The "intelligent" level, that it is important to add,
it has to operate so that to allow "to understand",
depending on the context, the parameters to
publish/subscribe. To add this level, the formal
ontologies are particularly useful. Not only the
ontology but also the technologies of reasoning that
allow from a side to define the context where the
system operates and from the other to individualize
the rules of action that each agent can take in order
to answer to a well defined event.
The comparison with this kind of system and the
e-business systems is immediate: every agent
represents every information system involved in the
information interchange. The "intelligent" level
allows subsequently the different information
systems to publish/subscribe in the DDS the data of
interest to an appropriate system of reasoning that
allows to identify the correct information.
In this paper we present a high-level architecture
designed in the research project: the architecture is
useful in order to allow the communication among
the involved agents. As it will be clearer
subsequently, the base idea will be that to make the
application domain independent in comparison to
the definition of the key elements that make possible
the communication (DDS). Besides particularly
interesting is the sharing of an only one knowledge
base among several agents of the system.
After having introduced in the section 2 the state
of the art to the multi-agent systems, in the section 3
will be introduced three alternatives considered as it
regards the positioning of the knowledge base on the
various agents that constitute the system. In the
section 4 we present the architecture of the multi-
agent system, finally, in the section 5, we present the
conclusions of the paper.
2 BACKGROUND
The e-business systems have not been thought and
developed, until now, according to a multi-agent
logic. It is useful, however, to present the state of the
art related to the multi-agent systems in order to
understand the problem list related to the
development of a system of this type.
In a multi-agent system we may speak about 4
main aspects:
decisional aspects related to the agent: what
actions an agent undertakes depending on the
external environment;
outsourcing of the execution: possibility that
more agents collaborate together, performing
elementary actions to complete a complex
task;
interactions among agents: information
interchange among agents in distributed
environment;
Evolution toward component more and more
endowed with autonomy: agents that in full
autonomy reach their own goals.
The multi-agent systems need a study of the
nature of the interactions. It results therefore
important the notions of collaboration and
cooperation.
Interesting, in this paper, to underline that for the
realization of a multi-agent architecture it is
necessary to have a layer of communication and a
layer of conceptual modelling.
Communication Layer: currently the most
qualified is standard seems to be FIPA-ACL
(FIPA, 2002) (Agent Communication
Language) created from the Foundation for
Intelligent Physical Agents (FIPA). This
standard is founded on the linguistic action
theory, elaborated by John Searle (Searle,
1969). An important implementation of the
FIPA standard is the framework JADE
(http://jade.cselt.it) an open source platform
for peer-to-peer agent based communication
developed by Telecom Italia Lab.
Conceptual Model Layer: it is very important
to identify the domain where agents operate
and it is important the dynamics of
interactions between agents.
Currently, particularly interesting within the
modelling of the multi agent systems it results:
The BDI Model (Beliefs-Desires-Intentions)
(Chang-Hyun, Guobin et. Al, 1969): it
considers the agent environment belief, which
is the result of its knowledge and perceptions,
and a set of Desires. Intersecting these two
sets, we obtain a new set of intentions, which
can become actions.
Tropos
(Bresciani, Perini et. Al, 2004) a
software development methodology founded
on concepts used to model early requirements.
In particular, the proposal adopts Eric Yu's
modelling framework, which offers the
notions of actor, goal and dependency, and
uses these as a foundation to model early and
late requirements, architectural and detailed
design. The language used in Tropos for the
conceptual modelling is formalized in a meta-
model described with a set of UML class
diagram.
In literature there are several papers that examine
the software engineering paradigm applied to the
multi-agents system, among them an article of Pratik
K. Biswas (Patrik, Biswas, 2007) describes
ICE-B 2008 - International Conference on e-Business
102