Conference PaperPDF Available

Reification of emergent urban areas in a land-use simulation model in Reunion Island

Authors:

Abstract and Figures

Emergent phenomena are often relevant for users and developers of simulation models. But the potential reification of these phenomena raises many questions, conceptually (should they be reified?) and technically (how to do it?). In this paper, we show that such a reification can be considered as an effective way to refine simulation models in which direct modifications, that are made laborious by the multiplicity of the entities and behaviors, often leads to the destabilization of the entire system. We propose a reification technique of the emergent phenomena that do emerge in an agent-based simulation. We illustrate this proposition through the reification of new urban areas, an emergent phenomenon observed in a model that we created to simulate land-use evolutions in Reunion Island.
Content may be subject to copyright.
Reification of emergent urban areas in a land-use
simulation model in Reunion Island
Daniel David1, Yassine Gangat2, Denis Payet2, Rémy Courdier2
1 CIRAD, UPR 78 Recyclage & Risque, Station de la Bretagne, Reunion Island
daniel.david@cirad.fr
2 Université de La Réunion, EA 2525 LIM IREMIA, Reunion Island
{yassine.gangat, denis.payet, remy.courdier}@univ-reunion.fr
Abstract. Emergent phenomena are often relevant for users and developers of
simulation models. But the potential reification of these phenomena raises many
questions, conceptually (should they be reified?) and technically (how to do
it?). In this paper, we show that such a reification can be considered as an effec-
tive way to refine simulation models in which direct modifications, that are
made laborious by the multiplicity of the entities and behaviors, often leads to
the destabilization of the entire system. We propose a reification technique of
the emergent phenomena that do emerge in an agent-based simulation. We il-
lustrate this proposition through the reification of new urban areas, an emergent
phenomenon observed in a model that we created to simulate land-use evolu-
tions in Reunion Island.
Keywords: Agent-Based Simulation, Emergence, Reification, Urbanization,
Land-use planning
1 Introduction
Emergence is a fascinating concept for scientists from different backgrounds. In the
context of modeling and simulation, it is often known as a concept encouraging the
choice of Multi-Agent Systems (MAS) in comparison to other existing techniques.
Thus, lots of works have allowed definitions and classifications of emerging phenom-
ena observed in a system, while some of them have tackled the question of their po-
tential reification.
We consider that today the problem is not really to succeed in reifying a known
phenomenon, but it is to answer to why its reification should be done or not, and what
are the steps that can lead such a reification. Therefore we can legitimately think that
there are some cases where it will be useful, while there are other cases where its
usefulness remains more uncertain. Consequently, the initial question is what are the
cases where the reification of potential emerging phenomena takes an interest, and we
believe that this knowledge relies on the context in which each study is done.
In this paper, we briefly present the DS Model, a model that simulates land-use
evolutions in Reunion Island. This model is the result of the work of many researchers
since 2007. We then present the new urban areas, an emergent phenomenon that is
regularly observed in our simulation results. We propose a general architectural
framework that will lead us to the reification of such a phenomena in Agent-Based
Simulations (ABS), and we illustrate this proposal with new urban areas in DS.
2 Reunion Island and the DS Model
Reunion Island is a French territory of 2500 km2 in the western Indian Ocean. It has a
strong growth in a limited area with an actual population of 800,000 inhabitants that
will probably be more than 1 billion in 2030. This simple demographic change opens
the door to many issues, including housing: even when assuming high densification
hypothesis, the demand for urban land will increase of several thousand hectares.
The evolution of this territory must then be done according to a clear urbanization
policy and planning documents regulating the evolution of urbanization of the island
should take into account these projections within the bounds of possibility, as it is
evident that a rule-less urbanization is not a viable long-term scenario. In addition,
100,000 hectares of natural areas of the island are since 2010 included in the
UNESCO World Heritage due to their beautiful landscapes and their amazing biodi-
versity potential. And 40,000 hectares, including historical sugarcane areas, are used
by agricultural activities that need to be at least preserved.
So, as noted in [1], in terms of land-use planning, Reunion Island must take up the
challenge of hosting a growing population while developing its agricultural land and
protecting its natural areas and outstanding landscape. In such a context, and in order
to fill the blank in terms of tools dedicated to land-use foresight [6], the implementa-
tion of the DS Model (named by the contraction of the names Domino and Smat, the
two projects that led to its realization) has been initiated in 2006 [3, 6]. This model is
fruit of a collaboration between many partners [2, 7], researchers (CIRAD, Reunion
Island University, IRD, ...) and decision-makers (Reunion Island Regional Council).
This model can simulate at the same time the evolution of the population and the
land-use (urban, agricultural, natural) changes on the island territory.
Fig.1. A simulation with the DS Model (Initial state in 2003 on the left, Final state in 2030 on
the right). Urban areas are in red, agricultural areas in yellow and natural areas in green.
Implemented on the simulation platform GEAMAS-NG, its successive develop-
ments have made it a model in which there are a large number of entities whose be-
haviors and interactions are rich and varied: hundreds of thousands of Parcels agents
(land units of approximately one hectare) live together with agents representing the
different institutional layers (1 Region agent, 4 Micro-regions agents, 24 Cities
agents, ...). Results from simulations are used to illustrate (e.g. in the form of carto-
graphic outputs such as in Fig. 1) locally large scenarios of land use.
3 Emergence of new urban areas
When the DS Model is used to perform some simulations, whatever the scenarios, the
first glance shows us the very important part of urbanization in the island. Indeed, the
population of Reunion Island is such increasing that even with strong assumptions of
housing densification (configurable in DS), the need for housing and various con-
structions related to this increase (business parks, commercial areas, ) inevitably
increase.
