Vigilance decrement in automated vs. manual driving: Insights from multimodal data

Abstract

Driver vigilance is a crucial factor in road safety. Vigilance decrement can pose a significant challenge under Level 2 driving automation, where drivers are relieved of active vehicle control but must still continuously monitor the system. This study aims to evaluate the dynamic changes in vigilance during prolonged automated driving using multimodal data. A simulated driving experiment was conducted in which participants drove in either manual or automated mode while responding to randomly presented traffic sign targets. Subjective vigilance ratings, target detection performance, and physiological signals, including eye movements, electroencephalogram (EEG), and electrocardiogram (ECG), were recorded and analyzed. The results revealed that the first significant vigilance decrement in automated mode occurred around 15 min into driving, with a second decrement occurring between 25 and 40 min. In contrast, the first significant vigilance decrement in manual driving was observed after 25 to 40 min of driving, later than in automated driving. Additionally, vigilance decrement was more pronounced in automated driving, as evidenced by lower subjective vigilance ratings, reduced fixation counts, and increased EEG wave powers. Through cross-validation of multimodal metrics, this study demonstrates that vigilance decrement occurs earlier and to a greater extent in automated driving compared to manual driving. These findings provide a valuable theoretical foundation for developing intervention strategies to enhance driver vigilance and contribute to the overall improvement in driving safety.