Multi-level agent-based modeling

A literature survey
arXiv:1205.0561v7 [cs.MA] 15 Nov 2013

Gildas Morvan

http://www.lgi2a.univ-artois.fr/~morvan/
[email protected]

Univ Lille Nord de France, F-59000 Lille, France

UArtois, LGI2A, F-62400, Béthune, France

November 18, 2013

1
Absract
During last decade, multi-level agent-based modeling has received significant and dra-
matically increasing interest. In this article we conduct a comprehensive and struc-
tured review of literature on this emerging research domain that aims at extending
the classical agent-based modeling paradigm to overcome some of its limitations. We
present the main theoretical contributions and applications with an emphasis on
social, flow, biological and biomedical models.

Main acronyms used in this paper

ABM According to the context, Agent-Based Model or Agent-Based Modeling

DEVS Discrete Event System Specification

EBM Equation-Based Model

ML-ABM According to the context, Multi-Level Agent-Based Model or Multi-

Level Agent-Based Modeling

Contents
1 Introduction 3
1.1 Agent-based modeling . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Multi-level agent-based modeling . . . . . . . . . . . . . . . . . . . . 3
1.3 Terminology issues . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Theoretical issues 6
2.1 Meta-models, simulation engines and platforms . . . . . . . . . . . . 6
2.2 Observation, detection and agentification of emergent phenomena . 8
2.3 Representation of aggregated entities . . . . . . . . . . . . . . . . . . 9

3 Application domains 10
3.1 Social simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Flow modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3 Biological and biomedical models . . . . . . . . . . . . . . . . . . . 14
3.4 Ecology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

4 Discussion 18
4.1 Level integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.2 Multi-level technical tools . . . . . . . . . . . . . . . . . . . . . . . 19

5 Conclusion 20

References 20

2
1 Introduction
1.1 Agent-based modeling
Agent-based modeling (ABM) is a computational modeling paradigm that allow to
simulate the interactions of autonomous agents in an environment. It has been
widely used to study complex systems in various domains (Epstein, 2006; Ferber,
1999; Gilbert, 2007; Railsback and Grimm, 2011; Resnick, 1994; Treuil et al., 2008).
However, it suffers from important known limitations that reduce its scope (Dro-
goul et al., 2003; Scerri et al., 2010). First, ABM is purely bottom-up: a microscopic
knowledge, i.e., related to system components, is used to construct models while a
macroscopic knowledge, i.e., related to global system properties, is used to validate
models.Therefore, it is not straightforward to explicitly introduce bidirectional rela-
tions between these two points of view in the general case. It becomes even harder
when different spatial or temporal scales or domains of interest are involved in a same
simulation. Moreover, agent-based models do not scale easily and generally require
large computational resources since many agents are simulated. Finally, most agent-
based simulation platforms lack tools to reify complex singular emergent properties:
human observation often remains the most efficient way to capture multi-level pat-
tern formation or crowd behavior.

1.2 Multi-level agent-based modeling

1.2.1 Definitions
The works surveyed in this paper aim at extending the classical ABM paradigm to
overcome these limitations. While they can be very different (in terms of goals,
technical approaches or application domains), these woks share a common idea: in-
troducing more levels of description in agent-based models. Therefore, we group
them under a common term: Multi-Level Agent-based Modeling (ML-ABM). In the
following we propose definitions of the concepts of level and multi-level agent-based
model that seem broad enough to encompass the different surveyed approaches (see
section 1.3).
Definition 1 A level is a point of view on a system, integrated in a model as a specific
abstraction.
The name refers to the familiar expressions levels of organization, observation,
analysis, etc.
Definition 2 A multi-level agent-based model integrates heterogenous (agent-based)
models, representing complementary points of view, so called levels, of the same sys-
tem.
In this definition, three concepts are highlighted and should themselves be de-
fined formally.
Definition 3 Integration means that the ABMs within a ML-ABM can interact and
share entities such as environments and agents.
Definition 4 Heterogeneity means that the ABMs integrated in a ML-ABM can be based
on different modeling paradigms (differential equations, cellular automata, etc.), use dif-
ferent time representation (discrete events, step-wise) and represent processes at different
spatio-temporal scales.

3
Definition 5 Points of view are complementary for a given problem since they can
not be taken in isolation to address it.

This idea is very important in the literature on complex systems (Morin, 1992).
Indeed, as Müller and Aubert (2011) note, "the global behavior of a complex system
cannot be understood without letting a set of points of view interact".

1.2.2 Types of problems solved using multi-level agent-based approaches

ML-ABM is mainly used to solve three types of modeling problems:
• the modeling of cross-level interactions, e.g., an explicit top-down feedback
control,
• the coupling of heterogeneous models,
• the (dynamic) adaptation of the level of detail of simulations, e.g., to save
computational resources or use the best available model in a given context.

In the first case, the different points of view always co-exist, as they integrate interde-
pendent models, while in the last ones, levels are (de)activated at run-time according
to the context, as they represent independent models designed for specific situations.
For instance, in flow hybrid models areas with simple topologies are handled
with an equation-based model (EBM) while others are handled with an ABM.

1.3 Terminology issues

Different terms have been used to describe what we call here a level: e.g., perspective,
interaction, layer or view (Desmeulles et al., 2009; Parunak et al., 2009; Torii et al.,
2005; White, 2007). Some are domain-specific; thus, in the flow modeling domain,
when two levels with static relations are considered, models are often described as
hybrid as shown in the section 3.2 (Burghout et al., 2005; El hmam et al., 2008,
2006a,b,c; Marino et al., 2011; Mathieu et al., 2007a; Nguyen et al., 2012b; Rejniak
and Anderson, 2011; Wakeland et al., 2007).
The term multi-scale — or multi-resolution (Jeschke and Uhrmacher, 2008; Zhang
et al., 2009a, 2011) — is often used but has a more restrictive meaning as it focuses
on the spatial and temporal extents of levels and not on their interactions and orga-
nization. Gil-Quijano et al. (2012, p. 622–623) pointed that the term multi-scale can
be misleading and advocated for using multi-level instead. Let take as an example the
Simpop3 model, described by their authors as multi-scale (Pumain and Louail, 2009).
Two levels are considered: the city level, representing the internal dynamics of a city,
and the system of cities level, representing the interactions between cities. However,
the idea of scale does not fit to describe the relation between them: one can easily
figure a city bigger (in terms of population, spatial extent, economic exchanges, etc.)
than a system of cities. In contrast, the idea of levels of organization in interaction
seems more appropriate. Furthermore, to extend the definitions 1 and 2 such mod-
els could be more precisely denoted as nested or hierarchical multi-level agent-based
models.

4
1.4 Bibliography
During last decade, ML-ABM has received significant and dramatically increasing
interest (fig. 1). In this article we present a comprehensive and structured review of
literature on the subject1 .
Another survey on the subject has been previously conducted by Gil-Quijano
et al. and published in different versions2 . While the present article aims at providing
an overview of the literature, Gil-Quijano et al. performed a comparative study of
three models (Gil-Quijano et al., 2008; Lepagnot and Hutzler, 2009) and (Pumain
and Louail, 2009). A similar survey, comparing four models, can be found in Vo
(2012, p. 28–34).

160 1200
140 per year per year
cumulated 1000 cumulated
120
publications

100
citations 800
80 600
60 400
40
20 200
0 0
1996 2000 2004 2008 2012 1996 2000 2004 2008 2012
year year
(a) publications (b) citations

10
citations/publications

1200 2012
1000
citations

8 800 2011

600
...

6 400
P

200
4 0
2006 2008 2010 2012 40 80 120 160
P
year publications
P P
(c) ratio of citations to publications (d) citations vs. publications

Figure 1: Bibliographical statistics on ML-ABM computed from author’s biblio-

graphic database and google scholar data on the Nov. 15, 2013

The paper is structured as follows. Section 2 introduces the main theoretical

contributions and section 3 presents the different application domains of ML-ABM,
with an emphasis on social, flow, biological and biomedical models. In section 4,
some issues about ML-ABM are discussed. Finally, the remaining sections conclude
1 The bibliographic database is available at the following URL: http://www.lgi2a.univ-artois.

fr/~morvan/Gildas_Morvan/ML-ABM_files/mlbib.bib.
2 This work was first published in french (Gil-Quijano et al., 2009), then extended (Gil-Quijano et al.,

2010),(Louail, 2010, p. 185–204) and finally translated into english (Gil-Quijano et al., 2012)

5
this paper by an analysis of the benefits, drawbacks and current limitations of the
existing approaches.

2 Theoretical issues
In the surveyed literature, three main theoretical issues have been addressed so far:
• the definition and implementation of meta-models and simulation engines,

• the detection and reification of emergent phenomena,

• and the definition of generic representations for aggregated entities.
They are described in the following sections.

2.1 Meta-models, simulation engines and platforms

Many meta-models and simulation engines dedicated to ML-ABM have been pro-
posed in the literature. They are are briefly presented in the following, in a chrono-
logical order.
Approaches based on DEVS have also been included (Zeigler et al., 2000). Indeed,
DEVS, as a generic event-based simulation framework, has been extended to support
ABM (Duboz, 2004; Duboz et al., 2003; Müller, 2009). A comprehensive survey of
the literature on multi-level DEVS extensions can be found in Duhail (2013).
GEAMAS (Marcenac et al., 1998a,b; Marcenac and Giroux, 1998) (GEneric Ar-
chitecture for MultiAgent Simulation) is a pioneering ML-ABM framework inte-
grating three levels of description (micro, meso, macro). Micro and macro levels
represent respectively agent and system points of view while the meso (or middle)
level represents an aggregation of agents in a specific context. Communication be-
tween levels is asynchronous. GEAMAS-NG is a newer version of the framework
providing tools to detect and reify emergent phenomena (David et al., 2011).
tMans3 (Scheutz et al., 2005) is a multi-scale agent-based meta-model and plat-
form. Unfortunately, the project seems to have died in the bud.
ML-DEVS (Steiniger et al., 2012; Uhrmacher et al., 2007) is an extension of
DEVS that allows the simulation of multi-scale models (and not only coupled models
in which the behavior of a model is determined by the behaviors of its sub-models).
Two types of relation between levels are defined: information propagation and event
activation. However, ML-DEVS focuses on multi-scale modeling and therefore, only
supports pure hierarchies of models: interaction graphs are viewed as trees (Maus et al.,
2008).
CRIO (Gaud, 2007; Gaud et al., 2008a,b) (Capacity Role Interaction Organiza-
tion) is an organizational meta-model dedicated to ML-ABM based on the concept
of holon (Koestler, 1967, 1978). It has been used to develop multi-scale simulations
of pedestrian flows (cf. section 3.2).
SPARK4 (Solovyev et al., 2010) (Simple Platform for Agent-based Representation
of Knowledge) is a framework for multi-scale ABM, dedicated to biomedical research.
Scerri et al. proposed a technical architecture to distribute the simulation of
complex agent-based models (Scerri et al., 2010). This approach aims at integrating
3 http://tmans.sourceforge.net/
4 http://www.pitt.edu/~cirm/spark/

