Modeling and Simulation of Agent-Based Complex Systems and Application to Natural Disasters

Karam Mustapha, Hamid Mcheick Université du Québec à Chicoutimi

555 Boulevard Université, G7H2B1, Québec, Canada

Telephone 001 418-545-5011 karam.mustapha1@uqac.ca, hamid.Mcheick@uqac.ca

Sehl Mellouli Université de Laval

2325 Rue de l'Université G1k 0A6, Quebec, PQ, Canada Telephone 001 418-656-2131

sehl.mellouli@fsa.ulaval.ca

ABSTRACT A number of modeling and simulation tools have been developed in the domain of Natural Disasters. In these situations, several research teams may make an intervention and that have to coordinate their activities in order to save the maximum number of lives. To this end, they have to define an organizational structure and adopt management policies to improve their performance. The organizational structure and the policies are important elements that have to be taken into account to simulate a real emergency activity. To facilitate the design of these simulations, an agent-based methodological framework for complex system (Supply Chain, Disaster Natural) is proposed. The main contribution of the framework is that it will reflect the organizational structure and policies within the simulation, and which involves the integration truly dynamic dimension of this organization. Also, we validate the proposed work on a case study more precisely on the fire building.

Categories and Subject Descriptors Simulation and Modeling [Computing Methodologies], Distributed Artificial Intelligence: Multi-Agent Systems.

Keywords Agent Based Simulation, Multi-Agent Systems, Natural Disaster, Organization, Modeling, Simulation.

1. INTRODUCTION The management of Natural Disasters (NDs) is becoming more complex since NDs are becoming more frequent and more powerful. Hence, emergency teams may find difficulties to rescue people. To this end, these teams will need tools to help them making efficient interventions. Simulations are one of the means that can be used by these teams to predict their behavior during a ND. Over the past few decades, ND such as droughts, floods, cyclones, earthquakes and volcanic eruptions have resulted in the mortality of approximately three million people and affected the lives of 800 million people worldwide1. These have caused diseases as well as serious economic losses and homelessness. Therefore, modeling and simulating the rescue procedure may

1 http://epress.anu.edu.au/narayanan/mobile_devices/

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from permissions@acm.org. PM2HW2N'13, November 3–8, 2013, Barcelona, Spain. Copyright © 2013 ACM 978-1-4503-2371-0/13/11…$15.00.

help to facilitate their management and limit their impact on the society. These simulations may improve the efficiency of the teams on the field which may lead to reducing losses and damage of goods, and saving lives. Multi-agent systems (MAS) are among the methods used for modeling and simulating ND emergencies. A MAS can model the behavior of a set of entities expert, more or less organized according to the laws of social type [5]. Agents have a degree of autonomy and are immersed in an environment in which and with which they interact [4]. There are several areas where MAS can be applied; they can act as a modeling paradigm or as a solution for software implementation. Modelers can also use MAS to create computer representations of dynamic events such as ND emergency. Therefore, the application of MAS in this area could help managers to experiment all possible scenarios of a disaster and assist them in making decisions. This approach involves the simulation of systems in terms of models and their use. Agent-Based Simulation (ABS) [24] has spread out into many areas, including sociology, biology, economics, physics, chemistry, ecology, industrial applications and ND. For most of these areas, the ABS is gradually replacing the micro-simulation techniques and object-oriented simulation. This is due to the ability of the ABS to capture different dynamic models which usually consist of simple entities (reactive agent: if the simple behavior is required, a type of stimulus-response behavior is sufficient) or more complex entities (deliberative agent: if decision-making and negotiation is needed then capacities will require a deliberative agent to perceive its environment and other agent behavior).

A multi-agent simulator used in the management of ND is a powerful tool that can: give an indication of the sequence of events that may occur; provide a clearer understanding of the wrong actions that have to be avoided; detail the precautions that have to be prepared effectively if future disasters occur; establish procedures for actions, and communicate and prepare actors to conduct these procedures. These actors will be ready to coordinate their activities to rescue people and adapt their strategy depending on the reactivity of the phenomena caused by the disaster.

