COORDINATION IN MULTI-AGENT DECISION
SUPPORT SYSTEM
Application to a Boiler Combustion Management System (GLZ)
Noria Taghezout
1
, Abdelkader Adla
2
and Pascale Zaraté
2
1
Department of Computer Science, University of Es-Senia Oran, BP 1524, El-M' Naouer, 31000, Oran, Algeria
2
IRIT, INPT – ENSIACET, 118 route de Narbonne 31062 Toulouse Cedex 9, France
Keywords: Multi-agent system, Group Decision Support System (GDSS), MADS (Multi-agent Decision support
system), Coordination agent.
Abstract: In Multi-agents systems, the cognitive capability present in an agent can be deployed to realize effective
problem-solving by the combined effort of the system and the user. It offers the potential to automate a far
wider part of the problem solving task than was possible with classical DSS. In this paper, we propose to
integrate agents in a group decision support system. The resulting system, MADS is designed to support
operators during contingencies. We experiment our system on a case of boiler breakdown to detect a
functioning defect of the boiler (GLZ: Gas Liquefying Zone) to diagnose the defect and to suggest one or
several appropriate cure actions. In MADS the communication support enhances communication and
coordination capabilities of participants. A simple scenario is given, to illustrate the feasibility of the
proposal.
1 INTRODUCTION
In Distributed Artificial Intelligence (DAI) systems,
problem solving agents cooperate to achieve the
goals of the individuals and of the system as a
whole. Each individual is capable of a range of
identifiable problem solving activities, has its own
aims and objectives and can communicate with
others (Jennings, 1993). Typically agents within a
given system have problem solving expertise which
is related, but distinct, and which has to be
coordinated when solving problems. Such
interactions are needed because of the dependencies
between agents’ actions, the necessity to meet global
constraints and because often no one individual has
sufficient competence to solve the entire problem.
In this paper, we propose to integrate agents in a
group decision support system. The use and the
integration of software agents in the decision support
systems provide an automated, cost-effective means
for making decisions. The agents in the system
autonomously plan and pursue their actions and sub-
goals to cooperate, coordinate, and negotiate with
others, and to respond flexibly and intelligently to
dynamic and unpredictable situations.
We argue that an agent-oriented approach is the
most natural and appropriate mean to achieve better
support for the group decisions, and we propose to
improve the coordination protocol.
The manufacturing process of the oil plant
(GLZ), selected as application domain in this study,
is split into two subdivisions: Utility subdivision and
Process subdivision. The Utility subdivision is
constituted of Pumps, Desalination Unit, boilers,
Turbo-generators and Air Compressors while the
Process subdivision concerns the tasks of
manufacturing of liquefied Gas. This subdivision is
composed of 6 strings where a string is group of
equipments. Every string contains 10 sections which
are going to be used to liquefy gas. The management
system of the boiler combustion is one of the most
critical systems for the good functioning of the plant
and has a high impact on the methods of cogitation
and apprehension of various problems related to
maintenance (see Figure 2). The exploiting staff is
often confronted with situations that impose a quick
reaction of decision-making. This requires
consequent human and material resources and
adapted skills (for more details see (Adla, 2007)), to
diagnose the defect and to suggest one or several
194
Taghezout N., Adla A. and Zaraté P. (2009).
COORDINATION IN MULTI-AGENT DECISION SUPPORT SYSTEM - Application to a Boiler Combustion Management System (GLZ).
In Proceedings of the 11th International Conference on Enterprise Information Systems - Artificial Intelligence and Decision Support Systems, pages
194-199
DOI: 10.5220/0001999301940199
Copyright
c
SciTePress
appropriate cure actions.
The reminder of the paper is organized as
follows: Section 2 describes our contribution. It is
followed by section 3 which covers the integration
of agent technology into a DSS. The multi-agent
architecture for group decision support systems and
the corresponding coordination protocol are
described in section 4 and section 5. We also present
an example of a scenario in section 6. Finally,
Section 7 gives some concluding remarks.
2 CONTRIBUTION
Research that studied group decision support
systems in the existing literature used mainly face-
to-face facilitated GDSS. Some of its results may not
apply to distributed teams (Chen, 2002) that, it is
difficult for distributed teams to arrange face-to-face
meetings or to meet at the same time virtually.
Moreover, although most presented GDSS
environments try to solve problems in the real world,
the lack of an integrated procedure, from decision
identification, basic information acquiring, to final
decision proposed, makes the systems only partially
supportive or even needful of outside assistance.
Still, despite the existence as well as the extensive
use of numerous general-purpose commercial
systems, it is our belief that these systems do not
readily fulfill the needs or operational usages of
specialists or experts in different organizations to
render their expertise in GDM processes.
