Coordinating Autonomous Agents Situated in Time-space
Mikhail Prokopenko
CSIRO Mathematical and Information Sciences
Tuesday 14 May at 11am
Abstract
In this talk we attempt to formally define various types of behaviour-based
agent architectures and encapsulate them in a hierarchical setting.
The developed framework (Deep Behaviour Projection) is capable
of capturing some desirable properties of situated and tactical
agents, such as robustness, real-time response and adaptive coordination.
A new work on multi-agent coordination is also presented, including
characterisations of some communication policies in terms of generic
information-theoretic properties. In particular, the relative
entropy of joint beliefs is suggested as an indicator of coordination
potential of multi-agent systems. The Deep Behaviour Projection
framework enables consistent reasoning about belief change, including
beliefs about other agents - this allowed us to examine some of
the identified communication policies empirically.
Most recently, the DBP framework has been used for design and
development of FLOCK - FLexible Organiser of Content and Knowledge
- a multi-agent system for adaptive e-mail categorisation.
Short resume
Mikhail Prokopenko is leading the multi-agent technology
efforts in the Intelligent Interactive Technology group at CMIS
Sydney. His long-term research goal is an effective integration
of real-time perception, reasoning and action in autonomous agents,
and better understanding of emergent behaviour in complex multi-agent
systems. With his co-workers, Mikhail has developed a flexible multi-agent
framework applicable to a number of application areas, as diverse
as adaptive digital appliances and robotic soccer. Mikhail Prokopenko
received his M.Sc.(Hon.) (1988) in Applied Mathematics from the
Azerbaijan Institute of Petroleum and Chemistry (USSR), and M.A.
(1994) in Economics from the University of Missouri - Columbia (USA).
He recently completed his PhD thesis at the Computing Department
of Macquarie University, Sydney.
Back to HAIL Home Page
|