JChoc DisSolver

Bridging the Gap Between Simulation and Realistic Use

I. Benelallam

1,2

, Z. Erraji

, G. Elkhattabi

, J. Ait Haddou

and E. H. Bouyakhf

LIMIARF − FSR,University Mohammed V, Rabat, Morocco

INSEA, Rabat, Morocco

Keywords:

Constraint Programming (CP), Multi-Agent Systems, Distributed Problem Solving, Agent Models and

Architectures, Distributed Constraints Reasoning, Realistic Use, Constraint Satisfaction Problem (CSP),

Distributed CSP (DisCSP).

Abstract:

The development of innovative and intelligent multiagent applications based on Distributed Constraints Rea-

soning techniques is obviously a fastidious task, especially to tackle new combinatorial problems (e.i. dis-

tributed resource management, distributed air trafﬁc management, Distributed Sensor Network (B

ejar et al.,

2005)). However, there are very few open-source platforms dedicated to solve such problems within realistic

uses. Given the difﬁculty that researchers are facing, simplifying assumptions and simulations uses are com-

monly used techniques. Nevertheless, these techniques may not be able to capture all the details about the

problem to be solved. Hence, transition from the simulation to the actual development context causes a loss

of accuracy and robustness of the applications to be implemented.

In this paper, we present preliminary results of a new distributed constraints programming platform, namely

JChoc DisSolver. Thanks to the extensibility of JADE communication model and the robustness of Choco

Solver, JChoc brings a new added value to Distributed Constraints Reasoning. The platform is user-friendly

and the development of multiagent applications based on Constraints Programming is no longer a mystery to

users.

A real distributed problem is used to illustrate how the platform can be appropriated by an unsophisticated

user and the experimental results are encouraging for more investigations.

1 INTRODUCTION

Since the onset of real time electronic devices, mo-

bile, ubiquitous, and intelligent computing, new com-

binatorial problems have emerged in the AI commu-

nity such as: distributed resource management, dis-

tributed air trafﬁc management, Distributed Sensor

Network (B

ejar et al., 2005), disaster rescue (Ki-

tano et al., 1999) and distributed Meeting Schedul-

ing Problems (SMP), for which it is not suitable to

collect all data of problem in one site, to solve it

by a centralized algorithm. The reasons are com-

munication time and cost of translation of each sub-

problem in a common format. In addition, to give

a single agent all data of the problem can also be

excluded for reasons of security and conﬁdentiality.

Therefore, some of the AI communities are motivated

to take an interest in Distributed Constraint Reason-

ing (DCR), giving birth to other distributed formal-

ism (Yokoo, 2001), whose work focused on develop-

ing techniques for modeling and solving distributed

combinatorial problems with or without optimization

criterion. Distributed Constraint Satisfaction Prob-

lems (DisCSP) and Distributed Constraint Optimiza-

tion Problems (DCOP) provide a useful framework

of multiagent systems for distributed resolution of

combinatorial problems (Yokoo and Hirayama, 1995;

Yokoo et al., 1992; Yokoo, 2001; Yokoo, 2000; Yokoo

et al., 1998).

In this context, an agent must have a commu-

nication platform that allows the exchange of in-

formation or dialogue to coordinate their decision-

making. This reliable communication tool allows

agents to send and receive messages according to a

given distributed protocol. However, various sophisti-

cated solvers have been developed: DisChoco (Wahbi

et al., 2011), Disolver (Hamadi, 2003), MELY (Gal-

ley, 2000), Frodo (Petcu, 2006). These solvers rely

on several algorithms for solving DisCSP problems

such as Asynchronous Backtracking (ABT (Yokoo

Benelallam I., Erraji Z., Elkhattabi G., Ait Haddou J. and Bouyakhf E..

JChoc DisSolver - Bridging the Gap Between Simulation and Realistic Use.