When studying more closely the simulation results, we realize that some specific
urban areas appear over time: areas that are urbanized rapidly, during just a few simu-
lation cycles and that correspond to territories very concentrated, once considered to
be natural or agricultural areas. Fig. 2 gives a visual example, illustrating such urbani-
zation. On this portion of island's territory, we can see the evolution of the Cambaie
zone, an area formerly preserved due to some regulations, and whose potential use is
such that it will surely be fully occupied in a few years by constructions.
Fig. 2. A simulation on the Cambaie area (Initial state in 2003 on the left, Final state in 2030 on
the right). Urban areas are in red, agricultural areas in yellow and natural areas in green.
In the simulation from which this figure was extracted, the Parcels of land (represent-
ed in DS by agents at a microscopic scale) are sized 4 hectares each. At the top is Le
Port area, already highly urbanized, and at the bottom is the area of Saint-Paul. We
can note that during the simulation these two areas continue to urbanize as red Par-
cels were added to existing ones. But these Parcels newly urbanized usually just fill
(restricted) spaces that were not urbanized yet. They were already in an urban envi-
ronment.
However, the case of urbanization of the central area called Cambaie is different,
since the urbanized cells were a minority in the beginning of simulations. But we can
see that they are urbanizing together in a restricted temporal space.
This kind of phenomena is locally well-known as many areas of this type can be
spotted on the island's territory. They generally correspond to areas for which regula-
tions have been modified in the planning documents: PLU (Local Urbanization Plans,
at the communal scale), SCoT (Territorial Coherence Schemes, at the micro-regions
scale) and SAR (Regional Planning Scheme, at the regional scale). This is especially
what happens when large tracts of agricultural land are degraded, making them build-
able for rapid urbanization, whether for homes designed to fill the need for new hous-
ing or facilities, warehouses, halls, which will appear for companies setting up in new
areas of activity.
Later in this paper, we will consider the virtual development of an area of Saint-
Pierre. This area, still virgin a few years ago, is now occupied by many commercial
buildings, and, according to our simulations, seems designated to experience greater
urbanization. A sample output of a simulation (from 2003 to 2030) over an area con-
taining this plot is given in Fig. 3.
Saint-Pierre Saint-Pierre
Fig. 3. A simulation on the Saint-Pierre area (Initial state in 2003 on the left, Final state in 2030
on the right). Urban areas are in red, agricultural areas in yellow and natural areas in green.
In this figure, there is a significant development of urbanization. But compared to the
Cambaie sample, it is more difficult here to visually distinguish and to define the
emerging phenomena corresponding to a new urban area (or even just to say whether
we are witnessing such a phenomenon). If we wish to observe such phenomena, it
requires the use of detection techniques other than the simple look of the experts who
focus on the results.
But when we faced this type of phenomena, we often have (legitimately?) the de-
sire to go beyond their “simple” detection and approach the broader issue of their
reification.
4 Reification of emerging phenomena
Reification of emergent phenomena is a subject often mentioned in works related
directly or indirectly to the emergence, particularly in the MAS community, but
which is rarely defined. In our case, we consider that the reification of an emergent
phenomenon in an ABS is a process that takes place in two phases (possibly dissoci-
ated): a detection phase and a phase of materialization [4, 5]. This process raises
many conceptual and technical questions.
4.1 To reify or not to reify?
We can legitimately wonder about the validity of the simple desire of reification of an
emergent phenomenon. Of course, experts and users of a model have nothing to lose
(but everything to gain?) when they hope to detect any emergent phenomena, because
they will improve their knowledge on the model studied and thereby on the real sys-
tem modeled. Moreover, as emergent phenomena are often considered part of the
expected results of a simulation, therefore they should be put forward.
We will not discuss here on the various software techniques that can detect emer-
gent phenomena, they are numerous (research techniques pattern [12], techniques
based on building of interaction graphs [8], techniques based on emergence laws and
emergence revelators [4, 5]) if we consider (as in our case) that an emerging phenom-
ena is only contingent of the eye that looks at it and the level of expertise associated
with it. For example, a given emergent phenomena would be obvious for a geographer
but would not even exist in the eyes of an economist (or vice versa, obviously). It will
be the same in virtual systems in which emergent phenomena can be detected easily
with the degree of knowledge of the system itself (or entities or mechanisms respon-
sible of that detection).
But regarding the materialization phase, which would fill out the reification of an
emergent phenomena, the question of its being arises for good reasons. Indeed, on a
philosophical level, if one seeks to give shape to an emergent phenomenon within a
system in which it would have emerged, doesn't it lose its emergent nature? Moreo-
ver, reifying an emergent phenomenon in a system also means that we tend to change
the original model with the risk of destabilizing it and lose its essence, this very one
which leads to the emergence of the phenomena considered.
The choice to complete the reification of emergent phenomena potentially detected
is a choice we should not trifle with. If this choice is made, we must ideally do this
reification without destabilizing the initial model produced and implemented, as long
as this implementation does not require a thorough look to make possible the desired
reification (which is often the case in large scale projects). In order to do this, we
propose the use of special emergent structures that allow materialization of the
emerging phenomena in an ABS.
4.2 Emergent structures
In general, emergent phenomena detected in the real world often manifest behaviors
that make the very existence of these phenomena affect the real world entities: some
of these entities may participate directly in the emergence of phenomena, while others
are influenced by these phenomena, and still others have their perception modified by
the presence of these phenomena. This is obviously the same in the virtual world that
we are handling in an ABS, since our goal is to reproduce the phenomena that occur
in systems or processes of the real world. Thus it is important to offer solutions in
order to represent these phenomena in the architecture of an ABS platform. That's
why we propose to use two types of emergent structures: emergence agents and in-
terposition elements, which are shown in Fig. 4.
An emergence agent is an intelligent agent that runs within an ABS platform. It
evolves in the same environment(s) as all other agents in the system and interacts with
them through mechanisms of influence and perception that underlie the host platform.
