Information Technology Reference
In-Depth Information
only control its effectors, but not the effects of its action in the
environment' [69, p. 9]. This distinction is especially 'important in
modeling situations, where an entity's controller might attempt to
achieve a state that is inconsistent with the environmental state' [114,
p. 1040].
Assumption of time-stepped execution
Although not always ex-
plicitly mentioned, most approaches make use of a time-stepped
execution with fixed time increments [24]. The definition of a simula-
tion of Scheutz and Schermerhorn explicitly covers this time-stepped
execution and further implies that all agents are executed in each
time step and that no action can happen in between these time steps.
Implicitly assumed in this approach is the fact, that all actions the
agents are able to execute always take exactly one time step duration
for execution.
Klugl defines a simulation quite similar in terms of model configura-
tions and 'the iterated application of the update functions' [69, p. 15].
Due to a missing formal description, the definition of a simulation
remains quite vague and the time-stepped execution is not mentioned
explicitly.
Weaknesses
Although the approaches presented have a lot in common, they also
have weaknesses and lack several features.
Unclear distinction between model and simulation engine
Many
approaches do not provide clear definitions of the model (i. e., en-
vironment, agents and their behavior), the simulation of a model
and the actual execution of the simulation by a specific simulation
engine. The approach of Scheutz and Schermerhorn as well as the
Agent-Object-Relationship Simulation are notable exceptions.
The absence of a clear distinction leads to fuzzy definitions and
hinders the future development as well as it puts unnecessary limit-
ations to the development of multi-agent simulations within (large)