A Survey of Model-Driven Approaches Applied to DEVS
A Comparative Study of Metamodels and Transformations
Stéphane Garredu, Evelyne Vittori, Jean-François Santucci and Bastien Poggi
Department of Computer Science, University of Corsica, Campus Grimaldi, Corte, France
Keywords: M&S, DEVS, MDE, MDA, M2M, M2T, Interoperability.
Abstract: Model-Driven Engineering (MDE) is a subset of Software Engineering (SE) which focuses on models.
MDE provides techniques and guidelines to create models (metamodeling) and to transform them onto other
models (including code). Recently, several MDE approaches have been successfully applied to the world of
Modeling and Simulation (M&S), of which DEVS (Discrete EVent system Specification) is one of the most
popular formalisms. The goal of those approaches is to increase DEVS interoperability. Many of them resort
to a metamodel to describe DEVS concepts. The most recent ones also provide automatic code generation
“Model-To-Text” (M2T) towards DEVS simulators (DEVS “internal” interoperability) and establish links
between DEVS and other formalisms, thanks to Model-To-Model (M2M) transformations (DEVS
“external” interoperability). The purpose of this paper is to give a state of the art of the MDE contributions
to DEVS formalism and to provide a comparative study of the most recent ones.
1 INTRODUCTION
Model-Driven Engineering (MDE) is a set of
methods, approaches and techniques inherited from
Software Engineering (SE). The common point
shared by all of the MDE-oriented approaches is
the use of models.
DEVS (Discrete EVent system Specification)
(Zeigler 1976) formalism relies on a strong
mathematical background, inspired by the set
theory, and enables to create models, which can be
interconnected, and to simulate them. To simulate a
DEVS model, it is needed to make a move from a
theoretical model into a concrete implementation.
There exist several different DEVS-oriented
frameworks, lying on different object-oriented
languages, used by several research teams in the
world: that induces a lack of interoperability
between DEVS models, which cannot be reused by
the same team on another DEVS-oriented platform,
and even less be shared among the whole DEVS
community. But this lack of interoperability
logically generates a need too. This need for
interoperability between DEVS implemented
models gave rise to several approaches, and a
significant part of them is inspired by MDE.
This paper is dedicated to those MDE
contributions to DEVS formalism. We chose to
highlight the approaches which propose a meta-
model for DEVS and involve transformation
mechanisms. We present a comparative study of
those approaches, focusing on three key aspects: the
way they handle the DEVS basic concepts, the
underlying meta-formalism, and the motivations of
the work (improving interoperability, code
generation…).
This paper is organized as follows: it starts with
a background section, dedicated to the DEVS
formalism, and the key elements of MDE. The
following section presents some of the MDE
approaches that have been applied to DEVS
formalism, and compare them. Finally, we conclude
with a short discussion on the actual and future
challenges in DEVS interoperability using MDE.
2 BACKGROUND
2.1 DEVS Formalism
Since the 1970s, formal approaches have been
proposed for the modeling and the simulation
discrete event dynamic systems, and the DEVS
formalism is a part of them. This formalism may be
defined as a universal and general methodology,
179
Garredu S., Vittori E., Santucci J. and Poggi B..
A Survey of Model-Driven Approaches Applied to DEVS - A Comparative Study of Metamodels and Transformations.
DOI: 10.5220/0005041001790187
In Proceedings of the 4th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH-2014),
pages 179-187
ISBN: 978-989-758-038-3
Copyright
c
 2014 SCITEPRESS (Science and Technology Publications, Lda.)

which
p
systems,
formalis
m
specific
a
modular
DE
V
p
latfor
m
2013) or
At t
h
and
p
lat
f
models.
one-dim
e
graphica
automat
o
tuples
a
necessar
y
most of
its imp
Several
a
standard
quote h
e
Interope
r
2008).
D
atomic
m
2.1.1
D
The bas
i
AM
=
•
•
•
•
•
•
•
2.1.2
D
The pu
r
describe
p
rovides to
o
whose beha
v
m
lies on the
a
tion of comp
l
and hierarch
i
V
S has bee
n
m
s, for instan
c
PyDEVS (B
o
h
is time, there
f
or
m
-indepen
d
Usually, a
s
e
nsional (q
u
a
lly represent
e
o
n. If it nee
a
re describe
d
y
, a pseudo-l
a
the time, a D
p
lemented f
o
approaches h
rep
r
esentati
o
ere the wor
k
r
ability Sta
n
D
EVS is co
m
m
odels and th
e
D
EVS Atom
i
i
c entity in D
E
=
< X, Y, S, t
a
X = {(p,v)|
p
input events
input ports
a
values for th
o
Y = {(p,v)|
p
output event
s
output ports
values for th
o
S is the state
s
ta: S → R
0
+
function (or
l
δ
int
: S →
S
function;
δ
ext
: Q ×
X
e∈[0,ta(s)]}
function;
λ: S → Y, w
i
v∈Y
p
} is the
D
EVS Coup
l
r
pose of a
D
a hierarchy:
o
ls to mode
l
v
io
r
is based
system theor
y
l
ex discrete e
v
i
cal way.
n
implemen
t
c
e DEVSJA
V
o
lduc et al. 2
0
exist no uni
q
d
en
t
) represe
n
s
imple DEV
S
u
alitative) st
a
e
d as a kind o
f
ds to be m
o
d
mathematic
a
nguage is us
e
EVS model
o
o
rm (objec
t
-
ave been try
i
o
n of DEVS
k
of the S
I
n
dards Orga
n
m
posed of t
w
e
coupled mo
d
i
c Model
E
VS is the ato
a
, δ
int
, δ
ext
, λ >
p
I
nputPorts
,
se
t
;
I
nputPo
a
nd X
p
is th
e
o
se input port
s
p
∈OutputPort
s
s
set; Output
P
and Y
p
is th
e
o
se output po
r
s
set of the sy
s
+
∪ +∞ is t
h
l
ifespan of a s
t
S
is the in
t
→ S with
is the ex
t
i
th Y = {(p,v
)
output functi
o
l
ed Model
D
EVS coupl
e
it has sub-m
o
l
and simu
l
on events.
