Level of automation: effects on train driver vigilance

In this experiment we investigated the impact of various levels of train automation on driver vigilance. Forty student participants were trained to drive a low physical and geo-specific fidelity computer simulated passenger train on an intercity route. Vigilance task performance was measured as the track distance between the train and a safety critical event at the point where emergency braking commenced. The vigilance measure was obtained while manipulating two IVs: (1) between-subjects variable - Level-Of-Automation (LOA), being either a Nil, low, intermediate, or high level of train automation; and (2) within-subjects variable - time-on-task, having two levels: safety critical event onset before 10 minutes (Early) and after 60 minutes (Late) of driving.Arail signal failurewas used as the safety critical event to be detected. Results revealed mean driver vigilance was poorest at the high level of train automation (auto-pilot). A trend of increased vigilance was apparent among participants in the low LOA group (equipped with in-cab signalling), when compared to the nil LOA group (manual). The vigilance of participants in the intermediate LOA group (equipped with in-cab signalling and over-speed intervention) was comparable to that of those in the nil LOA group. A trend emerged for a time-on-task related vigilance decrement for the auto-pilot train drivers only. The driving safety implications of these findings, in terms of potentially positive and negative effects of various levels of automation on driver vigilance, are discussed.