MULTICRITERIA DECISION AID USE FOR CONFLICTING

AGENT PREFERENCES MODELING UNDER NEGOTIATION

Noria Taghezout and Abdelkader Adla

Computer Science Department, University of Es-Senia Oran, BP 1524, El-M' Naouer, 31000, Oran, Algeria

IRIT - Paul Sabatier University, Toulouse, France

Keywords: Multi agent System (MAS), Negotiation, Decision Support System (DSS), ISP (Integrated Station of

Production), Dynamic scheduling, ELECTRE III.

Abstract: Individual decisions in multi-agent domains are rarely enough for producing optimal plans which satisfy all

the goals. Therefore, agents need to cooperate to generate the best multi-agent plan through sharing tentative

solutions, exchanging sub goals, or having other agents to satisfy their goals. In this paper, we propose a

new negotiation mechanism independent of domain properties to handle real-time goals. The mechanism is

based on the well-known Contract net Protocol. Integrated Station of Production agents will be equipped

with a sufficient behavior to carry out practical operations and simultaneously react to the complex

problems caused by the dynamic scheduling in real situations. These agents express their preferences by

using ELECTRE III method in order to resolve differences. The approach is tested for simple scenarios.

1 INTRODUCTION

The next generation manufacturing systems will be

more strongly time-oriented (or highly responsive),

while still focusing on cost and quality. So, it will be

necessary to satisfy requirements such as: (i) quick

response to external order changes and unexpected

disturbances from both internal and external

environments, and (ii) embodiment of human factors

into manufacturing systems.

As a consequence, a major research topic in

computer science over the past two decades has been

the development of tools and techniques to model

understand and implement systems in which

interaction is the norm.

Recently, agent technology has been considered

as an important approach for developing industrial

distributed systems. It has particularly been

recognized as a promising paradigm for next

generation manufacturing systems (Shen, 2006).

In distributed intelligent manufacturing systems,

agents can be applied and implemented in different

ways, the most interesting for our study are:

(i) Agents can be used to encapsulate

manufacturing activities in a distributed

environment by using a functional

decomposition approach. Examples of such

functional agents include order processing,

product design, production planning and

scheduling and simulation.

(ii) Agents can be used to represent negotiation

partners, either physical plants or virtual

players; they also can be used to implement

special services in multi agent systems like

facilitators and mediators. A good discussion on

agent technology can be found in (Shen, 2006).

This paper proposes a new negotiation mechanism

independent of domain properties to handle real-

time goals, and discusses some issues in

implementing agent coordination and agent

negotiation in real time scheduling. The proposed

model is based on multi agent architecture. Agents

provide some actions or services including the

collecting of data from heterogeneous sources,

information integration and analysis, real time

scheduling and decision making.

In order to implement Decision Making abilities,

the Electre III methodology is chosen for the

possibility given to decision makers to treat

imprecise and subjective data. The Contract Net

Protocol is used because of its facility to implement

negotiation protocols

The paper is organized as follows: Section 2

presents the related work. The DSS architecture and

the most important agents are given in Section 3 and

508

Taghezout N. and Adla A. (2008).

MULTICRITERIA DECISION AID USE FOR CONFLICTING AGENT PREFERENCES MODELING UNDER NEGOTIATION.

In Proceedings of the Tenth International Conference on Enterprise Information Systems - AIDSS, pages 508-512

DOI: 10.5220/0001701205080512

 SciTePress

Section 4. In Section5, we present the negotiation

protocol and its facilitating techniques. Section 6 is

devoted to the integration of the multicriterion

method ELECTRE III in the decision-making

processes implemented in the internal structure of

the negotiation agent (ISP): a scenario will be given.

And, finally Section 7 concludes the paper.

2 RELATED WORK

Yee-Ming and al (Chen, 2005) develop a

collaborative framework of a distributed agent-based

intelligence system with a two-stage decision-

making process for dynamic scheduling. Many

features characterize the framework; more precisely,

the two-stage decision-making process, the fuzzy

decision-making process and the compensatory

negotiation process are adequate for distributed

participants to deal with imprecise and subjective

information, to conduct practical operations.

The work presented in (Reaidy, 2007) uses the

PABADIS architecture to model a distributed

manufacturing system. Basic components in

PABADIS are agents and services; they work in

cooperation and perform distributed tasks in a

networked manufacturing plant.

3 THE APPROACH PROPOSED

IN THIS PAPER

In the resolution of real time production

management problems, each decision-making

process of piloting is generally a multicriteria

process (Taghezout, 2006): the task assignment for

example, is a decision-making process which results

from a study on criteria of costs production, time of

change of series, convoying time, production

quality, etc.

The multicriteria methodology exploitation

allows integrating the set of these constraints, in

particular by the fact that the assumptions on which

the latter are based are closer to reality than

optimization methods. In addition, the multicriteria

approach facilitates the integration of human

operator to DSS (Adla, 2006).

