Steffen Maurer, external phd candidate of the research group human-computer interaction, defended his phd thesis with the title Guardian Angel - A driver-vehicle interaction for oversteering the driver in a highly automated vehicle. His examiners were Prof. Dr Enrico Rukzio, Prof. Dr Timo Ropinski (both instiute of media informatics), Prof. Dr Martin Baumann und Prof. Dr. Anke Huckauf (both Ulm University), and Prof. Dr Jacobo Torán (head of the commity and minute taker, all Ulm University).
Congratulations!
Abstract
With the rise of automated driving technologies, the functionalities of automotive systems are getting more and more refined and capable of driving in new situations. As injuries caused by road traffic are high throughout the world, automated technologies will hopefully help to reduce the risk of accidents and casualties in road traffic.
Automation is already an integral part of other transportation modes, for example in aviation. The capabilities of automation allow for a continuous evaluation of the situational state and the input of the human operator and to intervene in problematic and risky situations.
As driving manually is still possible and sometimes even required until autonomous driving has been achieved, it is possible for the human driver to make mistakes. An automation that could intervene in critical situations would act as a guardian angel and help mitigating the consequences.
For this matter, this thesis will provide a taxonomy of automated systems and use it to show the absence of a guardian angel-like system in the automotive domain.
To get a general idea of the acceptance of such systems and the required human-machine interface a simulator study is conducted. This study presents to situations to the participants, a critical and a non-critical situation during driving where the automation impeaches the driver and takes over control of the vehicle. Qualitative as well as quantitative user data is gathered along with an ethical examination and yields ambiguous results. Most participants accept an intervention in critical situations but only half are willing to accept a forced take over by the automation in a non-critical situation, for example to avoid an unnecessary detour if the driver is about to miss a turn. An examination concerning the ethics of such an intervention shows a dissonant view, especially regarding autonomy and felt autonomy by the human driver.
For a better understanding of the situations in which the human driver would accept an intervention by the system a study is designed that uses a gamified approach. The study shifts the perspective of the participants, as they are not the driver in the examined scenario, but are acting as the automation impeaching the (simulated) driver and giving full control to the automation.
The results of all previous studies are then combined in another driving simulator study to evaluate a guardian angel-like system in situations that have been identified in the gamification study. The results show that a guardian angel-like system is highly useful and generally accepted in dangerous situations. However, the self-assessment of the driving behavior of the participant has a huge influence on the acceptance in non-critical situations, as it significantly correlates with the aggressiveness in driving of the respective participant.
The research for this thesis was conducted during the course of a publicly funded research project called KoFFI. For this project a software architecture was designed and implemented. This architecture also allows the integration of mechanisms that act as the described guardian angel and are able to intervene with the driving.
In summary, an automated system that is able to intervene in critical situations during manual driving can act like a guardian angel and mitigate risky situations. It is also perceived as a guardian angel in very critical situations but can be felt as an overcautious protector in nonhazardous situations.
If you are interested in joining the video conference, please contact Mrs Duschek.
