Enhancing Robustness of Information Systems through Distributed Adaptive
Coordination
NSF IRI-9812764
The goal of this research project is to create and disseminate a software
infrastructure for building complex and large-scale information systems
that are both distributed and robust. The path to such systems lies in
viewing distributed information systems as multi-agent systems:
computational organizations consisting of potentially hundreds of
intelligent real-time agents. Robust information systems will adapt both
their functions and performance in response to changes in system goals,
changes in performance requirements, and changes in the available software
and hardware resources. A multi-agent architecture appropriate for such
systems is being developed that contains component technologies for local
agent scheduling, multi-agent coordination, organizational design,
detection and diagnosis of inappropriate behavior in face of environmental
change, and on-line learning to improve the performance of the other
components. To evaluate this research, a multi-agent simulation testbed is
being utilized to conduct extensive evaluation of the individual
technologies, an integrated system prototype, and the issues of scalability
to large agent societies. The impact of this work is to provide software,
techniques and empirical studies that will facilitate the construction of
the next generation of robust information systems.
Work on this project has just started. The first output has been an
agent toolkit called DECAF. You can read more about this tookit
on the DECAF web page.
We have a few initial workshop papers describing some of our approach and where we are going.
Here's some pointers to some
previous papers in this general area.
-
Keith Decker and Katia Sycara. Intelligent
Adaptive Information Agents. Journal of Intelligent Information
Systems, 9 : 239--260, 1997.
-
Keith Decker, Katia Sycara, and Mike Williamson. Middle-Agents
for the Internet. In Proceedings of the 15th International Joint
Conference on Artificial Intelligence, Nagoya, Japan, August 1997.
-
Keith Decker and Jinjiang Li. Coordinated Hospital
Patient Scheduling. In Proceedings of the Third International Conference
on Multi-Agent Systems, Paris, France, July 1998.
-
Keith S. Decker and Victor R. Lesser. Designing a
Family of Coordination Algorithms. In Proceedings of the First
International Conference on Multi-Agent Systems, San Francisco,
July 1995. Selected for reprinting in Readings in Agents,
M. Huhns and M. Singh, eds. Morgan Kaufmann, 1997.
-
Keith S. Decker and Victor R. Lesser. An approach
to analyzing the need for meta-level communication. In
Proceedings of the Thirteenth International Joint Conference on
Artificial Intelligence, Chambéry, August 1993.
Also see my collegues' and my previous work at
And also check out, for example,