Agent systems
During last few years, agent systems, a serious development of
agent systems can be observed. Applications of agent systems
are usually difficult problems which solving requires use
of intelligent distributed computation.
This relatively new technology introduces many useful
features that can be used to solve complicated problems, for
which deterministic solution does not exist, and even if
it does, its computational complexity makes direct implementation
impossible.
Definition of an agent
Definition of agent itself is difficult to state, and is usually
avoided by the researchers, it is perceived as a primary notion.
The simplest and the shortest answer to the question ,,What is an agent?''
is ,,Agent is that agent does'' (Bradshaw). This answer describes the
object that is delegated by its creator to fulfill some specific demands.
More accurate attempt to define an agent is desribing it
as hardware or software compuer system that fulfills following
requirements (Jennings):
- autonomy - agents work without human intervention,
they have specific capabilities of controlling their work
and their inner state,
- cooperation - agents communicate with one another using
specific communication language, their work can be changed basing on
the outcome of the information gained,
- reactiveness - agents react to the changes of environment,
that can bed the real world, user interface or virtual world
of agent connections,
- proactivity - agent not only is able to react on changes of environment,
it can work on its own (basing on autonomical decisions).
More formalized definition of an agent is describing it as
a computer system placed in an environment (being a part of it)
which can act on this environment, and can be changed by it
(environment acts on agent), its work leads to fulfill certain
specific task.
Other agent attributes can be:
- mobility - agent can move inside computer network,
- reliability - agent can not intentionally pass false information,
- accuracy - agents have not contradictory tasks and every agent
will always try to do what it is asked to,
- racjonalno.ć - agent będzie działał w~celu osi.gnięcia zadania,
nie doprowadzi do uniemożliwienia jego wykonania postępuj.c w~sposób
nieracjonalny.
Considering multi-agent systems fulfilling requirements above,
comunicating with one another, acting in order to fulfill
required task, it can be stated that such system
comprised of intelligent agents can be probably good
tool to solve complex problems, and the distribution of
the system tasks among the agents (specialisation)
leads to simplification of implementation and
interpretation of results. Multi-agent system
comprising of specialised agents will surely
be better at solving complex problems than one
big application (Nwana).
Autonomy
Agent as a process - autonomuous object, can reveal basic
intelligent features such as ability to learn and to predict
the situation.
In the agent system every agents has its own personality described
by its attributes. Acting of the agent depends on this attributes.
Agent can acquire information from its neighbours, explore the
environment, work in order to fulfill its preassigned task.
Thanks to its autonomy and ability to act on its own inside
the system, agent becomes more and more similar to a human
that lives inside a population. Acting of human is constrained
by specific ethical, moral and social norms present in the
group in which he lives, and his character.
Agent is constrained by the parameters of its environment
and its features.
|
