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 119

Table 2. Composition task rules Table 3. Group of acts (AGroup) for the player agent

Task Id Set of Enabling Tasks AAct AAct Execution Setting Tasks

Random Walk <1,1, a> T1

T1 -

Play Neighbor <1,1, b>, with b<a T2, T6, T7

T2 {T1}

Update Age <1,1, c>, with c<a T3

T3 {T1}

Fission <1,1, d>, with d<c & d<b T4

T4 {T7}

Die <1,1, e>, with e<d T5

T5 {T3, T4}

T6 {T2}

T7 {T6}

In Table 3 the Acts derived for the Player SAgent along with the associated Tasks
(see Table 1 and 2) and Execution Settings are reported. As the AMF communication
mechanism among instances of an SAgent is based on access to the SAttributes of the
SAgent (see Section 3.3.1), a single AAct ( Play Neighbor ) is derived from tasks T2,
T6 and T7 which carried out this kind of communication ( guideline G1 ). Execution
Settings of the AActs in Table 3 are characterized by both startingTime and period
equal to one to guarantee that all the Player SAgents perform all their AActs in each
simulation step, and priorities are set ( guideline G2 ) on the basis of the Compositions
Task Rules (see Table 2). On the basis of the AAct Execution Setting (see
Section 3.3.1) in Table 3 the type of AActs is obtained ( guideline G4 ).
In Figure 7.a an example of a PIM model representation, obtained by exploiting the
visual and Eclipse-based modelling environment provide by AMF, is reported. More-
over, an AAct of the AInizialize type ( Inizialize ) has been introduced for setting up
the SAttributes of the DPDGame SContext and the Player SAgent (guideline G4) . In
Figure 7.b. the definition of the Random Walk and Update Age AActs is reported
where the actions associated to each AAct are defined by exploiting the wide set of
functions provided by AMF ( rule R5 ).
Starting from this definition of the PIM model, AMF is able to automatically gen-
erate the PSM models and the related code for the ABMS platforms which are cur-
rently supported: Repast Simphony [29], Ascape [35] and Escape [3]. The simulation
of the system can then be executed in a target simulation environment and simulation
results can be thoroughly analyzed by exploiting several analysis tools (as Matlab, R,
VisAd, iReport, Jung) which can be directly invoked from the environment.

(a) DPDGame model in AMF (b) Random Walk and Update Age AActs

Fig. 7. The AMF-based PIM model of the DPDGame