ProMAIS: A MULTI-AGENT MODEL FOR PRODUCTION
INFORMATION SYSTEMS
Lobna Hsairi, Khaled Ghédira
Laboratoire URIASIS, Institut Supérieur de Gestion
41, rue de la liberté-Cité Bouchoucha-Tunis, CP 2000 le Bardo, Tunisie
Faiez Gargouri
Laboratoire LARIS, Faculté des Sciences Economiques et de Gestion de
Sfax, Route de l'aérodrome km 4,Bp 1088-3018 Sfax, Tunisie
Keywords: Information System (IS), Cooperative Information System (CIS), Agent, Multi-Agent System (MAS),
Production system, ProMAIS, Manufacturing system.
Abstract: In the age of information proliferation and communication advances, Cooperative Information System (CIS)
technology becomes a vital factor for production system design in every modern enterprise. In fact, current
production system must hold to new strategic, economic and organizational structures in order to face new
challenges. Consequently, intelligent software based on agent technology emerges to improve system design
on the one hand, and to increase production profitability and enterprise competitive position on the other
hand. This paper starts with an analytical description of logical and physical flows dealt with manufacturing,
then proposes a Production Multi-Agent Information System (ProMAIS). ProMAIS is a collection of
stationary and intelligent agent-agencies with specialized expertises, interacting to carry out the shared
objectives: cost-effective production in promised delay and adaptability to the changes. In order to bring
ProMAIS’s dynamic aspect out, interaction protocols are specially zoomed out by cooperation, negotiation
and Contract Net protocols.
1 INTRODUCTION
The use of the Information System (IS) technology
today has already imbedded into every modern
enterprise’s core as a result of the quantity of
information and the need of sharing them.
Recent developments, driven by the proliferation
of parallel architecture, the advances in wide area
net-working and the Internet democratization, have
made the information systems parts of increasingly
complex systems. That’s how the classic IS are
becoming hard to generate necessitating to break
them up into sub-systems and distribute them. This
results in the birth of the Cooperative Information
Systems (CIS) technology. The CIS aim at
continued cooperativity among user groups through
componentized networks of information systems.
One of the most important classes of IS are
production systems. Such system consists in an
important domain from strategic, economic and
organizational points of view. A number of new
concepts of such systems has been proposed and
developed in the past few years, e.g., Virtual
Organizations, Virtual Enterprise, Supply Chain
Management and Electronic Commerce, etc. These
concepts are similar or overlapping.
Currently, the world is undergoing a Hi-Tech
industrial revolution with information technology as
its main feature. The production systems will be
characterized by intensively concurrent engineering
based on information technologies such as
digitalization, computer network, artificial
intelligence (Qiao et al, 1999).
To compete effectively in today’s markets,
manufacturers must be moving towards an open and
297
Hsairi L., Ghédira K. and Gargouri F. (2004).
ProMAIS: A MULTI-AGENT MODEL FOR PRODUCTION INFORMATION SYSTEMS.
In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 297-302
DOI: 10.5220/0002602502970302
Copyright
c
SciTePress
distributed architecture. Increasingly, traditional
centralized and sequential manufacturing or
production system decisions are being found
insufficiently flexible to respond to changing
production styles and highly dynamic variations in
production requirements.
The intelligent software agents technology
provides a natural way to overcome such problems.
In sharp contrast to traditional software programs,
software agents are programs that help people to
solve problems by collaborating with other software
agents and other resources in the network. That’s
how, the use of the agent-oriented approach is
imposed in contrast to the object-oriented approach.
Recently, agent technology has been considered
as an important approach for developing intelligent
manufacturing and production systems. Agents help
to capture individual interests, local decision-making
using incomplete information, autonomy,
responsiveness, robustness and modular, distributed,
reconfigurable organizational structures.
A Multi-Agent System (MAS), as a society of
autonomous agents, is an inherently open and
distributed system. It is formed by a group of agents
combined with each other through a network to
cooperatively solve a common problem. The
system’s inter-agent communication capability
provides the essential means for agent collaboration
that aids interoperability of the system.
There is a multifold advantage in using an
agent-based approach for manufacturing and
production systems. First, information for
production system is stored and processed in a
distributed manner in contrast to that stored in one
large program. Second, the agent-based approach
makes the incremental improvement of the
production system possible through learning and
cooperation between agents. Third, it provides a
promising method for enterprise integration
(Qiao et al 1999).
