February 25, 2005
4-5:00, LH 102

Combining Agent Architectures and Multi-Agent Systems: Steps towards Complex Cognitive Robots for Human-Robot Interaction

Matthias Scheutz

Abstract:
Complex robots that have to interact with people in real-time using natural language pose several design challenges, from the functional definition of the overall robotic architecture, to the integration of different modules that operate at different temporal scales using different representations, to the implementation in parallel hardware that has to be fault-tolerant and allow for high reactivity. Work on high-level cognitive agent architectures has traditionally focused on reasoning and planning in simulated agents, thus largely ignoring real-world, real-time constraints. Work in AI robotics has mainly been centered around lower level perceptual-motor couplings and behaviors, even though some attempts have been made at briding the "gap" between reactive and deliberative components using specific "hybrid systems".

In this talk, we argue that a different approach towards defining and implementing architectures for autonomous complex robots is needed for human-robot interaction and present our proposal for an architecture framework that allows for the full distribution of architectural components over multiple computers by enlisting the services of an underlying multi-agent system. We will demonstrate our approach with various examples from different subtasks of our implemented robotic system (such as face detection and people tracking, natural language parsing and understanding, affective speech production, and complex action sequencing).

Biography:
Matthias Scheutz received the M.Sc.E. degrees in formal logic and computer engineering from the University of Vienna and the Vienna University of Technology, respectively, in 1993, and the M.A. and Ph.D. of philosophy in philosophy at the University of Vienna, Austria, in 1989 and 1995 respectively. He also received the joint Ph.D. in cognitive science and computer science from Indiana University Bloomington in 1999. He is an Assistant Professor in computer science and engineering at the University of Notre Dame, where he heads the Artificial Intelligence and Robotics Laboratory. He has over 60 publications in artificial intelligence, robotics, artificial life, cognitive modeling and foundations of cognitive science. His current research interests include distributed agent architectures, cognitive robotics, affective agent control, agent-based modeling, and models of bilingual language processing. Dr. Scheutz is member of the AAAI, IEEE, and the cognitive science society.