DOI: 10.5220/0005208600660074

In Proceedings of the International Conference on Agents and Artiﬁcial Intelligence (ICAART-2015), pages 66-74

ISBN: 978-989-758-073-4

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

et al., 1992), ABT Family (Bessiere et al., 2005)),

Asynchronous Forward Checking (AFC) (Meisels

and Zivan, 2007) and Nogood-based Asynchronous

Forward-Checking (AFC-ng) (Ezzahir et al., 2009).

Asynchronous Distributed constraints OPTimization

(Adopt) (Modi et al., 2005), Asynchronous Forward

Bounding (AFB) (Gershman et al., 2009), Asyn-

chronous Branch-and-Bound (Adopt-BnB) (Yeoh

et al., 2008) and Dynamic Backtracking for dis-

tributed constraint optimization (DyBop) (Ezzahir

et al., 2008) were developed to solve DCOP prob-

lems. As well as the authors recognise that most

of these tools are specially developed for simulation

context. This fact can be clearly observed from its

experimental setups. Given the difﬁculty that re-

searchers are facing, they often make many simpli-

fying assumptions (i.e. simple agent (one variable

per agent), agents as multi-thread, single physical

platform, communication via simulated perfect FIFO

channels, etc...) about the underlying distributed

problem, which might affect the predictions obtained

from the simulation in non-trivial ways. Switching

from the simulation to the actual development prac-

tice often leads to loss of accuracy. Hence, bridging

the gap between simulation and actual development

and deployment within distributed constraints solvers

is the primary motivation for presenting the different

ideas discussed in the present paper.

In this paper we focus on the development of a

multiagent platform for Distributed Constraint Rea-

soning, namely JChoc DisSolver. This proposed plat-

form allows non-expert user to address and solve eas-

ily real Distributed Constraint Satisfaction Problems.

This document is organized as follows. Section

2 presents a brief deﬁnition of Distributed Constraint

Satisfaction Problem (DisCSP) and gives an exam-

ple. In section 3, we present related work . Section

4 presents the global architecture of JChoc platform.

In section 5, we show how use this platform in a dis-

tributed environment. And ﬁnally, in section 6 we

conclude the paper by experiment this platform within

a real Distributed Constraints Satisfaction Problem.

2 PRELIMINARIES

2.1 Distributed Constraint Satisfaction

Problems

Constraint Programming distinguishes between the

description of the constraints involved in a problem

on the one hand, and the algorithms and heuristics

used to solve the problem on the other hand. Mod-

eling and solving problems is through a very elegant

mathematical formalism, called the Constraint Satis-

faction Problems CSPs.

The Distributed Constraint Satisfaction Problem

(DisCSP) is represented by a constraint network

where variables and constraints are distributed among

multiple automated agents.

Deﬁnition. A ﬁnite DisCSP is deﬁned by a 5-

tuple(A, X, D,C, ψ), where:

• A = {A

, ..., A

} is a set of p agents.

• X = {x

, ..., x

} is a set of n variables such that

each variable x

is controlled by one agent in A.

• D = {D(x

), ..., D(x

)} is a set of current do-

mains, where D(x

) is a ﬁnite set of possible

values for variable x

• C = {c

, ..., c

} is a set of m constraints that

specify the combinations of values allowed for

the variables they involve. We note that the

constraints are distributed among the automated

agents. Hence, constraints divide into two broad

classes: inter-agent and intra-agent.

• ψ : X 7−→ A is a function that maps each variable

to its agent.

A solution to a DisCSP is an assignment of a value

from its domain to every variable of the distributed

constraint network, in such a way that every constraint

is satisﬁed. Solutions to DisCSPs can be found by

searching through the possible assignments of values

to variables such as ABT algorithm (Yokoo et al.,

1992).

2.2 Meeting Scheduling Problem as a

DisCSP

The Distributed Meeting Scheduling Problem (MSP)

is a real distributed problem where agents may not de-

sire to deliver their personal information to a central-

ized agent to solve the whole problem (Meisels and

Lavee, 2004; Wallace and Freuder, 2002).