T
y
and
p
ermits
v
ent systems
i
t
ed on sev
e
V
A (DEVSJA
V
0
01).
q
ue (standardi
z
n
tation of D
E
S
model, wit
h
a
te variable,
f
annotated fi
n
o
re accurate,
a
lly and, w
h
e
d. Neverthel
e
o
nly exists u
n
o
riented co
d
i
ng to propos
models: we
c
SO (Simula
t
n
ization) (S
I
w
o artifacts:
d
els.
m
ic model:
, where :
,
v∈X
p
} is
r
ts is the se
t
e
set of poss
i
s
;
s
, v∈Y
p
} is
P
orts is the se
t
e
set of poss
i
r
ts;
s
tem;
h
e time adva
n
t
ate);
t
ernal transi
t
Q = {(s,e)/s
t
ernal transi
t
)
|p∈OutputPo
o
n.
e
d model is
o
dels (which
c
l
ate
T
his
the
i
n a
e
ral
V
A
z
ed
E
VS
h
a
is
n
ite
i
t
s
h
en
e
ss,
n
der
d
e).
s
e a
c
an
t
ion
I
SO
the
the
t
of
i
ble
the
t
of
i
ble
n
ce
t
ion
∈S,
t
ion
o
rts,
to
can
b
e
e
the
m
C
M
W
h
2.2
Th
r
ide
a
for
DE
“in
t
of
t
ai
m
b
as
the
2.2
Ev
e
the
dif
f
DE
V
si
m
mo
d
are
(ac
c
int
r
int
e
a
D
Th
i
sol
v
e
ither atomic
m
. A coupled
M
= <
X
, Y, D,
h
ere
• X and
Y
• D is the
• M
d
is a
D
• EIC is
t
an ext
e
b
etwee
n
coupled
its sub-
m
• EOC
i
couplin
g
link bet
w
coupled
of its su
b
• IC is t
h
internal
the out
p
input p
o
• select is
DEVS I
n
r
oughout this
a
of interope
r
a simple cla
s
V
S interoper
a
t
ernal” or “e
x
t
he two diffe
r
m
to increase
ed or model-
second one.
.1 “Intern
a
Interop
e
e
n when it is
word inte
r
f
erently, depe
n
For instance,
V
S coupled
m
ulated toget
h
d
els, it’s not
not imple
m
c
ording to t
h
r
oduce here
e
roperability,
b
D
EVS world,
s is a typic
a
v
ed using
o
r coupled) a
n
model is for
m
{Md|d
∈ D},
E
Y
are the same
set of compo
n
D
EVS model
t
he set of ext
e
e
rnal input
n
the input
model and t
h
m
odels;
s the set
g
s; an extern
a
w
een the out
p
model and t
h
b
-models;
h
e set of i
n
coupling is
a
p
ut port of
a
o
rt of another
s
the tiebreake
r
n
teroperabi
l
article, we f
r
r
ability. We
b
s
sification of
t
a
bili
t
ies, in t
e
ternal”. Then
r
ent families
DEVS inter
o
b
ased. In thi
s
a
l” vs. “Ext
e
e
rability
employed wi
t
r
operability
n
ding on the
s
let us consid
e
models hav
e
e
r
. Although
possible to
s
m
ented in
t
h
e same si
m
the con
c
b
ecause this
p
between t
w
a
l case of an
a DE
V
n
d couplings
m
ally defined
b
E
IC, EOC, IC,
as in 2.2.1)
n
ent names,
d
(atomic or c
o
e
rnal input c
o
coupling is
port of the
h
e input port
o
of external
a
l output cou
p
p
ut port of th
e
h
e output po
r
n
ternal coupl
i
a
link which
a
sub-model
s
ub-model;
e
r (selection)
f
l
ity
fr
equently ref
e
b
egin with a
p
the different
e
rms of rang
e
n
we give an
o
of approach
e
o
perability: si
m
s
paper, we
f
e
rnal”
t
hin a DEVS
can be un
d
s
ituation.
er the ability
e
to be im
m
they both ar
e
s
imulate the
m
the same
l
m
ulator). He
n
c
ept of “
i
p
roblem takes
w
o DEVS si
m
issue whic
h
V
S simulat
o
between
b
y:
selec
t
>
d
∈ D; 
o
upled);
o
uplings;
a link
current
o
f any of
output
p
ling is a
e
current
r
t of any
i
ngs; an
involves
and the
f
unction.
e
r to the
p
roposal
kinds of
e
of use:
o
verview
e
s which
m
ulato
r
-
f
ocus on
context,
d
erstood
that two
m
ediately
e
DEVS
m
if they
l
anguage
n
ce, we
i
nternal”
place in
m
ulators.
h
can be
or
-based
SIMULTECH2014-4thInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
180

interope
r
model t
o
refers t
o
once ag
world.
T
oriented
success
fu
On t
h
a DEVS
model,
p
roblem
of rema
i
with tw
o
formalis
m
call “ex
t
and anot
h
Thos
issues h
a
and (Ga
r
2.2.2
S
S
During
t
made to
DEVS
fr
have be
e
in (Tour
a
The
interope
r
simulato
b
etween
where d
i
we can
oriented
b
ased
Architec
2.3 M
o
MDE
method
o
and tra
n
MDE is
of the
formalis
m
incarnat
i
(OMG)
M
2001).
