Predicting self-assessment of the out-of-the-loop phenomenon from visual strategies during highly automated driving

International audienceDuring highly automated driving, drivers do not physically control the vehicle anymore, but they still have to monitor the driving scene. This is particularly true for SAE level 3 (SAE International, 2016), as they need to be able to react quickly and safely to a take-over request. Without such an (even partial) monitoring, drivers are considered out-of-the-loop (OOTL) and safety may be compromised. This OOTL phenomenon may be particularly important for long automated driving periods. The current study aimed at scrutinizing driver’s visual behaviour for a long period of highly automated driving (18 minutes). Intersections between gaze and 13 areas of interest (AOI) were analysed, considering both static (percentage of time gaze spent in one single AOI) and dynamic (transitions from one AOI to another) patterns. Then, a prediction of the self-reported OOTL level (subjective assessment) from gaze behaviour was performed using Partial Least Squares (PLS) regression models. The outputs of the PLS regressions allowed defining visual strategies associated with good monitoring of the driving scene and paved the way for an online estimation of the OOTL phenomenon based on driver’s spontaneous visual behaviour