Information Technology Reference
In-Depth Information
4 Summary
Complex systems are characterized by multiple entities, often of het-
erogeneous nature, that are interacting in many ways. The causes
of complexity are manifold: relationships and dependencies of sys-
tem components, number and value range of model parameters, or
intricate behavior and interaction patterns of single entities may cause
complexity. Agent-based modeling is a modeling paradigm which is
well suited for representing such complex systems. The main benefit
of agent-based models is a close structural similarity of a system under
investigation and a corresponding simulation model.
Although a certain common understanding of agent-based modeling
exists, there is no exhaustive and generally accepted specification de-
fining what actually comprises an agent-based model. Several research
approaches exist which try to give such a definition of agent-based
modeling. Unfortunately, all approaches known so far are suffering
from various weaknesses. In summary, a well-specified definition of
agent-based modeling including a development framework is missing.
Existing approaches considering parallel execution of multi-agent
simulations always focus on a specific domain. Furthermore, existing
approaches are often developed with specific decomposition patterns
in mind. Generally applicable partitioning strategies or approaches for
parallel execution of agent-based models and simulations are hardly
available. Existing approaches are not only application dependent, but
- as simulation models and simulation engines are often not strictly
separated - the approaches for partitioning simulation models are
also tightly integrated into the simulation engines. The result is that
simulation models, simulation engines and parallelization approaches
are highly interwoven and hardly distinguishable in currently repor-
ted work. Static as well as dynamic parallelization approaches are