2.3.1 A
b
Metamo
d
differen
t
r
ability appr
o
o
be exporte
d
o
an “intern
ain, this pro
T
his is a t
y
interopera
b
fu
lly applied.
h
e other hand
,
model to be
it is obvio
u
is different f
r
i
ning in a D
E
o
different fo
r
m
s and not i
m
t
ernal” the in
t
h
er formalis
m
e internal
a
a
ve been exp
l
r
redu et al. 20
S
imulator-B
a
S
olutions
t
he last deca
d
fill the gap
b
fr
ameworks.
T
e
n used. An o
v
a
ille et al., 20
0
first one
r
ability from
rs, using, fo
r
at least two
d
i
fferent mode
l
quote (Seo,
proposal fo
r
interoperabili
ture.
o
del-Drive
n
is a gen
e
o
logy that fo
c
n
sforming m
o
seen as a mo
d
real world,
m
. One o
f
i
ons is the
M
odel-Drive
n
b
straction
L
d
els and mo
d
t
abstraction l
o
ach. The abi
l
d
onto sever
a
al” interope
r
blem remain
y
pical case
w
b
ility appr
o
,
if we consi
d
simulated w
u
s that this
r
om the previ
o
E
VS world,
w
r
malisms. H
e
m
plementatio
n
t
eroperability
m
.
a
nd external
l
ored in (Gar
r
13).
a
sed vs. Mo
d
e, several e
f
b
etween the
d
T
wo major k
i
v
erview of th
0
9).
aims to
the point
o
r
instance, st
a
d
ifferent sim
u
l
s are defined
.
2009) which
r
a better
D
ty using
S
n
En
g
ineer
i
e
ric softwar
e
c
uses on cre
a
o
dels: in fac
t
d
el, i.e. an ab
s
described
w
f
the most
Object Ma
n
n
Architectur
e
L
evels in M
D
d
els are linke
d
evels, in the
f
l
ity for a D
E
a
l platforms
a
r
ability beca
u
s in the D
E
w
here a mo
d
o
ach can
d
er the ability
ith a non-D
E
interoperabi
o
us ones. Inst
e
w
e have to
d
e
re we deal
w
n
. We propos
e
between D
E
interoperabi
l
r
edu et al. 20
del-Based
f
forts have b
e
d
ifferent exis
t
i
nds of soluti
o
e
m can be fo
u
increase
o
f view of
a
ndard messa
g
u
lation platfo
r
.
As an exam
p
is a simula
t
D
EVS simula
t
S
ervice-Orie
n
i
n
g
e
develop
m
a
ting, exploi
t
t
, everything
s
traction of a
p
w
ith a mode
l
popular M
D
n
agement Gr
o
e
(MDA) (O
M
D
E
d
to each othe
r
f
ollowing wa
y
E
VS
a
lso
u
se,
E
VS
d
el-
be
for
E
VS
lity
ead
d
eal
w
ith
e
to
E
VS
lity
12)
een
t
ing
ons
u
nd
the
the
a
ges
r
ms
p
le,
t
o
r
-
t
o
r
-
n
ted
m
ent
t
ing
in
p
art
l
ing
M
DE
o
up
M
G
r
at
y
: a
mo
d
lev
e
A
m
co
nf
me
t
“m
e
des
hig
h
an
d
sh
o
2.3
M
o
me
t
tra
n
(M
2
ge
n
fol
l
me
t
M
M
b
et
w
an
d
He
r
b
et
w
M
M
Ea
c
to
tra
n
me
t
or
tra
n
M
M
is
c
(M
2
d
el (M
1
lev
e
e
l), and it co
nf
m
etamodel de
nf
orms itself t
o
t
aformalism
i
e
ta” levels
cription and
h
er level tha
n
d
description
a
o
wn in Figure
Figure
1
.2 Model
T
del transfor
m
t
amodels, t
h
n
sformation i
s
2
) and execu
t
n
eral case fo
l
owing one: i
f
t
amodels, an
d
M
x, the ass
o
w
een MMx a
n
d
My. It trans
r
e, we assu
m
w
een the for
m
M
y, which is
n
c
h metamode
l
a differ
e
n
sformation i
t
t
aformalism t
o
MF
y
. If M
M
n
sfor
m
ation i
s
M
y are the sa
m
c
alled endoge
2
M) transf
o
e
l) describes
nf
orms to its
m
s
cribes a mo
d
o
a metaform
a
s located at
hierarchy: i
t
conforms t
o
n
M
3
. The co
n
a
re detailed i
n
1.
1
: Abstraction l
e
T
ransformat
m
ations are,
h
e key con
c
s
defined at
t
ed at the m
o
r
a model t
r
f
MMx and
M
d
Mx is a mo
d
o
ciated trans
f
n
d MMy and
e
f
orms Mx in
t
m
e that ther
m
alis
m
s descr
i
n
ot always the
l
, according t
o
nt metafor
m
t
self, as a
m
o
o, which ca
n
M
x and M
M
s
called exo
g
m
e metamod
e
n
ous. Usuall
y
o
rmation fo
l
the real w
o
m
etamodel (
M
d
eling formal
i
alism (M
3
le
v
the top of t
h
t
contains
i
o
itself: the
r
n
cepts of con
fo
n
(Bézivin, 2
0
l
evels in MDE.
t
ions in MD
E
with mod
c
epts of
M
the metamo
d
o
del level (
M
t
ransformatio
n
M
My are tw
o
d
el which con
f
ormation is
executed bet
w
t
o another m
o
r
e exists a
m
r
ibed by MM
x
e
case.
o Figure 1,
c
r
malism, a
n
m
odel, confor
m
n
be, in this c
a
M
y are diffe
r
g
enous. If
M
e
l, the transf
o
y
, a Model-T
o
llows the
o
rld (M
0
M
2
level).
