Information Technology Reference
In-Depth Information
(sometimes impossible) because of the possibility of emergent be-
havior. Due to this missing micro-macro-link, it is often dicult
to choose the right level of detail for a model. This may lead to
overly complex models requiring time-consuming and possibly erro-
neous calibration. Also, it is often unknown which behavior on the
micro-level yields a known or desired behavior on the macro-level.
2. Agent-based modeling and simulation typically requires a higher
computational effort (runtime, memory) than traditional methods.
This is due to the more detailed modeling of behavior associated
with microscopic models.
Weighing the pros and cons, agent-based modeling seems most
promising when dealing with complex systems characterized by a
large number of individually acting entities. The behavior of the
entities and their interaction may be translated quite directly into
an agent-based model. Compared to other modeling paradigms, this
similarity (of structure and interaction) between the system under
investigation and the model is probably the most notable aspect of
agent-based models. Finally, the practical (not conceptual) application
of multi-agent simulation is closely connected with the availability
of su
cient computational power allowing to simulate large-scale
microscopic models [83].
2.3 Related work
In the following subsections, currently existing approaches to (form-
ally) specify agent-based models and their simulation are presented.
Due to the amount of approaches and systems in the agent-based
modeling and simulation domain, it is not possible to exhaustively
present all of them. Instead, various selected approaches are presented
which represent different ways of describing agent-based models. The
first approach is a proposal by Klugl [69] for a formal framework
for multi-agent simulations. The second approach was developed by
Scheutz and Schermerhorn [114] and is especially interesting as it
