Information Technology Reference
In-Depth Information
phenomenon (and even desirable) in this system. However, at the same time
the stabilization of the system requires a limited number of agents (total, and
of certain types). Similarly, the agents (of a certain type and cumulative) should
not disappear in the system (unless such a scenario for some types of agents is
foreseen in a given situation).
The problem of scaling in the case of the multi-agent system involves choosing
the adequate number of agents according to the complexity of a task to be per-
formed. The agent systemallows for more precise scaling that consists in choosing
the adequate number of agents of particular types and establishing numerousness
of cooperating groups of agents. Specification of this numerousness should be
done on the basis of the complexity and character of tasks which the agent sys-
tems are entrusted with. It gives new possibilities of scaling—a kind of functional
scaling . It should be emphasized that the agent acting in the environment uses the
resources of the system for the realization of its tasks (computing power and the
memory of the processor). If there are more agents in the system than needed for
the realization of entrusted tasks then the excessive agents use the resources of the
system mainly (or even solely) for the tasks connected with their own existence
(they do not contribute to performing tasks the system is entrusted with).
Below, we present a couple of exemplary mechanisms which ensure the
stabilization and scaling of the agent system.
4.5.1 Keeping the Minimal Number of Agents
The solution to the problem of keeping the minimal number of agents of a particular
type is keeping the adequate number of agents which can generate them. For instance,
in the systemwe analyse, only one instance of the agent of type A g 0 exists in each node
of the environment, whereas their number is constant when the system is operating.
The agent of type A g 0 generates other types of agents ( A g 1 , A g 2 ) whenever possible,
which guarantees that they will not totally disappear in the system.
4.5.2 The Limitation of the Maximum Number of Agents
There are three types of agents in the system: A g 0 , A g 1 , A g 2 . The agents of type A g 0
exist in the number established at the moment the systemwas created (by the number
of nodes of the environment) and therefore constant, and there is no problem with
an excessive increase in their number. However, the number of agents A g 1 and A g 2
changes during the time of the operation of the system and may increase excessively,
thus being a danger to its operation (at least efficient).
The removal of excessive agents.
This may be realized in two ways:
 
Search WWH ::




Custom Search