The global objective of the present research is to give a focus on the impact of a ND organizational structure performance by providing a methodological framework which ranges from domain model analysis to running the simulation. In line with our previous works in the simulation, and in order to consider these organizational aspects of the ND, we have proposed a specific agent-based methodological framework [20] allowing, from modeling to simulation, the production of observables at different levels of details related to a ND organization. This framework aims to facilitate the realization of the simulation with gradual processes. It begins by defining the needs of the user prior to arriving to the implementation of the system while satisfying the initial requirements. The challenge of this research, in an agent-

75

oriented approach, is the ability to integrate dynamic organizational characteristics of ND in the modeling and simulation procedure. It will also include the specification of the translation process from generic model to specific model, to ensure the transition between the proposed models (presented in the methodological framework), as well as to facilitate the communication between the modeler and the developer (several solutions proposed to solve these problems in the paper [29]).

This paper is structured as follows: Section 2 defines the objectives of the research and related concerns modeling and simulation of ND. Section 3, introduces an organizational-oriented methodological framework, which is capable of taking into account the organizational aspects at both the conceptual and the operational abstraction levels. As a proof of concept, section 4 presents an illustrative example of the proposed methodological framework to the modeling of a building fire. Finally, section 7 is a conclusion and recommendations for future work.

2. AGENTS AND ORGANIZATION ORIENTED ND MODELING AND SIMULATION A multi-agent based system is a powerful modeling technique for simulating individual interactions in a dynamic system and is distinctive in its ability to simulate situations with unpredictable behavior [16].

Previous researches have focused on modeling of the rescue during NDs. However the current technological developments allow us to envision a systems approach that includes modeling of all aspects of the disaster event, its impact on the resources, population and the required response by the involved agencies. This is why we choose to use the ABS approach to model and simulate these aspects.

ABS allows a modeler to manipulate different levels of representations, such as individuals and groups of individuals. Agent-based modeling allows capturing the dynamic nature of the ND and facilitates the study of numerous resource coordination associated with the interaction of multiple teams [18]. Hence, multi-agent technology is widely used in various disaster management and community awareness issues. In the following section, a study on the proposed models in the field of ND is made.

2.1 Agent-Based ND Frameworks Even if, agents are used in the simulation of NDs, few researchers have proposed a framework to support both the design and the implementation of the ND simulation. Two studies are presented hereafter:

- ABDiSE (Agent-Based Disaster Simulation Environment) is a framework that provides model elements and tools to support the modeling and the simulation of different types of ND such as fires, floods and debris flows. This tool describes how agents move, attach, and interact with each other and with their environment [9].

- D4S2 (Dynamic Discrete Disaster Decision Simulation System) is a comprehensive decision support system to simulate the large-scale disaster responses. This model has a specialized architecture designed for decision makers who can be public safety service officials such as fire and police [25]. More precisely, the architecture integrates several models such as: an agent-based simulation model, a geographical information system (GIS) data bases, a rule based system for

responders and optimization modules to create a hybrid system of agent-based and discrete simulation components.

2.2 Agent Oriented Frameworks The organizational modeling in multi-agent systems is based on the management of a process metaphor that underrates the organizational structure [20]. A more general study of agent oriented software engineering methodologies, undertaken in order to find conceptual and operational solutions, has confirmed that organizational issues were to be added to the actor approach. This approach is the basis of a methodological framework for helping the domain experts to design their models in their own language, as well as transitional agent-based models which are used to produce the distributed simulation model on which experiments are conducted [15]. Methods like GAIA [26], CRIO [7], MOISE+ [10] or the Luis Antonio’s work [1], provide only a part of the solution for the required objectives. Most of these approaches use the notion of roles in order to promote the flexibility in the design process, even with different abstraction or hierarchical levels. As an abstract view of the distributed organization, roles can be combined and associated to the agents' specific architecture, from complex information processing units (i.e. with deliberating capacities) to a simple programmable units (reactive agents or state-machine like automata).

2.3 ND Specific Models There exist other ND models that do not use the agent approach as a method for modeling and simulating complex problems, such as emergency responses, evacuations, fires, traffic events, earthquakes and flooding, among those approaches:

- Emergency Response Framework [11]: this framework allows for integration of modeling, simulation, and visualization tools for emergency response. The development and implementation of this framework will significantly improve the nation’s capability in the emergency response area.

- For evacuation model: there are more than 26 models that have focused on simulating building evacuations. Many of these models used to simulate evacuation procedure from different types of structures. Featured models include: EVACNET4, WAYOUT, STEPS, PedGo, PEDROUTE, Simulex, GridFlow, ASERI, FDS+Evac, Pathfinder, SimWalk, PEDFLOW, buildingEXODUS, Legion, SpaceSensor, Evacuation Planning Tool (EPT), MassMotion, PathFinder, Myriad II, ALLSAFE, CRISP, EGRESS, SGEM, Evac/FDS, Massegress, Hidac [14]. In addition, VSM model (Vitae System model) is known to provide a systematic framework for the adaptive evacuation strategy in a life-critical situation by integrating its three fundamental elements: survivability, vitality and communication [2].

