Postural and physiological indicators of drowsiness at the wheel compared in partially and conditionally autonomous on-road driving

• Postural and physiological characterization of asleep drivers during on-road driving. • Level L3 induces drowsiness more quickly than level L2+ when driving on the road. • Level of autonomy influences postural behavior. • Level of autonomy affects drivers’ subjective feelings. With increasing numbers of partially and conditionally autonomous vehicles (i.e., levels L2 to L3) on the roads, altered driver states such as inattention or drowsiness at the wheel may be exacerbated. This new way of driving will, however, make the indicators traditionally used to detect altered states less relevant. New detection methods using other indicators, such as physiological and postural indicators, could be a solution. In this study, 21 participants drove on motorways in a dual control vehicle at two levels of automation (L2+ and L3). Seat pressure and physiological (i.e., cardiac and respiratory) indicators were continuously recorded and state of drowsiness was classified according to PERCLOS70. In defined states of drowsiness, different physiological and postural characteristics were identified. Results show that postural indicators are more sensitive to level of automation than physiological indicators, suggesting that posture may reflect behavior associated with a particular level of autonomous driving. This suggestion is supported by subjective feelings elicited from the drivers. The multiple postures observed to be associated with deep drowsiness (or falling asleep) point to inter-individual postural variability. Postural information could therefore reflect both active behavior and drowsiness, whereas physiological information appears only to reflect drowsiness. This study demonstrates the value of using such measurements jointly to detect drowsiness at the wheel.