The MSP involves a set of n agents having a per-

sonal private calendar and a set of m meetings each

taking place in a speciﬁed location. Each agent,

∈ A, knows the set of the k

among m meetings

he/she must attend. It is assumed that each agent

knows the traveling time between the locations where

his/her meetings will be held. The traveling time be-

tween locations where two meetings m

and m

will

JChocDisSolver-BridgingtheGapBetweenSimulationandRealisticUse

be hold is denoted by TravellingTime(m

, m

). Solv-

ing the problem consists in satisfying the following

constraints: (i) all agents attending a meeting must

agree on when it will occur, (ii) an agent cannot at-

tend two meetings at same time, (iii) an agent must

have enough time to travel from the location where

he/she is to the location where the next meeting will

be held.

We illustrate in Figure 1 the encoding of the in-

stance of the meeting scheduling problem in the dis-

tributed constraint network formalism. This ﬁgure

shows 4 agents where each agent has a personal pri-

vate calendar and a set of meetings each taking place

in a speciﬁed location. Thus, we get the following

DisCSP:

• A = {A

, A

} each agent A

corresponds to

a real agent,

• For each agent A

∈ A there is a variable m

, for

every meeting m

that A

attends,

• X = {m

, m

• D = {D(m

)|m

∈ X } where,

∗ D(m

) = D(m

) = {s |

s is a slot in calendar(A

)}.

∗ D(m

) = D(m

) = {s |

s is a slot in calendar(A

)}.

∗ D(m

) = D(m

) = {s |

s is a slot in calendar(A

)}.

∗ D(m

) = {s | s is a slot in calendar(A

)}.

• For each agent A

, there is a private arrival-time

constraint (c

intra-agent constraint) between ev-

ery pair of its local variables (m

, m

) (e.g. Omar

must attend tree meetings m

, m

and m

). For

each two agents A

, A

that attend the same meet-

ing m

there is an equality inter-agent constraint

i j

) between the variables m

and m

, corre-

sponding to the meeting m

on agent A

and A

(e.g. Omar and Jean participate in the same meet-

ing m

). Then, C = {c

, c

i j

}

Given this example, a Distributed Constraint Rea-

soning (DCR) platform must allow agents to have a

reliable communication tool that allows sending and

receiving messages, in order to ﬁnd the feasible solu-

tions.

3 RELATED WORK

Recently, B. Lutati and al. (Lutati et al., 2014) have

proposed a MAS platform, called AgentZero. This

Figure 1: Meeting Scheduling Problem modeled as DisCSP.

tool can be considered as a new addition to the avail-

able MAS tools in general and to the DCR research

ﬁeld in particular. The authors claim that AgentZero

is generic and applicable to many domains, speciﬁ-

cally introducing beneﬁts for the DCR simulation do-

main. However, the platform has been designed only

for simulation use and used only by researchers in

Distributed Constraint Reasoning. So developing and

setting computer software for real problems based on

DCR is not simple and remains a difﬁcult task for

users in general.

In (Petcu, 2006) A. Petcu. Proposes a Frame-

work for Open Distributed Optimization (FRODO).

The framework is implemented in Java, and simu-

lates a multiagent environment in a single Java vir-

tual machine. Each agent in the environment is exe-

cuted asynchronously in a separate execution thread,

and communicates with its peers through message ex-

change. FRODO comes with several built in algo-

rithms and a suite of problem generators for bench-

marking.

The authors of (Sultanik et al., 2007) proposed a

open-source tool for solving DCR, called DCOPolis.

DCOPolis is an open-source framework designed to

abstract algorithm implementation from the underly-

ing platform (i.e. hardware, network, operating sys-

tem). This allows a single implementation of an al-

gorithm to be run in simulation (i.e. on top of the

NS2 network simulator with AgentJ).DCOPolis dif-

fers from existing DCR frameworks and simulators,

however, it supports a novel type of simulation in

which the runtime of any distributed algorithm can be

accurately estimated on a single physical computer.

