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 117

G4. Tasks (or group of Tasks ) that must be executed at every simulation step are
mapped into ARules , except for Tasks containing a Do-While loop which should be
mapped into APerforms . Tasks that must be executed with a periodicity different from
a single simulation step should be mapped into ASchedule ; finally, Tasks that must be
executed once before any other Tasks should be mapped into AInizialize (an AAct of
type AInitialize should nevertheless be provided for setting the SAttributes ).

4 Exploiting the Proposed MDA-Based Process

In this section, the MDA4ABMS process is exemplified with reference to the well-
known Demographic Prisoner's Dilemma which was introduced by Epstein in 1998
[15] and is able to represent several social and economic complex scenarios in which
interesting issues regard the identification of starting configurations and conditions
that allow initial populations to reach stable configurations (in terms of both density
and geographic distribution). Specifically, in these scenarios k players are spatially
distributed over an n -dimensional toroidal grid. Each player is able to move to empty
cells in its von Neumann neighborhood of range 1 ( feasible cells), is characterized by
a fixed pure strategy ( c for cooperate or d for defect) and is endowed with a level of
wealth w which will be decremented or incremented depending of the payoff earned
by the player in each round of the Prisoner's Dilemma game played during its life
against its neighbors [15]. The player dies when its wealth level w becomes negative,
whereas, when w exceeds a threshold level w b , an offspring can be produced with
wealth level w 0 deducted from the parent and plays using the same strategy as the
parent unless a mutation (with a given rate m ) occurs. A player also dies if its age
exceeds a value age max randomly fixed at the player creation.

4.1 The CIM Model

For the Demographic Prisoner's Dilemma , the CIM model envisages a DPDGame
Society of k Player Agents which are situated in an Environment which includes a
Grid Resource constituted by an n -dimensional toroidal grid. Main Properties of the
DPDGame Society are Prisoner's Dilemma payoffs , initial and threshold wealth levels
( w 0 , w b ), and mutation rate ( m ), and those of the Player Agent are its wealth level w ,
age , and strategy . The Behavior of the Player Agent is obtained by composing the set
of Tasks reported in Table 1 according to the Composition Task Rules shown in
Table 2; corresponding UML Activity diagrams are reported in Figure 6.

4.2 The PIM Model

In this section, the transformation from the defined CIM to a PIM is detailed with
reference to a simulation scenario where all players are required to play exactly one
round in a simulation step.
The transformation from the CIM to a PIM is enabled by the mapping between the
CIM and PIM metamodels (see Sections 3.3.2, 3.3.3) which originates: the
DPDGame SContex from the DPDGame Society ( rule R1 ), the Player SAgent from
the Player Agent ( rule R2 ), the GameSpace SProjection from the Grid Resource