Analysis of drivers’ reaction to automation failures in a curve scenario

The last years the automotive industry experienced an enormous increase of Advanced Driving Assistance Systems installed in vehicles, moving the driver’s role into a more passive. Previous studies have shown delays during the takeover in case of automation failures and related them to the out of the loop performance. So far, though, the research focused on rear end collision scenarios whereas curve scenarios received little attention. The purpose of this study was to examine the drivers’ reaction and performance to an automation failure of SAE level 2 occurring in a curve scenario. The analysed data set stem from a moving based driving simulator study with 18 participants within the project SHADES. The findings of this thesis showed that the initial response to automation failures in curves is steering, however not all participants performed in the same manner. Two groups were identified who differed significantly in their reaction time and their following steering performance. The slow responding group with an average reaction time of 2.97 s showed a maximum lane deviation of 1.11 m and spent 39% of the time outside of the lane, resulting in an insufficient take over performance. The fast responding group in contrast, with a reaction time of 1.30 s drifted from the lane center only 0.52 m and therefore managed to stay in the lane. The steering wheel input and the maximum lateral jerk reached by the slow responding group showed higher values compared to the slow responding group (14 m/s3 vs. 7 m/s3). On the other hand, the participants of the fast responding group maintained significantly higher steering wheel control before the failure, compared to the slow responding group. Altogether, those results seem to indicate that the fast responding group executed rather lane correction whereas the slow responding group showed collision prevention performance and out of the loop behaviour. Besides, this study showed that the lateral position and the heading angle at the start of driver’s steering response is highly correlated to the maximum lane deviation reached by each participant during the takeover. Significant correlations between the drivers’ performance and age, gender, trust or technique interest could not be found so further research with a larger sample would be required to investigate if any of those variables could influence driver’s reaction. The findings from this study can be used for modelling the drivers’ reaction in the specific scenario and to provide information for the design of automated driving functions in critical scenarios