Researchers in DCR are concerned with devel-

oping new algorithms, and comparing their perfor-

mance with existing algorithms. Therefore, in (Wahbi

et al., 2011) the authors present an open source Java

library , called DisChoco which aims at implement-

ing DCR algorithms from an abstract model of agent

. DisChoco allows to represent both DisCSPs and

DCOPs, as opposed to other platforms. A single im-

ICAART2015-InternationalConferenceonAgentsandArtificialIntelligence

plementation of a DCR algorithm can run as simula-

tion on a single machine.DisChoco is a elegant plat-

form, but all the different issues of realistic uses and

actual deployment have not been addressed.

Developing intelligent software applications

based on DCR algorithms is a difﬁcult task, because

the programmer must explicitly juggle between many

very different concerns, including centralized pro-

gramming, distributed programming, asynchronous

and concurrent management of distributed structures,

communication concerns and others. In addition,

there are very few open-source tools for solving DCR

problems in a physically distributed environment.

In this paper we have been looking for a singular

platform that would possess not only simulation

qualities, but especially designed for realistic and

actual deployment. JChoc platform is a new added

value which allows bridging the gap between simula-

tion and realistic use. To our knowledge, this is the

ﬁrst DCR platform respecting FIPA standards and

speciﬁcations.

4 JCHOC PLATFORM

4.1 JChoc Description

The best way to prove the effectiveness of a proposed

distributed constraint reasoning algorithm, is to use

it in a realistic multi-platform agent. This is how

we can reduce the gap between theory and practice.

JChoc is a distributed constraints multiagent platform

proposed for solving combinatorial problems within a

speciﬁc distributed environment. It can also be used

to analyze and test the algorithms proposed by con-

straints programming community. This platform is

presented in the form of programming environment

(API) and applications to help different types of users.

Hence, JChoc can be easily appropriated by two main

actors:

• Developers to design and develop applications

(e.i. client application, web application, mobile

application, etc...) within distributed constraints

programming based on JChoc API;

• Non-expert user to interact directly with appli-

cations based on distributed constraints program-

ming.

This proposed platform has several advantages:

• A distributed constraints problem can be easily

addressed and solved in a realistic environment by

unsophisticated users;

• The performances of the proposed protocols (i.e.

ABT, AFC, Adopt, etc...) can be actually tested

and proved in a realistic communication channel

(i.e. WLAN WPAN WMAN WWAN);

• It offers a modular software architecture which

accepts extensions easily (i.e. security, conﬁden-

tiality, cryptography, etc...);

• Thanks to the extensibility of JADE communica-

tion model (JADE, 2013), JChoc allows the de-

velopment of multiagent systems and applications

consistent with Foundation for Intelligent Physi-

cal Agents (FIPA)

standards and speciﬁcations;

• Thanks to the the robustness of Choco plat-

form (Jussien et al., 2008), complex agent (i.e.

multiple variables per agent) can easily address

and solve its local sub-problem and use solutions

as a compiled domain.

This platform consists of several modules pre-

sented as services. The main constraint programming

services offered are based Distributed Constraint Rea-

soning Protocols (DCRP) and Choco Solver (CS).

Choco is a platform for research in centralized con-

straint programming and combinatorial optimization.

This choice of Choco enabled us to beneﬁt from the

modules already implemented in it. In the next sec-

tion, we will study the different elements of JChoc

platform.

4.2 JChoc Architecture

JChoc architecture is motivated by FIPA speciﬁca-

tions, it allows the development of multiagent systems

and applications conforming to MAS standards. It is

implemented in JAVA and provides classes that imple-

ment and inherit from JADE and Choco platforms to

deﬁne the behavior of speciﬁc agents. Figure 2 rep-

resents the main JChoc architectural elements. This

platform has ﬁve main modules.

