version française english version
logo Lirmm


logo um


logo cnrs


logo iut

Formalism, tools and methodological elements
for the modeling and simulation of
multi-agents systems

french document !

pdf file
for printer
(without colored hyperlinks)

pdf file
for screen



Jury
A. DROGOUL

Professeur, Université Paris VI

Reviewer
D.R.C. HILL

Professeur, Université Blaise Pascal, Clermont-Ferrand

Reviewer
H.V.D. PARUNAK

Chief Scientist, Altarum Institute, Ann Arbor, MI, USA

Reviewer
Y. DEMAZEAU

Chargé de recherche HDR, Institut IMAG, Grenoble

Chairman
J-P. MULLER

Directeur de recherche HDR, CIRAD, Montpellier

Examinateur
J. FERBER

Professeur, Université Montpellier II

PhD advisor


Abstract: This thesis addresses questions related to the modeling and simulation of complex systems which are based on the multi-agents system (MAS) paradigm. Multi-Agent Based Simulations (MABS) rely on the idea that it is possible to directly represent the behavior and the interactions of a set of autonomous entities situated in a common environment. One critical issue is the difficulty to accurately achieve replication of published models. This raises problems for verification and validation of MABS. In this thesis, we advocate that MABS verification and validation problems are mainly related to the gap that exists between models specifications and the computational structures which are used to execute them. The motivation of our work is to take steps toward an effective mapping between models specifications and concrete software structures considering two complementary approaches. The first approach proposes generic simulator engineering tools. The idea is to make explicit the computational structures which are used for the implementation. The second approach relies on the study of today MAS modeling means: we propose a reflexion about the MAS paradigm and we identify a set of modeling constraints which are related to it. We then propose modeling principles based on (1) the Influence/Reaction approach and (2) a formal model of the environment for MASs, namely MIC*. We show how our approach handles the various identified modeling constraints and we illustrate its feasibility through a concrete example.

Keywords: modeling and simulation, multi-agents systems, Influence/Reaction model, interaction