MASC : MAp Sectors Creator - A tool to help at the configuration of multi-agents systems for everyone

Abstract

The initialization at the beginning of all type of simulation is recurring. In the field of multi-agents, spatial environments on a micro or macro scale, or more abstract context, such as mapping out energy sources for example, need often to be modelled. The initialization of these environments wherein evolve agents is generally tedious and time-consuming depending on the fineness of the mesh of the cutting in case of a 2D spatial representation. In this paper, we present MASC (Map Sectors Creator), a user-oriented tool that permits easy initialization through the creation of a mesh directly usable in multi-agents systems simulations platforms. The automation of cutting out maps followed by the generation of directly usable code snippets optimize the working time on initialization and allow to focus on the simulations results and observations.

Full text

MASC : MAp Sectors Creator
A tool to help at the configuration of multi-agents systems for everyone
Aurélie Gaudieux1, Joël Kwan2, Yassine Gangat3 and Rémy Courdier3
1Centre of Economics and Management of the Indian Ocean, University of Reunion Island, Saint-Denis, Reunion Island
2Higher School Reunion Indian Ocean Engineering (ESIROI), Saint-Denis, Reunion Island
3Laboratory of Computer Science and Mathematics (LIM), University of Reunion Island, Saint-Denis, Reunion Island
{aurelie.gaudieux, joel.kwan, yassine.gangat,remy.courdier}@univ-reunion.fr
Keywords: AUTOMATION, INITIALIZATION, CONFIGURATION, MAP, SKETCHES, PLANS, CUTTING OUT,
EXTRACTION, GRID, BITMAP FILE, CODE SNIPPET, ONLINE, WEB BASED, USER-FRIENDLY
Abstract: The initialization at the beginning of all type of simulation is recurring. In the field of multi-agents, spatial
environments on a micro or macro scale, or more abstract context, such as mapping out energy sources for
example, need often to be modelled. The initialization of these environments wherein evolve agents is
generally tedious and time-consuming depending on the fineness of the mesh of the cutting in case of a 2D
spatial representation. In this paper, we present MASC (Map Sectors Creator), a user-oriented tool that permits
easy initialization through the creation of a mesh directly usable in multi-agents systems simulations
platforms. The automation of cutting out maps followed by the generation of directly usable code snippets
optimize the working time on initialization and allow to focus on the simulations results and observations.
1 INTRODUCTION
Computer modelling is the formal description of a
system that allows manipulation of a virtual copy
using a computer. To obtain this virtual
representation, a model must be implemented. The
process from a theoretical context to a more concrete
one involves the production of codes to be used to
have a final product in the form of an application.
Before running a simulation, the initialization
must be settled, which includes configuring the
simulation context. Knowing that the context has a
huge impact on the proceedings and the outcomes of
the simulation, we often give more attention to data
in particular and starting environment. Each
configuration leads, so to speak, to an almost unique
experience depending on the configuration settings
(Gangat, et al., 2009).
Optimization of working time issue arises when
handling complex systems, especially in multi-agents
systems where more attention must be paid to data
and different proceedings performed during
iterations.
To help solving this problem, we propose MASC
– Map Sectors Creator, an easy to use tool that can
provide an environment base represented by cutting a
base map, a diagram or a plan into a grid of sectors
directly usable in multi-agents simulation platforms
such as NetLogo (Wilensky, 1999) or the platform of
multi-agent simulation of IREMIA laboratory at the
University of Reunion Island, GEAMAS-NG
(GEneric Architecture for MultiAgent Simulations)
(Payet, Courdier, Sebastien, & Ralambondrainy,
2006).
Thus, MASC permits to save time and focus on
implementation of models. In addition, MASC
provides a simplified user experience through
opportunities brought by recent developments in Web
technologies. We want to propose a user-friendly tool
that could help anyone to optimize working time and
facilitate the configuration of the initialization. The
word “masc” means “men” in Portuguese.
This kind of tool is useful for several simulation
domains for example geographical, economic,
meteorological or epidemic ones.
2 MASC
2.1 The need
In the context of models based on real existing ones,
we perform simulations in which agents operate on a

