Automated Detection of Train Drivers’ Head Movements: A proof-of-concept study

Abstract

With increasing automation in the rail sector, the train driver's task changes from full control of the train to a supervisory position. This bears risk of monotony and subsequent changes in visual attention, possibly to the worse. Similar to concepts in car driving, one solution for this could be driver state monitoring with triggered interventions in case of declining task attention. Previous research on train drivers' visual attention has used eye-tracking. In contrast, head-tracking is easier to realize within the train driver cabin. This study set out to test whether head-tracking is a feasible alternative to eye-tracking and can provide similar findings. Based on previous eye-tracking research, we compared differences in head movements in automated vs. manual driving, and for different levels of driving speed and driving experience. We conducted a study with 25 active train drivers in a high-fidelity train simulator. Statistical analyses revealed no significant difference in vertical head movements between automation levels. There was a significant difference in horizontal head movements, with train drivers looking more to the right for manual driving. We found no significant influence of driving speed and experience on head movements. Safety implications and the feasibility of head-tracking as an alternative to eye-tracking are discussed.