In real time production management, the DSS

memorizes the current state-of the workshop. It

knows constantly the whole of the decisions and the

possible events involved. A detailed description of

the workshop’s state was given in our previous work

(Taghezout, 2007). We distinguish 3 contexts for the

decision-making aid: (1) Decision-making aid in the

context of an acceptable sequence; (2) Assistance for

the admissibility covering; and (3) Negotiation

support among different decision-making centres in

a dynamic context.

DSS gives the decision centers the opportunity to

make decisions in a dynamical context. A decision

aid is then increased by a negotiation support. The

system suggests the selected decision in the set of

planned solutions. As a conclusion, the proposed

DSS in this approach addresses the situations

described in levels 1 and 3.

The DSS architecture is composed of several

modules. Each module has its own functionalities

and objectives. The DSS architecture is described in

Figure 1. The analysis and reaction module is

developed thanks to a multi-agent technology. The

agent based system is decomposed into a supervisor

agent and several ISP agents. Each ISP agent has the

possibility to use resources. A detailed description is

given in (Taghezout, 2006) and (Taghezout, 2007).

4 DESCRIPTION OF THE

DECISION LEVELS

The decision-making takes place in two steps:

1. In the first step, ISP agents recognize the

encountered problems, and start the local decision-

making processes. In case of success, they adopt the

adequate behaviors. The basic principle of resolution

has been described in (Taghezout, 2007).

2. In the second step, delays in the planned task

execution or a conflicting situation cause a failure in

the complex problem resolution. ISP agents open

negotiation then. The protocol is based on the

classical contract Net approach. ISP agents express

their initial preferences, priorities and data in the

evaluation matrices. The decisional processes use

the multicriterion assistance method, ELETRE III.

ISP Agents, which are in several cases the most

important, correspond to:

• An ISP agent, which meets the problem during

its task execution, should make a decision in

collaboration with other ISP agents; it is called

the initiating ISP agent and is noted as IISP

(Initiating Integrated Station Production).

• An ISP agent, which undergoes the delay

consequences or disturbance in its task

execution because of a conflict on the common

resource or another unpredicted event, is called

participating ISP agent and is noted as PISP

(participating ISP).

MULTICRITERIA DECISION AID USE FOR CONFLICTING AGENT PREFERENCES MODELING UNDER

NEGOTIATION

509

Figure 1: Structure of negotiation Agent (ISP).

Figure 2: The negotiation model.

5 THE NEGOTIATION AID

Generally speaking, the outcome of a negotiation

depends on many parameters-including the agent’s

preferences, their reservation limits, their attitude

toward time and the strategies they used.

Although in most realistic situations it is not

possible for agents to have complete information

about each of these parameters for its opponent, it is

not uncommon for agents to have partial information

about some of them. The purpose of our study is not

to allow the agent select the optimal strategy, Some

Works do it for example the approach given in

(Shaheen,2001), but it helps to treat some

uncertainty.

6 DECISION AID THROUGH

ELECTRE III

Decision making is a difficult process due to factors,

such as information about incompleteness,

imprecision, and subjectivity factors which tend to

be resent in real life situations to lesser or greater

degree (Taghezout,2007). The multicriteria

methodology Electre III allows sorting out actions

likely to solve a decision problem, on the basis of

arguments on several criteria (Roy, 1977).

6.1 The Proposal of Negotiation Model

At the second stage of resource allocation decision-

making process, the IISP agent will open negotiation

Behaviours

Configuration of

resources

Data

Information

Subs

stem

Decision

Subsystem

Interface

Subsystem

Communicati

on Subsystem

Human

Operator

Expert

Interface

Total Diary

Control Subsystem

Analysis and Reaction

Module

Decision-

making aid

Generator of

proposal

Sub-problem

Negotiation

Coordinato

ISP Agent1

ISP Agent

Supervisor

ISP Agent2

ICEIS 2008 - International Conference on Enterprise Information Systems

510

with PISP agent that is concerned with the result of

ELECTRE III execution. The latter consists in

searching the best resource. The framework of the

negotiation model consists in various components

such as:

• The Alternatives: This component gathers

all the resources classified from the best to

the less good according to the classification

performed by ELECTRE III. It corresponds

to the multicriterion decision-making

process application in resolving the

problem of allocation of the best resource

in case of breakdowns.

• Criteria Updating: Each agent is equipped

with a module allowing it to calculate the

function production cost at any time.

• Selection Function: Each negotiation agent

possesses a selection function to evaluate

the proposals and counter-proposals for the

negotiation strategy.

Each negotiation agent needs to consult the

supervisor diary to know the state of execution of

the activities of each agent ISP. Agents execute

method ELECTRE III before making their strategic

and/or tactical decisions.

6.2 Listing Retained Criteria

The most relevant criteria in our study are gathered

in Table1 (Taghezout, 2006). We can consider that a

Failure or breakdown event is defined by the

following items in Figure 3.

Table 1: List of Criteria Retained for the Allocation.