space generated from spatial division of maps. E.g.:
SIEGMAS (Gaudieux, Gangat, Kwan, & Courdier,
2014).
Some existing tools already provide some parts of
solution to this problem. We can take the example of
XELOC Language, eXtensible Editing Language Of
Configuration, which is a support for the
configuration an initialization of multi-agents
systems. XELOC is based on the basics of XML and
requires the handlings of another programing
language. One of MASC objective is to repeal both
the acquisition processes and handling of a new tool.
The development of a tool like MASC was driven
by mainly two reasons.
Firstly, the need of a tool allowing the automation
of the process of cutting out maps to perform multiple
simulations of a same model with different
geographical areas. The tool aims to facilitate the
transition from map into an exploitable
representation. It is very interesting to incorporate
modelling with different configurations. Some
models can lead to significant results from the
abstract level to a less abstract one.
And secondly, the use of a user-friendly tool also
permits to effortlessly make rather redundant tasks of
creating an environment where agent evolves along
different simulations. Several configurations can then
be produced more efficiently in no time as revealed
in Figure 1.
2.2 The concept
MASC has been designed to increase productivity on
implementing a model so that it provides a
streamlined user experience. That is why the tool was
developed in order to ask the fewest operations and
minimum familiarization while providing enough
controls. The user interactions with MASC interface
are then based on the daily operations being done by
any computer user (drag and drop, file uploading,
form filling, mouse clicking). The graphical user
interface (Figure 2) is quite simple and reduces the
displayed information to only the useful ones.
The quantity of requested information to input is
also kept to a minimum, the rest being automatically
filled by calculations.
Figure 2 : MASC GUI
Figure 1: Using MASC for configuring multiple experimentations of a simulation

A question arose as to which file types would be
supported by the tools. The most used type of output
images are vector graphic files generated from GIS
softwares or bitmap files for several maps, plans,
drawings, sketches, or other digital copies created
from paper versions. Knowing that it is much less
restrictive to convert a vector image into bitmap on to
the opposite direction, we have decided that the first
versions of the tool would only support bitmap
images in input.
Moreover, we wanted the tool be designed so that
it is usable at any time by anyone with any computer
without altering the workspace of the user. That is
why we chose to make an online web based
application.
Regardless the necessity of a connection to the
server, the downside of this type of application is that
the power of calculation is limited by the capacity of
the hosting server and the technology used for the
different calculations and the allowed uploading file
size. For the first version of our application, we chose
to make it deployable to a maximum of easily
accessible servers. So we deployed it on a basic server
provided by a common hosting plan. And for a
maximum compatibility with available programming
languages on these type of hosting, we used PHP
programming language for the server side coupled
with HTML5 and JavaScript for the client side.
On the upside, anyone can access the application
with a simple connection to the server and a recent
Internet browser. In a future perspective, the server
side will be switched to a Java version that will have
a better performance calculations but require an
upgrade of the hosting plan.
2.3 Presentation
2.3.1 The ease of using MASC
Using MASC is divided around three main steps.
The first step consists of inputting all the
necessary information needed by the tool to correctly
extract a cutting grid space from an image that is
presented to it. This stage takes place mainly on the
client side that is to say on the user’s Internet browser
(shown highlighted in Figure 3). The information
required for the proper functioning of the tool are:
 A bitmap image file in JPEG, PNG, GIF, or
TIFF format, which is a representation of the
space where agents evolve during the
simulations.
 The size of the grid is simply summarized by
the number of sectors to cut out across the
width of the image. Because we are working
with a grid where sectors are square shaped, the
number of sectors in the height direction can be
automatically calculated.
 A list of colors selected from the preview of the
image taken by clicking on it.
In a second time, these three data are sent to the
server. Then the script on the server can manipulate
the bitmap file and can extract a grid that will serve
as a starting point for generating the code snippet used
for initialization of simulations. At this stage, the user
will have no operations to do, all the treatments will
be done in autonomy at the server level.
In the last step, new information resulting from
the extraction made on the server side is displayed
graphically to the user using the preview area of the
graphical user interface displayed on his browser. On
the one hand, the user has a preview and on the other
hand he can slightly edit the grid before final code
snippet generation according to the chosen
programming language. At this level, the user can still
change the color of the different sectors by clicking
on them.
If needed, the user can change the number of
given sectors across the width to change crisscross
scaling and then regenerate the code snippet.
2.3.2 Colors
Always with the idea of putting the user at the center
of the experience and increasing the productivity in
mind, MASC uses a list of colors selected from the
image by the user. This list is used in the second step.
Figure 3 : Details of MASC workflow

