Agent Versus Decomposition of an Algorithm - A Perspective on Agent Systems - page 39

algorithm which is to be decomposed

decomposition
based on the
concept of the
Cartesian product

decomposition
based on the
concept of the
division of sets

the application of the Cartesian product
to the category of sets results in non-
autonomous component algorithms, but
the realization is easier

subprogram

the application of the Cartesian
product to the category of algorithms
results in autonomous component
algorithms, but the realization is difficult

Step 1: the application of the
environment concept

Step2: access to the state
of another algorithm

object agent

the application of the
communication process

the application of the
observation process

Fig. 2.21 Schema of decomposition of an algorithm into autonomous component algorithms

the environment or changes that occur in that environment (Fig. 2.20 a). Under the
concept of the agent, the agent A g 1 incorporates the agent A g 2 into the area of the
environment it observes, and in particular:

• The agent A g 1 is able to observe not only the environment X 0 (Fig. 2.20 ), but also

the fact that the agent A g 2 exists in the environment,

• The agent A g 1 is able to observe changes in the environment caused by the agent

A g 2 , which means that the agent A g 1 is able to associate the event occurring in
the environment that changes the state of the environment with a specific agent),
the doer of this event (e.g., with the agent A g 2 ,Fig. 2.20 b). It can be said that the
agent A g 1 extends its own model of the environment it observe by incorporating
the agent A g 2 into it, as presented in the Fig. 2.20 .
Summing up the process of decomposition of a component algorithm that leads to
the concept of an object, and especially to the concept of an agent, we may use a
schema from Fig. 2.21 , which shows the following possibilities of transformation of
an algorithm:

• An algorithm we are to realize (design, program, activate) is too complicated and

has to be decomposed into more or less independent component algorithms. There
are two possible ways of decomposition: one based on the concept of the set
division, and on the other on the Cartesian product.

• The decomposition based on the division of sets leads to the concept of a sub-

program. This approach is often successfully applied in practice but not always
effectively.