• DCRP Distributed Constraint Reasoning

Protocols provides distributed constraints pro-

tocols as service. This element deﬁnes new types

of messages and implements the behavior of the

agent when receiving and sending a speciﬁc type

of information (e.i. ABT, AFC, Adopt, etc...);

• CS Choco Solver provides the ability to ad-

dress and resolve local CSP sub-problem;

• DF Director Facilitator provides a service of

”yellow pages” to the platform;

• ACC Agent Communication Channel man-

ages the communication between agents;

http://www.ﬁpa.org/

JChocDisSolver-BridgingtheGapBetweenSimulationandRealisticUse

• AMS Agent Management System oversees

the registration of agents, their authentication,

their access and the use of the system.

These ﬁve modules are activated at each time the plat-

form is started.

Figure 2: The JChoc Architecture.

The JADE agent is also a key player in our plat-

form. Thanks to JADE an Agent Identiﬁer (AID)

identiﬁes an agent uniquely.

JChoc uses extensively a snifﬁng tool for debug-

ging, or simply documenting conversations between

agents. The sniffer subscribes to AMS agent to be

notiﬁed of all platform events and of all message ex-

changes between a set of speciﬁed agents. When

the user decides to monitor an agent or a group of

agents, every message directed to, or coming from,

that agent/group is tracked and displayed in the snif-

fer GUI. The user can select and view the details of

every individual message, save the message or serial-

ize an entire conversation as a binary ﬁle.

5 USING JCHOC IN

DISTRIBUTED ENVIRONMENT

In this section we present how to use the JChoc plat-

form in real distributed environment. The MSP prob-

lem depicted in ﬁgure 1 is used to illustrate this pro-

posed platform. Initially we generate a sub-problem

for each agent involved in the global DisCSP prob-

lem, modeled by an expert as an XML ﬁle, which al-

lows standardizing the syntactic structure of the sub-

problems. A sub-problem containing only the infor-

mation necessary for a single agent, so he can partici-

pate in solving the global problem in a real distributed

environment.

Figure 3 shows an example of representation of

the MSP sub-problem deﬁned above in the XDisCSP

format. Each variable has a unique ID, which is

the concatenation of the ID of its owner agent and

index of the variable in the agent. This is neces-

sary when deﬁning constraints (scope of constraints).

For constraints, we used two types of constraints:

TKC for Totally Known Constraint and PKC for Par-

tially Known Constraint. Constraints can be de-

ﬁned in extension or as a Boolean function. Dif-

ferent types of constraints are predeﬁned: equal to

eq(M

, M

), different from ne(M

, M

), greater than

or equal ge(M

, M

), greater than gt(M

),etc. In

this sub-problem there is 1 complex agent A

which

controls exactly 3 variables. The domain of A

con-

Figure 3: Deﬁnition of DMS sub-problem in XDisCSP for-

mat.

ICAART2015-InternationalConferenceonAgentsandArtificialIntelligence

tain 14 values D

= {1...14}. There are three con-

straints of Arrival time ge(abs(sub(M

, M

)): the ﬁrst

constraint is between M

3.2

and M

3.3

the second one is

between M

3.3

and M

3.4

and the third is between M

3.2

and M

3.4

, three constraints of equality eq(M

, M

): be-

tween M

1.4

and M

3.4

, between M

1.3

and M

3.3

, between

2.2

and M

3.2

after deﬁning our sub-problem we can

conﬁgure our solver.

Once the sub-problem is generated, we can test the

functioning of the platform in a physically distributed

environment. So we chose machines that simulate the

different agents of the problem, and ﬁled each sub-

problem in a machine, before launching it.

Figure 4 shows how the master launches its com-

munication interface listening on the network. We

start with instantiate the dissolver object (line 6), This

class models the distributed problem when JChoc is

used to solve a problem in a real distributed environ-

ment. All information on distributed problem is en-

capsulated in this object (identities of agents, inter-

agent constraints, protocol, etc.). Then, we deﬁne the

type of master (line 7) (ABT in this case). Finally, we

trigger the container and we launch the master (lines

8-9).