It also allows to filter unwanted colors and to avoid to
keep annotations that are often merged on the image
itself. So the user only select interesting colors.
When extracting the colors, the image is divided
into sectors according to the given cutting grid. Then
each pixel is examined to determine if the pixel color
matches with one of the colors in the list. If it is the
case, the counter of the matching color in the sector
of the pixel is incremented.
In this process, the tool tolerates a degree of
shade. Thus, the color alterations brought by the
image compression does not prevent MASC to
correctly extract the desired colors in each area.
An array of colors grouped by sector is gradually
created regarding the progress of the review of the
image pixels. Then, the dominant color is selected to
be the color of the sector (Figure 4). By proceeding
pixel by pixel, the precision of the interpolation that
assesses a sector color is dependent on the cutting out
grid size. The more sectors across the width is given
in input, the greater the precision will be. There is no
minimum grid resolution, it all depends on the goal
and the need of the model.
To allow a certain degree of freedom and to leave
the choice of colors to prioritize on sectors, the user
can click on the preview grid cells to choose the
dominant color regarding the listed ones in the
working sector. These modifications are stored in the
client side to avoid the useless data transfers between
client and server. So directly after the adjustments
brought by the user, the code snippet containing the
representation of the space can be generated.
3 CASES STUDY
The team being at work on two ongoing projects.
SIEGMAS which use the simulation platform
NetLogo and EDMMAS, which use the GEAMAS-
NG, we would like to test the relevance of MASC in
these projects. In this part of the paper, we will
explain the advantage of using MASC with theses
example.
3.1 SIEGMAS
SIEGMAS (Stakeholders Interactions in
Environmental Governance by a Multi-Agent
System) is a modelling conceptualized in order to
study the deviances of the farmers against the forester
and natural resources protected by the management's
transfers or another text of law.
3.1.1 Context
In the context of this project, the economical and
classic steps being limited enough for the
conceptualization of the individuals' behaviors and
the process of the common pool resources, a process
of multi-agents systems allows some improvements
in the study of these phenomenon and the creation of
a sales tool for the decision
This modelling has been conceptualized in order
to study the deviances of the farmers against the
forester and natural resources protected by the
management's transfers or another text of law.
Therefore, this simulation allows to identify the
interactions between the actors of the governance
taking part in the management's transfers and to
propose some targeted measures in agreement being
able to reduce the merging anomalies. Thus, this
model reveals that the knowledge of the sanctions by
farmers does not prevent them to infringe the
legislative measures of protection of natural resources
coming from the management's transfers or located in
some few protected areas. (Brandouy, Mathieu, &
Venryzhenko, 2012).
3.1.2 SIEGMAS
Farmers (deviant or not) generally aim to make their
land as profitable as possible to cover several costs
Figure 4 – Colors interpolation