6
multiple ABMs in a single simulation, each ABM representing a specific aspect of the
problem. In their article, authors focus on the management of time in such simula-
tions. Therefore, the proposed platform provides two main technical services that
ensure the consistency of simulations: (1) a time manager that ensures that integrated
ABMs advance in time in a consistent way and (2) a conflict resolver that manages the
problematic interactions between agents and shared data (such as the environment).
Authors evaluate their approach on a modified version of Repast5 and show it can
scale large-scale models easily.
IRM4MLS6 (Morvan and Jolly, 2012; Morvan et al., 2011) (Influence Reaction
Model for Multi-Level Simulation) is a multi-level extension of IRM4S (Influence
Reaction Model for Simulation) (Michel, 2007), an ABM meta-model based on the
Influence Reaction model which views action as a two step process: (1) agents pro-
duce "influences", i.e., individual decisions, according to their internal state and per-
ceptions, (2) the system "reacts", i.e., computes the consequences of influences, ac-
cording to the state of the world (Ferber and Müller, 1996). The relations of per-
ception and influence between levels are specified with digraphs. IRM4MLS relies
on a generic vision of multi-level modeling (see section 1). Therefore, interactions
between levels are not constrained. It has been applied to simulate and control intel-
ligent transportation systems composed of autonomous intelligent vehicles (Morvan
et al., 2012, 2009; Soyez et al., 2013, 2011) (see section 3.2).
ML-Rules (Maus et al., 2011) is a rule-based multi-scale modeling language ded-
icated to cell biological systems. Rules, describing system dynamics, are described
in a similar way as in chemical reaction equations. ML-Rules has been implemented
within the simulation framework JAMES II7 . This approach does not refer explicitly
to ABM; however, multi-level rule-based languages seem a promising way to engineer
complex individual-based models.
Müller et al. developed an approach that consists in decomposing a problem
according to the complementary points of view involved in the modeling (Müller
and Aubert, 2011; Müller and Diallo, 2012; Müller et al., 2011). For instance in
their case study the problem is the relation between residential and scholar segrega-
tion. Three points of view are considered: the geographer, the sociologist and the
economist. Then, independent conceptual agent-based models are defined for each
point of view. As models share agents and concepts, the conceptual models can-
not be merged without some processing. Indeed, a same concept can have different
meanings according to the point of view. To solve this issue, Müller et al. adapt
a technique described in the modular ontology literature: defining bridge rules that
explicit the relations between concepts.
PADAWAN (Picault and Mathieu, 2011) (Pattern for Accurate Design of Agent
Worlds in Agent Nests) is a multi-scale ABM meta-model based on a compact ma-
tricial representation of interactions, leading to a simple and elegant simulation
framework. This representation is based on the meta-model of IODA (Interaction-
Oriented Design of Agent simulations) dedicated to classical (1-level) ABM (Kubera
et al., 2008).
GAMA8 (Drogoul et al., 2013; Taillandier et al., 2010, 2012) is an ABM plat-
form with a dedicated modeling language, GAML, that offers multi-level capabili-
ties. Moreover, it includes a framework (a set of predefined GAML commands) to
5 http://repast.sourceforge.net
6 http://www.lgi2a.univ-artois.fr/~morvan/Gildas_Morvan/IRM4MLS.html
7 http://www.jamesii.org
8 http://code.google.com/p/gama-platform/

7
Multi-perspective modelling of complex phenomena

Fig. 1 Modelling simple phenomena

Fig. 2 Modelling complex phenomena through multiple perspectives

Figure 2: The Seck and Honig (2012) approach

Now if the phenomena we try to model are complex, a reductionist formal sys-
tem can only be partially successful in describing the natural system (Agazzi 1991;
agentify emerging structures (Vo et al., 2012b). It is certainly the most advanced
Mikulecky 2001). By describing a natural system as a collection of perspectives,
platform, from an end-user point of view, that integrates a multi-level approach. The
though, where each perspective is associated with a unique formal system (having
multi-scale meta-model focuses on the notion of situated agent and therefore, top class
a unique decomposition) as shown in Fig. 2, we can model a system in an inher-
abstractions include geometry and topology of simulated entities (Vo et al., 2012a).
ently ‘richer’ way by having multiple non-isomorphic decompositions that may in-
The notion of level does not appear explicitly but the concept of species defines at-
fluence each other. Such multi-perspective models can indeed capture the tangledness
tributes and behaviors of a class of same type agents and the multi-scale structure of
of the systems that result when we observe the world from different perspectives. As
the model, i.e., how species can be nested within each other.
Morin puts it (Morin 1990), “we must found the idea of a complex system on a non-
Seck and Honig developed an extension of DEVS that allows the simulation of
hierarchical concept of the whole” (Morin 1990). In a similar way, Levins (2006) pro-
multi-level (i.e., non hierarchically coupled) models (Seck and Honig, 2012). The
poses the robustness methodology, which, in a sort of triangulation, invites to analyze
coupling between levels is done through regular DEVS models, named bridge mod-
and model systems with multiple conceptually independent tools, thus improving ac-
els (fig. 2).
curacy of the models by relating the outcomes obtained from different perspectives.
AA4MM (Camus et al., 2013, 2012; Siebert, 2011; Siebert et al., 2010) (Agent and
The relation between complexity and multiple perspectives has been acknowl-
Artifact for Multi-Modeling) is a multi-modeling (or model coupling) meta-model
edged by various authors. Kaufmann has stated that the number of possible theo-
applied to ML-ABM. Levels are reified by agents that interact trough artifacts. This
meta-model extends existing ones, see e.g., Bonneaud (2008); Bonneaud et al. (2007),
distributing the scheduling between levels.

2.2 Observation, detection and agentification of emergent phe-

nomena
An important issue concerning ML-ABM is to observe, detect and possibly reify (or
more precisely agentity) phenomena emerging from agent interactions. Of course,
the question is not to detect any emergent phenomenon but those of interest, in
order to e.g., adapt the level of detail of simulations, model cross-level interactions or
observe multi-level behaviors.

8
 Very different approaches have been proposed to solve this problem. The first
ones were of course exploratory. Therefore, they rely on dedicated methods related
to specific models. Newer works focus on generic methodologies and frameworks.
They are briefly presented in a chronological order.

2.2.1 Dedicated clustering methods

The pioneering RIVAGE project (Servat, 2000; Servat et al., 1998a,b) aimed "at mod-
eling runoff, erosion and infiltration on heterogeneous soil surfaces" (Servat et al.,
1998a, p. 184). At the microscopic level, water is viewed as a set of interacting water-
balls. An indicator characterizes waterball movements to detect two types of remark-
able situations: straight trajectories (corresponding to the formation of ravines) and
stationary particles (corresponding to the formation of ponds). Close agents sharing
such properties are aggregated in ravine or pond macroscopic agents.
Bertelle et al. (2002); Tranouez (2005); Tranouez et al. (2001, 2006); Tranouez
and Dutot (2009); Tranouez et al. (2003) aimed at changing the level of detail of
fluid flow simulations using the vortex method (Leonard, 1980). The goal was, as
in the RIVAGE model, to detect complex structures, i.e., clusters of particles shar-
ing common properties, and aggregate them. However, the detection of emergent
phenomena relies on graph-based clustering methods. Moncion et al. (2010) used a
similar approach to detect aggregations of agents in flocking simulations.
Gil-Quijano (2007); Gil-Quijano and Piron (2007) and Gil-Quijano et al. (2008)
used various clustering algorithms such as self-organizing maps, K-Means and particle
swarm algorithms, to detect group formations.

2.2.2 Generic frameworks

Chen (2009, 2013); Chen et al. (2010); Chen and Hardoon (2010); Chen et al. (2008a,b,
2009) proposed a formalism, named complex event types to describe multi-level behav-
iors in ABMs. "Conceptually, complex events are a configuration of simple events
where each component event can be located in a region or point in a hyperspace
that includes time, physical space and any other dimensions" (Chen et al., 2008b, p.
4). Using this approach, it is then possible to formally define and observe complex
phenomena at different levels.
David and Courdier (2008, 2009) developed a conceptual and technical frame-
work to handle emergence reification. It is implemented in the GEAMAS-NG plat-
form (cf. GEAMAS paragraph in the previous section) and has been used in a popu-
lation model of the Reunion Island, to detect and reify new urban areas (David et al.,
2012, 2011).
A similar framework has been integrated in the GAMA platform (see GAMA
paragraph in the previous section). It includes various clustering methods developed
in the literature (Vo et al., 2012b).
SimAnalyzer is a general-purpose tool, to detect and describe group dynamics in
simulations (Caillou and Gil-Quijano, 2012; Caillou et al., 2012).

2.3 Representation of aggregated entities

While developing generic representations for aggregated entities (also called group
abstractions) seems an important issue, to the best of our knowledge a few publica-
tions are available on the subject. The common idea in these works is to delegate

9
the computation of agent behavioral functions to less detailed agents, representing
groups of aggregated agents, in other levels in order to reduce the complexity of
interaction computing.
Sharpanskykh and Treur (2011a,b) proposed two approaches to group abstrac-
tion dedicated to models where agent state are represented by variables taking values
in {0, 1} or [0, 1]:
• weighting averaging: an aggregated state of a group is estimated by averaging
agent states, with a weighting factor related to the strength of influence of an
agent in the group (the stronger, the more important).
• invariant-based abstraction: this approach consists in determining an invariant
in a group of agents, i.e., a property that does not change in time and using it
as a conservation law.

Sharpanskykh and Treur applied these methods to a collective decision making model
of social diffusion. They performed a comparative study of the methods, focusing
on computational efficiency and approximation error.
Parunak (2012) introduced the notion of pheromone field (refering to the concept
of mean field in statistical physics) that "gives the probability of encountering an
agent of the type represented by the field at a given location" (Parunak, 2012, p.
115). In this approach, agents act according to their perceptions of pheromone fields
(but not of agents).
Navarro et al. (2012) proposed a generic approach based on the notion of meso-
scopic representation: agents sharing common properties (related to their physical
or mental states) delegate the computation of behavioral functions to a mesoscopic
agent. Authors developed this approach to reduce the computational cost of simula-
tions while guaranteeing accurate results.

3 Application domains
ML-ABM has been used in various fields such as
• biomedical research
– cancer modeling (Andasari et al., 2012; Brown et al., 2011; Deisboeck and
Stamatakos, 2010; Lepagnot and Hutzler, 2009; Olsen and Siegelmann,
2013; Paiva et al., 2009; Rejniak and Anderson, 2011; Sun et al., 2012;
Wang et al., 2013, 2008; Wang and Deisboeck, 2008; Wang et al., 2007;
Zhang et al., 2007, 2009a, 2011, 2009b),
– inflammation modeling (An, 2008; An et al., 2013; An and Wilensky,
2009; Kim et al., 2012; Scheff et al., 2012; Vodovotz et al., 2008; Wakeland
et al., 2007),
– arterial adaptation (Hayenga et al., 2011; Thorne et al., 2011),
– stent design (Tahir et al., 2011),
– vascular tissue engineering (Zahedmanesh and Lally, 2012),
– bone remodeling (Cacciagrano et al., 2010),

10
 • flow modeling of walking (and running) (Gaud et al., 2008a; Navarro et al.,
2011; Nguyen et al., 2011, 2012b; Xi and Son, 2012), driving (Bourrel, 2003;
Bourrel and Henn., 2002; Bourrel and Lesort, 2003; Burghout et al., 2005;
El hmam et al., 2008, 2006a,b,c; Espié et al., 2006; Magne et al., 2000; Mam-
mar and Haj-Salem, 2006; Morvan et al., 2012, 2009; Poschinger et al., 2002;
Sewall et al., 2011; Soyez et al., 2013, 2011; Wedde and Senge, 2012) or stream-
ing (Servat et al., 1998a,b; Tranouez et al., 2006) agents,
• biology (Adra et al., 2010; Biggs and Papin, 2013; Christley et al., 2007a,b;
Jeschke and Uhrmacher, 2008; Marino et al., 2011; Montagna et al., 2010a,b;
Seal et al., 2011; Shimoni et al., 2011; Smallwood et al., 2010; Smallwood and
Holcombe, 2006; Stiegelmeyer et al., 2013; Sun et al., 2009),
• social simulation (Conte et al., 2007; Conte and Castelfranchi, 1996; Das-
calu et al., 2009, 2011; Gil-Quijano et al., 2008; Hassoumi et al., 2012; Laper-
rière, 2012; Louail, 2010; North et al., 2010; Ozik et al., 2008; Parry and
Bithell, 2012; Pumain and Louail, 2009; Sawyer, 2001, 2003; Schaller et al.,
2012; Schillo et al., 2001; Seck and Honig, 2012; Squazzoni, 2008),
• ecology (Belem, 2009; Belem and Müller, 2009; Belem and Müller, 2013; Cheong
et al., 2012; Duboz, 2004; Duboz et al., 2003; Le et al., 2011; Marilleau et al.,
2008; Morvan et al., 2008, 2009; Prévost et al., 2004; Ratzé et al., 2007; Roun-
sevell et al., 2012; Schmidt et al., 2011; Seidl et al., 2012, 2010; Semeniuk et al.,
2011; Vincenot et al., 2011),
• military simulation (Mathieu et al., 2007a,b; Parunak et al., 2009).
An interesting comparative analysis of three of these models can be found in Gil-
Quijano et al. (2009, 2010); Gil-Quijano et al. (2012) and Louail (2010, p. 185–204).