Several agents can be created for reifying the same emergent phenomenon.
An interposition element is a structure allowing change of one or more agents
from one or more environments in which they operate. Such a structure modifies (as
appropriate by altering them, improving them, restricting them, etc.) mechanisms of
perception or influence used by the agents of the ABS.
Interposition
element
Emergence
agent
Environnement
Unmodified
Agent
modified
influences
perceptions unmodified
influences
perceptions
Unmodified
Agent
Unmodified
Agent
Fig. 4. Emergent structures (an emergence agent and an interposition element) with modified
perceptions/influences
Both types of structures, seen as complementary or independent, allow us to take into
account different types of phenomena that occur in the studied ABS. Thus, in an ex-
ample of intrinsic emergence (as defined in the classification of Boschetti [2]), like
the apparition of a school of fish. The school of fish will be represented directly in the
system by the emergence agent. And the different fishes that constitute the school of
fish will continue to move in their environment but will have their perceptions and
influences changed by interposition elements.
In the same way of thinking, we can include an example of low emergence (as it is
defined in [11]): twigs (objects in the environment), that have been stockpiled by
termites agents, making emergence of a pile of stick. Here the woodpile does not have
its own behavior, there is therefore no need to create an emergence agent to represent
it. However, if certain entities of the system must perceive the woodpile as such, this
will be possible through interposition elements that will change perceptions and influ-
ences of these entities.
One of the real benefits of this technique is that it does not require to modify the
code of the agents involved in our emergent phenomena. Changes of these agents’
behaviors are only a side effect due to the presence of emergent structures related to
them. Therefore, the agents that are not concerned by the emergence of a particular
phenomenon (which generally constitute the vast majority of agents in the ABS)
would never be destabilized.
4.3 In the GEAMAS-NG platform
To make possible the desired reification of emergent phenomena and to materialize
them in an ABS in the form of emergence agents and/or interposition elements, the
platform running the simulation should provide some services in order to facilitate the
work of designers and developers. We have identified three main services:
An observation service, which will help us to place various mechanisms that al-
low observation to obtain knowledge on the ABS and thus detect any emerging
phenomenon that have been described by experts.
A handling service for agents, which enables creation, suppression, and modifi-
cation of the agents' life cycle.
A handling service for object of the environment, which allows the establish-
ment of elements of interposition.
The general framework of this proposal allows the expression of behaviors associated
with emergent structures within the ABS. It refines the initial knowledge from the
simulated system and therefore, in a process of knowledge production, from the real
or theoretical complex system studied.
In the GEAMAS-NG platform (GEneric Architecture for MultiAgent Simulation
New Generation) we have established a clear separation between agents and environ-
ments, in which agents interact through mechanisms of influences and perceptions.
The platform was equipped with a set of services for the observation of the simula-
tion. Therefore it has all the desired characteristics for reification of emergence in
simulated models. The detection of wanted emergent phenomena is done in accord-
ance with the emergence laws described in [4] and integrating observation mecha-
nisms described in [9] and [10].
The potential creation and/or suppression of any emergence agent is done in the
platform's core. The establishment of any interposition elements was made possible
by the functionality offered by GEAMAS-NG which lets you define multiple envi-
ronments within a single simulation (as shown by our dynamic-oriented modeling
[9]). It allows us to use a particular environment to model each potential interposition
element associated to a given emergent phenomena.
5 The reification of urban areas in DS
The phenomenon of new urban areas previously described corresponds, in terms of
urbanization, to a real emergent phenomenon that requires further study. In particular,
it would be interesting to allow the detection of such phenomena in the DS Model.
Thus they would be noticed before the (meticulous and fastidious) analysis of experts
from the results and maps generated by the simulations. And it would also be interest-
ing to consider the emergence of these new urban areas in the system to test various
hypotheses of urbanization associated with them and the consequences they induce.
Obviously, although the Parcels agents that compose it have all an urban state,
every new urban areas are not considered in the same way. We can easily imagine
that their potential behaviors may differ depending whether the recent urbanization is
for example housing areas or business parks. We can therefore use the emergent
structures that we have defined in order to detect new emerging urban areas in the DS
Model and to materialize them, so we would be able to experiment different scenarios
and assumptions.
In the following part, we will therefore describe the detection and materialization
phases whose realization leads to the reification of the considered phenomenon.
5.1 Detection phase in the DS Model
The first stage of the reification process of new urban areas in the DS Model is to be
able to detect the formation of these zones during simulations. This is done through
platform mechanisms we have implemented in GEAMAS-NG to which we (as users
of the system) must give elements to describe the emergent phenomenon as a new
urban area. Thus, Fig. 5 illustrates a new urban area, appeared in the simulation
shown in Fig. 3, which is composed of thirty Parcels agents. All of them were detect-
ed using an indicator to detect the emergence of Parcels that are at least 5 in number,
in the same proximity, and had their urbanization performed in the same time period
of 5 years. Geographer experts and specialists in urbanization have indicated these
numerical values to us during the experimentation process.
City of Saint-Pierre
Saint-Pierre
Area of the emergent phenomenon
Fig. 5. The emerging phenomena “new urban areas” that has been detected (in pink) with the
actual aerial view of the area of our emerging phenomena (from Google Maps)
In this figure, we can notice the set of cells that are grouped together in the detect-
ed new urban area. This area is also well known locally, because it is a recent ZAC (a
local activities development zone) of Saint-Pierre: the Canabady ZAC). We can also
easily see the town of Saint-Pierre, already highly urbanized, and the little urbanized
area in which our emergent phenomenon will occur in our simulations (the circled
area), but in which there are already the first buildings of the Canabady ZAC on the
right part. In this aerial view, we can also note many farmland areas, cultivated, that,
according to the choices made in the different simulation scenarios, could be devoted
to urbanization in the coming years.
