When Driving Automation Fails. Drivers' Experiences and Interactions

Doctoral thesis, 2014

Drivers’ interactions with automation based on experiences from real driving, and results from driving simulators were under investigation in this thesis. The thesis had a mixed methods research approach and was exploratory. Special attention has been given to driver behavior on exposure to unexpected automation failures. Questions posed were: 1) How are end-users’ of ADAS experiencing automated driving? (Paper I, II, and III) 2) What is the influence of automation failure on driving performance? (Paper IV, V, and VI) 3) How does the extent of automation failure concerning deceleration effect driving performance? (Paper IV, V, and VI) 4) In what way does the level of automation influence driving performance in safety critical events due to automation failures? (Paper VI) The results are based on six papers, adopting an approach with a combination of qualitative and quantitative methods. In Paper I and II focus group interviews were conducted to address end-users’ experiences of vehicle automation, mainly ACC but additional systems were included as well. Paper III was a survey questionnaire addressing the same experiences as did Paper I and II, though with the addition of evaluation of usability and acceptance of ACC. In Paper IV and V a driving simulator experiment was conducted to address automated driving in situations with automation failures. Paper VI was also a driving simulator study with a focus on automated driving in safety critical events due to automation failures. Although Paper VI included an additional level of automation and focused on deceleration failures in which the automated system to some extent failed to brake sufficiently, while Paper IV and V had a broader failure focus including acceleration and speeding failures. The findings include notions on behavioral adaptations and monitoring inefficiencies for drivers facing failures. Implications for design, failure detection, and traffic safety are discussed. With regard to human-automation-interaction it is concluded that automation has effects on driver’s behavior, and that measures towards failure containment and driver control should be taken. Experience, mental models, and situation awareness are used to explain changes and differences in drivers’ behavior.