In this paper we present our research efforts and
experiences in analysing the information (logical)
and physical flows between the various divisions of
the enterprise, modeling an intra-organization
cooperative information system based on
decomposition into recursive agents, developing a
MAS architecture named Production Multi-Agent
Information Systems (ProMAIS). ProMAIS is an
ongoing project. It is an organizational structure
which draws descriptions of flows between actors in
the production systems. It allows the actors involved
to communicate with each other, to conserve the
autonomy of every one in making local decision, to
guarantee the adaptability and ability to quickly
respond to the changes and the new conditions of
markets and to react quickly to unexpected risks.
The remainder of this paper is structured as
follows: section 2 reviews research literature,
section 3 presents ProMAIS the Multi-Agent model
for Production Information Systems, section 4
presents a prototype implementation, section 5 gives
concluding remarks and perspectives.
2 RESEARCH LITERATURE OF
AGENT TECHNOLOGY IN THE
INDUSTRY
Recently, agent technology has been considered as
an important approach for developing industrial
distributed systems. A number of researchers have
attempted to play agent technology to industrial
enterprise integration. In this section, we briefly
review some interesting projects in this domain such
as MetaMorph II (Shen et al, 1998) and an Agent-
Based Intelligent Manufacturing System for the 21
st
Century (Qiao et al, 1999).
At the university of Calgary, the MetaMorph II
project start at the beginning of 1997. The aim of
this project was to integrate the manufacturing
enterprise’s activities such as design, marketing,
planning, scheduling, execution and product
distribution, with those of its suppliers, customers
and partners into an open distributed intelligent
environment. To do this, the researchers propose an
Agent-Based Manufacturing Enterprise
Infrastructure (ABMEI) combining and extending
the architecture used in previous projects
MetaMorph I (Maturana et al, 1996), ABCDE
(Balasubramanian et al, 1996) and DIDE
(Shen et al, 1996). In this Infrastructure, the system
is primarily organized at the highest level through
sub-system mediators which are connected via the
Internet/Intranet. Each sub-system is connected to
the system through a special mediator. Each
manufacturing enterprise must have at least one
Enterprise Mediator. It can be considered as the
administration center of the enterprise. All other
mediators should register with it.
Qiao shows in (Qiao et al, 1999) that the
intelligent manufacturing system is a solution of the
problems of the 21
st
century manufacturing industry.
In order to establish an Agent-Based Manufacturing
System, they propose an intelligent object called
Manufacturing Agent (MA). According to the
general structure of the manufacturing enterprise,
they propose the Agent-Based Manufacturing
System architecture, in which the whole operation
logic of a manufacturing enterprise is divided into
four parts: central part, management, planning and
production. Each part consists of a group of MAs.
ICEIS 2004 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS
298
All the MAs are connected to the distributed
enterprise network or Intranet. This MA network is
heterarchical in nature.
3 PRODUCTION MULTI-AGENT
INFORMATION SYSTEMS
(ProMAIS)
In order to present our ProMAIS model, we must
first exhibit how to ‘agentify’ the IS.
3.1 Information Systems as MASs
The definition of Hayes-Roth (Hayes-Roth, 1995)
advocates the most important features of an agent.
Therefore, intelligent agents continuously perform
three functions. First, the ability of perception of
dynamic conditions in the environment. Second, the
capability of action to affect conditions in the
environment. Finally, the ability of reasoning to
interpret perceptions, solve problems, draw
inferences, and determine actions. Autonomous
agents have the opportunity and ability to make
decisions of their own.
Nevertheless, in the case of an agentified IS, the
perception of dynamic conditions in the environment
refers to different kinds of incoming messages that
perform a communication event (such as receiving
an acknowledgement of a sales order or receiving a
request for a sales quotation) and environment
events (such as receiving a payment). The action to
affect conditions in the environment refers to
communication acts of the agentified IS (such as
acknowledging a sales order) and to physical acts
(such as making a payment or delivering goods).
Finally, the reasoning to interpret perceptions, solve
problems, draw inferences and determine actions
refers to the computational inference of correct
answers to queries, the proper processing of
incoming messages and the determination of proper
actions (such as checking all sales orders of a
customer whose credibility is in question or using an
alert when the fulfilment of a commitment is
overdue).