- For traffic model : there are three main approaches for the modeling and simulation of traffic:

o The macro simulation approach also referred to as macroscopic is one of the primary approaches for modeling complex systems. Macros models follow the top-down strategy, focusing on observable behavior of a system. They define and regenerate observable behaviors in terms of aggregate, abstract parameters and their probability distributions [8].

o The micro simulation approach also referred to as microscopy follows the bottom-up strategy where a complex system is regarded as a great set of interactive components. This approach focuses on the identification of components in a system and their

76

behaviors and interactions among them. The overall system is observed by local behavior of the various components and their interactions [2, 13].

o A less common approach is the mesoscopic approach [3]. This approach is widely used in the economic research and studying patterns of movement.

2.4 Limitations of Methods In the previous section (Section 2), research devoted to developing framework to guide designers in the development of multi-agent models was presented. It was noticed that the previous research work is suffering from a number of weaknesses, in terms of modeling and simulation. These weaknesses are briefly discussed in the following section:

- The first weakness in the modeling of ND is related to the fact that these methodological works do not take sufficient account of the purely organizational aspects of the ND, i.e. explicitly including the structure and organizational dynamics, particularly which related to the behavior of actors or agents, behaviors generally associated with multiple roles.

- The second weakness concerns not taking into account observable and indicators specific to the organization of the natural disaster. These observable and indicators are usually defined by the domain experts in a first modeling of the natural disaster. Observables and indicators are data and information ongoing decision processes, which need to be highlighted in the simulation results for particular study.

- The third weakness is related to the validation, it is one of the most important limitations of evacuation models. In evacuation modeling, validation refers to a systematic comparison of model predictions with reliable information [28]. Model predictions are dependent upon the data and codes of the evacuation model and the user of the evacuation model. The lack of suitable experimental data to feed the evacuation modeling causes a challenge to modelers.

- The fourth weakness of evacuation models related in the presentation of occupants in the models. Accurate occupants’ representation based on comprehensive anthropometric data and human performance and behavior should be used in evacuation modeling to provide additional level of validity to the models. Otherwise, building codes and standards should be reformed according to the dynamic changes of individuals’ ages and sizes.

- The fifth weakness is the interoperability between emergency response modeling and simulation applications is currently extremely limited (for example: the interoperability between different model such as fire model, evacuation model).

- The sixth weakness: the cost of transferring data between emergency response simulation software applications is often very high.

- The seventh weakness: the emergency response organizations usually do not have the technical expertise or the time for building simulation models [11].

Therefore, this study proposes a solution to overcome some limitations (presented above: organizational dynamics, interoperability, and transferring data) in order to allow modeling / simulation of more "corporate" management after a ND has occurred. Modeling and simulation of "dynamic organizational" means that organizational structures are explicitly incorporated in the management after ND (either temporary or permanent). This will take into account the associated indicators defined at different levels of detail of the organization.

3. A METHODOLOGICAL FRAMEWORK FOR ND ORGANIZATIONAL ASPECTS MODELING AND SIMULATION The complexity in the modeling and simulation process of ND as well as the implementation support led us to propose a modeling approach based on an incremental process. This modeling technique relies back on the gradual increasing of the level of detail in the model. The real system is represented by a domain model of ND (e.g. an UEML model -Unified Enterprise Modeling Language -www.ueml.org) to represent the organizational aspects.

In our previous work, an organizational methodological framework for modeling a complex system was proposed, which was according to two main abstraction levels: a conceptual and an operational level [20]. Using the domain model provided by the domain expert, a simulation model is built step by step. The conceptual level proposes concepts and models helping to grasp the complexity of the problem and its simulation objectives. Whereas the operational level involves the implementation of the simulation model which includes the software integration issues. The different models and the transition to agent-oriented modeling and simulation in our methodological framework are presented in Figure 1 (refer to [20] for further detail).