In our study we consider another gap: the
coordination problems when they occur have several
causes. Most of them are a consequence of
limitations in both the decision making processes
and the technological support for communication.
For this reason, the information and tasks related to
the decisions made in GDDS have to be visible to
other organizations to keep the relief effort
coordinated between the agents.
In addition, the quality of support received
during the decision making processes is the key to
reaching optimal decisions. Decisional guidance
mechanism provides the decision makers with step-
by-step guidance throughout the decision-making
process and allows them to evaluate more
alternatives. As a result, DSS users with decisional
guidance can easily come up with better decisions
than those with no decisional guidance.
Mahoney et al. (Mahoney, 2003) pointed out that
when faced with Complexities in a decision
situation, decisional guidance helps users to choose
among and interact with a system’s capabilities.
They argued that in less structured tasks that deal
with uncertainty and risk, users need more guidance
to choose among competing solution techniques or
among alternative methods of processing
information to structure an appropriate decision-
making process using the GDSS.
3 AGENT INTEGRATION IN DSS
We got inspired by two main research works.
Firstly, the main ideas resumed in the table
described in (Forth, 2006) were very interesting for
our study (see Table 1). It defines how the capability
of an agent may be utilised in a DSS application, and
also identifies alternative agent design architectures
suitable to underpin this. As a constituent part of
problem-solving in the domain, an agent may choose
particular sources of information to use. Data
Gathering may be a function within an agent
(sensing), or a dedicated activity of a specialised
information agent if the task is complex.
Secondly, the approach developed by Zamfirescu
(Zamfirescu, 2003) addressed the problem of self-
facilitation in GDSS by establishing a common and
meaningful high-level collaboration pattern among
the group members inspired from the SP theory. In
his GDSS approach, the main entities have been
defined: the personal assistant agents, the resource
agents and the plan agents.
Table 1: Mapping DSS functions to agent capabilities.
DSS Function Agent Function
Data collection
Knowledge acquisition
and assimilation
Model creation
Perception and knowledge
representation
Alternatives case
creation
Planning and reactivity
Choice Action selection
Implementation Action execution
4 THE MULTI-AGENT SYSTEM
Agents are used to collect information outside of the
organisation and to generate decision-making
alternatives that would allow the user to focus on
solutions that were found to be significant.
According to this a set of agents is integrated to the
system and placed in the DSS components (as shown
in Figure 1), we distinguish:
COORDINATION IN MULTI-AGENT DECISION SUPPORT SYSTEM - Application to a Boiler Combustion
Management System (GLZ)
195
Figure 1: Agent architecture for Individual DSS (Adla et al., 2007; Jennings, 1993).
Figure 2: A partial hierarchy of tasks and methods of the application ( A01, A05 , SD1 and A12: feasible methods;
Decompose1, 2: Decomposition methods).
The Interface Agent (IA) Continuously receives
data from the process – e.g. alarm messages about
unusual events and status information about the
process components.
A Decision Maker Agent (DMA) performs most
of the autonomous problem solving.
An Information Retrieval Agent (IRA)
primarily provides intelligent information services.
A Diagnosis Agent (DA) is activated by the
receipt of information from DMA which indicates
that there might be a fault.
Knowledge Management Agent (KMA)
comprises, manage and update knowledge base;
The Action Agent (AA) generates a plan of
action which can be used to repair the process once
the cause and location of the fault have been
determined. As described in Figure 3, a refined
representation of DA, AA and KMA agents is given.
The Coordinator Agent provides two services to
task agents: (i) it computes summary information for
hierarchical plans submitted by the task agents, and,
(ii) coordinates hierarchical plans using summary
information.
5 A COORDINATION
PROTOCOL
The problem solving mechanism is based on a set of
cycles until the entire problem is solved. Each cycle
consists of the following steps :(1) identifying
candidate methods; (2) identifying triggered
methods ;( 3) selecting a method; (4)assigning the
method to an agent; (5) executing the method; and
(6) Evaluating the task state.
ICEIS 2009 - International Conference on Enterprise Information Systems
196
5.1 Agents Structure
Clearly, all the modules representing the inner
structure of an agent may depend on each other.
This is especially true for the local problem solver
and the coordination module (as shown in Figure 3.)
which do not only exchange real time information
but, in addition, must coordinate their decision rules
and performance criteria. If we consider the
relationship between the coordination module, the
problem solver, and the knowledge base. We found
that they have to make sure the data needed
available. The communication between the agents
may roughly be described by the coordination
module and the interface component. They are
describing the way how agents may communicate.
5.2 Agent Communication Language
The agents needs are formalized as a set R of
requests r
R, each of which expresses a question
about agent’s data or an action to perform given a
set of parameters. These requests appear both in
human-agent and in interagent dialogues.