Therefore, an IS may be explicitly designed as
an agent (Wagner, 2000) by treating its information
items as its beliefs or knowledge, adding further
mental components like perceptions (as incoming
messages) memory and commitments and providing
support for agent-to-agent communication on the
basis of a standard Agent Communication Language
(ACL) (Labrou et al, 1999).
3.2 System design
The actors of an enterprise are collectively
responsible for carrying out a common objective.
The common objective consists in making cheaper
and more cost-effective products. However, ‘cost-
effective’ means the production with the best
quality, in the promised delay and with a much
lower price. The entities are very interdependent and
interact to respond to the same objective and to
assure the improvement of its performance. Each
entity operates under a number of constraints and
objectives and is responsible for its process and its
production. Hence, the performance of each entity
depends on the performance of the others and of
their good cooperation. Thus, each entity has its own
mission to contribute to the global goal of the net
and it is connected to other entities defined with hard
collabration relationship. Modeling such entities
with agent technologies provides a natural way to
cover such features. Our proposed model ProMAIS
is a way of modeling such entities.
ProMAIS is based on reflexivity and
heterogeneity. The reflexivity implies that an agent
is itself a MAS and inversely. Agents should have
capability of controling their interaction with the
external environment and their internal decision.
ProMAIS is composed of a number of agents
forming a MAS (see figure 1). The diversity of
agents implies the heterogeneity of the architecture
of knowldge representation and exchanged
messages.
Thus, our approach involves a number of agents
continually in interaction. ProMAIS is built on the
eco-resolution model proposed by
Ferber and Jacopin (Ferber & Jacopin, 1990) and
enriched by Ghedira (Ghedira, 1994). It’s consist in
a MAS in which each agent has a local memory that
consists of the static and the dynamic knowledge,
the acquaintance (the agents that it know and
communicate with) and a mailbox in which it
receives different kinds of messages, that it stores
and treats one by one according to their priority
degree. In addition, each agent has a behaviour
based on satisfaction (see figure 2).
While (MailBox not empty)
Withdraw_And_Treat_First_Message;
If Satisfy
Satisfaction_behaviour;
Else
Unsatisfaction_Behaviour;
EndIf
Figure 2: The general agent behaviour.
ProMAIS: A MULTI-AGENT MODEL FOR PRODUCTION INFORMATION SYSTEMS
299
In the ProMAIS organizational structure, we can
distinguish two environments: internal and external
one. The external environment includes all kinds of
external actors to the enterprise such as Partner
agents, Supplier agents, Customer agents and
Competitor agents. As part of our study, we take an
interest in Customer and Supplier agents ones. The
internal environment contains all entities of a
manufacturing enterprise. In our proposed model,
agents having a MAS architecture have a common
mailbox beside their individual ones. This common
mailbox belongs to a Messenger agent. This agent
has the role of communicating the received message
to the appropriete agent. Furthermore, it constitues
the interface between the local MAS agents and the
other agents in the global model.
The Monitor, the Design & Engeneering and the
Stock Management & Procurement agents have a
MAS architecture in ProMAIS. The organizational
structure of each one is shown in figures 3, 4 and 5
respectivly.
Figure 1: Organizational structure and Interaction model of ProMAIS.
Planning
Agent
Manufacturing
Agent
Scheduling
Agent
Stock Management
& Supply Agent
Agent
Quality Control
Agent
Design &
Engineering Agent
Internal Environment
External Environment
Caption
Cooperation & communication
Communication
Coordination & communication
Negotiation
Monitor
Agent
Customer Agents
Partner Agents
Competitor Agents
Supplier Agents
Figure 3: Organizational structure of Monitor Agent
.
Marketing
Agent
Finance &
Account
Agent
Research &
Development
Agent
M
anageme
nt
Agent
Messenger
Agent
Monitor Agent
Figure 4: Organizational structure of Design &
Engineering Agent.
Design
Agent
E
ngineerin
g
Agent
Messenger
Agent
Design &
Engineering Agent
Figure 5: Organizational structure of Stock
Mangement & Procurement Agent.
RM : Raw Material
SFP: Semi-Finished Product
FP : Finished Product
Procurement
Agent
FP Stock
Agent
Messenger
Agent
Stock Management &
Supply Agent
RM & SFP
Stock Agent
ICEIS 2004 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS
300
3.3 MAS Dynamic aspect
3.3.1 Communication protocol
A communication act consists of an information
transmission from a sender to a receiver. This
information is encoded into an ACL
(Agent Communication Language), and it is decoded
by the receiver (Labrou et al, 1999).