The Conceptual Organizational Modeling engages through a dialogue between the domain expert and an agent-knowledgeable modeler. An actor model is produced by identifying the active entities and their organization from the domain model according to the role concept. The modeler has to translate/abstract the domain model into a Conceptual Organizational Modeling based on (hierarchical) levels, actors, roles and groups named Conceptual Role Organizational Model (CROM). This stage highlights the organizational structure of the ND as wells as the structural and dynamic relations between the entities composing the ND. Following this conceptual, agent-based model is produced on the basis of observables that the user needs to obtain from the simulation, building up the route toward the implementation of the simulation.

This model is transposed into the agent world (at a conceptual level) concluding the phase of "specification" with a multi-agent and organization model named Conceptual Agent Organizational Model (CAOM) ready to be described at an architectural and software design level ([29] presents the transformation between the CROM model to CAOM Model using the MDA).

The important key of this step is to identify precisely the agents defined at the conceptual level in order to develop them adequately at the operational level. The Operational and Organizational Modeling provides a solution to implement an executable system to perform simulations based on the previous conceptual models. This step involves the choice of agent architectures, depending on the complexity of the behaviors needed to be simulated. This process is guided by the observables selected earlier by the domain expert.

The software designer detail the CAOM by associating a conceptual agent with a software agent architecture (e.g. BDI - Believe, Desire, Intention -[23]) and specifying their behaviors (e.g. an UML - “Unified Modeling Language” - activity diagram for a reactive agent) and interactions (e.g. AUML - “Agent Unified Modeling Language” - sequence diagram [21]), resulting in an Operational Agent Model (OPAM). The implementation of these models in a simulation(s) environment results in an ABS system which can be executed.

77

Tebtr

Taobeleovcsc

OmoCmD[wm

4FTbv

The refining proexperience with be defined anransformation en

The observables a real system areonly mentioned before the impleearlier in the moevel) as they m

organization moview and modelecan describe the state or activitycompany, a coop

Our work can bemethodological fof the translatioCROM modelsmethodological Driven Developm17], these probl

we validate the pmore precisely on

4. ILLUSTFIRE The example prbuilding fire (Fivalidates the fram

ocess currently fthe models incr

nd then can ngines.

that are related e not described in

in the multi-agmentation stageodeling process

may induce diffeodeling can be sed according to phenomena at d

y, a group of aperation process…

e summarized aframework, whicon process to ens, CAOM anframework moment [6] or MDems solved in thproposed work pn the fire buildin

RATIVE E

resented is concigure 2). This

mework describe

RCSta

UMmeAU

PrObReorof

Obse

Fig

follows the ad reases, the rule gbe automated

to the organizan the design mogent system moe. It is necessary

(at a conceptuaerent modeling rstudied along dthese observabl

different level ofactor i.e. a prod…).

s the enrichmench would includensure the transi

nd OPAM (podels) [20] usinDA (Model Drihe paper [29]. Inpresented in [29ng.

EXAMPLE:

cerned with theproof of conce

ed above.

CA/ atechart

ML etamodel/ UML

roblems bservable elated potentially to thrganizational structure f the real system

ervable

Exp

lora

tion

sC

once

ptu

al M

odel

ing

Op

erat

ion

al M

odel

ing

ure 1.A Method

hoc rules. As thgeneralization ca

through mod

tional structure del [15]. They a

odel, i.e. one stey to describe theal and operationrequirements. Thdifferent points es, as observablf the ND (an actduction cell or

nt of the proposee the specificatioition between throposed in th

ng MDD (Modven Architecturn the next sectio9] on a case stud

BUILDING

e description of ept illustrates an

Exp

MultInfrastructure

e

R

ObserInput

D

CAOM-Conceptu

D

Transformat

CROM-Conceptu

Transformat

OPAM – Op

dological Frame

he an

del

of are ep

em nal he of

les tor

a

ed on he he

del re) on, dy

G

f a nd

4.1 IframThis seMethodobtainephases Orientethe Contransfor(OPAMmethod

In the fmodelinbuildingeneratThere relationcommupresent

perimentations

ti-agent simulationMulti-Agen

Real system

Outprvable

Development

ual Agent Organizati

Domain Model

tion model « Using

ual Role Organizati

tion model « Using

perational Agent

ework for ND

F

Implementaework ection illustratedological frameed by presentin

of design. Wed Model (CROnceptual Agent rm CAOM M

M). The results odological framew

figure below (Fing and simulate

ng fire. During te the traffic, this a relationsh

nship), also we uunication betweting simple and c

t System

Performanc

Complex Organiz

Tests

put

ional Model

Simulation

MDD»

ional Model

MDD»