3.1 Social simulation

Social simulation is defined by Squazzoni (2008, p. 4) as "the study of social out-
comes, let us say a macro regularity, by means of computer simulation where agents’
behavior, interactions among agents and the environment are explicitly modeled to
explore those micro-based assumptions that explain the macro regularity of interest".
Major social theories developed in the second half of the twentieth century, e.g.,
structuration (Giddens, 1987) and habitus (Bourdieu, 1994) theories9 , share a common
ambition: solving the micro/macro (so called agency/structure) problem that can
be summarized by the following question: To understand social systems, should we
observe agent interactions (micro level) or structures emerging from these interactions
(macro level)? Such theories tend to consider altogether agent positions in the social
space (objective facts) and goals (subjective facts) to explain their beliefs and actions.
Their answer to the previous question could be: social systems can only be understood
by considering simultaneously agent interactions and structures in which they occur:

social structures
social practices

agent interactions.
9 These theories are described by some sociologists as hybrid (Sawyer, 2001).

11
 A key concept used by social theorists and modelers to understand downward (or
top-down) causation in social systems, i.e., how social structures influence agents, is
reflexivity. It can be defined as the "regular exercise of the mental ability, shared by all
normal people, to consider themselves in relation to their (social) contexts and vice
versa" (Archer, 2007, p. 4). Thus, social systems differ from other types of systems,
by the reflexive control that agents have on their actions: "The reflexive capacities of
the human actor are characteristically involved in a continuous manner with the flow
of day-to-day conduct in the contexts of social activity" (Giddens, 1987, p. 22). Two
very different approaches, both from technical and methodological perspectives, can
be considered to simulate systems composed of reflexive agents:
• a purely emergentist approach, only based on the cognitive capabilities of
agents to represent and consider themselves in relation to the structures emerg-
ing from their interactions — e.g., Conte and Paolucci (2002); Gilbert (2002),

• a multi-level approach based on the cognitive capabilities of agents and the

dynamic reification of interactions between social structures and agents, i.e.,
processes that underlie social practices — e.g., Gil-Quijano et al. (2009); Pumain
et al. (2009).
According to Giddens (1987), two forms of reflexivity can be distinguished: practical
(agents are not conscious of their reflexive capabilities, and therefore, are not able to
resonate about them) and discursive (agents are conscious of their reflexive capabili-
ties) reflexivity. These two forms are respectively related to the ideas of immergence
in which agent interactions produce emergent properties that modify the way they
produce them (Conte et al., 2007) and second order emergence in which agent interac-
tions produce emergent properties that are recognized (incorporated) by agents and
influence their actions (Gilbert, 2002).
ML-ABM can also be viewed as a way to link independent social theories (and
therefore concepts) defined at different levels (fig. 3) (Sawyer, 2001, 2003; Seck and
Honig, 2012). Thus, Seck and Honig (2012) proposed a model of social conflicts
integrating agent behaviors and social laws. Gil-Quijano (2007); Gil-Quijano et al.
(2007, 2008) developed a multi-scale model of intra-urban mobility. The microscopic
level represents households and housing-units, the mesoscopic one, groups of micro-
agents and urban-sectors and the macroscopic one, the city itself. A clustering algo-
rithm is applied to detect and reify groups of households and housing-units. Pumain
and Louail (2009) developed Simpop3, a multi-scale model based on two previously
developed single-scale models: Simpop nano, simulating the internal dynamics of a
city and Simpop2, simulating city interactions.
Readers interested in a more comprehensive presentation of these questions may
refer to Sawyer (2003); Schillo et al. (2001); Squazzoni (2008) and Raub et al. (2011) .

3.2 Flow modeling

A flow of moving agents can be observed at different scales. Thus, in traffic modeling,
three levels are generally considered: the micro, meso and macro levels, representing
respectively the interactions between vehicles, groups of vehicles sharing common
properties (such as a common destination or a common localisation) and flows of
vehicles. Each approach is useful in a given context: micro and meso models allow
to simulate road networks with complex topologies such as urban area, while macro

12
 macrosociology large-scale social structures

mesosociology community organization

microsociology agent

Figure 3: ML-ABM in social simulation as a link between concepts defined at differ-

ent levels

models allow to develop control strategies to prevent congestion in highways. How-

ever, to simulate large-scale road networks, it can be interesting to integrate these
different representations (fig. 4). The main problem is to determine an appropriate
coupling between the different representations, i.e., that preserves the consistency of
simulations (Davis and Hillestad, 1993).

simulated entities control strategies

macro flow of vehicles ramp metering

meso group of vehicles variable-message panels

micro vehicle vehicle instrumentation

Figure 4: ML-ABM and control in traffic simulation

3.2.1 Micro-macro models

An interesting comparison of existing micro-macro traffic models can be found in El
hmam (2006, p. 31–44) (table 1). The coupling methods rely on the following idea:
creating a virtual section which is both macro and micro at level connections. This
virtual section is used as a buffer to generate appropriate micro or macro data ac-
cording to the type of connection. On this basis, El hmam (2006); El hmam et al.
(2008, 2006a,b,c) proposed a generic coupling method between agent-based micro-
scopic models and widely used macroscopic models such as LWR, ARZ and Payne.
While micro-macro flow models were essentially developed in the traffic domain,
other applications such as crowd simulation emerged in recent years (Nguyen et al.,
2011, 2012b).
However, all the surveyed hybrid models share the same limitation: connections
between levels are fixed a priori and cannot be changed at runtime. Therefore, to

13
be able to observe some emerging phenomena such as congestion formation or to
find the exact location of a jam in a large macro section, a dynamic hybrid modeling
approach is needed (Sewall et al., 2011).

model micro model macro model

Magne et al. (2000) SITRA-B+ SIMRES
Poschinger et al. (2002) IDM Payne
Bourrel and Lesort (2003) LWR
optimal velocity
Mammar and Haj-Salem (2006) ARZ
Espié et al. (2006) ARCHISM SSMT
El hmam (2006) LWR, ARZ, Payne
generic ABM
Sewall et al. (2011) ARZ

Table 1: Main micro-macro traffic flow models, adapted from El hmam (2006, p. 42)

3.2.2 Micro-meso models

This kind of models is often used to reduce the complexity of agent interactions.
Agents sharing common properties can be aggregated to form up a higher level
(mesoscopic) agent and then, save computer resources or describe group dynamics
such as in the already mentioned RIVAGE (Servat, 2000; Servat et al., 1998a,b) and
DS (David et al., 2011) models (cf. section 2.2). Conversely, mesoscopic agents can
be broken up into lower level agents if related structures vanish.
Morvan et al. (2012, 2009) introduced an multi-level approach to solve the dead-
lock problem in field-driven autonomous intelligent vehicle systems. These sys-
tems generally rely on self-organization to achieve their goals, but AIVs can remain
trapped into dead-locks. When such a situation is detected (using a similar approach
than Servat (2000); Servat et al. (1998a,b)), it is agentified to solve the problem using
hierarchical control.
Flacher et al. (2012); Navarro et al. (2012, 2013, 2011) proposed an innovative
framework for such models: (de)aggregation functions rely not only on the observ-
able state of simulations (the environment) but also on the internal state of agents. It
has been applied to pedestrian flow simulation. The proximity between agent states
(external and internal) is computed by an affinity function.
Soyez et al. (2013, 2011) extended this framework on the basis of IRM4MLS.
Agents are "cut" into a a set of physical parts (bodies), situated in different levels,
and a non-situated part (mind) (see fig. 5). Therefore, these different parts can be
(de)aggregated independently. This approach has been applied to dynamically adapt
the level of detail in a port operations simulator.

3.3 Biological and biomedical models

A biological system can be considered at different levels of organization:
... → molecule → cell → tissue → organ → ... ,
that basically correspond to the segmentation of biological research into specialized
communities:
... → molecular biology → cell biology → histology → physiology → ... .

14
 conceptualAgent environment
1 1
1 1
1 1..n 0..n 0..n
spiritAgent bodyAgent level
1 1..n 0..n 1

Figure 5: Mind/bodies separation in the Soyez et al. (2013) model

Each research area has developed its own ontologies and models to describe the
same reality observed at different levels. However, this reductionist approach fails
when addressing complex issues (Schnell et al., 2007). Thus, it has been shown that
living systems are co-produced by processes at different levels of organization (Mat-
urana and Varela, 1980). Therefore, an explanatory model of such systems should
account for the interactions between levels.

3.3.1 Cell biology

At least two levels are explicitly represented in cell biology models: the macroscopic
one, representing the extracellular environment and the interactions between cells,
and the microscopic one, representing the intracellular environment and the inter-
actions between cell components such as signaling pathways and gene regulatory
networks. A major modeling issue is that these two levels continuously influence
each other. It leaded to the development of multi-scale models.
Montagna et al. (2010a,b) developed a model of morphogenesis in biological sys-
tems, in particular for the Drosophila Melanogaster species. Maus et al. (2011) pro-
posed a model of Schizosaccharomyces pombe (a species of yeast) cell division and
mating type switching based on the ML-Rules approach (cf. section 2.1).

