ARe you watching the road? Comparing Gaze Behaviour and Hazard Prediction between Heads-Up and Heads-Down NDRT Presentation in Augmented Reality

Driver eye movements are commonly used as a predictor of visual attention and to estimate situational awareness of a road scene. However, these gaze patterns will change in automated vehicles, where drivers can engage in distracting non-driving related tasks (NDRTs). There is some suggestion that using augmented reality (AR) to present NDRTs at eye-level with the road might be beneficial. This paper evaluates gaze behaviour and situational awareness of experienced drivers (N = 22) while performing an NDRT presented in AR. Eye-tracking data were collected during an experiment where participants were simultaneously performing a Hazard Prediction task and an AR keypad dialling task, presented either as a Heads-Up Display (HUD) or Heads-Down Display (HDD). The results suggest that eye movements were significantly affected when engaged with an NDRT, resulting in an impaired ability to predict hazards. This was also evident even when attention was explicitly cued towards the area of the road where a hazardous event was occurring. The results from this study indicate that driver fixations alone do not suggest they have situational awareness, in particular when interacting with a distracting NDRT. From this, we provide considerations for designing AR interfaces to be displayed to drivers of conditionally automated vehicles.