How do novice and expert drivers prepare for takeover when they are drivengers of a level 3 autonomous vehicle? Investigation of their visual behaviour

By Sharon Ouddiz, Pierre-Vincent Paubel, Céline Lemercier
English

In the very near future, cars will be in autonomous mode most of the time, allowing drivengers (i.e., users of Level 3 automated vehicles) to do what they want until a system inoperability. This implies for the drivenger to disengage from a NDRT (Non-Driving-Related Task) to engage in the driving task within a limited time interval. This takeover period has given rise to a great deal of concern, as studies have highlighted a lack of ability to safely resume the driving activity. In addition, the increase in automation is predicted to be accompanied by a reduction in driving skills and experience. The present study looked at how drivengers prepare for takeover according to their level of driving experience. A group of drivengers (n = 25) were told to watch a movie during a 30-minute autonomous driving phase and prepare for takeover during the 18 seconds of a takeover request. A group of drivers (n = 29) performed the same drive in manual driving mode without any NDRT. We focused on the amount of attention the drivengers allocated to the resumption of the driving task versus the continuation of the NDRT, by investigating their visual behaviour between the start of the takeover request and the system being deactivated. We also evaluated their mental workload via their pupil size. We compared their eye-tracking data with those of the drivers in the non-autonomous condition. Results indicated poorer visual behaviour among drivengers than among drivers. The novice drivengers’ visual behaviour was the most impaired during this critical period. There was also a tendency for drivengers to experience greater mental workload than drivers during the takeover period. These results suggest that drivengers do not have sufficient attentional capacity to cope with the high demand imposed by the takeover, particularly when they are inexperienced.

  • road safety
  • virtual reality
  • autonomous cars
  • crossing decisions
  • message content
  • pedestrian