A Process Based on the Model-Driven Architecture to Enable the Definition of Platform-Independent Simulation Models - Simulation and Modeling Methodologies, Technologies and Application - page 115

a single iteration, whereas AActs of type APerform , in each iteration in which they are
scheduled, are over and over again executed until their escape conditions are met.
With respect to the communication among SAgents, since the SAttributes of an
SAgent can be freely accessed by all the instances of the SAgent, and the SAttributes
of an SContext by all the instances of all the SAgents in the SContext, communication
among instances of the same SAgent (intra-agent communication) can exploit SAgent
SAttributes whereas communication among instances of different SAgents (inter-
agent communication) can be enabled by SContext SAttributes .
Finally, the design of SAgent communications should take into account how ran-
dom choices among the enabled AAct (see Figure 4) affect the values of the SAttri-
butes on which the communication is based.
3.3.2 Mapping from CIM to PIM Metamodels: Mapping Rules
The automatic generation of a PIM starting from a given CIM is enabled by the
QVT/R-based representation of mapping rules [32, 33]. Specifically, due to the dif-
ferent abstraction level between the concepts of the reference CIM and PIM metamo-
dels (see Section 3.1 and 3.2), the mapping rules introduced in this Section along with
the QVT/R-based representation allow to obtain a preliminary PIM which needs to be
refined by applying additional guidelines (see Section 3.3.3).
The preliminary transformation of a CIM into a PIM, which involves the definition
of instances of concepts of the PIM metamodel from instances of concepts of the CIM
metamodel by exploiting the mapping rules among the corresponding concepts, con-
sists in the following steps which are listed in the order they should be performed:
R1. each Society is transformed into a Simulation Context ( SContext ) and any en-
closed Society into a (sub) SContext of the corresponding enclosing Society ; SAttri-
butes of each SContext are, then, originated by the Properties of the corresponding
Society ;
R2. each Agent belonging to a Society is transformed into an SAgent of the corres-
ponding SContext, generating the SAgent SAttributes on the basis of the Agent Prop-
erties , and introducing the SAgent AGroup which groups the AActs constituting its
behavior;
R3. on the basis of the set of Resources , which compose the Environment in which
Agents are situated, a set of SProjections , whose types (SNetwork, SGrid, SSpace,
SGeography, SValueLayer) depend on the characteristics of the mapped Resources ,
are then introduced in the corresponding SContext ;
R4. AActs associated to each SAgent are to be defined on the basis of the behavior of
the corresponding Agent which is composed by a set of Tasks organized according to
Composition Task Rules ; this transformation is not direct as requires to take into ac-
count the specific aspects of both an AMF-based PIM (see Section 3.3.1) and the
simulation scenarios to be represented;
R5. Actions which constituted the Tasks mapped into an AAct have to be properly
realized by exploiting the wide set of predefined functions provided by AMF [3].
With reference to the above introduced rules, in the following, some guidelines are
provided which address some relevant issues related to: (i) the different communica-
tion mechanisms adopted by CIM and PIM metamodels, the former based on