i
sm, and
v
el). The
h
e MDE
i
ts own
e is no
for
mance
0
04) and
E
e
ls and
M
DE. A
d
el level
M
1
). The
n
is the
o
distinct
forms to
defined
w
een Mx
o
del My.
m
apping
x
and by
c
onforms
n
d the
m
s to a
a
se, MFx
r
ent, the
M
Mx and
o
rmation
o
-Model
schema
ASurveyofModel-DrivenApproachesAppliedtoDEVS-AComparativeStudyofMetamodelsandTransformations
181

p
resente
d
(M2T)
transfor
m
its impl
e
turns a
resortin
g
model t
r
(Mens e
t
F
2.3.3
A
The M
D
2004) is
Figure
2
MOF (
M
p
opular
approac
h
(UML)
(
N
ev
e
drawbac
k
express
e
solution
s
UML (
p
metamo
d
used in
p
One
former
i
Langua
g
consider
latter of
f
often m
o
instead
o
MD
A
CIMs (
C
and PS
M
Some
tools/la
n
dedicate
d
(Jouault
d
in figure
2
transformati
o
m
ation (from
e
mentation i
s
model Mx
g
to a targe
t
r
ansformatio
n
t
al. 2006).
F
igure 2: A cla
s
A
Particular
D
A approach
very similar
2
. At the top
M
eta-Object
metamodel
s
h
es is the
U
(
OMG UML
2
e
rtheless, the
a
k
when sp
e
e
d. To solve
t
s
: use the s
p
p
rofiles), or
d
el (conformi
n
p
lace of UML
of the most
i
s the OMG
g
e) (OMG S
y
ed as an exte
n
f
ers a larger s
o
re complicat
e
o
f using an ex
i
A
uses three
C
omputation
M
s (Platform
I
famous M
D
n
guages are
A
d
to M2T
et al. 2006)
d
2
. Even if a
o
n is, in t
h
a MDE-
b
ase
d
s
often simpl
into sourc
e
t
metamodel.
n
techniques
s
sic MDE tran
s
MDE Inca
r
(OMG 2001
to the archi
t
level lies th
e
Facility). O
n
s
used in
U
nified Mo
d
2
011) (Rumb
a
a
daptability
o
e
cific conce
p
t
his, the O
M
p
ecialization
define a
c
n
g to the MO
.
famous ap
p
SysML (SY
S
y
sML 2012)
,
n
sion of a sub
et of possibil
i
e
d to create a
i
sting one.
different k
i
Independent
I
ndependent/
S
D
A-inspired
A
cceleo (A
C
transformati
o
d
edicated to
M
Model-To-
T
h
eory, a m
o
d
point of vie
ified: it dire
c
e
code, wit
h
Many exis
t
are detailed
s
formation.
r
nation: M
D
) (Kleppe et
t
ecture show
n
e
metaformal
i
n
e of the
m
MDA-orie
n
d
eling Langu
a
a
ugh et al. 20
0
o
f UML can
b
p
ts have to
G proposes
t
mechanisms
c
ompletely
n
F) which wil
l
p
lications of
S
tems Mode
l
,
which can
set of UML.
T
i
ties, even if
i
new metam
o
i
nds of mod
e
Models), P
I
S
pecific Mod
e
transforma
t
C
CELEO 20
1
o
ns, and
A
M
2M. They b
T
ext
o
del
e
w),
c
tly
h
out
t
ing
in
D
A
al.
n
in
i
s
m
m
ost
n
ted
age
0
5).
b
e a
be
t
wo
of
n
ew
l
be
the
l
ing
be
T
he
i
t is
o
del
e
ls:
I
Ms
e
ls).
t
ion
1
3),
A
TL
b
oth
are
(E
M
3
In
t
the
a s
t
b
e
e
Th
e
rep
r
rat
h
do
m
nec
me
t
ne
w
a
m
the
of
t
ato
m
tha
t
ap
p
mo
d
20
0
the
on
e
3.1
Th
o
int
e
exi
s
me
t
de
n
ap
p
(ab
s
not
a
mo
d
we
l
is t
h
fro
m
al.
tec
h
to
D
Sta
t
mo
d
sa
m
p
ro
p
Sta
t
available
a
M
F) (Steinber
g
MDE A
P
TO DE
V
t
his section, t
h
background
s
t
ate of the art
e
n applied to
e
main pur
p
r
esent DEV
S
h
er than code.
We have pr
e
m
ain concepts
essary to re
s
t
amodels co
u
w
ones could
b
The biggest
p
m
etamodel for
structure (co
u
t
he behavior
m
ic function
s
t
the first (
u
p
lied to DE
V
d
els (see 3.1)
0
7) are based
moment, the
e
s.
UML-L
i
o
se MDE
a
e
roperability
o
s
ting metam
o
t
amodel: U
M
n
ominator.
T
p
roaches is t
h
s
tract syntax)
a
tions as an a
b
d
els written
i
l
l-documente
d
h
at it is often
m
those UM
L
2007) who
h
nological sp
a
One of the f
i
D
EVS was (S
c
t
echarts for
m
d
els that we
r
m
e philosop
h
p
osed to u
s
t
echarts for
m
s Eclipse
M
g
et al. 2009)
p
P
PROAC
H
V
S FORM
A
h
e ideas and
c
s
ection are pu
t
of the MDE
DEVS, in a
p
ose of tho
s
S
models, f
o
e
viously state
d
had to be ex
p
ort to a met
a
l
d be reused
b
e created.
p
roblem that r
e
DEVS is no
t
u
pled model
s
(atomic mo
d
s
. In broad o
u
u
ntil 2007)
M
V
S were bas
e
,
while the m
o
on new met
a
s
e seem to b
e
i
ke Approa
c
a
pproaches t
a
o
f the DEVS
o
dels, or mo
L. Hence, U
M
T
he main
a
h
e use of U
M
,
so that the
u
b
stract synta
x
i
n UML are
d
. However, a
n
difficult to o
b
L
models, ex
c
use tools r
a
ce.
i
rst UML-Li
k
c
hulz et al. 2
0
m
alism (Har
e
r
e equal to
D
h
y, (Risco-
M
s
e XML in
m
alism (kno
w
M
odeling Fr
a
plugins.
