Information Technology Reference
In-Depth Information
micro-level [150, p. 5]. Besides in the social sciences, such emergent
phenomena are often observed in a business context when analyzing
flows, markets, organizations and diffusion processes [22]. Nowadays
agent-based models are used in a wide variety of domains, for example
in logistics [28], trac analysis [18], distributed systems [60, p. 278]
and various other business applications [93].
Despite this widespread use, the foundations of agent-based model-
ing are much less profound than other established modeling paradigms
like discrete-event simulation. Regarding agent-based modeling and
simulation 'there exists neither a unified formal framework for multi-
agent models nor a widely accepted methodology for developing multi-
agent simulations' [71][11, 139, 137]. Given this lack, model developers
as well as developers of simulation engines have just little guidance
and are left with the inherent complexity mostly on their own.
The lack of a solid theoretical foundation of agent-based modeling
and the missing engineering approach for model development causes
many diculties. Ecient model development is hampered, verific-
ation and validation activities are made di
cult and reusability of
model components is hardly possible.
1.2 Goals of this thesis
With special focus on agent-based modeling and simulation, this thesis
aims at two goals (see Figure 1.2):
1. Improving the effectivity and eciency of model development.
2. Improving the effectivity of model execution.
The following sections provide detailed descriptions of these two goals.
Effective and e
cient model development
Model development often involves the participation of many - possibly
geographically distributed - parties, for instance domain experts,
model developers, software engineers and analysts [12]. Each party has
