Synchronized Petri Net: A Formal Specification
Model for Multi Agent Systems
Sofia Kouah
University of Oum ElBouaghi, Algeria
MISC Laboratory, University Mentouri Constantine, Algeria
Email: kouah@miscumc.org
Djamel Eddine Saïdouni
MISC Laboratory, University Mentouri Constantine, Algeria
Email: saidouni@miscumc.org
Jean Michel Ilié
Lip6 Laboratory, University of Paris 6, Paris, France
Email: jeanmichel.ilie@upmc.fr
—
! "
! !
# $
% !
$
! !
&
! ’ !
!
—(
!
I. INTRODUCTION
Multi Agent Systems (MAS for short) form a powerful
paradigm to design complex software [7][22][33].
Generally, it resolves complex problems where reactivity,
mobility [30], dynamicity and adaptation of the system to
uncertain or unpredictable factors should be considered.
MAS may be seen as societies made up of autonomous
and independent entities, called agents. These agents
interact together in order to solve a specific problem or to
achieve collectively a common task. Agent is viewed as a
computer system situated in some environment and that is
capable of executing flexible autonomous actions in this
environment in order to meet its design objectives [35]
[56].The large majority of applications basedagents are
designed by [48]:
Using methodologies based on results of Object
Oriented SoftwareEngineering [5][9][34];
Highlighting organizational aspects [17] or
relations between various aspects of MAS
[10][40][44];
Applying construction techniques of expert
systems [19];
Or using formal reasoning, based on Z language
[43] or on temporal logic [1].
Then, the domain for designing MAS is becoming very
attractive. However, several functionalities of MAS like
parallelism, dynamicity and communication may not be
easily specified using existing specification models. In
fact, these models are not defined specifically for MAS.
Therefore, new specification models, new paradigms and
new tools are required. Designing safe and sound MAS
whose behaviors could be checked before its
achievement, calls for a rigorous specification step,
assisted by a formal specification model. This model
should have a welldefined semantics and being able to
take into account all MAS functionalities. Previous
efforts toward modeling and verifying MAS can be found
in the work of [29] [54]. They are based on reachability
graph generation. In our work, this issue can be addressed
likewise, by generating reachability graph from SyPN
model. Such generation can be easily built by generating
all possible firing sequences as follow:
Step 1: start from initial marking.
Step 2: fire all enabled transitions from the current
marking (i. e. apply firing rules).
Step 3: reiterate the second step from the new
marking until no transition can be enabled.
In this context, we are mainly interested by formal
specification model of MAS. In this field some basic
questions arise:
How to preserve agent’s proprieties, such as
autonomy, sociability, awareness...etc.
How to model dynamically agents’ interactions.
Asynchronous is a ubiquitous aspect
characterizing concurrent interactions that must be
straightforwardly modeled.
JOURNAL OF SOFTWARE, VOL. 8, NO. 3, MARCH 2013 587
© 2013 ACADEMY PUBLISHER
doi:10.4304/jsw.8.3.587-602