Figure 4: How the master launches its communication in-

terface.

Figures 5-8 show how to launch JChoc agents. We

start with instantiate the DisSolver object (line 6), fol-

lowed by the agent and distributed sub-problem dec-

laration which speciﬁes the resolution algorithm to

be used (line 7-8). Next, the declaration of the con-

tainer containing the master with its IP address (line

9). Eventually, we launch the agent (line10).

Figure 5: How to implement and launch JChoc DisSolver

in Omar agent (A1).

Figure 6: How to implement and launch JChoc DisSolver

in Jean agent (A2).

Figure 7: How to implement and launch JChoc DisSolver

in Yun agent (A3).

Figure 8: How to implement and launch JChoc DisSolver

in Mamadou agent (A4).

The master waits for the conﬁrmation of creation

of all agents before ordering the start of the search.

Thus, the problem can be solved using a speciﬁed im-

plemented protocol (ABT for example).

6 EXPERIMENTAL RESULTS

6.1 Conﬁguration Example

To experiment the JChoc platform in a physically dis-

tributed environment, we chose ﬁve machines with

features 2.93 GHz, CORE(TM) 2 duo with 2 GB

RAM that simulate agents. These machines are con-

nected via the WLAN of our laboratory. We also

chose ABT algorithm to solve Meeting Scheduling

problems (MSP). In ﬁgure 1 above, we depict an ex-

ample of problem solved by this platform in a live

distributed environment network. This ﬁgure illus-

JChocDisSolver-BridgingtheGapBetweenSimulationandRealisticUse

trates an instance of MSP viewed as DisCSP where

each agent has a personal private calendar and a set

of meetings each taking place in a speciﬁed location.

In that example, there are four agents, A

, A

and

, and four meetings, m

, m

and m

. Each agent

has its own calendar divided into 14 slots. The time

required for traveling among places where meetings

can be scheduled is 2 slots.

We have intentionally limited the number of

agents to 4 for this problem needs, but the number

of the agents can be easily extended to N4 for the

neediest problems.

Figures 9 and 10 show the GUI of the sniffer agent

at the start and the end of ABT resolution. The canvas

provides a graphical representation of the messages

exchanged between sniffed ABTagents, where each

arrow represents a message and each color identiﬁes

a type of conversation. For example agent A

sends an

OK? message to informs A

that he has done a new

assignment m

1.1

:1 (line 5).

Figure 9: The start on sniffer agent GUI.

Figure 10: The ﬁnish on sniffer agent GUI.

If no new consistent value is found (line 10), A

generates a new nogood m

1.3

:3 ∧ m

1.4

:5 ⇒ m

2.2

6= 5

by the resolution of existing nogoods. Eventually, the

system can stabilize in a state where each agent has

a value and no constraint is violated. This state is

a global solution and the network has reached quies-

cence, meaning that no message is traveling through

it (lines 37, 40, 43, 46). Once the solution is found,

the master should be advised to spread the stop order

to all agents (lines 49-52).

A solution to this example is :

−→(m

1.1

: 3; m

1.3

: 7; m

1.4

: 1), A

−→(m

2.1

3; m

2.2

: 5), A

−→(m

3.2

: 5; m

3.3

: 7; m

3.4

: 1),

−→(m

4.4

: 1).