This detection of new urban areas that are likely to appear in the simulations with
the DS Model, allows us to provide support to experts specialists dealing with the
analysis of the simulation results. In that sense, this experiment is therefore an im-
portant proposition of progress by the possibilities offered by the stable and utilized
versions of the model. But to go beyond the “simple” assisted detection, we will now
show how the new urban areas that have emerged can be materialized in the ABS.
5.2 Materialization phase in the DS Model
The first thing to be done, in order to materialize new urban areas that emerge in a
simulation, is to examine how this phenomenon will be integrated into the ABS,
through emergent structures. We can also ask ourselves, according the value we want
to give to the phenomenon, if it must be materialized using only interposition ele-
ments, if we should use an emergence agent only, or if we should move towards a
joint use of both types of structures.
In our example of new urban areas, we began by analyzing how the system entities
are involved in this phenomenon. Naturally, there are objects of the environment and
agents (Parcels, Cities, Micro-regions, …) which are within its geographical area and
are directly concerned by the emergence of a new urban area, while agents and ob-
jects that are quite far from it and are not directly involved in its emergence. If we
want the materialization of new urban areas in the ABS to be useful, it's evident that
we must at least establish interposition elements with Parcels agents who are con-
cerned with the emergent phenomenon. This will allow us to test different assump-
tions affecting the evolution of these agents that were internal to the DS Model, with-
out editing them directly.
But we should bear in mind that the smallest entities are not the only ones affected
by the emergence of a new urban area. Indeed, the DS Model is composed of different
levels of agents and each new urban area emerges within a particular City, in particu-
lar Micro-region, and inside a global Region itself. So we must take this into account
for the corresponding agents and implement elements of interposition for the Region
agent, for each Micro-region agent and for each City agent concerned by the emergent
phenomenon.
Finally, we can materialize our new urban areas by establishing for each new urban
area, an emergence agent that will assign a specific behavior to the phenomenon and
that can interact with the environment of the DS Model via its own influences and
perceptions. Again, this will allow us to test various hypotheses of evolution.
In our experiment, we chose to reify each new urban area that would emerge in a
simulation using:
An emergence agent.
An interposition element used for Micro-regions agents (in our example only for
the Southern Micro-region agent, but it is possible that a new emerging urban
area emerges on the territories of several Micro-regions agents).
An interposition element used for Cities agents (in our example only the Saint-
Pierre City agent, but it is possible that a new emerging urban area emerges on
the territories of several Cities agents).
These interposition elements are enough in our experiments, as we consider reasona-
ble to assume that all agents of the same kind that are involved in an emerging phe-
nomenon will be affected the same way. However, it seems obvious that agents of
different types (and scales) will be affected differently. In our example, these choices
can be explained because the Southern Micro-region agent and the Saint-Pierre City
agent are those in middle of the urbanization process for the new urban area detected
at Saint-Pierre.
Regarding the emergence agent representing the phenomenon, it will help us,
through the behavior that will be given, to test various hypotheses in order to refine
the general behavior of the DS Model in relation to the emergence of this particular
phenomenon. This intelligent will be the Urbanization Manager agent of the area.
5.3 Main Results
The main interest that emerges from the reification process of new urban areas is to
help users of the DS Model. They could test different hypotheses that can refine the
behavior of the model in order to reflect the specific requirements of the phenomena
that have been put forward.
Indeed, the DS Model allows, in its original version, to take into account the be-
havior at the scale of the Parcel, the Region and the Micro-regions. It is clear that the
hundreds of thousands of small Parcels agents, that all have specific characteristics
related to their location, have therefore a sufficient precision to assume that the treat-
ments they perform are adequate on these specificities. But for agents of larger scale,
such as Cities agents, which are in the middle of the hierarchy of Par-
cels/Cities/Micro-Regions/Region agents, the initial expected behavior in the DS
Model are sometimes too general to consider specificities such as new urban areas.
With the emergent structures we established, we can indirectly alter the evolution
of the entities of the DS Model. In our example, the conducted experiments allow us
to test hypotheses in order to refine the behavior of the Saint-Pierre City agent in the
area of our new urban area. Concretely, if we want to make this area a commercial
area rather than a housing area, the interposition elements used with the Saint-Pierre
City agent and the Southern Micro-region agent allow us to hide from them the Par-
cels agents affected by the new urban area. For example, they can no longer consider
them when they are looking forward to allocate the new population calculated on the
territory. And the Urbanization Manager agent of the new urban area can change the
objects of the environment that are in the territory of the new urban area, for example
to increase the parameter of attractiveness of the land associated with them, which
would then simulate a faster urbanization.
Fig. 6 illustrates the results of two simulations where the exact same rapid urbani-
zation has been observed and where we have detected the new urban area present in
the city of Saint-Pierre. But we materialized them in two different ways. The two
images have a gradation of red to represent the population density observed at the end
of both simulations:
In the first case (on the left), the new urban area is materialized by considering
that it would be mostly dedicated for residential units.
In the second case (on the right), the new urban area is materialized by consider-
ing that it would be mostly dedicated to commercial and business buildings.
Area of the emergent phenomenon
Saint-Pierre Saint-Pierre
Fig. 6. The population density observed in the results of both simulations (on the left the new
urban area is materialized by considering that it is for housing, on the right it is materialized by
considering that it is for business premises)
We can note the difference in shades of red in the encircled area, indicating logically
that people began to be distributed over the area when its urbanization in the first case
but not in the second one, where for which, disturbed by elements of interposition, the
Southern Micro-region agent and the Saint-Pierre City agent have not assigned new
population to Parcels agents contained in the area. In all this experiment, the behavior
of the system has change while the behaviors (and so the code) of the Parcels, Micro-
Regions, Cities, and Region agents have never been modified directly.
