changes in workload and psychophysiological stress. This study examined the effects of NDRT packages, takeover request (TOR) modalities, and TOR lead time (TORlt) on cognitive workload, cardiac responses, gaze behavior, and vehicle-control performance. A within-subject experiment used EEG, eye movements, cardiac interval measures, NASA-TLX, and post-TOR driving indicators. News reading and video watching produced larger post-TOR changes in selected physiological indicators and greater vehicle-control effort than the phone call task, while video watching also resulted in higher subjective workload. Visual TORs were associated with higher subjective workload, altered cardiac interval responses in the primary analysis, and greater control fluctuation than auditory and auditory-visual TORs. Shorter TORlt was associated with greater changes in cardiac interval measures, whereas its relationship with steering corrections was less robust across sensitivity analyses. Physiological and behavioral responses were not always aligned, as drivers sometimes maintained relatively stable vehicle control despite elevated workload or psychophysiological stress. These findings support the value of multimodal driver-state monitoring during takeover transitions. Under the present simulator conditions, auditory and auditory-visual TORs may warrant further evaluation when drivers are engaged in screen-based NDRTs or when lead time is limited. Overall, the results provide preliminary evidence for condition-specific adaptive warning design and suggest that combining physiological, visual-attention, subjective, and vehicle-control indicators may improve the characterization of driver state during automated driving takeovers.