H
ES APP
L
A
LISM
concepts pre
s
t
together. W
e
approaches t
h
model-orien
t
s
e approach
e
o
cusing on
c
d
that, when
p
ressed (i.e.
w
amodel), we
l
(and specia
l
e
mains when
t
the represen
t
s
), but the de
s
d
el), with th
e
u
tline, it can
M
DE-like ap
p
e
d on existi
n
ost recent on
e
a
models (see
e
the most
pr
c
hes (before
a
ckled the
formalism b
y
o
re exactly
o
M
L is their
c
a
dvantage o
f
M
L as a m
e
u
ser can use
g
x
. Globally, t
h
well-represe
n
n
important
d
o
btain simulat
i
c
ept (Risco-
M
r
elated to t
h
k
e approache
s
0
00). It resort
e
e
l 1987) t
o
D
EVS model
s
M
artin et al
.
order to
w
n as U
M
a
mework
L
IED
s
ented in
e
present
h
at have
t
ed way.
e
s is to
c
oncepts
specific
w
hen it is
l
l-known
l
ized) or
defining
t
ation of
s
cription
e
DEVS
be said
p
roaches
n
g meta-
e
s (since
3.2). At
r
omising
2007)
lack of
y
reusing
o
nly one
c
ommon
f
those
e
tamodel
g
raphical
h
e DEVS
n
ted and
rawback
i
on code
M
artin et
h
e XML
s
applied
e
d to the
o
create
s
. In the
.
2007)
t
ranslate
M
L state
SIMULTECH2014-4thInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
182

machine diagram) into DEVS. This goes back to
create DEVS models using Statecharts.
Some researches proposed to use more than one
UML diagram. For instance, (Zinoviev 2005)
claims that DEVS suffers from a lack of graphical
representation and uses UML to represent DEVS
models. Atomic models are specified using UML
state machine diagrams, while coupled models are
described with UML component diagrams. This
approach was not given any concrete
implementation. On the other hand, (Mooney et al.
2009) proposed a complete environment that allows
the execution of DEVS models specified with
UML: this approach tackles the temporal aspects,
which are more present in DEVS than they are in
UML.
Other approaches rely on the UML extension
mechanisms. Some use existing UML profiles, this
is the case for (Nikolaidou et al. 2008). The authors
employ SysML to graphically represent DEVS
models created with DEVSML. Others create
special profiles, for instance (Nikolaidou et al.
2008). The authors also provide a partial code
generation (skeleton).
3.2 Recent MDE Approaches (since 2007)
Usually, all the MDE approaches we study here
follow the same pattern, inherited from MDE
philosophy:
 Define a metamodel for DEVS
 Create M2T transformations (and/or)
 Define M2M mappings
The definition of a metamodel for DEVS is the
first step, it makes one able to define DEVS atomic
and coupled models. To create it, the starting point
is the basic definition of DEVS formalism (2.1);
In the second step, links between the DEVS
metamodel and DEVS simulators (code) are
established. The goal here is to reach several DEVS
frameworks from only one DEVS platform-
independent model (internal interoperability).
Hence, a single model would be simulated on
several platforms (M2T);
The third step establishes links between the
DEVS metamodel and other formalisms. Those
links tackle the external DEVS interoperability
problem using M2M techniques.
Regardless of the technique used, the most
essential element that directly influences the
viability, the power of expression and the accuracy
of the DEVS models is the quality of the metamodel
they lie on.
The approaches we study in the next subsections
take place within different technological spaces
(Bézivin et al. 2005). The metaformalisms used by
the following approaches are: XML Schema, MOF-
Ecore, EBNF (Extended Backus-Naur Form), E/R
Diagrams.
3.2.1 XML-Based Approaches
DEVSML (Mittal et al. 2007) is a metamodel used
for the description of DEVS models within Service-
Oriented Architectures (SOA). This approach
enables to specify DEVS models but it is not totally
platform-indepndent: the models can only be used
with Java-based DEVS platforms, because
DEVSML inherits from a XML-like language used
to describe Java programs, JavaML (Badros 2000).
More recently, (Mittal et al. 2012) introduced
the “DEVSML framework” where it is possible to
create Domain-Specific Languages (DSLs) and
associate them to the DEVSML language. It is also
possible to generate DEVSJAVA (DEVSJAVA
2013) simulation code. This MDE-oriented
approach is implemented within the Eclipse
Modeling Framework (EMF) but uses the EBNF
metaformalism.
There also exists a similar approach: SimStudio
(Touraille et al. 2011) based on a metamodel named
DEVS Markup Language (DML). The idea is to
improve DEVS internal interoperability using
“hybrid code”: some same parts of the code are
written in different languages so they can be chosen
during the code generation.
DEVSML has been used for M2M
transformations towards DEVS (Risco-Martin et al.
2007) (see 3.1). The M2T implementations use the
language Xtend while all the transformations in
SimStudio use XSLT (eXtensible Stylesheet
Language Transformations).
3.2.2 E/R Diagrams-Based Approaches
The metamodeling environment AToM
3
(Lara et al.
2002) has been used by (Posse et al. 2003) and after
that by (Levytskyy et al. 2003) to create a DEVS
metamodel based on the E/R Diagrams
metaformalism. Those approaches use the abilities
of AToM
3
in order to generate a graphical modeling
environment based on the DEVS meta-model. All
the DEVS states are enumerated (sequential states).