The request performative rp can take the following
values, for example see Figure 4, Figure 5 and
Figure 6):
Ask-for requests that represent questions about
the agent’s data.
Assert-is for assertions about agent’s data values
(to answer to Ask-for requests).
Order for requests that represent actions to
perform.
Affirm for acknowledgment when an action has
been performed (in answer to an Order request).
Assert-cannot to express that the agent is unable
to perform a command.
Assert-can to express that the agent can perform
a command, but misses a field value to be able
to do so.
Unknown for asserting that the agent doesn’t
know a field or the overall action to perform.
5.3 Communication between Agents
The requests structure presented above is also used
by agents in their communications. Indeed, these
requests formalize the contents of the messages
exchanged by agents, each of which is structured as
follows: m = [id, C, sender, receiver, agenda].
Where id is the message id, C is the message
content, sender and receiver are the sender and
receiver agents’ ids and the agenda term for
indicating the memory of the message associated to
the initial agent request.
Figure 3: Representation of DA, AA, and KMA agents.
5.4 Discussion
When an agent receives:
An Order request, and if it misses a value to
perform it, it builds an Assert-can answer.
An Assert-can request, it replaced the required
field stated in the assert-can answer by its value
if owned.
An unknown request, it looks in the
corresponding field in the Ask-for request.
An assert-is request, when answering an Ask-for
request, it can trigger it by using the information
received trough the Assert-is answer.
Figure 4: Ask-for protocol.
Coordination
module
Knowledge
Base
Problem-solver
Interface
COORDINATION IN MULTI-AGENT DECISION SUPPORT SYSTEM - Application to a Boiler Combustion
Management System (GLZ)
197
Figure 5: Unknown protocol.
Figure 6: Order protocol (fr: final response).
Figure 7: A coordination scenario diagram (AUML).
6 A COORDINATION SCENARIO
When the task management agent (DMA) receives a
task from an interface agent (IA), it decomposes the
task based on the domain knowledge it has and then
delegates the primitive tasks to the other agents
(IRA, MA, KMA, DA or AA). The task
management agent will take responsibility for
retrieving data, modelling, diagnosing fault,
planning action, resolving conflicts, coordinating
among the related agents and finally reporting to the
interface agent which conveys the results to the user.
ICEIS 2009 - International Conference on Enterprise Information Systems
198
As described in Figure 7, the DMA agent first
gets input data through the interface agent. Next, the
modelling agent searches for rules to select a
suitable model and to execute the model to get
analytical results. Additionally, all the parameters
values needed by the models are retrieved from the
database via the information retrieval agent. After
finishing model analysis, the diagnosing and the
action agents use the results of the model analysis to
identify the fault causes and to perform a suggested
action plan.
7 CONCLUSIONS
MAS paradigm offers a new dimension with respect
to GDSS integration with complementary services,
making it easier to build complex and flexible
architectures suitable to organizational settings
(Zamfirescu, 2003).
In this paper, we have integrated agents into
GDSS for the purpose of automating more tasks for
the decision maker, enabling more indirect
management, and requiring less direct manipulation
of the DSS. In particular, agents were used to collect
information and generate alternatives that would
allow the user to focus on solutions found to be
significant. We expect to finish the system
implementation that supposes all the decisional tools
and validation with other oil plants which are
geographically dispersed.
REFERENCES
Adla, A., Soubie, J-L., and Zarate, P., 2007. A Co-
operative Intelligent Decision Support System for
Boilers Combustion Management based on a
Distributed Architecture. In Journal of Decision
Systems, Lavoisier, Vol. 16, pp.241-263.
Chen, M., 2002. TeamSpirit: Design, implementation, and
evaluation of Web-based group decision support
system. In Decision Support System, Lavoisier B. V.,
43. pp. 1186-1202.
Filip, F.G, 2008. Decision support and control for large-
scale complex systems. Annu Rev Control, doi:
10.1016/j.arcontrol.03.002.
Forth, J., Statis, K., and Toni, F., 2006. Decision Making
with a KGP Agent System. In Journal of Decision
Systems, Lavoisier, vol. 15, pp. 241- 266.
Jennings, N. R., 1993. Coordination Through Joint
Intentions In Industrial Multi-Agent Systems. AI
MagaZine, 14 (4). pp. 79-80.
Mahoney, L. S., Roush, P. B. and Bandy, D., 2003. An
investigation of the effect of decisional guidance and
cognitive ability on decision making involving
uncertainty data. In Information and Organization 13
(2), pp. 85–110.
Zamfirescu, C.B. , 2003. An agent – oriented approach for
Supporting Self-Facilitation in Group Decisions. In
Studies in informatics and Control, 12(2), pp. 137-148
COORDINATION IN MULTI-AGENT DECISION SUPPORT SYSTEM - Application to a Boiler Combustion
Management System (GLZ)
199