Ex

pert o

f do

ma

inM

od

elerP

rog

ram

mer

Model

Figure 2: Buildin

ation of t

es the implemeework on ND ng models that

We describe firOM), secondly w

Oriented ModeModel to the O

of various modework leading to t

igure 3) we illuse the complex sy

the fire, we he emergency rehip between thused the multi-aeen these modecomplex entities

ce Indicator

zational Structure

Organizational

Specification

Ob

servable

Consid

ering the organization

ng fire

the method

entation of Agecase. This illare derived fr

rstly the Concewe transform thel (CAOM), andOperational Ageling phases assothe conduct of si

trate the differenystem, more speneed different

esponse and the ese models, (c

agent system to fels, and then ts.

dological

ent-Oriented lustration is rom various eptual Role

his model to d finally we gent Model ociated with imulation.

nt models to ecifically the

services to evacuation.

collaborative facilitate the to facilitate

78

ASpe

-

- - -

Figu

A building fire cSix main roles proceeding furtheach role:

Detailed strresponse). Detailed builEstimate the Traffic in the

ure 3: Illustrate

consists of sevewere identified

her into the mod

reet map (inc

lding fire and firnumber of peop

e local area.

the different m

eral roles withinand need to b

del ND. Below i

luding location

resides. ple in the buildin

models

n the organizatiobe defined befois a description

n of emergenc

ng.

Figure 4. A

on. ore of

cy

- Hofor

This romodel buildinevacuat

4.2 AThe foldescribthis exconsistthe firsare fivethese adifferen

For exathe N1securitytake shenact thon theManageFiremaActor 3same ro

CROM model o

spital location the emergency r

ole can give us and simulate a

ng fire, Inclution and traffic).

An Illustratllowing example

be the dynamic xample, the strus of one or morest level, Police, e actors connectactors is represent hierarchal lev

ample, Actor Po1 level, whiley/fire fighting ca

hort term decisiohese decisions ae third level. Iement Inventory

an as a Police A3 belongs to Poole in the dynam

of a ND

(availabilities aresponse (police

the ability andproblem such a

uding emergen.

tive CROM e illustrates howorganizational s

ucture is dividede group (structurFireman, First ated by a collabo

ented by a groupvels.

olice can negotat level N2

apacities with thons on this leveland thus control In the given ey) process is desActor 3 when nolice (structural mic group.

and capacities).e, fireman...).

d also the help as a building firency response,

model of a w a CROM modestructure of a gid into 4 levels. ral or dynamic) aid, Evacuationoration relationshp of actors play

tiate with Actor it checks their

heir manager. Thl but they are rein real time the

example, a VMscribed where P

needed. Whereasrelationship) th

.Availability

we need to e (modeling

firefighter,

ND el is used to iven ND. In Each level of actors. In and Traffic

hip; each of ying roles at

Fireman at r respective he last actors sponsible to

eir execution MI (Vendor Police uses a s the Police

hat plays the

79

4Tmr

Tm

InEagVth

4Ine

Aaecime

WeaTW

4.3 An IllusThe following emodel describingeactive and delib

The structure is dmore groups (stru

n the first level, Evacuation Agenan interaction (cogroup of agents pVMI (Vendor Mhe agents depen

4.4 An Illusn this model, th

environment for

Agents present inachieve their goenvironment andcan play severmplementation

environment act

We can separate environment. Foragents and PoliceThese agents areWe can find in

strative CAexample, given g the organizatioberative agents.

divided into 4 levuctural or dynam

Police Agent, Fnt and Traffic Aomplex). Each oplaying roles at danagement Invend on each other.

strative OPAhe structure is divcognitive agents

n the cognitive als. They have

d have reasoninral roles in th

of multiple pin response to en

our agent preser example: Police Agent 1, Police connected by

n our model a

OM model in Figure 5, illuonal structure of

vels. Each level mic) of agents.

Fireman 2 Agentgent are five ag

of these agents isdifferent hierarcntory) process is

AM model ovided into two es and the other fo

environment actan explicit repr

ng abilities. Thehe multi-agent plans. Agents nvironmental sti

ented in the CAOce, Police chief ice Agent 3 in than interaction thybrid agent si

Figure 5. A

of a ND ustrates a CAOf a given ND wi

consists of one

t, First Aid Agenents connected bs represented by

chal levels. Also,s described, whe

of a ND environments, onor reactive agent

t independently resentation of th

e cognitive agensystem by thin the reactiv

imuli.

