Information Technology Reference
In-Depth Information
Various approaches and formal models have been developed and
proposed in the past (see Chapter 2.3). Each approach usually focusses
on some specific objective and therefore often trades generally accepted
principles and ideas for the advancement in this specific area (e.g.,
unclear distinction between a model and its simulation). Furthermore,
existing approaches of suitable generality often simplify key aspects
in an unacceptable way (e.g., actions without duration). Accordingly,
more precise definitions of key notions and concepts of agent-based
modeling as a prerequisite to establish a common understanding are
necessary [60, 91, 13, 46].
Resulting from unclear key notions and imprecisely defined basic
concepts of agent-based modeling, the
simulation
of agent-based
models is often equally weak defined. This unsatisfactory state was
summarized by Muller who observed that 'most, if not all, existing
multi-agent simulation platforms produce simulation results which do
not depend only on the model but on the way the model is implemented
and the scheduling ordered, this ordering being at worst arbitrary and
at best randomized' [89], see also [101, p. 148].
The lack of a well-defined foundation as a basis for developing spe-
cific simulation engines is acknowledged independently by Wagner and
Himmelspach. Wagner observed that many agent-based simulation
systems do not have a theoretical foundation and advocates a simu-
lation metamodel as well as an abstract simulator architecture and
execution model [139, 137]. Himmelspach recalls that a lot of simula-
tion systems were developed in past years and almost all of them focus
on some specific formalism, language and execution paradigm (e.g.,
sequential, distributed). In this context, Himmelspach observes that
repeated re-development is very time-consuming and usually distracts
from the intended investigations. This disadvantage becomes even
more obvious against the background of the (admittedly subjective)
statement of Minar that 'unfortunately, computer modeling frequently
turns good scientists into bad programmers' [88].