Our MAS communication protocol is based on
an intentional point-to-point mode (between two
agents), a broadcast mode (one to all agents), or a
multicast mode (to a selected group of agents). It
uses the folowing two message primitives:
SendMsg (Sender, Receiver, “Message”).
GetMsgs (MailBox) to extract all current
messages from the agent mailbox and reconstitue the
message list according to the priority degree of each
message and the agent internal believes.
In order to capture more semantics of MAS
dynamic aspects, it is necessary to make an
ontological distinction between the simple
communication, the coordination, the cooperation
and the negotiation (see figure 1).
3.3.2 Cooperation protocol
Agents involved in ProMAIS take shape of MAS.
They cooperate in order to satisfy the customer, to
be more flexible and open to different kinds of
orders and to respond quickly to the dynamic
changes of external environment to take a position in
the market. Thus, we distinguish the main
cooperation situations between one or several
agencies of agents and we consequently distinguish
the cases of emergency orders and those of
unexpected risks.
In case of emergency orders, we draw apart the
orders hand over by a faithful customer or
characterized by a cash payment and the immediate
orders. In the first case, the cooperation takes place
between the Monitor (in particular the Marketing
agent), the Planning, the Scheduling and the
Manufacturing agents. While, in the case of
immediate orders, in addition to the agents involved
in the last case, the Stock Management &
Procurement (SMP) agent (in particular the Finished
Product Stock agent) is added.
In the case of the unexpected risks, we draw
apart the cases of a strike, a delay in carrying out a
task or in the arrival of raw materials, on the one
hand and those of machine breakdown, on the other
hand. All these kinds of unexpected risks necessitate
a cooperation between different agents involved. In
the first case, the cooperation takes place first
between the Manufacturing and Scheduling agents.
If the problem is not solved, the Planning agent is
then involved to cooperate with already montioned
agents. If the problem is still not solved (the
replanning operation introduces a delay in the
deadline of some orders), the Monitor agent is
involved to cooperate with the others, with the aim
to make decisions like the use of the subcontracting,
the additional hours or, if it is necessary, the interim
hours. In such a case, the strategic decisions already
made can be modified. In case of a machine
breakdown, the cooperation takes place between the
Manufacturing, the Maintenance and the Scheduling
agents if the state of the machine is adjustable.
However, if the breakdown is serious and cannot be
adjusted in a convenient time, the Monitor agent is
involved with the aim to make quickly the decision
to overcome this kind of problem like the use of the
subcontracting.
3.3.3 Negotiation protocol
In ProMAIS, we can distinguish two cases of
negotiation. The first one appears at the time of the
hand-over of an order from a Customer agent. The
negotiation will take place between the Customer
agent and the Monitor agent regarding the fixing
order price. The second one appears at the time of
fixing the estimated price of a new product that is
conceived by the Design & Engeneering agent. This
negotiation will take place between the Design &
Engineering agent and the Monitor agent.
3.3.4 Contract Net protocol
The Contract Net is a negotiation protocol proposed
by Smith (Smith, 1980). It facilitates distributing
subtasks among various agents. The agent wanting
to solve the problem broadcasts a call for bids, waits
for a reply for some time, and then awards a contract
to the best offer(s) according to its selection criteria.
The Contract Net protocol used in our proposed
model ProMAIS consists in a process for selecting a
supplier for a purchasing operation. For this, the
SMP agent broadcasts a call for bid for all Supplier
agents. Each Supplier agent thus becoming a bidder
sends its bid as regarding price and delay. Even the
Supplier agents can’t satisfy the order send the same
bid message giving the value of (–1) to the price and
delay variables. This message is necessary for the
learning of the SMP agent. For example, if it
receives at many times a refusal message from the
same Supplier agent, the SMP agent eliminates the
latter from the list of Suppliers.
After receiving the propositions (agreement and
refusal bids) the SMP agent verifies the fund
availability through a cooperation with the Monitor
agent and it will select a Supplier agent to perform
the purchasing operation according to its criteria
ProMAIS: A MULTI-AGENT MODEL FOR PRODUCTION INFORMATION SYSTEMS
301
(mainly a price and a fund availability for this
purchasing operation), and award a contract to it.
4 PROTOTYPE
IMPLEMENTATION
The proposed model ProMAIS has been
implemented in a simulation form.