OM model into in the deliberative reactive Agenthrough messageince it plays tw

CAOM model o

M ith

or

nt, by y a , a

ere

ne ts.

to he nts he ve

to ve ts. es. wo

differenother re

5. CONaturalpeople These and hoare mos

In thismethodsimulatdifferenaccordiis struabstracon a Cthen re(CAOMthe Ope

This frorganizND perND, an(CAOMwhich concep

of a ND

nt roles, one coeactive in the rea

ONCLUSIOl Disasters haveand affected th

have caused disomelessness. Thestly neglected w

s paper, we hadological framtion of ND organt level of deting to desired ob

uctured accordinction levels. At tConceptual Role efined into a

M). At the operaerational Agent

ramework allowzational structurrformance. Base

n organization/roM) conceptual mwill reproduce

ptual models are

ognitive in the active environm

ON AND FUe resulted in thehe lives of 800 seases as well e organizational

while simulating

ave presented mework, which anizational aspectail while reprobservables. Thisng to a concethe conceptual l

Organizational Conceptual Agational level, mModel (OPAM)

s the study of thre and its relateed on a ND expole oriented (CR

model help in dethe ND global

e defined indepe

cognitive envirent.

UTURE WOe mortality of thmillion people

as serious econl structure and a real emergency

an organizationpermits mod

cts. It allows oboducing the N

s methodologicaleptual and an level, the model Model (CROM

gent Organizatiomodeling is main

).

he impact of a ed management ert modeling of

ROM) and an agsigning a simulaand local behav

endently of part

onment and

ORK hree million worldwide.

nomic losses the policies y activity.

nal oriented deling and servables of

ND behavior l framework operational

ing is based M), which is onal Model

nly based on

specific ND policies on

f a particular gent-oriented ation model, viors. These ticular agent

80

architecture or even on specific software architecture but propose transitional steps to guide their development.

Future work will propose an open software architecture supporting the transformation of the conceptual model into an operational model by generalizing the previous “hard wired” architecture inspired by previous agent-based integration framework. This architecture can be seen as the interaction between different simulation platforms, an interaction that can be resolved in two ways: by means of signals and a mediator.

Different types of agents will be used: deliberative agents and reactive agents. Their development will be based on the interaction between the Anylogic or JADE platform (for the reactive agent) and the JASON platform (for the deliberative agent).

6. REFERENCES [1] Antonio, L., D”Amours, S., Frayret, J.M., “A methodological

framework for the analysis of agent-based supply chain planning simulations”, SpringSim '08: Proceedings of the 2008 spring simulation multiconference, Society for Computer Simulation International San Diego, CA, USA, (2008).

[2] Ben-Akiva, M.E, Koutsopoulos, H. and Mukundan, A. 1994 A dynamic traffic model system for ATMS/ATIS operation. IVHS Journal.. 2. 1-19. MITTNS.

[3] De Palma, A. and Marchal, F. and Nesterov, Y., 1996. METROPOLIS : a modular system for dynamic traffic simulations.

[4] Erceau, J. et Ferber, J. (1991) ‘L’Intelligence Artificielle Distribuée’, La recherche, vol. 22, pp. 750-758.

[5] FERBER, J. (1995) Les Systèmes Multi-Agents, vers une intelligence collective, InterEditions.

[6] France, R., Ghosh, S., Dinh-Trong, T. and Solberg., A., Model- Driven Development Using UML 2.0: Promises and Pitfalls. Computer, 39(2), Feb. 2006.

[7] Gaud, N., Galland, S., Koukam, A., Towards a Multilevel Simulation Approach based on Holonic Multi-agent. Published in the 10th International Conference on Computer Modeling and Simulation (EUROSIM/ UKSiM‟08), pp. 180–185, England. April 1–3, (2008).

[8] Helbing, D. and Treiber, M., Numerical simulation of macroscopic traffic equations, Computing in Science and Engineering 1, 89–99 (1999).

[9] Hsu T. L., Liu J. W. S. An Agent-Based Disaster Simulation Environment. RITMAN Workshop 2012, December 19, 2012, Taipei, Taiwan

[10] Hübner J. F, Sichman J.S., et Boissier O., (2007), Developing Organised Multi-Agent Systems Using the Moise+ Model: Programming Issues at the System and Agent Levels, Int. J. Accounting, Auditing and Performance Evaluation,1(3/4):370–395

[11] Jain, S. and C.R. McLean, 2003, “A Framework for Modeling and Simulation of Emergency Response,” Proceedings of the 2003 Winter Simulation Conference, Dec. 7-10, New Orleans, Louisiana, 1068-1076.