3.3.2 Cancer modeling

Cancer is a complex spatialized multi-scale process, starting from genetic mutations
and potentially leading to metastasis. Moreover, it has multi-scale (from from genetic
to environmental) causes. Therefore, it can be studied from various perspectives
from the intracellular (molecular) to the population levels.
ML-ABM is a promising paradigm to model cancer development (Wang and Deis-
boeck, 2008). Indeed, as Schnell et al. (2007, p. 140) note, "a multi-scale model would
allow us to explore the effect of combination therapies, approaches that attempt to
stop cancer in its tracks by barricading multiple pathways. Most present models,
focusing on processes at a single scale, cannot provide this comprehensive view."
Zhang et al. developed an ML-ABM of a brain tumor named Glioblastoma Multi-
forme (GBM) (Zhang et al., 2007, 2009a, 2011, 2009b). This model explicitly defines
the relations between scales and uses different modeling approaches: ordinary dif-
ferential equations (ODE) at the intracellular level, discrete rules typically found in
ABM at the cellular level and partial differential equations (PDE) at the tissue level
(see fig. 6).
Moreover, this model also relies on a multi-resolution approach: heterogenous
clusters, i.e., composed of migrating and proliferating cells are simulated at a high

15
resolution while homogenous clusters of dead cells are simulated at a lower resolu-
tion. In short, "more computational resource is allocated to heterogenous regions of
the cancer and less to homogenous regions" (Zhang et al., 2011, p. 6). This model
has been implemented on graphics processing units (GPU), leading to an efficient
parallel simulator (Zhang et al., 2011).

tissue tissue evolution (PDE)

cell’s pathway receptor

diffusion of the chemoattractants

cellular cell — cell interactions (discrete rules)

cell’s phenotype

intracellular gene — protein interactions (ODE)

Figure 6: ML-ABM in brain tumor modeling (Zhang et al., 2007, 2009a, 2011, 2009b)

Sun et al. (2012) also developed a brain tumor ML-ABM available as a MATLAB
library called ABM-TKI10 . It is based on a 4 level architecture (tissue, microenviron-
mental, cellular, modelcular).
Lepagnot and Hutzler (2009) model the growth of avascular tumors to study the
impact of PAI-1 molecules on metastasis. To deal with the problem complexity (a
tumor may be composed of millions of cells) two levels are introduced: the cell and
the tumor’s core levels (fig. 7). Indeed, such cancers are generally structured as a ker-
nel of necrosed or quiescent cells surrounded by living tumor cells. As necrosed and
quiescent cells are mostly inactive, tumor’s core is reified as a single upper-level agent,
interacting with cells and PAI-1 molecules at its boundary. A more comprehensive
analysis of this model can be found in Gil-Quijano et al. (2012).

meso tumor’s core

micro cell PAI-1 molecules

Figure 7: ML-ABM in avascular tumor growth modeling (Lepagnot and Hutzler,

2009)

3.4 Ecology
Ecologists study processes that can have very different spatio-temporal dynamics.
Then, characterizing their interactions is a complicated problem and traditional
bottom-up or top-down approaches do not seem relevant: ABMs tend to be too
10 https://sites.google.com/site/agentbasedtumormodeling/home

16
complex, requiring a lot of computational resources11 while EBMs cannot deal with
complex heterogenous environments (Shnerb et al., 2000).
Ecological systems are generally described as hierarchies (Müller et al., 2005;
Ratzé et al., 2007). Thus the hierarchy theory is "a view of ecological systems, which
takes the scales of observation explicitly into account and which tries to conceptual-
ize the phenomena at their proper scale" (Ratzé et al., 2007, p. 14). ML-ABM seems
a interesting way to implement this concept. Different modeling issues in Ecology
have been solved by ML-ABM.
Duboz (2004); Duboz et al. (2003) proposed the scale transfer approach to link
microscopic and macroscopic models: the state of the system is computed by an
ABM and is used to parametrize an EBM describing population dynamics. This
EBM can then be used to parametrize the ABM environment (fig. 8).

emergent computation

ABM EBM

environment parametrization

Figure 8: The scale transfer approach (Duboz, 2004; Duboz et al., 2003)

Marilleau et al. (2008) introduced a efficient method to represent complex soils,

named APSF (Agent, Pores, Solid and fractal). Traditionally, the environment is
viewed as a regular grid, discretized into cells. A cell can represent a pore, i.e., a part
of a soil cavity, a solid or a fractal. The idea is that a cell is not necessarily an atomic
element describing an homogenous area but can be fractal, i.e., composed of smaller
pore, solid or fractal cells with a self similar structure. Fractal cells are instantiated
at run time, generating finer representations of the environment when it is needed.
Thus, this approach based on self-generation allows to represent complex multi-scale
environments at a minimal computational cost. It has been used in the SWORM
(Simulated WORMS) model that studies the relation between earthworm activity
and soil structure (Blanchart et al., 2009; Laville et al., 2012).
Diptera larvae have a complex gregarious behavior that lead to the formation of
large groups in which individuals regulate the temperature to optimize their devel-
opment speed. This phenomenon can be described by a mesoscopic equation-based
model (knowing the mass of the group and and the external temperature), while the
crowding behavior of larvae can be modeled by an ABM. Moreover, the thermal
dynamics of the cadaver can be modeled by a Cellular Automaton (CA). Morvan
et al. (2008, 2009) integrated these different models in a ML-ABM to perform more
accurate forensic entomology expertises. In this model, the EBM is parametrized ac-
cording to the ABM state. It computes the increase of temperature caused by Diptera
interactions at the group level and send it to the CA model that is used as an envi-
ronment for the ABM (fig. 9). The environment can thus be viewed as an artifact,
used to synchronize the different models.
11 An interesting solution to this problem is to reduce the complexity of agent interactions using esti-

mation algorithms such as the fast multipole method (Razavi et al., 2011).

17
 meso thermal dynamics (CA) maggot mass effect (EBM)

micro Diptera larvae (ABM)

Figure 9: Levels of organization in a ML-ABM of necrophagous Diptera develop-

ment (Morvan et al., 2008, 2009)

4 Discussion
In this section two issues are discussed:
• the different forms of level integration,
• the use of ML-ABM to solve technical problems.

4.1 Level integration

In the introduction, ML-ABM has been defined as integrating heterogenous ABMs
in a single model. Following the approach of Michel et al. (2003) on interaction, one
can distinguish at least two forms of integration:
• weak integration: levels share objects, e.g., environment properties, but not
agents,
• strong integration: levels share objects and agents.
Weak integration can be regarded as a form of multi-modeling (or model cou-
pling) where levels represent different models interacting through shared variables
called artifacts (Camus et al., 2012; Seck and Honig, 2012; Siebert et al., 2010).
Weaker forms of integration are not regarded as multi-level modeling. Thus, in
the SWARM platform (Minar et al., 1996), integration can be described as bottom-up
or isotropic (information flows in one direction). An agent is designed as a russian
doll and its behavior at a given level depends on the lower ones.
An simple example of strong integration is given by Picault and Mathieu (2011,
p. 334): "a membrane protein, which has an end inside the cell, and the other end
outside".
In the meta-models presented in section 2.1, IRM4MLS and PADAWAN are,
to the best of our knowledge, the only ones able to simulate strongly integrated
levels. It is not surprising as they are based on formal interaction models (respectively
IRM4S and IODA) that differentiate between agent influences and level reaction and
therefore, are able to represent strong interaction (Michel et al., 2003). In short,
agents are strongly interacting if the interaction output depends on the influences of
each agent. Thus, STRIPS-like action models (i.e., that view action as a change of
the state of the world), used in most of agent-based simulation platforms, are unable
to represent such interactions. Yet, (Michel et al., 2003) showed that modeling a
strong interaction as a weak one leads to arbitrary implementation choices and result
interpretation issues.

18
 In ML-ABM, the problem is similar since levels can be seen are strongly interact-
ing entities. Thus, we can conclude that a modeling formalism capable of represent-
ing strong interaction12 can be extended to a multi-level one.

4.2 Multi-level technical tools

Although considering cross-level interactions is usually related to the application
domain as shown previously in this article, it can also be viewed as a technical tool:
• an ABM (microscopic level) can be used to parametrize an equation based
model (macroscopic level) (Duboz, 2004; Duboz et al., 2003; Nguyen et al.,
2012a),
• levels can be created at run-time by other levels to generate fractal environ-
ments (Marilleau et al., 2008),
• a mesoscopic level can be viewed as a controller (in the control theory mean-
ing) of group-related properties (Morvan and Jolly, 2012).
• automated observation and analysis tools can be introduced at levels not ex-
plicitly present in the model to detect and study (multi-level) emergent phe-
nomena13 (cf. section 2.2). However, reified emergent phenomena cannot be
considered as model entities since they are not re-injected in the simulation (see
fig. 10),
• a radical interpretation of ML-ABM is the concept of multi-future (Parunak
et al., 2010; Parunak and Brueckner, 2010). The possible trajectories of agent
actions are computed by "ghosts" as a pheromone field and agents act according
to it, selecting the most probable one.
As these ideas are domain-independent, they could be implemented as a generic
library, providing such services to a classical ABM or ML-ABM framework or plat-
form.

Inputs
Agentification

Simulation
emergent phenomena
Detection

Visualization tools

Figure 10: Two main uses of detected emergent phenomena: visualization or re-
injection in the simulation as agents

12 Some authors described these approaches as "interaction-based modeling", by opposition to the

term "individual-based modeling", as that they focus on interactions rather than on individual behav-
iors (Desmeulles et al., 2009; Kubera et al., 2008).
13 Indeed, as An (2008, p. 4) notes about ABM, "since the models rely on an ill-defined principle of

’emergence’ in order to transcend the epistemological boundaries represented by the multiple hierarchies
of system organization, their behavior is difficult to characterize analytically".

19
5 Conclusion
An important challenge for the scientific community is to find ways to deal with
the complexity of natural and artificial complex systems. This issue led to the devel-
opment of dedicated modeling paradigms and engineering principles that focus on
interaction and organization. We strongly believe that such techniques will play an
important role in the future.
In this article, we conducted a comprehensive analysis of the bibliography avail-
able on one of them: multi-level agent-based modeling.
As we shown, many papers focus on the application of this technique and then,
are published in domain-related journals and conferences. However, a dedicated
venue for theoretical or methodological papers is lacking. Such forum would al-
low to unify the vocabulary and concept definitions, discuss the main issues of this
approach and more generally, share ideas with the interested communities.

Acknowledgments Belem, M. and Müller, J.-P. (2009). Toward a conceptual

framework for multi-points of view analysis in com-
plex system modeling: Orea model. In 7th Interna-
I am particularly grateful to my col- tional Conference on Practical Applications of Agents and
leagues Shahin Gelareh, Yoann Kubera Multi-Agent Systems (PAAMS 2009), volume 55 of Ad-
vances in Intelligent and Soft Computing, pages 548–556.
and Jean-Baptiste Soyez for their help Springer.
and support. I also would like to thank
Belem, M. and Müller, J.-P. (2013). An organizational
Chih-Chun Chen, Jean-Pierre Müller model for multi-scale and multi-formalism simulation:
and Sebastian Senge for suggesting new Application in carbon dynamics simulation in west-
african savanna. Simulation Modelling Practice and The-
entries in the bibliographical database. ory, 32:83–98.

Bertelle, C., Jay, V., Lerebourg, S., Olivier, D., and

References Tranouez, P. (2002). Dynamic clustering for auto-
organized structures in complex fluid flows. In ESS
Adra, S., Sun, T., MacNeil, S., Holcombe, M., and Small- 2002 Conference.
wood, R. (2010). Development of a three dimensional
multiscale computational model of the human epider- Biggs, M. B. and Papin, J. A. (2013). Novel multiscale mod-
mis. PloS one, 5(1):e8511. eling tool applied to pseudomonas aeruginosa biofilm
formation. PloS one, 8(10):e78011.
An, G. (2008). Introduction of an agent-based multi-scale
modular architecture for dynamic knowledge represen- Blanchart, E., Marilleau, N., Chotte, J., Drogoul, A.,
tation of acute inflammation. Theoretical Biology and Perrier, E., and Cambier, C. (2009). SWORM: an
Medical Modelling, 5(11). agent-based model to simulate the effect of earthworms
on soil structure. European Journal of Soil Science,
An, G., Wandling, M., and Christley, S. (2013). Agent- 60(1):13–21.
based modeling approaches to multi-scale systems bi-
ology: An example agent-based model of acute pul- Bonneaud, S. (2008). Des agents-modèles pour la modélisa-
monary inflammation. In Systems Biology - Integrative tion et la simulation de systèmes complexes: application
Biology and Simulation Tools, pages 429–461. Springer. à l’écosystémique des pêches. PhD thesis, Université de
Bretagne occidentale, Brest.
An, G. and Wilensky, U. (2009). Artificial life models in
software. In From Artificial Life to In Silico Medicine: Bonneaud, S., Redou, P., and Chevaillier, P. (2007). Pat-
NetLogo as a Means of Translational Knowledge Represen- tern oriented agent-based multi-modeling of exploited
tation in Biomedical Research, pages 183–212. Springer.
ecosystems. In Proceedings of the 6th EUROSIM
congress.
Andasari, V., Roper, R., and ans M.A.J. Chaplain, M. S.
(2012). Integrating intracellular dynamics using com-
pucell3d and bionetsolver: Applications to multiscale Bourdieu, P. (1994). Raisons pratiques: sur la théorie de
modelling of cancer cell growth and invasion. PLoS l’action, volume 331 of Points. Éditions du Seuil.
ONE, 7(3):e33726.
Bourrel, E. (2003). Modélisation dynamique de l’écoulement
Archer, M. (2007). Making our Way through the World: Hu- du trafic routier : du macroscopique au microscopique.
man Reflexivity and Social Mobility. Cambridge Univer- PhD thesis, Institut National des Sciences Appliquées
sity Press. de Lyon, Lyon.