6.2 Platform Scalability

The scalability of JChoc is the ability of the system,

network, and process to handle a growing amount of

work in a capable manner and its ability to be enlarged

to accommodate that growth. In order to experiment

our platform, we consider a large number of MSP in-

stances. These Meeting Scheduling Problem are char-

acterized by < m, p, n, d, h,t, a >, where m is the num-

ber of meetings, p is the number of participants, n is

the number of inter-agent constraints d determines the

number of days. Different time slots are available for

each meeting, and h is the number of hours per day,

t is a duration of the meeting and a is the percent-

age of availability for each participant. We present

our results for the class < m, p, n, 5, 10, 1, 90% > and

we vary three parameters : m, p, n (each agent has 2

meetings):

#p #m #n #messages Time (ms)

4 8 3 11 17070

5 10 5 11 17204

6 12 6 14 16144

7 14 7 14 17073

8 16 8 19 19180

9 18 9 24 20210

10 20 10 22 18294

11 22 11 32 20197

12 24 12 27 18516

13 26 15 30 20370

14 28 33 51 26073

15 30 35 105 31103

16 32 29 69 28914

17 34 33 175 38324

18 36 35 139 43172

19 38 38 141 37121

20 40 43 94 33457

Figure 11: Performance of JChoc platform using ABT pro-

tocol on the Meeting Scheduling Problem (MSP).

ICAART2015-InternationalConferenceonAgentsandArtificialIntelligence

As shown in experimental results, in ﬁgure 7, the

performance of our platform is measured in terms of

network load (number of messages) and run-time ex-

ecution. From these preliminary results we see that

JChoc platform performs rapidly in small instances

(#p ∈ [4, 14]). The number of messages increases for

#p ∈ [15, 18] and reduces for #p > 18. This scalabil-

ity behavior is due to complexity of MSP problems.

When the instance is hard the problem can be solved

rapidly.

7 CONCLUSION

In this paper, we have proposed a modular, reli-

able, deployable and distributed software architec-

ture, called JChoc DisSolver, which can be used eas-

ily for several real combinatorial problems. The main

purpose of our platform is to break down the barriers

to building new and innovative applications. The pos-

sibility of combining the expressiveness of Choco, the

extensibility of JADE and our powerful Distributed

Constraint Reasoning Add-On bring a strong added

value in the development of innovative applications

based on Constraints Programming paradigm. The

JChoc platform presented in this paper has been de-

signed to support extensions: security, cryptography.

In this work, we have implemented ABT protocol

and solved the Meeting Scheduling problem (MSP)

in a real distributed environment. We found that by

using this platform we can adopt easily any proposed

protocol for solving distributed constraint problem in

such environment.

Future activities are focusing on enhancing the

platform by the implementation of other DCR algo-

rithms and to enrich the graphical user interface to

make it easier to use for researchers. Another direc-

tion of improvement is to allow JChoc platform to be

suitable to mobile devices. We plan also to implement

new approaches of conﬁdentiality.

REFERENCES

ejar, R., Domshlak, C., Fern

andez, C., Gomes, C., Kr-

ishnamachari, B., Selman, B., and Valls, M. (2005).

Sensor networks and distributed csp: communication,

computation and complexity. In Artiﬁcial Intelligence

161:1-2, 117-148.

Bessiere, C., Brito, I., Maestre, A., and Meseguer, P. (2005).

Asynchronous backtracking without adding links: A

new member in the abt family. In Artiﬁcial Intelli-

gence,161:724.

Ezzahir, R., Bessiere, C., Benelallam, I., Bouyakhf, E., and

Belaissaoui, M. (2008). Dynamic backtracking for

distributed constraint optimization. In Proceeding of

the 2008 conference on ECAI 2008, 901-902. Amster-

dam, The Netherlands, The Netherlands, IOS Press.

Ezzahir, R., Bessiere, C., Wahbi, M., Benelallam, I.,

and Bouyakhf, E. (2009). Asynchronous interlevel

forward-checking for discsps. In Principles and Prac-

tice of Constraint Programming (CP-09).

Galley, M. (2000). Distributed constraint pro-

gramming platform using sjavap. In

http://cs.ﬁt.edu/Projects/asl/#MELY.

Gershman, A., Meisels, A., and Zivan, R. (2009). Asyn-

chronous forward bounding for distributed cops. In

Journal of Artiﬁcial Intelligence Research, 34, 61-88.

