Information Technology Reference
In-Depth Information
task, in absence of tools enabling (semi)automatic transitions, can be quite difficult
due to the consistent gap between the design and the operational model of the system.
3. AOSE Processes and Methods for Agent-based Modeling and Simulation.
Processes and methodologies for the analysis, design and implementation of agent-
based models, can be derived from the AOSE domain and possibly adapted for
ABMS. Specifically, among the several available AOSE methodologies (such as
PASSI [13], PASSIM [14], ADELFE [7], GAIA [43] and GAIA2 [44], TROPOS [8],
SONIA [2], SODA+zoom [27], MESSAGE [9], INGENIAS [36], O-MaSE [16],
SADDE [39], and Prometheus [34]), some of these, such as GAIA2 [44],
SODA+zoom [27] and MESSAGE [9], provide processes, techniques and/or abstrac-
tions which are particularly suited for the ABMS context; moreover, specific ABMS
extensions of AOSE methodologies can be found in [23, 37, 40].
Although these proposals can represent reference methods for guiding domain ex-
perts through the different phases of an ABMS process, only few of them go beyond
the high level design phase and deal with detailed design and model implementation
issues. As a consequence, they fail in supporting domain experts in the definition of
agent-based models which can be directly and effortless executed on ABMS plat-
forms able to fully handle the phases of simulation and result analysis. In fact, the
adaptation between the models obtained and the target simulation models requires
significant efforts which are time-consuming, error-prone and demands advanced
programming skills.
4. Model-driven Approaches for ABMS. To fully support and address not only the
design but also the implementation of simulation models on available ABMS plat-
forms, some Model-Driven approaches for ABMS have been proposed [18, 22, 28].
However, as they refer to specific ABMS platforms, their exploitation is strongly
related to the adoption of these platforms (e.g. Repast Simphony for [18], BoxedE-
conomy for [22], MASON for [28]).
With reference to other MDA-based approaches, which aim to provide a methodo-
logical support for the design of agent-based distributed simulations compliant to the
High Level Architecture (HLA) [12, 41], in the ABMS context a still debated issue
[26] concerns the trade-off between the overhead which the HLA layer introduces and
the provided distribution and interoperability benefits. Specifically, some approaches
conceive HLA as the PSM level of an MDA Architecture and provide a process for
transforming a System PIM, based on UML, in a HLA-based System PSM [12]. On
the contrary, HLA is conceived as the PIM level in [41] where the Federation Archi-
tecture Metamodel (FAMM) for describing the architecture of a HLA compliant
federation is proposed to enable the definition of platform-independent simulation
models (based on HLA) and the subsequent code generation phase.
3
The MDA4ABMS Process
This section describes the proposed MDA4ABMS process which combines the Mod-
el-Driven approach and the exploitation of Platform-Independent Metamodels so
making available in the ABMS context the benefits of both exploited approaches. The
MDA4ABMS process relies on the Model-Driven Architecture (MDA) [32] and the
Search WWH ::




Custom Search