Information Technology Reference
In-Depth Information
domains (e.g. financial, economic, social, logistics, chemical, engineering) [30], these
approaches are slow to be widely used as the obtained agent-based simulation models
are at a too low-abstraction level, strongly platform-dependent, and therefore not easy
to verify, modify and update [17, 22, 28]; moreover, significant implementation ef-
forts which are even more for domain experts, typically lacking of advanced pro-
gramming skills, are required [30]. In particular, agent-based simulation models can
be currently obtained mainly through either direct implementation or manual adaption
of a conceptual system model for a specific ABMS platform. The former approach
inevitably suffers from the limitations and specific features of the chosen platform,
whereas the latter requires additional adaptation efforts, the magnitude of which in-
creases depending on the gap between the conceptual and implementation models of
the system.
To overcome these issues, solutions based on approaches well-established in con-
texts other than the ABMS can be exploited; in particular: (i) approaches based on
Platform-Independent Metamodels, which enable the exploitation of more high-level
design abstractions in the definition of Platform-Independent Models and the subse-
quent automatic code generation for different target platforms [1]; (ii) Model-Driven
approaches, which enable the definition of a development process as a chain of model
transformations [4]. Therefore, some solutions for the ABMS context currently ex-
ploit either the approach based on Platform-Independent Metamodels [3, 21, 30] or
that based on Model-Driven [18, 22, 28]. The former approach makes available in this
context the benefits of exploiting the high level abstraction typical of Platform-
Independent Models which also enables the exchange of models regardless of the
specific platform used for the simulation; in addition, Platform-Independent Models
can be reviewed by domain experts working on different target platforms (possibly on
the basis of the simulation result obtained), and then shared with other domain ex-
perts. The latter approach enables the definition of complete and integrated processes
able to guide domain experts from the analysis of the system under consideration to
its agent-based modeling and simulation. In fact, according to the Model-Driven pa-
radigm, the phases which compose a process, the work-products of each phase and the
transitions among the phases in terms of model transformations are fully specified; in
addition, as the Model-Driven paradigm makes it possible the automatic code genera-
tion from a set of (visual) models of the system, the focus can be geared to system
modeling and simulation analysis rather than to programming and implementation
issues.
Under these considerations, this paper proposes the jointly exploitation of both
Platform-Independent Metamodels and Model-Driven approaches as a viable solution
able to fully address the highlighted issues so to promote a wider adoption of the
ABMS especially in those domains that would benefit more from it. In particular, the
paper proposes a Model-Driven process [4] able to guide and support ABMS practi-
tioners in the definition of Platform-Independent Models starting from a conceptual
and domain-expert-oriented modeling of the system without taking into account
simulation configuration details. The proposed process conforms to the OMG
Model-Driven Architecture (MDA) [32] and then allows to (automatically) produce
Platform-Specific simulation Models (PSMs) starting from a Platform-Independent
simulation Model (PIM) obtained on the basis of a preliminary Computation Indepen-
dent Model (CIM).
Search WWH ::




Custom Search