Belem, M. (2009). A conceptual model for multipoints of Bourrel, E. and Henn., V. (2002). Mixing micro and macro
view analysis of complex systems. Application to the analy- representations of traffic flow: a first theoretical step.
sis of the carbon dynamics of village territories of the West Proceedings of the 9th Meeting of the Euro Working Group
African savannas. PhD thesis, AgroParisTech. on Transportation.

20
Bourrel, E. and Lesort, J. (2003). Mixing micro and macro Chen, C., Nagl, S., and Clack, C. (2009). A formal-
representations of traffic flow: a hybrid model based ism for multi-level emergent behaviours in designed
on the LWR theory. 82th Annual Meeting of the Trans- component-based systems and agent-based simulations.
portation Research Board. In Aziz-Alaoui, M. and Bertelle, C., editors, From
System Complexity to Emergent Properties, volume 12
Brown, M., Bowring, E., Epstein, S., Jhaveri, M., Mah- of Understanding Complex Systems, pages 101–114.
eswaran, R., Mallick, P., Mumenthaler, S., Povinelli, Springer.
M., and Tambe, M. (2011). Applying multi-agent tech-
niques to cancer modeling. In Sixth Annual Workshop Cheong, S., Brown, D., Kok, K., and Lopez-Carr, D.
on Multiagent Sequential Decision Making in Uncertain (2012). Mixed methods in land change research: to-
Domains (MSDM-2011). wards integration. Transactions of the Institute of British
Geographers, 37(1):8–12.
Burghout, W., Koutsopoulos, H., and Andreasson, I.
(2005). Hybrid mesoscopic-microscopic traffic simu- Christley, S., Newman, S. A., and Alber, M. S. (2007a).
lation. Transportation Research Record: Journal of the Agent-based model for developmental pattern forma-
Transportation Research Board, 1934(1):218–255. tion with multiscale dynamics and varying cell geome-
try. In Mathematical Modeling of Biological Systems, vol-
Cacciagrano, D., Corradini, F., and Merelli, E. (2010). ume 1, pages 149–161. Springer.
Bone remodelling: A complex automata-based model
running in bioshape. Cellular Automata, pages 116– Christley, S., Zhu, X., Newman, S. A., and Alber, M. S.
127. (2007b). Multiscale agent-based simulation for chon-
drogenic pattern formation in vitro. Cybernetics and
Caillou, P. and Gil-Quijano, J. (2012). Simanalyzer : Au- Systems: An International Journal, 38(7):707–727.
tomated description of groups dynamics in agent-based
simulations. In Proc. of 11th Int. Conf. on Autonomous Conte, R., Andrighetto, G., Campennì, M., and Paolucci,
Agents and Multiagent Systems (AAMAS 2012). M. (2007). Emergent and immergent effects in com-
plex social systems. In Proceedings of AAAI Symposium,
Caillou, P., Gil-Quijano, J., and Zhou, X. (2012). Auto- Social and Organizational Aspects of Intelligence, Wash-
mated observation of multi-agent based simulations: a ington DC.
statistical analysis approach. to appear in Studia Infor-
matica Universalis, 10(3):62–86.
Conte, R. and Castelfranchi, C. (1996). Simulating multi-
agent interdependencies. a two-way approach to the
Camus, B., Bourjot, C., and Chevrier, V. (2013). Multi- micro-macro link. In Social Science Microsimulation,
level modeling as a society of interacting models. In pages 394–415. Springer.
SpringSim’13, ADS Symposium-Spring Simulation Multi-
Conference, Agent-Directed Simulation Symposium-2013,
Conte, R. and Paolucci, M. (2002). Reputation in Artificial
volume 1, pages 15–22.
Societies. Kluwer Academic Publishers.
Camus, B., Siebert, J., Bourjot, C., and Chevrier, V. (2012).
Modélisation multi-niveaux dans AA4MM. In Actes Dascalu, M., Gheorghe, S., and Zafiu, A. (2009). Multi-
des 20èmes Journées Francophones sur les Systèmes Multi- level simulator for artificial societies. Romanian Journal
Agents ( JFSMA). of Information Science and Technology, 12(4):466–484.

Chen, C. (2009). Complex Event Types for Agent-Based Sim- Dascalu, M., Stefan, G., Zafiu, A., and Plavitu, A. (2011).
ulation. PhD thesis, University College London. Applications of multilevel cellular automata in epi-
demiology. In Proceedings of the 13th WSEAS interna-
Chen, C. (2013). A theoretical framework for con- tional conference on Automatic control, modelling & sim-
ducting multi-level studies of complex social systems ulation, pages 439–444. World Scientific and Engineer-
with agent-based models and empirical data. CoRR, ing Academy and Society (WSEAS).
abs/1302.4774.
David, D. and Courdier, R. (2008). Emergence as meta-
Chen, C., Clack, C., and Nagl, S. (2010). Identifying multi- knowledge: refining simulation models through emer-
level emergent behaviors in agent-directed simulations gence reification. In Proceedings of ESM’2008, Le Havre,
using complex event type specifications. Simulation, France, pages 25–27.
86(1):41–51.
David, D. and Courdier, R. (2009). See emergence as a
Chen, C. and Hardoon, D. (2010). Learning from multi- metaknowledge. a way to reify emergent phenomena in
level behaviours in agent-based simulations: a systems multiagent simulations? In Proceedings of ICAART’09,
biology application. Journal of Simulation, 4(3):196– pages 564–569, Porto, Portugal.
203.
David, D., Gangat, Y., Payet, D., and Courdier, R. (2012).
Chen, C., Nagl, S., and Clack, C. (2008a). A method for Reification of emergent urban areas in a land-use sim-
validating and discovering associations between multi- ulation model in reunion island. In ECAI workshop
level emergent behaviours in agent-based simulations. on Intelligent Agents in Urban Simulations and Smart
In N.T. Nguyen, G.S. Jo, R. H. and Jain, L., editors, Cities (IAUSSC2012), pp. 28-32, Montpellier, pages 28–
Agent and Multi-Agent Systems: Technologies and Appli- 32, Montpellier, France.
cations, volume 4953 of Lecture Notes in Computer Sci-
ence, pages 1–10. Springer. David, D., Payet, D., and Courdier, R. (2011). Réifica-
tion de zones urbaines émergentes dans un modèle sim-
Chen, C., Nagl, S., and Clack, C. (2008b). Multi-level be- ulant l’évolution de la population à la réunion. In Actes
haviours in agent-based simulation: colonic crypt cell des 19èmes Journées Francophones sur les Systèmes Multi-
populations. In Proceedings of the Seventh International Agents ( JFSMA’11), pages 63–72, Valenciennes, France.
Conference on Complex Systems. Cépaduès.

21
Davis, P. K. and Hillestad, R. (1993). Families of models Ferber, J. (1999). Multi-Agent Systems: An Introduction
that cross levels of resolution: issues for design, calibra- to Distributed Artificial Intelligence. Addison-Wesley
tion and management. In Proceedings of the 25th confer- Longman Publishing Co., Inc.
ence on Winter simulation, WSC ’93, pages 1003–1012,
New York, NY, USA. ACM. Ferber, J. and Müller, J.-P. (1996). Influences and reaction:
a model of situated multiagent systems. In 2nd Inter-
Deisboeck, T. and Stamatakos, G. (2010). Multiscale cancer national Conference on Multi-agent systems (ICMAS’96),
modeling, volume 34. CRC Press. pages 72–79.
Desmeulles, G., Bonneaud, S., Redou, P., Rodin, V., and Flacher, F., Navarro, L., and Corruble, V. (2012). Multi-
Tisseau, J. (2009). In virtuo experiments based on the agent simulation system. European Patent EP2518711,
multi-interaction system framework: the réiscop meta- Thalès S.A.
model. Computer Modeling in Engineering and Sciences,
47(3):299–329.
Gaud, N. (2007). Systèmes multi-agent holoniques : De
Drogoul, A., Amouroux, E., Caillou, P., Gaudou, B., Grig- l’analyse à l’implantation. Méta-modèle, méthodologie, et
nard, A., Marilleau, N., Taillandier, P., Vavasseur, M., simulation multi-niveaux. PhD thesis, Université de
Vo, D.-A., and Zucker, J.-D. (2013). GAMA: A spa- Technologie de Belfort-Montbéliard.
tially explicit, multi-level, agent-based modeling and
simulation platform. In Advances on Practical Applica- Gaud, N., Galland, S., Gechter, F., Hilaire, V., and
tions of Agents and Multi-Agent Systems, pages 271–274. Koukam, A. (2008a). Holonic multilevel simulation of
Springer. complex systems : Application to real-time pedestrians
simulation in virtual urban environment. Simulation
Drogoul, A., Vanbergue, D., and Meurisse, T. (2003). Modelling Practice and Theory, 16:1659–1676.
Multi-agent based simulation: Where are the agents?
In Sichman, J., Bousquet, F., and Davidsson, P., editors, Gaud, N., Galland, S., and Koukam, A. (2008b). Towards a
Multi-Agent-Based Simulation II, volume 2581 of Lecture multilevel simulation approach based on holonic mul-
Notes in Computer Science, pages 1–15. Springer. tiagent systems. In UKSIM, pages 180–185, Cambridge,
UK.
Duboz, R. (2004). Intégration de modèles hétèrogenes pour la
modélisation et la simulation de systèmes complexes. PhD Giddens, A. (1987). La constitution de la société: éléments de
thesis, Université du Littoral. la théorie de la structuration. Presses Universitaires de
France.
Duboz, R., Ramat, É., and Preux, P. (2003). Scale transfer
modeling: using emergent computation for coupling Gil-Quijano, J. (2007). Modèles d’auto-organisation pour
an ordinary differential equation system with a reac- l’émergence de formes urbaines à partir de comportements
tive agent model. Systems Analysis Modelling Simula- individuels à Bogota. PhD thesis, Université Paris VI.
tion, 43(6):793–814.

