Marcel Walch, Forschungsgruppe Weber, schreitet zur Verteidung seiner Dissertation mit dem Titel Driver-Vehicle Interaction in Automated Driving: Overcoming System Boundaries via Driver Involvement. Sein Promotionsausschuss setzt sich aus Prof. Dr.-Ing. Michael Weber (UUlm, Medieninformatik), Prof. Dr. Wendy Ju (Cornell University, USA), Prof. Dr. Enrico Rukzio (UUlm, Medieninformatik), Prof. Dr. Martin Baumann (UUlm, Human factors, Wahlmitglied) und Prof. Dr. Timo Ropinski (UUlm, Medieninformatik, Vorsitz und Protokoll) zusammen.
Interessenten werden gebeten, sich bis zum 14. 9. 2021 bei Herrn Walch direkt anzumelden.
Wir wünschen gutes Gelingen!
Abstract: Automated driving reaches public roads. However, there are still operational limits of automated vehicles, for instance, they can only operate when certain circumstances like good lighting and weather conditions are met. What happens when such a vehicle reaches a system boundary and which role a human driver plays in these situations depends on the level of automation and the interaction paradigm. In SAE Level 3, drivers have to respond appropriately to a takeover request at any given time. In contrast, in SAE Level 4, they do not have to be available as a fallback. In my research, I investigate how drivers can help the automated system to overcome its boundaries in critical handover situations (SAE Level 3) as well as cooperative uncritical situations (SAE Level 3 and 4).
First, I present a study investigating unplanned handovers at a system boundary: a generic handover process was suggested and evaluated with three different takeover request implementations. The study has shown that the participants were able to take over control within three seconds and were able to deal with a broken-down car in dense fog shortly after the takeover. However, handovers can be problematic due to human factor issues like unstable manual control that arise with the use of automation - even when they are planned. In many situations in which an automated vehicle reaches a system boundary it just lacks some information, consequently, an entire shift of the control is not necessary. I developed interaction concepts for such scenarios that are characterized by driver involvement on higher abstraction levels of the driving task: users do not actually steer the vehicle. Several driving simulator studies have been conducted to explore driver-vehicle cooperation in different use cases: object and situation recognition, pedestrian intention prediction, and maneuver approval. These studies provide evidence regarding the feasibility of driver-vehicle cooperation and high usability of the developed user interfaces in these use cases. Moreover, several insights for the future implementation of such concepts are derived.