[12] Jouault, F., Kurtev, I., Transforming models with ATL, in: Proceedings of the Model Transformations in Practice Workshop at MoDELS 2005, MontegoBay, Jamaica, 2005.

[13] Kosonen,I. and Pursula, M. 1991 A simulation tool for traffic control planning. IEEE Conference Publication Number 320. Third international Conference on Road Traffic Control. vol 320 pp 72-76.

[14] Erica D. Kuligowski, Richard D. Peacock. A Review of Building Evacuation Models, Fire Research Division Building and Fire Research Laboratory, 2006.

[15] Labarthe, O., Espinasse, B. ,Ferrarini, A., Montreuil B., Toward a Methodological Framework for Agent-Based Modeling and Simulation of Supply Chains in a Mass Customization Context, in: Simulation Modeling Practice and Theory International Journal (SIMPAT), vol. 15, n° 2, pp. 113-136, February (2007).

[16] Lampert, R., 2002 Agent-based modeling as organizational and public policy simulators. Proceedings of the National Academy of Sciences of the United States of America, 99:7195-196.

[17] MDA: Model Driven Architecture Guide Version http://www.omg.org/cgibin/doc?omg/03-06-01, 12 juin 2003.

[18] Monteiro T., Anciaux D., Espinasse B., Ferrarini A., Labarthe O., Roy D. , “Chapter 6. The Interest of Agents for Supply Chain Simulation”, in: Wiley-ISTE (Ed.), ``Simulation for Supply Chain Management'', C. Thierry – A. Thomas – G. Bel, septembre (2008).

[19] Montagna, S., Ricci, A., et Omicini, A., (2008) A&A for modeling and engineering simulations in Systems Biology, International Journal of Agent-Oriented Software Engineering- Vol. 2, No.2 pp. 222 – 245.

[20] Mustapha, K., Tranvouez, E., Espinasse, B. et Ferrarini, A. (2010) An Organization-oriented Methodological Framework for Agent-Based Supply Chain Simulation. Fourth International Conference on Research Challenges in Information Science. May 19-21, 2010, Nice, France.

[21] Odell, J., Parunak, H.V.D., Bauer, B. „Representing agent interaction protocols in UML‟, Proceedings of the First International Workshop on Agent- Oriented Software Engineering, CIANCARINI, P. and WOOLDRIDGE, M. (Eds), (2001).

[22] Piunti, M., A. Ricci, Boissier, O. et Hübner, J.F. (2009) Accéder à une organisation multiagent par l’environnement des agents, JFSMA.

[23] Rao A. S., M. P. Gorgeff. Modeling rational agents within BDI-Architecture.in J. Allen & al Ed., Proceedings of the 2nd International Conference on Principles of Knowledge Representation and Reasoning. San Mateo, USA, Morgan Kaufmann, Pub, p. 473-484, (1991)

[24] Russel, S. et Norvig, P. (2003). Artificial Intelligence A Modern Approach. Pearson Education, Upper Saddle River, New Jersey, second edition

[25] Wu, S., Shuman, L.J., Bidanda, B., Kelley, M., Sochats, K., and Balaban, C. 2007c. System implementation issues of Dynamic Discrete Disaster Decision Simulation System

81

(D4S2) - Phase I. In the Proceedings of the 2007 Winter Simulation Conference, 1127-34.

[26] Zambonelli, F., Jennings N., Wooldridge, M.. Developing multi-agent systems: the GAIA methodology. ACM Trans. on Software Engineering and Methodology, 12(3), (2003).

[27] Okada N., City and Region Viewed as Vitae System for Integrated Disaster Risk Management, Annuals of Disas. Prev. Res. Inst., Kyoto Univ., No. 49 B, pp 131- 136, 2006.

[28] Galea, E. & Gwynne, S. (2005). Principles and Practices of Evacuation Modeling. London, UK: CMS Press.

[29] Mustapha, K., Mcheick, H., Mellouli, S., Modeling and Simulation Agent-Based of Natural Disaster Complex Systems. The 4th International Conference on Emerging Ubiquitous Systems and Pervasive Networks (EUSPN-2013)

82