SELMAS (SELf-Adaptive MAS)
Header : Zahia Guessoum
A multi-agent system is a set of autonomous and
interactive entities called agents. An important issue in building multi-agent
systems is how to observe, understand and control coordination in a dynamic
organization of agents. These systems have several characteristics which make
them more difficult to monitor and control than conventional single agent systems.
The first problem is the element of distribution: Each agent is an autonomous
and pro-active entity, and coordination is strictly limited to what can be done
through message passing.
This project deals with the problem of how
to allow such a multi-agent system to observe and monitor its own behavior to:
- build a model of its organization
so that it can foresee its future behavior and plan the solution of a problem
(e.g. failure of an agent t)while taking its agents' capabilities
into account,
- monitor and oversee the problem-solving
process,
- etc.
The aim of our proposal is to introduce a dynamic
and self-adaptive multi-agent model. We focus our work on the way that a multi-agent
system adapts its structures to changes of its environment and produces the emergence
of the organizational structures. These structures define the interdependence
among
a population of agents.