Another DEVS metamodel, introduced in (Song
2006), allows specifying state variables as
attributes. The transition functions are specified
with text blocks. For all of those approaches, it is
ASurveyofModel-DrivenApproachesAppliedtoDEVS-AComparativeStudyofMetamodelsandTransformations
183

possible to generate Python code for the simulator
PyDEVS (Bolduc et al. 2001) using M2T
mechanisms. Moreover, some M2M approaches
were implemented, for instance (Borland 2003)
proposed a transformation from Statecharts to
DEVS. Code generation and model transformations
are performed using the Python language.
3.2.3 MOF-Based Approaches
Those approaches are the most recent ones. They
are located in the object-oriented technological
space, and use MOF (and its subset EMF Ecore) as
a metaformalism: in other words, they can be
considered as MDA approaches.
(Lei et al. 2009) use a M2M transformation
from DEVS to SMP2 (Simulation Model Portability
2). To do so, they resort to two packages. One of
them contains the definition of the DEVS
formalism. The DEVS functions are taken into
account but the programmer needs to fill them:
otherwise, they remain empty. The states are
explicitly enumerated, under a textual form.
On the contrary, EMF-DEVS (Sarjoughian et al.
2012) is only centered on the DEVS formalism and
its internal interoperability (EMF-DEVS does not
provide any M2M transformation). The Ecore
metamodel proposed by this approach enables to
generate Java code using the native M2T EMF
mechanisms. The atomic functions are abstract, and
must be implemented manually. Only the coupling
functions are automatically generated.
MDD4MS (Cetinkaya et al. 2012) was
originally based on the GME environment, but it is
now fully implemented within EMF Ecore. The
authors follow a MDA-oriented approach to define
a DEVS metamodel. The atomic functions are
specified using a platform-independent pseudo-
code, which seems to be described itself by a
metamodel, linked to the DEVS metamodel. The
states are handled by state variables, they can be
typed and they also can be affected an initial value.
A total code generation is provided, towards Java
platforms (M2T). A M2M transformation from
BPMN to DEVS is also proposed.
MetaDEVS (Garredu et al. 2012) is a
metamodel for DEVS based on a MDA approach,
implemented within the Eclipse EMF framework. It
specifies the DEVS states using typed variables that
can be either enumerated or not. The functions are
defined in a platform-independent way, using the
DEVSRule concept based on Conditions and
Actions. Code generation mechanisms use Acceleo.
A generic approach form M2M transformations
have also been proposed in (Garredu et al. 2013)
and applied to a transformation between FSMs and
DEVS.
3.2.4 Comparison of the MDE Approaches
Those approaches can be compared following
several criteria. As far as we are concerned, the
most significant ones are the solutions chosen to
express states and functions. All of them favor
finite and enumerated states and some ones also use
state variables: (Song 2006), (Mittal et al. 2012),
(Touraille et al. 2010), (Lei et al. 2009), (Cetinkaya
et al. 2012), (Garredu et al. 2012). The advantage of
taking into account the state variables in addition to
enumerated states is that multidimensional states
can be specified. In our opinion, these solutions
appear to be complementary.
Modeling the behavioral functions is also a
criteria to evaluate the DEVS metamodels. Three
different approaches are used.
The first one is proposed by the approaches that
chose to represent the states in a finite way: (Posse
et al. 2003), (Song 2006), (Mittal et al. 2012), (Lei
et al. 2009). In this case, the transition functions are
enumerated and explicitly specify the state changes
that must occur. The time advance and output
functions are instantiated with each state: therefore,
a state contains its lifespan and the associated
output. The main advantage of the approach is that
it leads to a complete automatic code generation,
while its main drawback is that it is not able to
specify more complex logical behaviors.
The second one, chosen by (Song 2006),
(Sarjoughian et al. 2012) and (Risco-Martín et al.
2007) is depending on the platforms. It defines the
elements that must be specified, or completed, by
the programmer. Those elements depend on the
target platform. Hence, some approaches consider
the atomic functions are considered as abstract
methods, textual meta-attributes, or code blocks. To
provide a complete code generation, the same
function must be written in several languages, but
the logic they specify is not platform-independent
(Touraille et al. 2010).
Finally, the third one uses specifics metamodels
in order to allow creating behavioral rules that
define all the functions in a platform-independent
way. The goal is to perform automatic code
generations by limiting as much as possible the
intervention of the programmer. The two main
solutions aim to use a pseudo language (Cetinkaya
et al. 2012) or a behavioral logic within the
SIMULTECH2014-4thInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
184

Table 1: Comparison of the recent DEVS-MDE approaches.
Table 2: Transformations associated to some recent DEVS-MDE approaches.
associated DEVS metamodel (Garredu et al. 2012).
Table 1 summarizes our comparison.
Moreover, some of the approaches were used in
M2M and/or M2T contexts. Some approaches are
located in the OMG MDA technological space and
use M2M transformation languages as ATL: (Lei et
al. 2009), (Cetinkaya et al. 2012), (Garredu et al.
2012). Other approaches use different
transformation techniques (because of the
technological spaces they belong to), such as XML
for (Risco-Martín et al. 2007) or Python for
(Borland 2003). Some approaches do not perform,
at this time, M2M transformations.
The transformations associated to the
approaches we have presented are shown in Table
2. Some information about their M2M (DEVS
external interoperability) and M2T (DEVS internal
interoperability) aspects is provided.
4 CONCLUSION
In this article, an overview of some significant
MDE-oriented approaches linked to DEVS was
presented. Some of them were detailed and
compared. Using a MDE approach in a DEVS
context increases the lifetime of the models,
improves the way they are defined, makes them be
reusable, and enhances their interoperability with
other platforms and even other formalisms. Many
MDE-oriented tools have been developed, in
Approach  Metaformalism
Enum.
