Home // ICAS 2011, The Seventh International Conference on Autonomic and Autonomous Systems // View article

Self-adaptive tuning of dynamic changing problem solving: a first step to endogenous control in Multi-Agents Based Problem Solvers

Authors:
Gael Clair
Frédéric Armetta
Salima Hassas

Keywords: multi-agent system; self-organization; endogenous control

Abstract:
In this paper, we propose a new model, EC4MAS, for endogenous control in self-organizing system for problem solving. We first present the characteristics of exogenous and endogenous control and draw the differences between them. Problem solving methods are often faced with the exploitation/exploration dilemma. We propose in this work an approach that tries to find a good coupling between structural/topological characteristics of a problem and the local associated behaviors that compose the solving method. More precisely we organize the multi-agents based solver into a social organization that represents the organization of the different local behaviors and a spatial organization that represents the different characteristics of the problem. The objective of the system is thus to find a good coupling between these 2 organizations. We illustrate our approach on a graph coloring problem, show that our model can solve the exploitation/exploration dilemma in this case and how it can provide a mean to build a more robust and efficient tuning of the solving process on dynamic changing problems.

Pages: 88 to 93

Copyright: Copyright (c) IARIA, 2011

Publication date: May 22, 2011

Published in: conference

ISSN: 2308-3913

ISBN: 978-1-61208-134-2

Location: Venice/Mestre, Italy

Dates: from May 22, 2011 to May 27, 2011