Code

indicator

Entitled Signification

Axe

Min/Max

C1 Production cost Cost Min

C2 Time of a

resource

preparation of

an operation

Delay Min

C3 Potential

Transfer Time

Delay Min

C4 Next date of

availability

Delay Min

C5 Machine

reliability

indicator

Delay Max

C6 Attrition rate

Quality Max

Characteristic

tool

Quality Max

Level of

specialization

Quality Max

Figure 3: A breakdown event.

6.3 Negotiation Scenarios

We use the resource allocation problem to

demonstrate how the agents solve problems by

interactions among agents.

Scenario: Breakdown of a Resource

1. Resource n°1 controlled by agent 3 breaks down.

The analysis and reaction module discerns this event

and triggers off the associated behavior; if the

process fails; ISP agent n°3 re-redirects the re-

affectation request to the supervisor. This triggers

off behavior named as the second behavior at the

supervisor level.

2. The agent supervisor transmits the request

towards other ISP agents (ISP 1, ISP 2) and treats

the received answers to choose the best substitution

machine.

3. The result will be announced to the chosen ISP

agent as well as to the ISP agent applicant.

4. The ISP1 Agent answers favorably the supervisor

request (end of the first phase of the decision-

making process).

5. The required resource is also programmed for the

ISP4 agent according to the initial production

planning, ISP3 and ISP4 agents are found in a

conflicting situation.

6. The negotiation is then open: ISP3 agent becomes

ISSP

and ISP4 agent becomes PISP.

7. IISP agent activates the proposal generator, and

formulates a new contract proposal. It sends the

latter to PISP agent.

8. The agent formulates its contract, evaluates the

received proposals simultaneously thanks to the set

of preferences and priorities, contained initially in

the evaluation matrix (the decision-making module

presented in Figure 2 intervenes in the realization of

this step). The proposal or counter-proposal

evaluation is made by ELECTRE III. This is

according to the outclass algorithm.

MULTICRITERIA DECISION AID USE FOR CONFLICTING AGENT PREFERENCES MODELING UNDER

NEGOTIATION

511

7 CONCLUSIONS AND FUTURE

WORK

In this paper, we have addressed an agent

architecture-based model in order to present a

multicriteria DSS which can be applied to solve

some uncertainty problems in dynamic production

system scheduling. The established negotiation

contract thus deals with certain exceptions; it is

based on the agent concept. The major advantage

with this modeling consists in facilitating access to

the executed tasks carried out by entities ISP.

ELECTRE III is a tool that allows learning

information about the opponent’s preferences and

their relative weights.

For the future, the research will be extended

according to three important directions. Firstly,

given that scheduling robustness characterizes its

performance, we would like to extend the decision

support given by this approach to treat uncertainty.

Secondly, we propose to integrate software agents

and Web services at both the design level and

implementation level. Thus, we can treat a Web

service as a semi-autonomous agent. Finally, we

intend to integrate the human operator in the

decision support, several co-operation modes have

been defined with the decision support system.

REFERENCES

Adla, A., 2006. A Cooperative Intelligent Decision

Support System for Contingency Management. In

Journal of Computer Science 2 (10): 758-764, ISSN

1549-3636.

Chen, Y.-M., Wang, S.-C., 2005. Framework of agent-

based intelligence system with two stage decision

making process for distributed dynamic scheduling.

In Applied Soft Computing “, pp. 229-245.

Reaidy, J., Massote, P., Diep, .D. 2007. Comparison of

Negotiation protocols in dynamic agent-based

manufacturing Systems. In International Journal of

Production Economics 99 (26) 117-130.

ROY, B.: Electre III, un algorithme de classement fondé

sur une représentation floue des préférences en

présence de critères multiples, Rapport de recherche

du Lamsade 81, Paris (1977) 3-24.

S. Shaheen, F., Wooldridge, M., Jennings, N., 2001.

Optimal Negotiation Strategies for Agents with

Incomplete Information, Revised Papers from the 8th

International Workshop on Intelligent Agents VIII,

p.377-392, August 01-03.

Shen, W., YoonH-J., Norrie D-H., 2006. Applications of

agent-based systems in intelligent manufacturing“: An

updated review. In advanced engineering

INFORMATICS, pp. 415-431.

Taghezout, N., 2006. Expérimentation et Intégration de la

méthode Electre I dans système d’aide à la décision

appliqué aux SAP. In SNIB’06, 5eme Séminaire

National en informatique de Biskra, Vol.1, pp 196-

206.

Taghezout, N., Riad, A., Bouamrane, K., 2007.

Negotiation Strategy For a Distributed Resolution of

Real Time Production Management Problems. In

ACIT. LATTAKIA SYRIA pp 367-374.

Taghezout, N., Zaraté, P., 2007. A Multi agent Decision

Support System For Real Time Scheduling. In The 4th

International Workshop on Computer Supported

Activity Coordination (CSAC 2007), In Conjunction

with the ICEIS2007, Funchal, Madeira, Portugal, p.

55-65.

ICEIS 2008 - International Conference on Enterprise Information Systems

512