States
State
Vars
DEVSFunctions( δ
int
– δ
ext
– λ –ta) 
AToM
3
DEVSV1(Posseetal.2003)  ERDiagrams YES NO Finite 
AToM
3
DEVSV2(Song2006)  ERDiagrams YES YES Finiteorcodeblocks 
DEVSMLV2(Mittaletal.2012)  EBNF YES YES Finite(DEVSStateMachines) 
DEVSMLV1(Risco‐Martínetal.2007)  XMLSchema YES NO CodeBlocks 
SimStudio/DML(Tourailleetal.2010) XMLSchema YES YES PartiallyDescribed(Hybridcode)
DEVStoSMP2(Leietal.2009)  MOFEcore YES YES Finite 
MDD4MS(Cetinkayaetal.2012)  MOFEcore YES YES Platform‐IndependentPseudo‐Code
EMF‐DEVS(Sarjoughianetal.2012)  MOFEcore YES NO AbstractMethods 
MetaDEVS(Garreduetal.2012)  MOFEcore YES YES Platform‐IndependentMeta‐Classes

M2MTransformation M2TTransformation
Approach Nature
Transformation
Approach
DestinationPlatform
Transformation
Approach
AToM
3
DEVSV1(Posseetal.2003)
and(Borland2003)
SC→DEVS
Graph
(Python)
PyDEVS
ModelParsing
(Python)
DEVSMLV2(Mittaletal.2012) DSLs→DEVSML Xtend DEVSJAVA Xtend
DEVSMLV1(Risco‐Martínetal.2007) SC→DEVS‐SM XMLParser DEVS‐XML XMLParser
SimStudio/DML(Tourailleetal.2010) x x DML‐Lang&DML XMLParser
DEVStoSMP2(Leietal.2009) DEVS→SDML ATL x x
MDD4MS(Cetinkayaetal.2012) BPMN→DEVS ATL DEVSJAVA ModelParsing(Java)
EMF‐DEVS(Sarjoughianetal.2012) x x DEVS‐Suite ModelParsing(Java)
MetaDEVS(Garreduetal.2013) FSM→DEVS ATL PyDEVS&Other Acceleo(template)
ASurveyofModel-DrivenApproachesAppliedtoDEVS-AComparativeStudyofMetamodelsandTransformations
185

particular for the Eclipse EMF platform, as
additional plug-ins.
However, the metamodels that have been proposed
for DEVS face a difficult issue: the definition of the
states and the transition functions in a platform-
independent way. Doing so highly reduces the
power of expression of the metamodel. Some
research need to be done in order to increase the
power of expression of the metamodels, maybe with
a combination of graphical and textual notations.
An important criteria, which was not evaluated here,
is linked to the semantics of the metamodels: indeed,
a metamodel only specifies an abstract syntax and
needs semantics to be more accurate. MDAbased
meta-models usually resort to Object Constraint
Language (OCL) to express those semantics.
However, the power of a metamodel’s semantics
remain hard to evaluate.
REFERENCES
ACCELEO2013. ttp://www.eclipse.org/acceleo/.
Badros, G., 2000. “JavaML: A Markup Language for Java
Source Code.” Proceedings of the 9th International
World Wide Web Conference (Amsterdam,
Netherlands, May. 15-19),159- 7.
Bézivin J., « Sur les principes de base de l’ingénierie des
modèles », RSTI-L’Objet, 10(4):145–157, 2004.
Bézivin,J and Kurtev, I. : Model-based Technology
Integration with the Technical Space Concept. In
Metainformatics Symposium, Esbjerg, Denmark,
2005. Springer-Verlag.
Bolduc, J.S., Vangheluwe, H. A modelling and simulation
package for classical hierarchical DEVS. MSDL
technical report MSDL-TR-2001-01, McGill
University, June 2001
Borland, S., Transforming Statechart models to
DEVS,2003.
Cetinkaya D., Verbraeck A., and Seck M. D., Model
transformation from BPMN to DEVS in the
MDD4MS framework, Proceedings of the 2012
Symposium on Theory of Modeling and Simulation -
DEVS Integrative M&S Symposium, Orlando,
Floride, 2012
DEVSJAVA2013. http://www.acims.arizona.edu/
SOFTWARE/software.shtml#DEVSJAVA.
Garredu, S., Vittori, E., Santucci, J.-F., and Bisgambiglia,
P.-A., A Meta-Model for DEVS - Designed following
Model Driven Engineering Specifications,
Proceedings of the 2nd International Conference on
Simulation and Modeling Methodologies,
Technologies and Applications, Simultech 2012,
Rome, Italy, 28 - 31 July, 2012.
Garredu, S., Vittori, E., Santucci, J.-F., and Bisgambiglia,
P.-A., (In Press) From State-Transition Models to
DEVS Models - Improving DEVS external
interoperability using MetaDEVS: a MDE approach,
Proceedings of the 3rd International Conference on
Simulation and Modeling Methodologies,
Technologies and Applications, Reykjavik, Simultech
2013, Iceland, 29 - 31 July, 2013.
Harel D., Statecharts : A visual formalism for complex
systems, Science of Computer Programming,
8(3):231-274, 1987.
Jouault, F. and Kurtev, I. (2006) On the architectural
alignment of ATL and QVT. In: Proceedings of the
2006 ACM symposium on Applied computing, Dijon,
France.
Kleppe, A., Warmer, S., Bast, W., "MDA Explained. The
Model Driven Architecture: Practice and Promise",
Addison-Wesley, April 2003.
Lara J., Vangheluwe, H., “Using AToM as a Meta CASE
Tool”, 4th International Conference on Enterprise
Information Systems, Universidad de Castilla-La
Mancha, Ciudad Real (Spain), 3-6, April 2002.