This virtual system incorporates heterogenous
agents. It is implemented within a multi-agent
platform called AgentBuilder Pro 1.3
(Reticular, 2000) which is an integrated tool suit for
constructing intelligent software agents. It consists
of two major components: the Toolkit and the
Run-Time system. All agents are implemented using
Java programming language.
Communication among agents was realized
using RMI protocol and the inter-agent messages
were formatted in KQML (Knowledge Query and
Manipulation Languages) format.
In this simulation, the agent execution cycle
consists of the following steps: processing new
messages, determining which rules are applicable to
the current situation, executing the actions specified
by these rules and updating the mental model in
accordance with these rules.
5 CONCLUSION AND FUTURE
WORKS
Manufacturing systems are organizations composed
of heterogeneous entities involved in the production
and the delivery of finished product or services.
Nowadays, the IS for such organization can be
viewed as a collection of sub-systems distributed
between the different entities. This results in the
cooperative information system technology. The
entities communicate via computerized data. The
manufacturing organization must follow the
dynamic of the market and respond quickly to the
customers requirements.
The choice of intelligent software agent
technology provides a natural way to design such
systems because the intrinsic feature of MAS
correspond to those to be preserved in the hoped
production systems. Autonomy, heterogeneity,
openness, cooperation, dynamicity, commitment
etc… are at the heart of our reflection.
ProMAIS, provides the integration of different
entities (divisions) in manufacturing and production
systems. In fact, a cooperative MAS allows each
agent to communicate and cooperate with others
while conserving its autonomy.
ProMAIS is an ongoing project. A major
short-term research goal is to study the position of
humans in the system. The long-term research goal
tends towards fixing the distribution of the global
databases, study the interaction of agents with
databases, the resolution of different kinds of
conflict resulting from the heterogeneity of data
sources and then to implement this approach in a
real manufacturing enterprise.
REFERENCES
Balasubramanian, S., Maturana, F.P., and, Norrie, D.
1996. Multi-Agent Planning and Coordination for
Distributed Concurrent Engineering. International
Journal of Intelligent and Cooperative Information
Systems. Special Issue on Agent Based Information
Management.
Ferber, J., Jacopin, E., 1990. The framework of Eco
Problem Solving. In Proceedings of the 2
nd
European
Workshop MAAMAW'90. pp. 103-114
Ghedira, K., 1994. Distributed Simulated Re-annealing for
Dynamic Constraint Satisfaction Problems. In
Proceedings of the 6
th
IEEE International Conference
on Tools with Artificial Intelligence (TAI94).
Hayes-Roth, B., 1995. An architecture for adaptive
intelligent systems. Artificielle Intelligent, pp.
329-365.
Labrou, Y., Finin, T., Peng, Y., 1999. Agent
Communication Languages: the current Landscape.
Intelligent Agents IEEE. pp. 45-52.
Maturana, F. and Norrie, D. H., 1996. Multi-Agent
Mediator Architecture for Distributed manufacturing.
Djournal of Intelligent Manfacturing. pp. 257-270.
Qiao, B., Zhu, J., 1999. Agent-based Intelligent
Manufacturing System for the 21
st
Century.
Mechatronic Engineering Institute. University of
Aeronautics and Astronautics.
Reticular, S., 2000. AgentBuilder: An Integrated Toolkit
for Constructing Intelligent Software Agents:
Reference Manual. Version 1.3 Rev. 0. April 11, 2000
Reticular Systems, Inc. http://www.agentbuilder.com
.
Shen, W., and Norrie, D. H., 1998. A Hydrid Agent-
Oriented Infrastructure for Modeling Manufacturing
Entreprises. Division of manufacturing engineering.
The University of Calgary, Canada.
Shen, W., Barthès, J.P., 1995. DIDE: A Multi-Agent
Environment for Engineering Design. In Proceedings
of the First International Conference on Multi-Agent
Systems. San Francisco, CA, The AAAI press/The
MIT press.
Smith, R.G., 1980. The Contract Net Protocol: High-Level
Communication and Control in a Distributed Problem
Solver. IEEE Transactions on Computer. pp. 1104-
1113.
Wagner, G., 2000. Agent-Object-Relationship Modeling.
In Proceedings of Second International Symposium
“From Agent Theory to Agent Implementation”
(AT2AI-2), in conjunction with EMCRS 2000, Vienna.
ICEIS 2004 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS
302