Information Technology Reference
In-Depth Information
•
Self—liquidation of the agent. It may be realized by creating a mechanism iden-
tifying a lack of success of a given agent in performing the task it was entrusted
with. One of the simple solutions is to implement the concept of life energy of the
agent. The energy is used during the agent's activity and increases when success
is achieved. If the level of life energy decreases below a certain established level,
the agent undergoes self-liquidation.
•
Liquidation of the agent by other agents. To this purpose it is necessary to identify
and remove excessive agents. In order to do that it is necessary to:
- Set a criterion of identification (classification) of agents to be removed.
- Set the number of agents to be removed.
In the simplest case, the choice of type of agents to be removed takes place as a
result of the assessment of the number of agents of a given type, i.e. those agents are
chosen which are excessive in the system (e.g. the type of agents that outnumber
others). Afterwards, the number of agents to be removed is established (e.g. by
comparing a chosen type of agents to other types). This is done by choosing a
certain number of agents of a given type and removing them from the system. The
process of agents' removal is a solution burdened with certain disadvantages:
- The criterion for choosing the type of agents to be liquidated does not have
to mean that these agents are in fact unnecessary for realizing the tasks of
the system. It may happen tha a greater number of agents of a given type are
necessaryt for the realization of specific tasks.
- The liquidated agent is entrusted with some task to perform. As a result of the
liquidation of the agent, the task cannot be completed. It may disturb the correct
realization of the basic functions of the system.
In the example of balancing the resource in the graph environment, the method of
self-liquidation and liquidation of agents by other agents was used. Let us consider a
case of the application of the self-liquidation mechanism when there is an underflow
of the resource in the whole system. Consequently, in individual nodes there is a
local underflow of the resource and in these nodes agents of type
A
g
1
are generated
to search for the resource in other nodes. On being generated each agent
A
g
1
receives
a certain amount of life energy that is used while conducting a search. It allows for
self-liquidation of a given agent when the realization of its mission is futile and
overruns, and the life energy is used up.
We may also use the mechanism of removing the agent by other agents. If the
number of searching agents
A
g
1
in the system is too large, then their increased
number may be observed in individual nodes. This phenomenon may be observed by
the agent of type
A
g
0
residing in a given node (and managing this node). This agent
may establish that the number of the agents
A
g
1
existing in a given node is over a
certain limit, and that the tendency lasts for some time, or is growing. It may signify
that there are too many agents of type
A
g
1
in the system. The agent
A
g
0
may make
