Information Technology Reference
In-Depth Information
(result-)equivalence of sequential and parallel execution of a multi-
agent simulation is highly desirable for various reasons and should be
regarded as essential when parallelizing any simulation.
2.3.3 Ferber, Muller (1996) and Weyns, Holvoet (2003)
Motivation
Moving from multi-agent simulation to multi-agent systems,
situated-
ness
and autonomy remain two of the main aspects of any such system.
Resembling the current state in formalizing a multi-agent simulation,
Weyns and Holvoet state that 'comparatively few work has been done
to specify conceptual models for multi-agent systems in a formalized
way' [143].
Based on the model by Ferber and Muller which separates what
agents do and how the environment reacts upon this, Weyns and Hol-
voet point out the limitations and come up with an own approach. In
the following, the model of Ferber and Muller as well as its limitations
are described, followed by the approach of Weyns and Holvoet.
Influence-reaction approach of Ferber and Muller
The influence-reaction approach of Ferber and Muller was introduced
in 1996. Its key idea is a 'clear distinction between influences, which
are produced by agents' behaviour, and the reaction of the environ-
ment' [33]. Influences generally result from an agents internal behavior
and are attempts to execute some specific task. On the other hand,
reactions are the product of the combined influences of all agents
within a certain area of the environment.
A similar approach is proposed by Goodwin [46] who distinguishes
between commands and interactions. Whereas commands are totally
under control of an agent and represent its attempt to alter the envir-
onment, interactions represent the actual outcome in the environment.
Although the basic principle is very similar, Goodwin mentions only
vaguely the use of commands and interactions to model failing actions
or conflicting actions of multiple agents.