Duhail, N. (2013). DEVS et ses extensions pour des simula- Gil-Quijano, J., Hutzler, G., and Louail, T. (2009). De
tions multi-modèles en interaction multi-échelles. Mas- la cellule biologique à la cellule urbaine : retour sur
ter’s thesis, Université de Rennes 1 — Telecom Bre- trois expériences de modélisation multi-échelles à base
tagne. d’agents. In Actes des 17èmes Journées Francophones sur
les Systèmes Multi-Agents ( JFSMA’09).
El hmam, M. (2006). Contribution à la modélisation et à la
simulation hybride du flux de trafic. PhD thesis, Univer- Gil-Quijano, J., Hutzler, G., and Louail, T. (2010).
sité d’Artois. Accroche-toi au niveau, j’enlève l’échelle: Éléments
d’analyse des aspects multiniveaux dans la simula-
El hmam, M., Abouaissa, H., and Jolly, D. (2008). Con- tion à base d’agents. Revue d’Intelligence Artificielle,
tribution à la modélisation et la simulation du flux de 24(5):625–648.
trafic: approche hybride basée sur les systèmes multi-
agents. Actes INRETS, pages 93–111. Gil-Quijano, J., Louail, T., and Hutzler, G. (2012). From
biological to urban cells: Lessons from three multi-
El hmam, M., Abouaissa, H., Jolly, D., and Benasser, A. level agent-based models. In Desai, N., Liu, A., and
(2006a). Macro-micro simulation of traffic flow. In Winikoff, M., editors, Principles and Practice of Multi-
Proceedings of the12th IFAC Symposium on Information Agent Systems, volume 7057 of Lecture Notes in Com-
Control Problems in Manufacturing (INCOM06), pages puter Science, pages 620–635. Springer.
351–356, Saint Etienne FRANCE.
Gil-Quijano, J. and Piron, M. (2007). Formation au-
El hmam, M., Abouaissa, H., Jolly, D., and Benasser, A.
tomatique de groupes d’agents sociaux par techniques
(2006b). Simulation hybride de flux de trafic basée sur
d’apprentissage non supervise. In Actes de EGC’07 Ate-
les systèmes multi-agents. In 6e Conférence Francophone
lier Fouille de Données et Algorithmes Biomimétiques.
de MOdélisation et SIMulation-MOSIM.

El hmam, M., Jolly, D., Abouaissa, H., and Benasser, A. Gil-Quijano, J., Piron, M., and Drogoul, A. (2007).
(2006c). Modélisation hybride du flux de trafic. Revue Vers une simulation multi-agent de groupes d’individus
électronique Sciences et Technologies de l’Automatique. pour modéliser les mobilités résidentielles intra-
urbaines. Revue internationale de géomatique,
Epstein, J. (2006). Generative social science: Studies in agent- 17(2):161–181.
based computational modeling. Princeton University
Press. Gil-Quijano, J., Piron, M., and Drogoul, A. (2008). Social
Simulation: Technologies, Advances and New Discover-
Espié, S., Gattuso, D., and Galante, F. (2006). Hybrid traf- ies, chapter Mechanisms of Automated Formation and
fic model coupling macro- and behavioral microsimula- Evolution of Social-Groups: A Multi-Agent System to
tion. In 85th Annual Meeting of Transportation Research Model the Intra-Urban Mobilities of Bogota City. IGI
Board, Washington D.C. Global.

22
Gilbert, N. (2002). Varieties of emergence. In Agent 2002 Mammar, S.and Lebacque, J. and Haj-Salem, H. (2006).
Conference: Social agents: ecology, exchange, and evolu- Hybrid model based on second-order traffic model.
tion, Chicago. 85th Annual Meeting of Transportation Research Board,
1(06-2160).
Gilbert, N. (2007). Agent-Based Models. Number 153 in
Quantitative Applications in the Social Sciences. Sage Marcenac, P., Courdier, R., and Calderoni, S. (1998a). To-
Publications. wards a toolkit to study emergence in complex systems
simulations. http://personnel.univ-reunion.
Hassoumi, I., Lang, C., Marilleau, N., Temani, M., fr/courdier/pdfs/int_iea1998_MC98.pdf.
Ghedira, K., and Zucker, J. D. (2012). Toward a
spatially-centered approach to integrate heterogeneous Marcenac, P., Courdier, R., Calderoni, S., and Soulie,
and multi-scales urban component models. In Advances J. (1998b). Towards an emergence machine for com-
on Practical Applications of Agents and Multi-Agent Sys- plex systems simulations. In Tasks and Methods in Ap-
tems, volume 155 of Advances in Intelligent and Soft plied Artificial Intelligence, volume 416, pages 785–794.
Computing Volume, pages 81–86. Springer. Springer.
Hayenga, H., Thorne, B., Peirce, S., and Humphrey, Marcenac, P. and Giroux, S. (1998). Geamas: A generic
J. (2011). Ensuring congruency in multiscale mod- architecture for agent-oriented simulations of complex
eling: towards linking agent based and continuum processes. Applied Intelligence, 8(3):247–267.
biomechanical models of arterial adaptation. Annals
of biomedical engineering, pages 1–14. Marilleau, N., Cambier, C., Drogoul, A., Chotte, J., Per-
rier, E., and Blanchart, E. (2008). Multiscale mas mod-
Jeschke, M. and Uhrmacher, A. (2008). Multi-resolution elling to simulate the soil environment: Application to
spatial simulation for molecular crowding. In Proc. soil ecology. Simulation Modelling Practice and Theory,
of the Winter Simulation Conference, pages 1384–1392. 16(7):736–745.
IEEE.
Marino, S., El-Kebir, M., and Kirschner, D. (2011). A hy-
Kim, M., Christley, S., Alverdy, J., Liu, D., and An, G. brid multi-compartment model of granuloma forma-
(2012). A multi-scale agent-based model of necrotizing tion and t cell priming in tuberculosis. Journal of Theo-
enterocolitis integrating oxidative stress, inflammation, retical Biology.
and microvascular thrombosis. Journal of Surgical Re-
search, 172(2):196. Mathieu, J., James, J., Mahoney, P., Boiney, L.,
Hubbard, R., and White, B. (2007a). Hy-
Koestler, A. (1967). The Ghost in the Machine. Hutchinson.
brid systems dynamic, petri net, and agent-based
Koestler, A. (1978). Janus: A summing up. Random House. modeling of the air and space operations cen-
ter. http://www.mitre.org/work/tech_papers/
Kubera, Y., Mathieu, P., and Picault, S. (2008). Interaction- tech_papers_07/07_0455/07_0455.pdf.
oriented agent simulations: From theory to implemen-
tation. In Proc. of the18th European Conf. on Artificial Mathieu, J., Melhuish, J., James, J., Mahoney, P.,
Intelligence (ECAI’08), pages 383–387. Boiney, L., and White, B. (2007b). Multi-
scale modeling of the air operations center.
Laperrière, V. (2012). Un cadre de modélisation dy- http://www.mitre.org/work/tech_papers/
namique multiniveau pour la peste bubonique des tech_papers_07/06_1497/06_1497.pdf.
hautes terres de madagascar. Bulletin de l’Association
de géographes français, 89(2):278–288. Maturana, H. R. and Varela, F. J. (1980). Autopoiesis and
Cognition The Realization of the Living, volume 42 of
Laville, G., Mazouzi, K., Lang, C., Marilleau, N., and Boston Studies in the Philosophy of Science. Springer
Philippe, L. (2012). Using gpu for multi-agent multi- Berlin Heidelberg.
scale simulations. In Distributed Computing and Arti-
ficial Intelligence, volume 151 of Advances in Intelligent Maus, C., John, M., Röhl, M., and Uhrmacher, A. (2008).
and Soft Computing, pages 197–204. Springer. Hierarchical modeling for computational biology. In
Bernardo, M., Degano, P., and Zavattaro, G., editors,
Le, Q., Seidl, R., and Scholz, R. (2011). Feedback loops Formal Methods for Computational Systems Biology, vol-
and types of adaptation in the modelling of land-use ume 5016 of Lecture Notes in Computer Science, pages
decisions in an agent-based simulation. Environmental 81–124. Springer.
Modelling & Software.
Maus, C., Rybacki, S., and Uhrmacher, A. (2011). Rule-
Leonard, A. (1980). Vortex methods for flow simulation. based multi-level modeling of cell biological systems.
Journal of Computational Physics, 37(3):289–335. BMC systems biology, 5(1):166–186.

Lepagnot, J. and Hutzler, G. (2009). A multiscale agent- Michel, F. (2007). The IRM4S model: the influ-
based model for the simulation of avascular tumour ence/reaction principle for multiagent based simula-
growth. Journal of Biological Physics and Chemistry, tion. In Proc. of 6th Int. Conf. on Autonomous Agents
9(1):17–25. and Multiagent Systems (AAMAS 2007), pages 1–3.

Louail, T. (2010). Comparer les morphogénèses urbaines Michel, F., Gouaïch, A., and Ferber, J. (2003). Weak in-
en Europe et aux États-Unis par la simulation à base teraction and strong interaction in agent based simu-
d’agents – Approches multi-niveaux et environnements de lations. In Multi-Agent-Based Simulation III, volume
simulation spatiale. PhD thesis, Université d’Evry-Val 2927 of Lecture Notes in Computer Science, pages 43–56.
d’Essonne. Springer.

Magne, L., Rabut, S., and Gabard, J. (2000). Towards an Minar, N., Burkhart, R., Langton, C., and Askenazi, M.
hybrid macro-micro traffic flow simulation model. Pro- (1996). The swarm simulation system: A toolkit for
ceedings of the INFORMS Salt Lake City String 2000 Con- building multi-agent simulations. Technical Report 96-
ference. 06-042, Santa Fe Institute.

