Effects of non-driving related tasks in prolonged conditional automated driving – A Wizard of Oz on-road approach in real traffic environment

Abstract

ObjectiveThe aim of this study was to investigate the effects of non-driving related tasks on the drivers’ fatigue in prolonged conditional automated driving (CAD) in on-road real traffic environment. BackgroundIn CAD, the driving task is executed by the system. This leads to a monotonous situation for the driver as he just has to intervene if requested by the system. Monotony and increasing automation are known causations for fatigue and drowsiness. In previous studies (mostly conducted in driving simulators) an impaired take-over performance due to emerging fatigue and drowsiness could be observed. In the driving simulator studies a rapid increase in drowsiness and fatigue could be observed. To investigate if similar results occur in real traffic environment on-road a Wizard of Oz approach was used. MethodForty-two participants experienced prolonged conditional automated rides on-road in real traffic environment. To provoke fatigue one part of the participants had to engage in a monotonous monitoring task. A control group had no requirements according to the NDRT and had free choice of their activity. Effects on fatigue were measured using percentage of eye-lid closure over time (PERCLOS) and subjective Karolinska Sleepiness Scale (KSS). ResultsProlonged CAD in real traffic environment and simultaneously engaging in a monotonous monitoring task negatively affected the drivers’ state. PERCLOS and subjective KSS significantly increased compared to the control-group. ConclusionFatigue in CAD can emerge in real traffic environment as fast as in driving simulator environments. Especially when participants had to engage in a monotonous monitoring task PERCLOS and KSS rose. ApplicationThe results of this study demonstrate that due to increasing monotony fatigue can emerge in CAD within a one hour drive. The development of fatigue is comparable to the development of fatigue in the driving simulator, where an impaired take-over performance due to fatigue could be observed. Therefore, a monitoring of the driver state and adapted assistance in a take-over situation seems to be a good opportunity to ensure safety in CAD.