Information Technology Reference
In-Depth Information
of an agent are combined to sensor-effector-chains and define the
behavior of an agent. Agents can only try to execute a specific sensor
or effector action. The actual outcome is determined by taking into
account the current model state and all concurrent actions of all
agents. As the actions of two or more agents may be in conflict with
each other, the result of each action depends on whether this action
is in conflict with any other action. Constraints define exactly under
which circumstances the execution of an action is not successful.
In order to provide a solid foundation, the GRAMS reference model
separates strictly between an agent-based model itself and its simula-
tion which is executed by appropriate simulation engines. The ability
to execute an agent-based model using different simulation engines,
each producing identical results, is a central aspect of the GRAMS
reference model.
The possibility of using different simulation engines executing the
same model is also crucial for exploiting hardware-given parallelism.
In this thesis, parallelization of simulation engines is defined on three
levels - namely cluster-, node-, and processor-level - with multi-level
parallelization referring to the simultaneous application of paralleliz-
ation on multiple levels. Depending on the underlying hardware, a
simulation engine may execute a given model and exploit multi-core
processors as well as make use of multiple computing nodes.
The applicability of the GRAMS reference model is demonstrated
in two ways. At first, by providing an example implementation of
the GRAMS reference model including different simulation engines.
Secondly, by evaluating the GRAMS reference model with regard to
practical applicability using three case studies.
11.2 Scientific contributions and applicability
This thesis provides two main contributions to current research in the
area of agent-based modeling and simulation:
1. The GRAMS reference model provides solid and clear definitions
of key aspects of agent-based modeling and simulation. It provides
