Information Technology Reference
In-Depth Information
3. The evaluation of the possibility of realizing of the goal with the use of strategies
available for a given agent and choosing the best strategy realizing the goal
q
,
i.e., the optimal strategy
s
∗
.. A choice of the optimal strategy
s
∗
. is made with
the use of the evaluation to what extent the application of the strategy
s
∗
.ina
specific situation (defined with the use of the model
m
) leads to the realization of
the goal
q
. The search for the optimal strategy
s
∗
may come down to the search
for the (global) extreme.
s
∗
(
q
(
m
,
m
))
=
max
s
∈
S
q
(
m
,
s
(
m
))
(3.7)
Continue to step 4.
4. The realization of the chosen strategy
s
∗
in the specific environment
s
∗
,
v
:
X
(
m
,v)
=
v
. Continue to step 2.
The above algorithm is illustrated in a schematic form in Fig.
3.5
.
3.5 The Concept of the Agent's Profile and the Model
of the Multi-profile Agent
Considering the activity of an agent in multi-agent systems, we may notice that
the behaviour of the agent, its way of perceiving its surrounding environment and
affecting this environment changes significantly, depending on the kind of environ-
ment and its state at a particular moment. The differences in the algorithm of the
agent's activity are connected with the way the model of its surrounding environ-
ment is constructed (the knowledge of the agent
M
and the observation function
I
), the way of choosing strategies (a set of strategies
) and the evaluation of the
applied strategy (the objective function
q
and the set of the objective functions
S
).
Consequently, the way of the agent's activity (role) presented in Fig.
3.5
(with new
parameters) changes. According to various approaches this fact is considered in such
a way that a given agent plays different roles according to circumstances.
For the purpose of finding a more precise solution to the above problem, we
introduce the notion of the multi-profile agent. The multi-profile agent is presented
in a schema in Fig.
3.6
. The multi-profile agent consists of the finite number of
profiles. Each profile realizes a particular version of the algorithm of the agent's
activity similar to the one presented in this chapter (Fig.
3.5
).
In each profile an appropriate strategy is formulated on the basis of the observation
of the environment. Strategies from all profiles are used for formulating one strategy
which is realized in the surrounding environment. The result strategy is mapped out
on the basis of the state of the multi-profile agent which is based on the informa-
tion obtained from the models of the surrounding environment created in particular
profiles.
Q
