Prof. Dr.-Ing. Barbara Deml

Prof. Dr.-Ing. Barbara Deml studierte an der Universität Regensburg Psychologie und Sprecherziehung. Sie promovierte im Anschluss daran in einem Sonderforschungsbereich der Deutschen Forschungsgemeinschaft (SFB 453 – Wirklichkeitsnahe Telepräsenz und Teleaktion) an der Universität der Bundeswehr (UniBw) München zum Doktor der Ingenieurwissenschaften. In ihrer Dissertation beschäftigte sich mit der Gestaltung haptischer Mensch-Maschine-Schnittstellen; die Arbeit wurde mit dem Forschungspreis der Universität der Bundeswehr ausgezeichnet. Nach einer Forschungstätigkeit an der TU München und der Carnegie Mellon University in Pittsburgh ist sie an die UniBw zurückgekehrt. Sie erhielt dort einen Ruf auf eine Juniorprofessor für Kognitive Ergonomie und folgte dann einem Ruf an die Otto-von-Guericke-Universität Magdeburg, um den Lehrstuhl für Arbeitswissenschaft und Arbeitsgestaltung zu leiten. Frau Deml hat 2012 einen Ruf an die Universität Ulm abgelehnt und in diesem Jahr einen Ruf an das Karlsruher Institut für Technologie (KIT) angenommen. Seit dieser Zeit leitet sie das Institut für Arbeitswissenschaft und Betriebsorganisation am KIT. Daneben ist Frau Deml in vielen Gremien aktiv, so als Vizepräsidentin der Gesellschaft für Arbeitswissenschaft.

Prof. Dr.-Ing. Barbara Deml
Karlsruher Institut für Technologie
Institut für Arbeitswissenschaft und Betriebsorganisationsation (IFAB)
Institutsleiterin
Engler-Bunte-Ring 4, Geb. 10.91, Raum 007
76131 Karlsruhe
Telefon: +49 721 608-44250
www.ifab.kit.edu

Synopsis

On the design of artifical intelligence that understands human behavior – an example within the domain of autonomous driving

Prof. Dr.-Ing. Dipl.Psych. Barbara Deml*. Prof. Dr.-Ing. Fernando Puente León**.M. Sc. Jonas Imbsweiler*. M. Sc. Hannes Weinreuter**

Karlsruhe Institute of Technology (KIT). Institute of Human and Industrial Engineering (ifab)*. Institute of Industrial Information Processing (IIIT)**

At a first sight, there is no need for an artificial intelligence to behave like a human being. Of course, its perceptual system is based on different sensors as well as its information processing and decision taking differ from human cognition. However, an artifical intelligence is not an end in itself, but it is intended to assist a human user. And due to that reason, it is not sufficient to design only an artificial system that is intelligent – what, however, is difficult enough. But it is needed to design a system, which is intelligent and which is able to understand, at least particially, human behavior and human experience. Besides, it has to be able to behave and to act so that a human user accepts it.

Within this work an example of such an artificially intelligent system is provided that refers to the domain of autonomous driving, which has been discussed so often in recent days. There are numerous traffic situations in which a blockade of the traffic flow occurs when there is no cooperative behavior between the road users. Such situations cannot always be regulated by traffic rules, as is the case, for example, when road users approach a constriction with obstacles on both sides of the road, a T-junction, or a X-junction simultaneously. Within this work, a consortium of measurement and control engineers as well as ergonomics investigates how automated vehicles can interact intelligently with non-automated road users in such cooperative traffic scenarios.To this end, an empirical study with a test vehicle was carried out, in which the interaction of a human car driver acting as the test subject is examined in several representative cooperative scenarios with other, instructed road users. Based on this study, it was analyzed how human beings communicate in such situations, and a discrete-event model is developed to describe the underlying cooperative decision-making behavior. Next it is to be examined whether the optimized behavior is understood and accepted by humans. To this end, in a second survey in a driving simulator it will be analyzed both whether a human passenger of the automated vehicle agrees with the vehicle's decisions, and whether a human road user is able to implicitly and explicitly communicate with the automated vehicle in a familiar way.The results of this project will contribute to design automated vehicle behavior to seamlessly fit in the human traffic environment at the time of their practical application.

This work is carried out within the DFG Priority Programme SP1835 – Cooperative Interacting Vehicles