6 Conclusion
In this paper, we studied the problem of reification of phenomena that emerge in an
ABS. To that end, we presented the DS Model, a model that allows us to simulate
land-use evolution in Reunion Island. We focused on the study of new urban areas, a
particular phenomenon that emerges in many simulations. As an experiment, we have
detailed how it was possible to make the process of reification of these new urban
areas by relying on the use of emergent structures that we have defined and mecha-
nisms we have implemented in the GEAMAS-NG platform.
These proposals are not aimed to deliver the solution to take into account any
emergent phenomenon in an ABS, because to achieve the reification of a phenome-
non, we must, as we have seen through the experimentation of new urban areas and
particularly during their phase of materialization, go through stages (which are some-
times fastidious) of analysis, modeling and programming of the emergent structures
constituted by interposition elements and emergence agents.
But the sequence of this experiment shows that our approach allows integrating the
consideration of emergent phenomena in simulation models in which it was not antic-
ipated. And we can extend the functionality of models like the DS Model (whose
complex structure makes difficult any changes of behaviors of certain entities, with-
out causing a global imbalance) to refine the general behavior, in order to reflect new
specificities. This puts us therefore in the middle of the processes of injection and
production of knowledge in and through a simulation.
1. Agorah, Agence d'urbanisme de La Réunion, Une politique foncière, une des clefs pour
aménager le territoire (2006)
2. Boschetti, F., Gray, R., A Turing test for emergence, Advances in applied self-organizing
systems, pp. 349-369, (2007)
3. Botta, A., Daré, W., Antona, M., Leclerc, G., Integration of multi-scale stakes in govern-
ance by applying companion modelling to land-use foresight, MODSIM'09 International
Congress on Modelling and Simulation, pp. 2377-2383 (2009)
4. David, D., Courdier, R., See Emergence as a Metaknowledge, a Way to Reify emerging
Phenomena in Multiagent Simulations?, International Conference on Agents and Artificial
Intelligence (ICAART'09), pp. 564-569 (2009)
5. David, D., Courdier, R., Emergence as Metaknowledge: Refining Simulation Models
through Emergence Reification, European Simulation and Modelling Conference
(ESM'08), pp. 25-27 (2008)
6. David, D., Prospective territoriale par simulation orientée agent, PhD Thesis, Université de
La Réunion (2010)
7. Lagabrielle, E., Botta, A., Daré, W., David, D., Aubert, S., Fabricius, C., Modelling with
stakeholders to integrate biodiversity into land-use planning - Lessons learned in Réunion
Island, Environmental Modelling and Software, vol. 25, n. 11, pp. 1413-1427 (2010)
8. Moncion, T., Amar, P., Hutzler, G., Automatic characterization of emergent phenomena in
complex systems, Journal of Biological Physics and Chemistry, vol. 10 (2010)
9. Payet, D., Courdier, R., Sébastien, N., Ralambondrainy, T., Environment as support for
simplification, reuse and integration of processes in spatial MAS, Information Reuse and
Integration Conference (IRI'06), pp. 127-131 (2006)
10. Ralambondrainy, T., Observation de simulations multi-agents à large échelle}, PhD The-
sis, Université de La Réunion (2009)
11. Stephan, A., Emergence, Encyclopedia of cognitive science, Vol. 1, pp. 1108-1115 (2009)
12. Tranouez, P., Bertelle, C., Olivier, D., Changing levels of description in a fluid flow simu-
lation, Emergent Properties in Natural and Artificial Dynamical Systems. Understanding
Complex Systems series, pp. 1108-1115 (2006)
... Notre plateforme de base a servi à de nombreuses expérimentations et donné lieu à de multiples enrichissements et validations, intégrés ensuite dans ses nouvelles versions. Les travaux de recherche de l'équipe SMART se sont focalisés sur trois axes : l'observation dans la SOA [Ralambondrainy et al., 2006, Ralambondrainy, 2009, la distribution et la parallélisation de SOA [Sébastien et al., 2008, Sébastien, 2009], la gestion de l'émergence dans la SOA [David et al., 2009, David, 2010, David et al., 2012. ...
... 5 : 230 -239. • 3 articles publiés puis présentés dans des conférences internationales : -[Gangat et al., 2012a]Y. Gangat, D. Payet and R. Courdier. Another step toward reusability in agent-based simulation : Multi-behaviors & MVC, 24th IEEE International Conference on Tools with Artificial Intelligence (ICTAI 2012), Athens, Greece, nov.7-9, 2012.-[David et al., 2012] D. David, Y. Gangat, D. Payet and R. Courdier. Reification of emergent urban areas in a land-use simulation model in Reunion Island, ECAI 8.2. Perspectives utilisation, nous avons proposé l'approche méthodologique MMC. Si nous reprenons l'état de l'art présenté par les auteurs de [Pokahr et Braubach, 2009] (voir la première ligne de la ...
