Cognitive Load During Automation Affects Gaze Behaviours and Transitions to Manual Steering Control

Abstract

Automated vehicles (AVs) are being tested on-road with plans for imminent large-scale deployment. Many AVs are being designed to control vehicles without human input, whilst still relying on a human driver to remain vigilant and responsible for taking control in case of failure. Drivers are likely to use AV control periods to perform additional non-driving related tasks, however the impact of this load on successful steering control transitions (from AV to the human) remains unclear. Here, we used a driving simulator to examine the effect of an additional cognitive load on gaze behavior during automated driving, and on subsequent manual steering control. Drivers were asked to take-over control after a short period of automation caused trajectories to drift towards the outside edge of a bending road. Drivers needed to correct lane position when there was no additional task (“NoLoad”), or whilst also performing an auditory detection task (“Load”). Load might have affected gaze patterns, so to control for this we used either: i) Free gaze, or ii) Fixed gaze (to the road center). Results showed that Load impaired steering, causing insufficient corrections for lane drift. Free gaze patterns were influenced by the added cognitive load, but impaired steering was also observed when gaze was fixed. It seems then that the driver state (cognitive load and gaze direction) during automation may have important consequences for whether the takeover of manual vehicle control is successful.