Information Technology Reference
In-Depth Information
3.3.1 The Multi-agent System Activity—
The Basic Assumptions
It seems that the problem of creating agent systems is so important and extensive that
it should be treated not as an approach to a single case, a certain set of computing
tools or a specific algorithm, but as a general point of view or the way or reasoning.
Therefore, we should aim to develop the agent approach to the problem, the agent
analysis, the agentmethodology of designing systems and their programmingwith the
use of “agent” methods, algorithms, languages, i.e., the appropriate tools. Accepting
an informal, intuitive point of view, in order to create the multi-agent model of a
given problem we should define first:
•
A certain space constituting the area of agents' activity which enriched with
appropriate resources may create their environment. It is necessary to define this
environment taking account of features of the task to be solved with the use of the
system.
•
A set of agents (of different kind as the need arises) which will exist and operate
in the environment. This set should be defined by grouping agents according to
already established types, aiming at distinguishing common features of all agents
and characteristic properties for particular groups.
•
Mutual relationships between the defined agents as well as between the agents and
the environment. It should be taken into consideration that the possibilities of a
given agent (observation and action) are limited and a given agent may cooperate
in given conditions only with a certain group of agents and certain part of the
environment.
The rule that proves right in practice, suggested as a starting point for the analysis and
object design (see [67]) contained in a laconic saying “perform the role of the object”,
which means assume that you are an object—act, behave like a given object—may
be applied to agent programming: “you are an agent, and you must manage to exist
in the environment just like an agent”.
According to this rule, we accept that the observation of the world which is useful
for the analysis and agent design will be performed from the point of view of the
agent.
Therefore, the boundary line should be drawn between what the environment is
and what it is not (and possibly what it is?) through accepting the point of view of
the agent (this division should be universal to such an extent that it could possibly be
applied to all agents). It can easily be noticed that the division between the elements
of the agent-environment system does not have to be quite very simple and the sets
of those notions are not completely separable as they would seem to be. It may not
be the case that a given agent may treat certain elements (parts) as other elements of
the environment. Therefore, it seems to be more correct to ask whether there are any
elements of the system which cannot be treated by a given agent in the same way as
the elements of the environment.