(e.g. investment, maintenance). A farmer can choose
to spread legally his cultivation via the purchase of
new plots or reaching the legally maximum allowed
use of natural resources. In fact, the management's
transfers grant the farmer through the basis
community some right of cultivation of some natural
resources. However, the farmer can be a deviant by
farming illegally some plots of natural and forester
resources.
This model is mainly focused on the farmers and
the government and their interactions in relation to
the forester and natural resources as well as the
resources produced on the farms.
The model SIEGMAS has been implemented
under the platform of the programming called
NetLogo. The farmers can respect or infringe the
legislation towards the protection of natural
resources. The controls or the rupture can identify the
transgressions. Authorities grant some individual
sanctions to the deviants. Thus, not any collective
sanction impacts the respectful farmers of the
legislation. The farmers set themselves up in order to
increase their productivity and to perpetuate the trust
that the state authorities grant (or to avoid the
sanctions for everything by having a nearby context
in favor of their activity (Wade, 1987).
For the sanctions, the deviants want to exploit
randomly the lands being around his exploitations,
even those created illegally. At each iteration, the
government notes a random number of lands illegally
exploited and looks for the one who created the
exploitation. After launching several times the
simulation with different values of configurations, it
appears by observing the variation of the rate of
deviants that if the government controls the
management's transfers correctly, the sanctions are
well applied. The presence of the deviants increases
land deterioration. Some deviants however manage to
avoid sanctions and run some untaxed incomes.
Nevertheless, the application of a sanction to the
deviants rebalances the system so that the deviants
cannot get any profits from the illegal practice of
agriculture or the exploitation of natural resources.
Therefore, individual behaviors impact the collective
behaviors if a farmer knows the machinations of his
neighboring. The farmer generally grants his trust to
his neighbor as long as it doesn’t involve being a
deviant. However, if he knows the acts of deviance of
his neighbor, he will be more suspicious against him
and will denounce him.
This modelling is quite difficult to implement and
require a lot attention. For testing the model, several
configurations will be applied to multiple
simulations. Usually, these tasks take a lot of time to
prepare. Not to waste time, the creation of these initial
states should be automated.
3.1.3 MASC with SIEGMAS
Figure 5 : Analamanga region map
Figure 6 : Preview of Analamanga region map generated by
MASC

In this case study, we are trying to model the
interactions between stakeholders in a precise
geographical area. To do so we have to create a
representation of a region in a simulation platform. In
this specific example, the region of Analamanga
(Madagascar) which directly usable map data are
quite difficult to find, is used. So we go from a map
extracted from a digital document that was provided
to us (Figure 5).
Without MASC, we would have been constraint
to crisscross the image in software image processing
and then to manually select and assess the color of
each sector of the grid. Thanks to MASC, these tasks
were simply handled by giving the image file, the
number of sectors to cut in the width direction and a
color swatches that we simply took from the legend
of the map by clicking on it. Afterwards, a preview
was displayed to us (Figure 6). We decided then to
prioritize the colors of the sectors or not and generate
the code snippet regarding the chosen programming
language, in this case NetLogo.
Part of generated code snippet for Analamanga
region map:
; Permits to create a representation of
the map
to create-world
clear-all
set-patch-size 20
resize-world 0 39 0 50
ask patch 0 50 [
set pcolor [186 158 74
]
ask patch 1 50 [
set pcolor [186 158 74]
]
ask patch 2 50 [
set pcolor [255 252 217]
]
ask patch 3 50 [
set pcolor [255 252 217]
]
[...]
ask patch 36 0 [
set pcolor [255 255 255]
]
ask patch 37 0 [
set pcolor [255 255 255]
]
ask patch 38 0 [
set pcolor [255 255 255]
]
ask patch 39 0 [
set pcolor [255 255 255]
]
end
The advantage of using MASC in this context was
to more or less finely crisscross the map to translate
from the bitmap file to an exploitable representation
of this map inside a simulation platform. As a result
we can have easily different grid resolutions of the
map with no effort and can deploy rapidly different
base maps for different configurations.
This methodology also allows model
transposition to different geographical regions, for
different simulations executions on Reunion Island
and Mauritius regions and comparing obtained
results.
3.2 EDMMAS
EDMMAS (Energy Demand Management by Multi-
Agent Simulation) is a prototype based on the model
of the same name for simulating the production and
consummation of energy according to the evolution
of the population over the years.
3.2.1 Context
The issue in the EDMASS project is the energy
development of a territory using a reproducible
modelling. Reunion Island provides a favorable
research field in this area thanks to its isolations
because its small area (2.500 km²) promotes the quest
of new forms of energy development and prohibits
electricity production from nuclear power.
EDMMAS is a tool for helping experts to make
decisions through ABS, an overview of the evolution
of electricity consumption by 2030 and to provide
new facilities.
3.2.2 EDMMAS
EDMMAS is and evolution of DS model used to
explore various scenarios of land usages and
conservation planning that simulates on the whole
island, the interactions between the three types of land
use (natural, agricultural and urban), in order to allow
observation of the changes induced. The energy layer
has been added to allow this (Figure 7).
Figure 7 : EDMMAS dynamics