23
Moncion, T., Amar, P., and Hutzler, G. (2010). Automatic Müller, J.-P., Ratzé, C., Gillet, F., and Stoffel, K. (2005).
characterization of emergent phenomena in complex Modeling and simulating hierarchies using an agent-
systems. Journal of Biological Physics and Chemistry, based approach. In Proceedings of the MODSIM 2005 In-
10:16–23. ternational Congress on Modelling and Simulation, pages
1631–1638.
Montagna, S., Donati, N., and Omicini, A. (2010a).
Navarro, L., Corruble, V., Flacher, F., and Zucker, J.-D.
An agent-based model for the pattern formation in
(2012). Mesoscopic level: A new representation level
drosophila melanogaster. Artificial Life XII, pages 110–
for large scale agent-based simulations. In SIMUL 2012,
117.
The Fourth International Conference on Advances in Sys-
tem Simulation, pages 68–73.
Montagna, S., Omicini, A., and Ricci, A. (2010b). A multi-
scale agent-based model of morphogenesis in biological Navarro, L., Corruble, V., and Zucker, J.-D. (2013). A
systems. In Omicini, A. and Viroli, M., editors, Pro- flexible approach to multi-level agent-based simulation
ceedings of 11 ◦ Workshop nazionale "Dagli Oggetti agli with the mesoscopic representation. In Proc. of 12th
Agenti" WOA 2010, Rimini, Italy. Int. Conf. on Autonomous Agents and Multiagent Systems
(AAMAS 2013).
Morin, E. (1992). Method: Towards a Study of Humankind,
Navarro, L., Flacher, F., and Corruble, V. (2011). Dynamic
volume 1. Peter Lang Pub Inc.
level of detail for large scale agent-based urban simula-
tions. In Proc. of 10th Int. Conf. on Autonomous Agents
Morvan, G., Dupont, D., Soyez, J.-B., and Merzouki, R. and Multiagent Systems (AAMAS 2011), pages 701–708.
(2012). Engineering hierarchical complex systems: an
agent-based approach – the case of flexible manufactur- Nguyen, N., Taillandier, P., Drogoul, A., and Auger, P.
ing systems. In Service Orientation in Holonic and Multi (2012a). Inferring equation-based models from agent-
Agent Manufacturing Control, volume 402 of Studies in based models: a case study in competition dynamics.
Computational Intelligence. Springer. In Principles and Practice of Multi-Agent Systems, volume
7455 of Lecture Notes in Artificial Intelligence, pages
Morvan, G. and Jolly, D. (2012). Multi-level agent-based 413–427. Springer.
modeling with the Influence Reaction principle. CoRR,
abs/1204.0634. Nguyen, T., Zucker, J.-D., Nguyen, D., Drogoul, A.,
and Vo, A. (2011). Hybrid equation-based and agent-
based modeling of crowd evacuation on road network.
Morvan, G., Jolly, D., Veremme, A., Dupont, D., and In Eighth International Conference on Complex Systems
Charabidze, D. (2008). Vers une méthode de modéli- (ICSS), pages 558–570.
sation multi-niveaux. In Actes de la 7ème Conférence
de Modélisation et Simulation MOSIM, Paris, France, vol- Nguyen, T., Zucker, J.-D., Nguyen, D., Drogoul, A., and
ume 1, pages 167–174. Vo, A. (2012b). A hybrid macro-micro pedestrians
evacuation model to speed up simulation in road net-
Morvan, G., Veremme, A., and Dupont, D. (2011). works. In Dechesne, F., Hattori, H., ter Mors, A.,
IRM4MLS: the influence reaction model for multi-level Such, J., Weyns, D., and Dignum, F., editors, Advanced
simulation. In Bosse, T., Geller, A., and Jonker, C., Agent Technology, volume 7068 of Lecture Notes in Com-
editors, Multi-Agent-Based Simulation XI, volume 6532 puter Science, pages 371–383. Springer.
of Lecture Notes in Artificial Intelligence, pages 16–27.
Springer. North, M., Macal, C., Aubin, J., Thimmapuram, P., Bra-
gen, M., Hahn, J., Karr, J., Brigham, N., Lacy, M., and
Hampton, D. (2010). Multiscale agent-based consumer
Morvan, G., Veremme, A., Dupont, D., and Jolly, D. market modeling. Complexity, 15(5):37–47.
(2009). Modélisation et conception multiniveau de sys-
tèmes complexes : stratégie d’agentification des organi- Olsen, M. M. and Siegelmann, H. T. (2013). Multiscale
sations. Journal Européen des Systèmes Automatisés, 43(4- agent-based model of tumor angiogenesis. In Interna-
5):381–406. tional Conference on Computational Science, ICCS 2013,
volume 18 of Procedia Computer Science, pages 1026–
Müller, J.-P. (2009). Towards a formal semantics of event- 1035. Elsevier.
based multi-agent simulations. In David, N. and Sich-
man, J., editors, Multi-Agent-Based Simulation IX, vol- Ozik, J., Sallach, D. L., and Macal, C. M. (2008). Model-
ume 5269 of Lecture Notes in Computer Science, pages ing dynamic multiscale social processes in agent-based
110–126. Springer. models. IEEE Intelligent Systems, 23(4):36–42.

Paiva, L., Binny, C., Ferreira Jr., S., and Martins, M.

Müller, J.-P. and Aubert, S. (2011). L’ontologie pour con- (2009). A multiscale mathematical model for oncolytic
struire une représentation multi-niveau de et par les virotherapy. Cancer Research, 69:1205–1211.
systèmes sociaux. In Rencontres interdisciplinaires de
Rochebrune sur les systèmes complexes naturels et artifi- Parry, H. and Bithell, M. (2012). Agent-based models of geo-
ciels. graphical systems, chapter Large scale agent-based mod-
elling: A review and guidelines for model scaling, pages
Müller, J.-P. and Diallo, A. (2012). Vers une méthode 271–308. Springer.
multi-point de vue de modélisation multi-agent. In
Parunak, H. (2012). Between agents and mean fields.
Actes des 20èmes Journées Francophones sur les Systèmes
In Multi-Agent-Based Simulation XII, volume 7124 of
Multi-Agents, pages 33–42. Cepadues Editions.
Lecture Notes in Artificial Intelligence, pages 113–126.
Springer.
Müller, J.-P., Diallo, A., François, J.-C., Sanders, L.,
Waldeck, R., and Mathian, H. (2011). Building ontolo- Parunak, H., Bisson, R., and Brueckner, S. (2010). Agent
gies from a variety of points of view. In Proceedings of interaction, multiple perspectives, and swarming simu-
the 7th European Social Simulation Association Confer- lation. In Proc. of 9th Int. Conf. on Autonomous Agents
ence. and Multiagent Systems (AAMAS 2010), pages 549–556.

24
Parunak, H. and Brueckner, S. (2010). Multi-perspective, Sawyer, R. (2003). Artificial societies, multiagent systems
multi-future modeling and model analysis. In Proceed- and the micro-macro link in sociological theory. Socio-
ings of ICCDM, Miami, FL. logical Methods & Research, 31(3):325–363.

Parunak, H., Sauter, J., and Crossman, J. (2009). Multi- Scerri, D., Drogoul, A., Hickmott, S., and Padgham, L.
layer simulation for analyzing IED threats. In Technolo- (2010). An architecture for modular distributed simula-
gies for Homeland Security, 2009. HST’09. IEEE Confer- tion with agent-based models. In Proc. of 9th Int. Conf.
ence on, pages 323–330. IEEE. on Autonomous Agents and Multiagent Systems (AAMAS
2010), pages 541–548. International Foundation for Au-
Picault, S. and Mathieu, P. (2011). An interaction-oriented tonomous Agents and Multiagent Systems.
model for multi-scale simulation. In Walsh, T., ed-
itor, Twenty-Second International Joint Conference on Schaller, N., Lazrak, E., Martin, P., Mari, J., Aubry, C.,
Artificial Intelligence, pages 332–337, Barcelona, Spain. and Benoît, M. (2012). Combining farmers’ decision
AAAI Press. rules and landscape stochastic regularities for landscape
modelling. Landscape Ecology, 27(3):1–14.
Poschinger, A., Kates, R., and Keller, H. (2002). Coupling
of concurrent macroscopic and microscopic traffic flow Scheff, J., Mavroudis, P., Foteinou, P., An, G., Calvano, S.,
models using hybrid stochastic and deterministic disag- Doyle, J., Dick, T., Lowry, S., Vodovotz, Y., and An-
gregation. Transportation and Traffic Theory for the 21st droulakis, I. (2012). A multiscale modeling approach
century. to inflammation: A case study in human endotoxemia.
Journal of Computational Physics.
Prévost, G., Tranouez, P., Lerebourg, S., Bertelle, C., and
Olivier, D. (2004). Ecosystem complexity described
with ontological tool for a multi-scale, multi-model ap- Scheutz, M., Madey, G., and Boyd, S. (2005). tMANS -
proaches in distributed environment. In Jordan Inter- the multi-scale agent-based networked simulation for
national Conference on Computer Sciences and Engineer- the study of multi-scale, multi-level biological and so-
ing, Al-Salt, Jordan. cial phenomena. In Spring Simulation Multiconference.

Pumain, D. and Louail, T. (2009). Interaction des ontolo- Schillo, M., Fischer, K., and Klein, C. (2001). The micro-
gies informatique et géographique pour simuler les dy- macro link in DAI and sociology. In Multi-Agent Based
namiques multiscalaires. In XVIèmes rencontres inter- Simulation, volume 1979 of Lecture Notes in Artificial
disciplinaires de Rochebrune, Ontologies et dynamique des Intelligence, pages 133–148. Springer.
systèmes complexes, perspectives interdisciplinaires.
Schmidt, S., Picioreanu, C., Craenen, B., Mackay, R.,
Pumain, D., Sanders, L., Bretagnolle, A., Glisse, B., and Kreft, J., and Theodoropoulos, G. (2011). A multi-
Mathian, H. (2009). The future of urban systems: Ex- scale agent-based distributed simulation framework for
ploratory models. In Lane, D., Pumain, D., der Leeuw, groundwater pollution management. In 15th Interna-
S. V., and West, G., editors, Complexity Perspectives in tional Symposium on Distributed Simulation and Real
Innovation and Social Change, volume 7 of Methodos Se- Time Applications (DS-RT), pages 18–27. IEEE.
ries, pages 331–360. Springer.
Schnell, S., Grima, R., and Maini, P. (2007). Multiscale
Railsback, S. and Grimm, V. (2011). Agent-Based and modeling in biology. new insights into cancer illustrate
Individual-Based Modeling: A Practical Introduction. how mathematical tools are enhancing the understand-
Princeton University Press. ing of life from the smallest scale to the grandest. Amer-
ican Scientist, 95(2):134–142.
Ratzé, C., Gillet, F., Müller, J.-P., and Stoffel, K. (2007).
Simulation modelling of ecological hierarchies in con- Seal, J., Alverdy, J., Zaborina, O., An, G., et al.
structive dynamical systems. Ecological Complexity, (2011). Agent-based dynamic knowledge representa-
4(1–2):13–25. tion of pseudomonas aeruginosa virulence activation in
the stressed gut: Towards characterizing host-pathogen
Raub, W., Buskens, V., and van Assen, M. A. (2011). interactions in gut-derived sepsis. Theoretical Biology
Micro-macro links and microfoundations in sociology. and Medical Modelling, 8(1):33.
The Journal of Mathematical Sociology, 35(1-3):1–25.
Seck, M. and Honig, H. (2012). Multi-perspective mod-
Razavi, S., Gaud, N., Mozayani, N., and Koukam, A. elling of complex phenomena. Computational & Math-
(2011). Multi-agent based simulations using fast mul- ematical Organization Theory, 18(1):128–144.
tipole method: application to large scale simulations of
flocking dynamical systems. Artificial Intelligence Re-
Seidl, R., Rammer, W., Scheller, R., and Spies, T.
view, 35(1):53–72.
(2012). An individual-based process model to simulate
Rejniak, K. and Anderson, A. (2011). Hybrid models of landscape-scale forest ecosystem dynamics. Ecological
tumor growth. Wiley Interdisciplinary Reviews: Systems Modelling, 231:87–100.
Biology and Medicine, 3(1):115–125.
Seidl, R., Rammer, W., Scheller, R., Spies, T., and Lexer,
Resnick, M. (1994). Turtles, termites, and traffic jams: Ex- M. (2010). A mechanistic, individual-based approach
plorations in massively parallel microworlds. Mit Press. to modeling complexity and dynamics of forest ecosys-
tems across scales. The 95th ESA Annual Meeting.
Rounsevell, M., Pedroli, B., Erb, K., Gramberger, M.,
Busck, A., Haberl, H., Kristensen, S., Kuemmerle, T., Semeniuk, C., Musiani, M., and Marceau, D. (2011). Biodi-
Lavorel, S., Lindner, M., et al. (2012). Challenges for versity, chapter Integrating Spatial Behavioral Ecology
land system science. Land Use Policy. in Agent-Based Models for Species Conservation, pages
3–26. InTech.
Sawyer, R. (2001). Simulating emergence and downward
causation in small groups. In Multi-agent-based simu- Servat, D. (2000). Modélisation de dynamiques de flux par
lation, Lecture Notes in Artificial Intelligence, pages agents. Application aux processus de ruissellement, infil-
49–67. Springer. tration et érosion. PhD thesis, Université Paris VI.

25
Servat, D., Perrier, E., Treuil, J.-P., and Drogoul, A. Soyez, J.-B., Morvan, G., Dupont, D., and Merzouki, R.
(1998a). When agents emerge from agents: Introduc- (2013). A methodology to engineer and validate dy-
ing multi-scale viewpoints in multi-agent simulations. namic multi-level multi-agent based simulations. In
In Multi-Agent Systems and Agent-Based Simulation, vol- Multi-Agent-Based Simulation XIII, volume 7838 of Lec-
ume 1534 of Lecture Notes in Computer Science, pages ture Notes in Artificial Intelligence, pages 130–142.
183–198. Springer. Springer.