Hamadi, Y. (2003). Disolver : A distributed constraint

solver. In Technical Report MSR-TR-2003-91, Mi-

crosoft Research.

JADE (2013). Java agent developpement framework. In

URL http://jade.tilab.com/.

Jussien, N., Rochart, G., and Lorcal, X. (2008). Choco:

an open source java constraint programming library.

In CPAIOR’08 Workshop on Open-Source Software

for Integer and Contraint Programming(OSSICP’08).

France, Paris.

Kitano, H., Tadokoro, S., Noda, I., Matsubara, H., Taka-

hashi, T., Shinjou, A., and Shimada, S. (1999).

Robocup rescue: Search and rescue in large-scale dis-

aster as a domain for autonomous agents research. In

IEEE International Conference on System, Man, and

Cybernetics.

Lutati, B., Levit, V., , and Meisels, A. (2014). Agentzero:

A framework for simulating and evaluating multiagent

algorithms. In Agent-Oriented Software Engineering.

Springer.

Meisels, A. and Lavee, O. (2004). Using additional in-

formation in discsp search. In 5th workshop on dis-

tributed constraints reasoning, DCR’04.

Meisels, A. and Zivan, R. (2007). Asynchronous forward-

checking for discsps. In Constraints 12, 131-150.

Modi, P., Shen, W., Tambe, M., and Yokoo, M. (2005).

Adopt: asynchronous distributed constraints opti-

mization with quality guarantees. In Artiﬁcial Intel-

ligence 161:1-2, 149 180.

Petcu, A. (2006). Frodo: a framework for open/distributed

optimization. In Technical Report Technical Re-

port EPFL:2006/001, LIA, EPFL, CH-1015 Lau-

sanne,http://liawww.epﬂ.ch/frodo/.

Sultanik, E., Lass, R., and Regli, W. (2007). Dcopolis: A

framework for simulating and deploying distributed

constraint optimization algorithms. In CP-DCR.

Wahbi, M., Ezzahir, R., Bessiere, C., and Bouyakhf, E.

(2011). Dischoco 2: A platform for distributed con-

straint reasoning. In DCR’11, pages 112-121. URL

http://www.lirmm.fr/coconut/dischoco/.

Wallace, J. R. and Freuder, C. E. (2002). Constraint-

based multi-agent meeting scheduling: effects of

agent heterogeneity on performance and privacy loss.

In 3rd workshop on distributed constrait reasoning,

DCR’02, pages 176-182.

Yeoh, W., Felner, A., and Koenig, S. (2008). Bnb-adopt: an

asynchronous branch and bound dcop algorithm. In

JChocDisSolver-BridgingtheGapBetweenSimulationandRealisticUse

AAMAS08: Proceedings of the 7th international joint

conference on Autonomous agents and multiagent sys-

tems, 591-598.

Yokoo, M. (2000). Algorithms for distributed constraint sat-

isfaction problems: A review. In Autonomous Agents

and multiagent Systems 3, 185-207.

Yokoo, M. (2001). Distributed constraint satisfaction. In

:Foundation of Cooperation in multiagent Systems.

Springer.

Yokoo, M., Durfee, E., Ishida, T., and Kuwabara, K. (1992).

Distributed constraint satisfaction for formalizing dis-

tributed problem solving. In International Conference

on Distributed Computing Systems, 614-621.

Yokoo, M., Durfee, E., Ishida, T., and Kuwabara, K. (1998).

The distributed constraint satisfaction problem: for-

malization and algorithms. In IEEE Transactions on

Knowledge and Data Engineering, 10(5):673-685.

Yokoo, M. and Hirayama, K. (1995). Distributed breakout

algorithm for solving distributed constraint satisfac-

tion problems. In Proceedings of the First Interna-

tional Conference on MultiAgent Systems. MIT Press.

ICAART2015-InternationalConferenceonAgentsandArtificialIntelligence