EDMMAS was designed on the GEAMAS-NG
simulation platform. This platform has been created
with several powerful tools to reduce the gap that lies
between the real and complex system model. Among
the innovations of this platform, we notice in
particular XELOC. XELOC is a support for
configuration and initialization of multi-agent
systems. It is established on the basis of XML,
enriched by a semantic script language. Thus, while a
configuration for defining the form of a conventional
XML description, XELOC offers the possibility of
synthesizing a complex configuration by means of a
dynamic process. In addition, XELOC incorporates a
rich initialization technique using semantic maps
provided by mathematicians. All this makes XELOC,
scalable and accessible by nature, a generic support
used in many contexts of implementation. A XELOC
interpreter has been developed within the GEAMAS-
NG platform.
3.2.3 MASC with EDMMAS
In EDMMAS, we handle maps containing
information on energy sources available in Reunion
Island. We tried to use MASC to complement
XELOC Language and facilitate simulation
initialization.
We take a map of average annual global radiation
to the horizontal of Reunion Island in 2009 (Figure 9)
for the practical case.
Using MASC, we can easily retrieve a XELOC
code snippet corresponding to the map with the
correct color codes without using multiple tools.
When we manipulate codes, we usually don’t want to
lose time with redundant tasks such as picking all
wanted color codes in an image. Thus we benefit of
using advantages of MASC and XELOC Language.
MASC allows color selection and generates
effortlessly code snippets in XELOC language.
Part of generated XELOC code snippet of the
example:
<forMap name="OurMapName"
url="./myMapFile.jpg"
resolution="16">
<legend name="3classes">
<label name="rgba(255,18,56)"
r="255" g="18" b="56" />
[...]
<label name="rgba(233,254,4)"
r="233" g="254" b="4" />
<label name="rgba(229,255,0)"
r="229" g="255" b="0" />
</legend>
<cell legend="10classes"
operator="average"
mode="closer">
<case label="level_0">
<!-- INSTRUCTION FOR #0 -->
</case>
[...]
<case label="level_9">
<!-- INSTRUCTION FOR #9 -->
</case>
</cell>
</forMap>
Using MASC for generating XELOC code
snippets does not require to send the image on server
for a post treatment. In this case, MASC is just an
advanced color picker which allows to generate
XELOC code snippet as output. We also note that if
we need the reuse of this map for NetLogo, we have
just to set the grid resolution, ask for extraction, and
then change the language choice and regenerate the
Figure 8 : Preview generated by MASC of the average
annual global radiation of Reunion Island
Figure 9 : Reunion Island average annual global radiation
(2009)

code snippet that enables us to have an equivalent
map in patches for NetLogo (Figure 8).
Part of generated NetLogo code snippet for
Reunion Island average annual radiation (2009):
;Permits to create a representation of
the map
to create-world
clear-all
set-patch-size 20
resize-world 0 59 0 44
ask patch 0 0 [
set pcolor [255 255 255]
]
ask patch 1 0 [
set pcolor [255 255 255]
]
ask patch 2 0 [
set pcolor [255 255 255]
]
ask patch 3 0 [
set pcolor [255 255 255]
]
[...]
ask patch 19 44 [
set pcolor [254 91 117]
]
ask patch 20 44 [
set pcolor [254 91 117]
]
ask patch 21 44 [
set pcolor [254 122 135]
]
ask patch 22 44 [
set pcolor [254 122 135]
]
ask patch 23 44 [
set pcolor [254 122 135]
]
[...]
ask patch 59 0 [
set pcolor [255 255 255]
]
end
We notice in this example of type of use that an
advanced version of MASC would be very useful if
the tool can also permit to set the location of some
agents on the map. In the context of EDMMAS, we
can take for example the localization of power plants
in Reunion Island (Figure 10) that will be the agents
of the model. Keeping in mind that MASC is
designed to facilitate the configuration of simulations,
one of the future goals of MASC is to allow the
placement of agents in addition of creating the
environment in which they evolve.
4 FUTURE WORKS
MASC is already a fully functional tool and responds
more or less to the initial requirements mentioned in
above parts of this paper. However, during our
experiencing phases, we note that the tool can be
improved to improve its effectiveness.
On the server side, an upgrade of the hosting plan
would permit a usage of another programming
language for example Java which allows to have a
compiled server side application to improve the
calculation time instead of an interpreted script. With
Java, we could also ignore the passage of data through
Apache server and prevent the flow of information
between several services. The client side will directly
communicate with the server side of the application.
The client side can also be optimized by
modifying some functions and not using frameworks
as it is now. The advantage of using frameworks is
the time gained at developing the tools. Once the
proof of concept is acquired, the next step of MASC
is optimizations.
Then we can also add more feature on this tool.
Actually, the tools only manipulate the space of the
simulations. It could also permit to prepare the
initialization of multi-agents such as auto-detecting
and extracting positions of agents based on pattern on
the image, setting up clusters of patches or boundaries
Figure 10 - Reunion Island power plants (2007)

