Einladung zum Vortrag im Kolloquium Technische Kybernetik

Distributed and embodied intelligence methods in robotics

Prof. Dr. Paul Levi

Institut für Parallele und Verteilte Systeme
Universität Stuttgart

    Zeit: Dienstag · 31. 5. 2005 · 16:00 Uhr
    Ort: Raum V 9.31 · Pfaffenwaldring 9 · Campus Stuttgart-Vaihingen

Abstract

The foundations and the issues of artificial intelligence and cognition are thoroughly investigated since about 40 years by different disciplines. In the early days of AI (Artificial Intelligence) the scientific community characterized AI not only by typical applications like chess playing, but they also claimed that the used methods are the same as the ones which are used by men.
In the field of the classical paradigm of cognition - which is the basic block of intelligence - abilities like perception, generation of language, communication, learning, planning etc. are considered. Hereby there was no embodiment, no direct interaction with the environment. A classical example of such a connection of non embodied cognition and AI is a chess programme. Although the human is losing, he still feels that he is intelligent and the computer is stupid, because the chess computer represents a virtual world with precisely defined states and legal moves, and operates very different than a man is doing it. To remedy such deficiencies there exist the new assertion that intelligence requires a body (embodiment of intelligence). This statement means that we have to close the sense-plan-act-observe-cycle by real world applications. Here robots (macro- and micro- sized) and autonomous cars are showcases to implement and validate this new approach. The artificial creatures which behave in a real world are autonomous agents. They define the basic model for investigation of distributed Intelligence (DAI) and even for swarm intelligence.
In more details this talk is devoted to the presentation of the use of different agent-based methods of distributed and embodied intelligence like conflict handling by negotiations, cooperative planning, dynamic task assignment, organization forms and "Partially Observable Markov Decision Process" (POMDP). The tasks which have to be performed by these methods are applied for collective localization, mapping, and navigation, for collective perception, for collective learning, and for collective production planning and exception handling.

Biographical Information

1972	Dipl.-Phys. (M. S.): Institute for Theoretical Physics, University of Mainz, Germany
1972-77	Research Scientist in the Kernforschungszentrum, Karlsruhe (center for nuclear research)
1976	Dr. rer. nat. (Ph. D. ): Institute for Nuclear Physics, University of Karlsruhe, Germany
1977-83	Senior Research Scientist in the Institute for Informatics and Robotics of University of Karlsruhe, Germany
1984-88	Head of the department of Technical Expert Systems and Robotics in the Forschungszentrum für Informatik (Center for Informatics Research), Karlsruhe, Germany
1988	Dr. rer. nat. habil (D. sc.) in Computer Science: University of Karlsruhe, Germany
1988-92	Professor for Informatics (AI and Computer Vision) in the Institute for Informatics of the Technical University of Munich
1992-now	Professor for Informatics (Distributed AI, Computer Vision and massive parallel Algorithms) in the Institute for Parallel and Distributed Systems of the University of Stuttgart, Germany
1996-now	Member of the Management Board of the Centre for Computer Science (FZI), Director of the Division: Mobility Management and Robotics, Karlsruhe, Germany