Lei, Y., Wang, W., Li, Q., and Zhu, Y., A transformation
model from DEVS to SMP2 based on MDA,
Simulation Modelling Practice and Theory, Vol. 17,
Nr. 10 (2009) , p. 1690-1709.
Levytskyy, A., Kerckhoffs, E. J., Posse, E. and
Vangheluwe, H., “Creating DEVS components with
the meta-modelling tool AToM
3
” in 15
th
European
Simulation Symposium (ESS), A. Verbraeck and V.
Hlupic, Eds. Society for Modeling and Simulation
International (SCS), October 2003, pp. 97 – 103, delft,
The Netherlands.
Mens, T., Czarnecki, K., and Van Gorp, P., A Taxonomy
of Model Transformations, Electronic Notes in
Theoretical Computer Science (ENTCS) Volume 152,
March, 2006, pp. 125-142.
Mittal, S., Martín. J. L. R., Zeigler, B.P., « DEVSML:
automating DEVS execution over SOA towards
transparent simulators », Proceedings of the 2007
ACM Spring Simulation Multiconference, March 25-
29, 2007, Norfolk, VA, USA, Vol. 2, pp. 287-295.
Mittal, S., Douglass, S.A., DEVSML 2.0: The Language
and the Stack, DEVS Symposium, Spring Simulation
Multiconference 2012, Orlando.
Mooney, J. and Sarjoughian, H.S., A Framework for
executable UML models, In High Performance
Computing & Simulation Symposium, Spring Simulation
Conference, pages 1-8, 2009.
Nikolaidou, M., Dalakas, V., Kapos, G.-D., Mitsi, L. and
Anagnostopoulos, D., « A UML 2.0 profile for DEVS:
Providing code generation capabilities for simulation
» in Proceedings of 16th International.
Conference on Software Engineering and Data
Engineering (SEDE-2007), Las Vegas, USA, July
2007 (Invited paper).
Nikolaidou, M., Dalakas, V., Mitsi, L., Kapos, G.-D.,
Anagnostopoulos, D. « A SysML Profile for Classical
DEVS Simulators » (Conference Paper) Proceedings
of the 2008 The Third International Conference on
Software Engineering Advances, 978-0-7695-3372-8,
Pp 445-450, 2008, 10.1109/ICSEA.2008.24, IEEE
Computer Society.
SIMULTECH2014-4thInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
186

OMG 2001. Model Driven Architecture homepage -
http://www.omg.org/mda/.
OMG UML 2011. Unified Modeling Language
specifications-http://www.omg.org/spec/UML/2.4.1
OMG SysML 2012. Systems Modeling Language -
http://www.omg.org/spec/SysML/1.3/.
Posse, E., Bolduc, J.S., Vangheluwe, H., Generation of
DEVS Modelling & Simulation Environments. In
Proceedings of the 2003 Summer Computer
Simulation Conference SCSC 2003.
Risco-Martin, J.L., Mittal, S., Zeigler, B.P., Cruz, J.L,
"From UML Statecharts to DEVS State Machines
using XML", Multi-paradigm Modeling, IEEE/ACM
International Conference on Model Driven
Engineering Languages and Systems , 2007.
Rumbaugh, J., Jacobson, I. and Booch, G., “The unified
modeling language reference manual”, The Addison-
Wesley object technology series, Addison-Wesley,
2005.
Sarjoughian H.S. and Markid, A.M., 2012. EMF-DEVS
modeling. In Proceedings of the 2012 Symposium on
Theory of Modeling and Simulation - DEVS
Integrative M&S Symposium (TMS/DEVS '12).
Society for Computer Simulation International, San
Diego, CA, USA.
Schulz, S., T. C. Ewing, and J. W. Rozenblit. 2000.
Discrete event system specification (DEVS) and
statemate statecharts equivalence for embedded
systems modeling, Proceedings of the 7th IEEE
International Conference and Workshop on the
Engineering of Computer Based Systems, Edinburgh,
Scotland, 2000.
Seo, C. "Interoperability between DEVS Simulators using
Service Oriented Architecture and DEVS
Namespace", Ph.D. dissertation, Electrical and
Computer Engineering Dept., University of Arizona,
Spring 2009.
SISO 2008. Simulation Interoperability Standards
Organisation, SISO-REF-019-2008: Discrete-Event
Systems Specification (DEVS).
http://www.sisostds.org/ProductsPublications/ReferenceD
ocuments.aspx.
Song, H., Infrastructure for DEVS Modelling and
Experimentation. Master's thesis. McGill University.
School of Computer Science. (2006)
Steinberg, F.D., Budinsky, F., Paternostro, M., and Merks,
E. Eclipse Modeling Framework 2
nd
Edition, Addison
Wesley, 2009.
Touraille, L., Traoré, M.K., Hill, D., "On the
interoperability of DEVS components : On-Line vs.
OffLine Strategies.", 2009, UMR CNRS 6158,
LIMOS/RR-09-04, 13 p.
Touraille, L., Traoré, M.K., and Hill, D. R. C., 2011. A
model-driven software environment for modeling,
simulation and analysis of complex systems. In
Proceedings of the 2011 Symposium on Theory of
Modeling & Simulation: DEVS Integrative M&S
Symposium (TMS-DEVS '11). Society for Computer
Simulation International, San Diego, CA, USA, 229-
237.
Zinoviev, D., "Mapping DEVS Models onto UML
Models," Proc. of the 2005 DEVS Integrative M&S
Symposium, San Diego, CA, April 2005, pp. 101-106.
Zeigler, B.P., Theory of Modeling and Simulation, New-
York: Wiley-Interscience, 1976.
ASurveyofModel-DrivenApproachesAppliedtoDEVS-AComparativeStudyofMetamodelsandTransformations
187