for agents. We could also expand the list of available
multi-agents simulation platform languages to
generalize the use of the tool and permit more
members of the community of the domain of multi-
agent systems to use it.
5 PERSPECTIVES AND
CONCLUSIONS
The redundant problem of creating representation of
a space in multi-agents systems leads us to find some
useful tools to be more productive. We often use
maps from GIS softwares, paper supports and other
electronic documents that we want to set up in our
simulations.
In this article, we presented MASC, a support tool
for configuring multi-agent systems based on the
information extracted from colors of bitmap images.
MASC aims to provide a simplified user-friendly
service for the acquisition of code snippets ready to
use with multi-agent systems simulations platforms.
Thereby, creating multiple configurations for a same
modelling can be done quickly and easily, while
allowing the users to focus on the data and the results
of simulations. Moreover, MASC allows to
effortlessly create several of more or less finely
crisscrossed map for observing the evolution of
agents on different scales and precisions. We could
gain more time on configuration if the tool permits
also the placement of the agents if needed.
Our purpose in this article was not to bring out the
possibility to automate part of the implementation of
a modelling, but rather to highlight the lack of user-
friendly tools to simplify the initialization of a
simulation, which we assessed in example and
present to community a more or less adequate
solution.
REFERENCES
Brandouy, O., Mathieu, P. & Venryzhenko, I., 2012.
Optimal Portfolio Diversification ? A multi-agent
ecological competition analysis. PPAMS (Special
Sessions), pp. 323-331.
David, D., Payet, D., Courdier, R. & Gangat, Y.,
2009. XELOC : un support générique pour la
configuration et l’initialisation de systèmes multi-
agents. Journées Francophones Systèmes
MultiAgents, JFSMA’09.
Gangat, Y., 2013. Architecture Agent pour la
modélisation et simulation de systèmes complexes
multidynamiques : Une approche
multicomportementale basée sur le pattern
"Agent MVC", Saint-Denis: s.n.
Gangat, Y., Courdier, R. & Payet, D., 2009.
Démonstration : Aménagement énergétique d’un
territoire - une approche par simulation multi-
agents. Journées Francophones Systèmes Multi-
Agents (JFSMA), p. 237–240.
Gaudieux, A., Gangat, Y., Kwan, J. & Courdier, R.,
2014. Multi-agents model for the study of
interactions between the stakeholders in the
common pool resources : application to the
district of Miarinarivo ( Madagascar).
Payet, D., Courdier, R., Sebastien, N. &
Ralambondrainy, T., 2006. Environment as
support for simplification, reuse and integration of
processes in spatial. MAS. IEEE International
Conference on Information Reuse Integration, p.
127–131.
Wade, R., 1987. The management of common
property resources : collective action as an
alternative to privatisation or state régulation.
Cambridge Journal of Economics, Volume 11,
pp. 95-106.
Wilensky, U., 1999. NetLogo. Center for Connected
Learning and Computer-Based Modeling.
Northwestern University, Evanston, IL.
APPENDIX
A version of MASC is in public access at
http://labs.jkwan.com/masc/