.png){kind=icon}
.png){kind=icon}
Return
Takeover performance in level 3 autonomous driving: the effect of lead time and non-driving related task type
.png){kind=icon}
PRE
AI
S
J
Q
DOI：10.1016/j.trf.2026.103593.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Level 3 automated driving systems permit drivers to engage in non-driving-related tasks (NDRTs), making the safety of the takeover process critical. This study proposes a three-stage model of the takeover process (Arousal, Decision-Making, and Execution) and investigates how two distinct cognitive mechanisms (perceptual modal conflict and cognitive saturation) affect different stages. Through two simulated driving experiments, this study investigated how NDRT type and takeover lead time (TLT) jointly influence driver takeover performance. Experiment 1 manipulated NDRT modality (manual-visual vs. manual-auditory vs. no task) and TLT (3 s vs. 5 s). Results showed a significant main effect of NDRT type on reaction time (p < 0.001) and a significant NDRT & times; TLT interaction (p = 0.046). The manual-visual task caused the most severe delays, being 0.11 s slower than manual-auditory. At the 3 s TLT, the no-task condition elicited significantly faster reactions than both NDRT conditions (ps < 0.01), which themselves did not differ significantly. In contrast, at the 5 s TLT, the manual-visual task resulted in significantly longer reaction times than both the manual-auditory and no-task conditions (ps < 0.05), while the latter two showed no significant difference. Experiment 2 held visual modality constant while varying cognitive load (low vs. high). A significant main effect of NDRT type on accuracy was found (p = 0.010), with both low-load (81.67%) and high-load (80.00%) tasks reducing accuracy versus no-task (91.67%), while showing no significant difference between them. And the interaction between NDRT type and TLT on accuracy was non-significant. These results demonstrate a double dissociation: modal conflict in the Arousal Phase primarily delays reaction time, while cognitive saturation in the Decision-Making Phase impairs accuracy. The findings provide a mechanistic framework for resolving previous contradictions and suggest that safety interventions should implement two-stage warning systems and cross-modal alerts to address modal conflict, while providing cognitive decision support to mitigate saturation.
Keywords:
Autonomous driving
Non-driving related task
Takeover lead time
Journal
IF:
4.4
Papers:
3.3K
Citations:
1.3W
Organization
Cited Papers