Thesis
Full-text available
La co-construction et la réutilisation de modèles font l'objet de plusieurs travaux dans le domaine de la simulation. Cependant, dans le domaine plus spécifique de la Simulation Orientée Agent (SOA), nous pouvons constater un manque sur ces deux points malgré un besoin fort de la part des thématiciens. La co-construction est essentielle pour optimiser la mise en commun du savoir de différents experts, mais nous faisons souvent face à des divergences de points de vue. Les méthodologies existantes pour la co-construction en SOA ne permettent qu'un faible niveau de collaboration entre thématiciens durant la phase initiale de modélisation, ainsi qu'entre les des thématiciens avec les modélisateurs ou les modélisateurs-informaticiens... Pour faciliter cette co-construction, nous proposons de suivre une méthodologie de conception favorisant cette collaboration. La réutilisation de modèle octroie un gain de temps significatif, une amélioration du modèle et l'apport de nouvelle connaissance. Les méthodologies en SOA dans ce domaine existent. Cependant, dans le spectre de réutilisation, elles sont souvent limitées au niveau du modèle complet ou de l'agent avec l'impossibilité de "descendre" plus bas. L'expérience de EDMMAS, un cas concret d'un modèle issu de trois réutilisations successives, nous a permis de constater une nouvelle complexité qui découle de la démultiplication des comportements des agents et crée un décalage conséquent entre le modèle opérationnel et le modèle conceptuel. Notre objectif est de promouvoir la réutilisation aussi bien des modèles, que des agents et de leurs comportements.Pour répondre à ces questionnements, nous proposons dans ce manuscrit une manière de codifier et d'intégrer la connaissance provenant de disciplines différentes dans le modèle, tout en utilisant des modules "composables" qui facilitent la réutilisation. Nous proposons (i) une nouvelle architecture Agent (aMVC), appliquée dans un cadre multidynamique (DOM), avec l'appui (ii) d'une approche méthodologique (MMC) basée sur la décomposition et réutilisation des comportements. Cet ensemble de propositions, (i) et (ii), permet de conduire un projet pluridisciplinaire de SOA avec un grand nombre d'acteurs, facilitant la co-construction des modèles grâce à l'instauration de nouvelles synergies entre les différents acteurs participant à la modélisation. Les concepteurs pourront travailler de manière autonome sur leur dynamique et la plateforme fera l'intégration de ces dernières en assurant la cohésion et la robustesse du système. Nos contributions offrent la capacité de créer les briques élémentaires du système de manière indépendante, de les associer et de les combiner pour former des agents, selon des dynamiques conformément à l'approche DOM. Elles permettent ainsi de comparer la logique selon différentes possibilités pour une même dynamique et d'ouvrir la perspective d'étudier un grand nombre d'alternatives de modélisation d'un même système complexe, et de les analyser ensuite à une échelle très fine. Co-building and reuse of models are at the center of several studies in the field of simulation. However, in the more specific field ofMulti-Agent Based Simulation (MABS), there is a lack of methodology to resolve these two issues, despite a strong need by experts.Model co-building is essential to optimize knowledge sharing amongst different experts, but we often face divergent viewpoints. Existing methodologies for the MABS co-building allow only a low level of collaboration among experts during the initial phase of modeling, and between domain experts with modelers or computer scientists... In order to help this co-building, we propose and follow a methodology to facilitate this collaboration. Model reuse can provide significant time savings, improve models’ quality and offer new knowledge. Some MABS methodologies in this area exist. However, in the spectrum of reuse, they are often limited to a full model’s reuse or agent’s reuse with the impossibility of reusing smaller parts such as behaviors. The EDMMAS experiment was a concrete case of three successive model reuses. It allowed us to observe new complexity arising from the increase of agents’ behaviors. This creates a gap between operational model and conceptual model.Our goal is to promote the reuse of models, agents and their behaviors.To answer these questions, we propose in this thesis a new way to codify and integrate knowledge from different disciplines in the model, while using "composable"modules that facilitate reuse.We propose (i) a new agent architecture (aMVC), applied to a multidynamical approach (DOM), with the support (ii) of a methodology (MMC) based on the decompositionand reuse of behaviors.Proposals (i) and (ii) allow us to lead a multidisciplinary MABS project with a large number of actors, helping the co-building of models through the introduction of synergies among the different actors involved in the modeling. They can work independently on their dynamics and the platformwill integrate those, ensuring cohesion and robustness of the system. Our contributions include the ability to create the building blocks of the system independently, associate and combine them to formagents. This allows us to compare possibilities for the same dynamic and open the prospect of studyingmany alternate models of the same complex system, and then analyze at a very fine scale.
... Notre plateforme de base a servi à de nombreuses expérimentations et donné lieu à de multiples enrichissements et validations, intégrés ensuite dans ses nouvelles versions. Les travaux de recherche de l'équipe SMART se sont focalisés sur trois axes : l'observation dans la SOA [Ralambondrainy et al., 2006, Ralambondrainy, 2009, la distribution et la parallélisation de SOA [Sébastien et al., 2008, Sébastien, 2009], la gestion de l'émergence dans la SOA [David et al., 2009, David, 2010, David et al., 2012. ...
... 5 : 230 -239. • 3 articles publiés puis présentés dans des conférences internationales : -[Gangat et al., 2012a]Y. Gangat, D. Payet and R. Courdier. Another step toward reusability in agent-based simulation : Multi-behaviors & MVC, 24th IEEE International Conference on Tools with Artificial Intelligence (ICTAI 2012), Athens, Greece, nov.7-9, 2012.-[David et al., 2012] D. David, Y. Gangat, D. Payet and R. Courdier. Reification of emergent urban areas in a land-use simulation model in Reunion Island, ECAI 8.2. Perspectives utilisation, nous avons proposé l'approche méthodologique MMC. Si nous reprenons l'état de l'art présenté par les auteurs de [Pokahr et Braubach, 2009] (voir la première ligne de la ...