Soyez, J.-B., Morvan, G., Merzouki, R., Dupont, D., and

Servat, D., Pierrer, E., Treuil, J., and Drogoul, A. (1998b).
Kubiak, P. (2011). Multi-agent multi-level modeling – a
Towards virtual experiment laboratories : How multi-
methodology to simulate complex systems. In Proceed-
agent simulations can cope with multiple scales of anal-
ings of the 23rd European Modeling & Simulation Sympo-
ysis and viewpoints. In Virtual Worlds, volume 1434
sium.
of Lecture Notes in Computer Science, pages 205–217.
Springer. Squazzoni, F. (2008). The micro-macro link in social sim-
ulation. Sociologica, 1.
Sewall, J., Wilkie, D., and Lin, M. C. (2011). Interactive
hybrid simulation of large-scale traffic. ACM Trans- Steiniger, A., Kruger, F., and Uhrmacher, A. M. (2012).
action on Graphics (Proceedings of SIGGRAPH Asia), Modeling agents and their environment in multi-level-
30(6). DEVS. In Proc. of the Winter Simulation Conference,
pages 1–12. IEEE.
Sharpanskykh, A. and Treur, J. (2011a). Group abstrac- Stiegelmeyer, S. M., Giddings, M. C., et al. (2013). Agent-
tion for large-scale agent-based social diffusion models. based modeling of competence phenotype switching in
In International conference on Privacy, security, risk and bacillus subtilis. Theoretical Biology and Medical Mod-
trust and international conference on social computing, elling, 10(1):23.
pages 830–837. IEEE.
Sun, T., Adra, S., Smallwood, R., Holcombe, M., and Mac-
Sharpanskykh, A. and Treur, J. (2011b). Group abstrac- Neil, S. (2009). Exploring hypotheses of the actions of
tion for large-scale agent-based social diffusion models tgf-β1 in epidermal wound healing using a 3d computa-
with unaffected agents. In Agents in Principle, Agents tional multiscale model of the human epidermis. PloS
in Practice, volume 7047 of Lecture Notes in Artificial one, 4(12):e8515.
Intelligence, pages 129–142. Springer.
Sun, X., Zhang, L., Tan, H., Bao, J., Strouthos, C., and
Zhou, X. (2012). Multi-scale agent-based brain cancer
Shimoni, Y., Nudelman, G., Hayot, F., and Sealfon, S. modeling and prediction of tki treatment response: In-
(2011). Multi-scale stochastic simulation of diffusion- corporating egfr signaling pathway and angiogenesis.
coupled agents and its application to cell culture simu- BMC bioinformatics, 13(1):218.
lation. PloS one, 6(12):e29298.
Tahir, H., Hoekstra, A., Lorenz, E., Lawford, P., Hose, D.,
Shnerb, N., Louzoun, Y., Bettelheim, E., and Solomon, Gunn, J., and Evans, D. (2011). Multi-scale simulations
S. (2000). The importance of being discrete: Life al- of the dynamics of in-stent restenosis: impact of stent
ways wins on the surface. Proceedings of the National deployment and design. Interface Focus, 1(3):365–373.
Academy of Sciences, 97(19):10322–10324.
Taillandier, P., Vo, D.-A., Amouroux, E., and Drogoul, A.
(2010). GAMA: bringing GIS and multi-level capabil-
Siebert, J. (2011). Approche multi-agent pour la multi- ities to multi-agent simulation. In European Workshop
modélisation et le couplage de simulations. Application à on Multi-Agent Systems.
l’étude des influences entre le fonctionnement des réseaux
ambiants et le comportement de leurs utilisateurs. PhD Taillandier, P., Vo, D.-A., Amouroux, E., and Drogoul,
thesis, Université Henri Poincaré-Nancy I. A. (2012). GAMA: A simulation platform that inte-
grates geographical information data, agent-based mod-
Siebert, J., Ciarletta, L., and Chevrier, V. (2010). Agents eling and multi-scale control. In Desai, N., Liu, A., and
and artefacts for multiple models co-evolution. build- Winikoff, M., editors, Principles and Practice of Multi-
ing complex system simulation as a set of interacting Agent Systems, volume 7057 of Lecture Notes in Com-
models. In Proc. of 9th Int. Conf. on Autonomous Agents puter Science, pages 242–258. Springer.
and Multiagent Systems (AAMAS 2010), pages 509–516.
Thorne, B., Hayenga, H., Humphrey, J., and Peirce, S.
(2011). Toward a multi-scale computational model of
Smallwood, R., Adra, S., and Khodabakshi, G. arterial adaptation in hypertension: verification of a
(2010). Multiscale, multi-paradigm modelling multi-cell agent based model. Frontiers in Physiology,
of tissues: embedding development in tissue be- 2.
haviour. http://www.dcs.shef.ac.uk/~rod/
integrative_biology/Smallwood_Cellular_ Torii, D., Ishida, T., Bonneaud, S., and Drogoul, A. (2005).
dynamics_and_tissue_behaviour.pdf. Layering social interaction scenarios on environmental
simulation. In Multi-Agent and Multi-Agent-Based Simu-
Smallwood, R. and Holcombe, M. (2006). The epithe- lation, volume 3415 of Lecture Notes in Artificial Intelli-
liome project: multiscale agent-based modeling of ep- gence, pages 78–88. Springer.
ithelial cells. In Biomedical Imaging: Nano to Macro,
Tranouez, P. (2005). Contribution à la modélisation et à la
2006. 3rd IEEE International Symposium on, pages 816–
prise en compte informatique de niveaux de descriptions
819. IEEE.
multiples. Application aux écosystèmes aquatiques (Peni-
cillo haere, nam scalas aufero). PhD thesis, Université
Solovyev, A., Mikheev, M., Z., L., Dutta-Moscato, J., du Havre.
Ziraldo, C., An, G., Vodovotz, Y., and Mi, Q.
(2010). SPARK: A framework for multi-scale agent- Tranouez, P., Bertelle, C., and Olivier, D. (2001). Chang-
based biomedical modeling. International Journal of ing the level of description of a fluid flow in agent-based
Agent Technologies and Systems, 2(3):18–30. simulation. In ESS Conference.

26
Tranouez, P., Bertelle, C., and Olivier, D. (2006). Emer- Wang, Z., Birch, C., and Deisboeck, T. (2008). Cross-
gent properties in natural and artificial dynamical sys- scale sensitivity analysis of a non-small cell lung cancer
tems, chapter Changing levels of description in a fluid model: linking molecular signaling properties to cellu-
flow simulation, pages 87–99. Springer. lar behavior. Biosystems, 92(3):249–258.

Tranouez, P. and Dutot, A. (2009). Different goals in mul-

Wang, Z. and Deisboeck, T. (2008). Computational mod-
tiscale simulations and how to reach them. In Complex
eling of brain tumors: discrete, continuum or hybrid?
Systems and Self-organization Modelling, Understanding
Scientific Modeling and Simulation, 15(1):381–393.
Complex Systems, pages 29–39. Springer.

Tranouez, P., Lerebourg, S., Bertelle, C., and Olivier, D. Wang, Z., Zhang, L., Sagotsky, J., and Deisboeck, T.
(2003). Contribution à la représentation multi-échelle (2007). Simulating non-small cell lung cancer with a
des écosystèmes aquatiques. Technique et Science Infor- multiscale agent-based model. Theoretical Biology and
matiques, 22(4):255–259. Medical Modelling, 4.
Treuil, J.-P., Drogoul, A., and Zucker, J.-D. (2008). Modéli-
sation et simulation à base d’agents. Dunod. Wedde, H. and Senge, S. (2012). 2-way evaluation of
the distributed BeeJamA vehicle routing approach. In
Uhrmacher, A. M., Ewald, R., John, M., Maus, C., Proceedings of the IEEE Intelligent Vehicles Symposium,
Jeschke, M., and Biermann, S. (2007). Combining mi- pages 205–210.
cro and macro-modeling in DEVS for computational
biology. In Proceedings of the 39th conference on Winter White, B. (2007). On interpreting scale (or view) and emer-
simulation, pages 871–880. gence in complex systems engineering. In Systems Con-
ference, 2007 1st Annual IEEE, pages 1–7. IEEE.
Vincenot, C., Giannino, F., Rietkerk, M., Moriya, K., and
Mazzoleni, S. (2011). Theoretical considerations on the
combined use of system dynamics and individual-based Xi, H. and Son, Y.-J. (2012). Two-level modeling frame-
modeling in ecology. Ecological Modelling, 222(1):210– work for pedestrian route choice and walking behav-
218. iors. Simulation Modelling Practice and Theory, 22:28–
46.
Vo, A. (2012). An operational architecture to handle mul-
tiple levels of representation in agent-based models. PhD Zahedmanesh, H. and Lally, C. (2012). A multiscale
thesis, Université Paris VI. mechanobiological modelling framework using agent-
based models and finite element analysis: application
Vo, D.-A., Drogoul, A., and Zucker, J.-D. (2012a). An
to vascular tissue engineering. Biomechanics and Model-
operational meta-model for handling multiple scales in
ing in Mechanobiology, 11:363–377.
agent-based simulations. In International Conference on
Computing and Communication Technologies, Research,
Innovation, and Vision for the Future (RIVF), pages 1–6, Zeigler, B., Kim, T., and Praehofer, H. (2000). Theory of
Ho Chi Minh City. IEEE. Modeling and Simulation. Academic Press, 2nd edition.

Vo, D.-A., Drogoul, A., Zucker, J.-D., and Ho, T.-V. Zhang, L., Athale, C., and Deisboeck, T. (2007). Devel-
(2012b). A modelling language to represent and spec- opment of a three-dimensional multiscale agent-based
ify emerging structures in agent-based model. In Desai, tumor model: Simulating gene-protein interaction pro-
N., Liu, A., and Winikoff, M., editors, Principles and files, cell phenotypes and multicellular patterns in
Practice of Multi-Agent Systems, volume 7057 of Lecture brain cancer. Journal of Theoretical Biology, 244(1):96–
Notes in Computer Science, pages 212–227. Springer. 107.
Vodovotz, Y., Csete, M., Bartels, J., Chang, S., and An, G.
(2008). Translational systems biology of inflammation. Zhang, L., Chen, L., and Deisboeck, T. (2009a). Multi-
PLoS Comput Biol, 4(4):e1000014. scale, multi-resolution brain cancer modeling. Mathe-
matics and computers in simulation, 79(7):2021–2035.
Wakeland, W., Macovsky, L., and An, G. (2007). A hybrid
simulation model for studying acute inflammatory re-
Zhang, L., Jiang, B., Wu, Y., Strouthos, C., Sun, P., Su, J.,
sponse. In Proceedings of the 2007 spring simulation mul-
and Zhou, X. (2011). Developing a multiscale, multi-
ticonference, volume 2 of SpringSim ’07, pages 39–46.
resolution agent-based brain tumor model by graphics
Society for Computer Simulation International.
processing units. Theoretical Biology and Medical Mod-
Wang, J., Zhang, L., Jing, C., Ye, G., Wu, H., Miao, elling, 8(1):46.
H., Wu, Y., and Zhou, X. (2013). Multi-scale agent-
based modeling on melanoma and its related angiogen- Zhang, L., Wang, Z., Sagotsky, J., and Deisboeck, T.
esis analysis. Theoretical Biology and Medical Modelling, (2009b). Multiscale agent-based cancer modeling. Jour-
10(1):41. nal of Mathematical Biology, 48(4–5):545–559.

27