Article
Full-text available
Co-building and reuse of models are at the center of several studies in the field of simulation. However, in the more specific field ofMulti-Agent Based Simulation (MABS), there is a lack of methodology to resolve these two issues, despite a strong need by experts.Model co-building is essential to optimize knowledge sharing amongst different experts, but we often face divergent viewpoints. Existing methodologies for the MABS co-building allow only a low level of collaboration among experts during the initial phase of modeling, and between domain experts with modelers or computer scientists... In order to help this co-building, we propose and follow a methodology to facilitate this collaboration. Model reuse can provide significant time savings, improve models’ quality and offer new knowledge. Some MABS methodologies in this area exist. However, in the spectrum of reuse, they are often limited to a full model’s reuse or agent’s reuse with the impossibility of reusing smaller parts such as behaviors. The EDMMAS experiment was a concrete case of three successive model reuses. It allowed us to observe new complexity arising from the increase of agents’ behaviors. This creates a gap between operational model and conceptual model.Our goal is to promote the reuse of models, agents and their behaviors.To answer these questions, we propose in this thesis a new way to codify and integrate knowledge from different disciplines in the model, while using "composable"modules that facilitate reuse.We propose (i) a new agent architecture (aMVC), applied to a multidynamical approach (DOM), with the support (ii) of a methodology (MMC) based on the decompositionand reuse of behaviors.Proposals (i) and (ii) allow us to lead a multidisciplinary MABS project with a large number of actors, helping the co-building of models through the introduction of synergies among the different actors involved in the modeling. They can work independently on their dynamics and the platformwill integrate those, ensuring cohesion and robustness of the system. Our contributions include the ability to create the building blocks of the system independently, associate and combine them to formagents. This allows us to compare possibilities for the same dynamic and open the prospect of studyingmany alternate models of the same complex system, and then analyze at a very fine scale.
... In order to facilitate landscape visualization (Sheppard and Cizek 2009), Ocelet models results can be exported as kml or shp files with temporal attributes, which makes possible to display and compare simulations scenarios over time with popular software like Google Earth. Beyond the usual monitoring of selected indicators, this interactivity adds value to models that aim at providing support to discussion, knowledge integration, and land-use foresight as in current participatory modeling projects in Reunion Island (Lestrellin et al. 2013, David et al. 2012). ...
Technical Report
Full-text available
Full project title : « MOZALINK - Linking marine science, traditional knowledge and cultural perceptions of the sea in the Mozambique Channel to build tomorrow’s marine management using spatial simulation tools and educational game » This final report was jointly developed by the Research Institute for Development (IRD, France) with the contributions from CORDIO (Kenya), University Eduardo Mondlane (Mozambique), University of Dar es Salam (Tanzania) and Institut Halieutique et des Sciences Marines (Madagascar) in February 2018.
... Furthermore, in order to facilitate landscape visualization (Sheppard & Cizek, 2009), Ocelet allows exporting model results as Keyhole Markup Language (kml) and shapefiles (shp) with temporal attributes e making it possible to display and compare simulations with popular software like Google Earth. This flexibility and interactivity can constitute significant assets for a participatory approach to modelling (Augusseau et al., 2013;David, Gangat, Payet, & Courdier, 2012). ...
Article
Abstract This paper reflects on collaborative landscape research conducted in Reunion Island, an outermost region of the European Union. On this 2,500 km2 tropical island also considered a major international biodiversity hotspot, land-use planners must address important challenges, especially growing population densities and urban sprawl that cause important pressure on agricultural land and natural ecosystems. While progress has been made towards land-use zoning and planning at the island scale, entrenched interests and a lack of communication between the agricultural, urban and environmental sectors continue to hinder the design and implementation of integrated land-use plans at the local level. This paper presents an approach to territorial foresight where urban development scenarios and spatial models were co-constructed with a collective of institutional actors in order to facilitate dialogue on future urbanization patterns and impacts on landscapes. It describes how spatially explicit models and simulations of urban development, first used as demonstrators, have raised individual interests and expectations and facilitated the structuring of a collaborative research network. Models and scenarios were then questioned, redesigned collectively and used as boundary objects to facilitate a shift away from statistical and sectorial readings towards more territorialized and integrated perspectives. Analysing inputs, reactions and feedback from the actors involved in the research, this paper discusses the role and potential value of landscape modelling and simulation in mediating debates among planning stakeholders and creating social learning situations.
... En croisant l expertise des partenaires du projet DESCARTES sur les processus d urbanisation à La Réunion (Lajoie, 2007 ;David, 2010 ;David et al., 2012) et les connaissances issues des interactions avec les acteurs de l aménagement sur le TCO, nous avons entrepris la co-construction d'un modèle permettant de simuler l évolution de l urbanisation et le processus de consommation des terres agricoles sur ce territoire. La version courante de ce modèle est le fruit d'évolutions successives. ...
Data
Full-text available
Article
During last decade, multi-level agent-based modeling has received significant and dramatically increasing interest. In this article we present an exhaustive and structured review of literature on the subject. We present the main theoretical contributions and application domains of this concept, with an emphasis on social, flow, biological and biomedical models.
Article
Full-text available
This paper deals with the important concept of emer-gence in complex systems and in multiagent simulation. Research in this area yield to several definitions and classifications of emergent phenomena, but only a few of them offer a solution for emergence reification. As we know, this kind of notion does not have yet, for-mal definition, if any could be expressed, and we need to progress on the conceptual meaning, leading to more global definitions but allowing to give a general concep-tual framework for emergence manipulation. We define emergence as a metaknowledge controlled by emergence laws, and we present such a framework in which emergence is reified through emergent structures. Within this framework, emergent phenomena can be de-tected and injected in simulation systems to be manip-ulated like any other entity.
Article
Full-text available
Decentralization of land use governance creates new challenges for participatory approaches, including the involvement of highly diverse participants and the search for coherence among multiple regulations. In France, the 2000 law of Urban Reform and Solidarity ("Solidarité et Renouvellement Urbain") provides a legal framework to land use decentralization. It requires the planning design process to be participative and to involve civil society through consultation phases. It also addresses the issue of coherence among legal planning documents along the scale hierarchy, the larger scale plan conditioning the others. In the Reunion Island, the multi-level institutional system includes a region, its four micro-regions and 26 districts; each having their own land use plan. The revision of the regional plan ("Schéma d'Aménagement Régional", SAR) was the opportunity to revisit the various plans to make them more coherent across scales. This paper presents how research was included in this political process. The SAR revision was initially thought as a one-scale regular participatory foresight process in three stages, i.e. (1) a land use assessment (diagnostic); (2) the definition and discussion of contrasted scenarios; (3) the development of the final land use plan. The overall consultation process involved a large group of participants, including members of various institutions and of the civil society. They defined the logic of four contrasted land use scenarios, and sorted key challenges for the future of the Reunion Island. Stage 2 was however handled in an innovative way due to the collaboration with a research project called DOMINO- Reunion. The DOMINO-Reunion Project put together a team of researchers from various disciplinary backgrounds, and members of extension and support services for rural and agricultural development involved in the various debate on land use at each scale. Together, the team followed a companion modelling approach to develop a prototype of a dynamic model meant to assist the main players in building various land use scenarios and simulating their mid- to long-term consequences on urban, agricultural and natural stakes. A subset of participants of both processes, the SAR revision and DOMINO-Reunion, collaborated to adapt the model prototype to the SAR scenarios and feed the regional land use debate. This paper analyses how our companion modelling approach in two steps has helped integrating multi- scale stakes in the SAR revision process. It discusses specific challenges for participatory modelling, which are linked to power balance such as stakeholder's exclusion, over representation of specific interests and political co-optation, focussing on two specific processes: • the integration of multi-scale negotiation processes and indicators in the evolving model; • the implication of stakeholders, which are involved in various land use decision-making arena, in the participatory modelling of the system. We showed that companion modelling helped increase the representation and the weight of agricultural issues in the debate, although urban considerations still prevailed in the regional arena. The paper concludes on the benefits and drawback of integrating progressively different decision-marking scales in a participatory process.
Article
Full-text available
One of the main characteristics of complex systems is that the interrelations between the entities compos-ing the system are not permanently established but evolve along time. As opposed to complicated sys-tems, the structure of complex systems also evolve in a dynamic organizational process. When study-ing complex systems, self-organization and emergent phenomena must therefore be taken into account and studied carefully. In this paper, we propose to provide tools in order to automatically detect and characterize the emergent phenomena occurring in agent-based simulations. To this end, we consider the interactions between the entities at the lower level as the main organizational forces that shape the structure of the system at a higher level. These interactions are detected during the simulation and represented as dynamic graphs. Measures can then be made on various properties of the graph so as to detect the occurrence of structuring processes. Groups detection and tracking techniques are then introduced so as to characterize more precisely the exact nature of these processes.
Article
Full-text available
This paper considers participatory modelling to integrate biodiversity conservation into land use planning and to facilitate the incorporation of ecological knowledge into public decision making for spatial planning. Réunion Island has experienced rapid urban and agricultural expansion, which threaten its unique biodiversity. In this context, we designed three participatory modelling sequences, involving overall 24 multidisciplinary researchers and stakeholders. The sequences aimed: (1) to map land-use and biodiversity, (2) to develop a conservation plan following systematic conservation planning principles using a spatial optimization tool (MARXAN) and (3) to simulate coupled land-use/conservation scenarios using a multi-agent system (MAS). The conservation plan confirms that priority areas for biodiversity protection are located on the coast where rapid land-use changes occur. Nevertheless, stakeholders from the urban and agricultural sector did not participate to this sequence. Indeed, conservation planning tools are useful to locate conservation priorities but they have to be designed with stakeholders to be accepted as negotiation tool. Besides, the researchers engaged in this second sequence were perceived as conservation stakeholders rather than holders of scientific knowledge. In the third sequence, the researchers involved adopted the stance of facilitating the elicitation of each stake and gathered trust from stakeholders. Overall, we conclude that the participatory development of land-use simulation models should be promoted to explore alternative scenarios for biodiversity conservation with stakeholders. In a situation of land-use conflict, a gradual and sequential participatory modelling approach should be implemented to fit into public decision-making processes.
Conference Paper
Full-text available
Emergence is a fascinating concept for most scientists, and multiagent simulations are known to allow and facilitate its representation. Research in this area yield to several definitions and classifications of emergent phenomena, but only a few of them offers a solution for a concrete reification of emergence in simulation. This paper deals with this important notion of emergence reification that, as we know, does not have yet formal mathematic definition, if any could be expressed. We need to progress on the conceptual meaning, leading to more global definitions but allowing to give a general conceptual framework that makes possible the reification of emergent phenomena in multiagent simulations. We define emergence as being a metaknowledge and we present a conceptual framework in which emergent phenomena can be detected and injected into simulation systems and be handled like other entities.
Conference Paper
Full-text available
The development of solutions to complex problems requires finding answers to several questions which are obstacles on the road leading to the solution. To perpetuate progress, the possibility to reuse and integrate these intermediate answers in the other constructions of complex solution becomes a requirement In this article we consider the context of spatial multi-agent systems and we propose a modeling method that allows to structure models in order to make their components autonomous and, thus, to allow their reuses and their integrations in the design of future models. This method is based on the definition of a principle that permits to differentiate the model components, and, on the use of the environment (which is the location in which agents evolve) as the coupling element of these components
Article
Full-text available
this article some simple but powerful cases will be used to illustrate what has been learned from the techniques of experiment. As we shall see, however, the institutions that survive are not born equal on any one measure of performance: efficiency, price volatility, demand responsivity, dependence on external information channels and the (transactions) cost of participation. Rather, each seems to be an adaptation to environmental wrinkles, or niches, that are not evident to the naked eye
Une politique foncière
  • Agence Agorah
  • De La Réunion
Agorah, Agence d'urbanisme de La Réunion, Une politique foncière, une des clefs pour aménager le territoire (2006)
A Turing test for emergence, Advances in applied self-organizing systems
  • F Boschetti
  • R Gray
Boschetti, F., Gray, R., A Turing test for emergence, Advances in applied self-organizing systems, pp. 349-369, (2007)
Prospective territoriale par simulation orientée agent
  • D David
David, D., Prospective territoriale par simulation orientée agent, PhD Thesis, Université de La Réunion (2010)