TELKOM
NIKA
, Vol.11, No
.1, Janua
ry 2013, pp. 103
~11
4
ISSN: 2302-4
046
103
Re
cei
v
ed Se
ptem
ber 28, 2012; Revi
se
d No
vem
ber
22, 2012; Accepted Novem
ber 29, 20
12
Hardwa
re-in-the-loop and Parallel Simulation
Architecture for Wireless Sensor Network
Shihong Du
an*, Yadong
Wan, Peng
Meng, Qin Wang
Schoo
l of Com
puter & Comm
unic
a
tion E
ngi
neer
ing
No 30
Xu
e
y
u
a
n
Roa
d
, Hai
d
ia
n District, Beiji
ng, 137
01
32
31
76/
*corres
pon
di
ng
author, e-mai
l
: duans
h@ust
b
.edu.cn
A
b
st
r
a
ct
Discrete
event
-base
d
si
mu
lat
i
on
is co
mmo
n
ly us
ed to
e
v
alu
a
te res
ear
ch on
W
i
rel
e
s
s
Senso
r
Netw
orks (W
SNs). How
e
ver, highly acc
u
ra
te simu
latio
n
mo
de
ls are re
quir
ed in rec
e
nt advanc
es o
n
w
i
reless co
mmu
n
ic
ation
tec
hno
logy, w
h
ic
h resu
lts i
n
a
steep
incr
ea
se in
si
mu
lati
on c
o
mpl
e
xity
an
d
runti
m
e. T
h
e c
ontrib
u
tions
of
this p
a
p
e
r for t
he
are tw
ofol
d, on
e is
to
pres
ent a
g
ener
al
l
a
yer structur
e f
o
r
hardw
are-
in-th
e
-lo
op e
m
ulati
on a
nd W
S
N s
i
mulati
on
e
m
b
edd
ed w
i
th i
m
ple
m
entatio
n o
f
mod
e
ls, suc
h
as
ener
gy mod
e
l
and l
i
nk
mo
de,
to introd
uce th
e distrib
u
t
ed
n
odes i
n
to the s
i
mulati
on fra
m
ew
ork; the oth
e
r is
to buil
d
a p
a
ra
llel
i
z
e
d s
i
mul
a
tion b
a
se
d on
mu
lti-pr
oc
esso
r computers a
s
the
de-facto defau
lt
hardw
a
r
e
platfor
m
a
nd
p
o
w
e
rful priv
ate
computi
ng cl
u
s
ters to mi
rr
or
the rea
l
W
S
N
mor
e
cl
osely.
T
he w
o
rk in t
h
is
pap
er is rea
l
i
z
e
d
an
d use
d
to simu
late i
ndustri
a
l W
S
N
s
for describi
ng an
d verifyi
ng the d
e
tail
an
d
meth
od
olo
g
y o
f
W
S
Ns.
Key
w
ords
: E
v
ent-ba
s
ed
si
m
u
lation, hardwa
r
e
-
in-th
e
-l
oop em
ulatio
n, paralleli
zati
on, energy
m
odel, link model
Copyrig
h
t
©
2013
Univer
sitas Ahmad
Dahlan. All rights res
e
rv
ed.
1. Introduc
tion
The mo
del
s in sim
u
latio
n
of wi
rele
ss sen
s
o
r
net
works b
e
co
me mo
re
co
mplex to
accurately describ
e wirel
e
ss cha
nnel
cha
r
a
c
teri
stics, node
s’ op
eration a
nd the pro
pertie
s
of
transmissio
n
techn
o
logie
s
,
whi
c
h typi
cally lead
to
a high
comp
utational lo
a
d
an
d exten
s
ive
simulation runtimes. Also, the pr
actical wirel
e
ss sensor
net
work
of
large
scalbilit
y includes m
o
re
than 100
0 no
des, a
s
obje
c
ts in simulatio
n
, to consum
e con
s
id
era
b
l
e
memory an
d CPU time. So
redu
cin
g
ru
ntimes with b
u
il
ding inten
s
iv
e node mo
del
and enviorm
ent model in
simulatio
n
is the
key issue of this pa
per.
Curre
n
tly, wireless
sen
s
o
r
netwo
rk
sim
u
lation tool
s a
r
e divide
d int
o
proto
c
ol
si
mulator
and p
r
og
ram
cod
e
sim
u
lat
o
r. The fo
rme
r
is rep
r
e
s
ent
ed by OM
Net
++, OP
NET a
nd NS
-2, whi
c
h
norm
a
lly defines the p
r
ot
ocol
stack in
a node
an
d
emulates th
e data tran
smissi
on in th
e
netwo
rk. T
h
e
latter is
rep
r
ese
n
ted by T
O
SSIM,
which de
scribe
s
operation
s
of
singl
e nod
e
in
high-preci
s
io
n to accurate
ly imitate applicatio
n beh
aviors of no
d
e
s. The featu
r
es of fre
que
ntly
use
d
si
mulati
on tool
s a
r
e l
i
sted in t
able
1. Co
m
p
a
r
at
ively spea
kin
g
, OMNet++, whi
c
h i
s
o
p
e
n
-
sou
r
ce suit
s
for
larger-scale
net
wo
rk simulation, sup
port
s
sim
u
lating network p
r
oto
c
ol
and
debu
gging
code. So
sim
u
lation tool
s
in this p
ape
r are
develo
p
ed ba
se
d on
OMNet++ a
nd
extended
with model
s and
layed archite
c
ture of n
ode
and net
work.
Table 1. The
Comp
aration
of Simulation Tools
T
ools Accur
a
cy
of
net
w
o
rk
Accur
a
cy
of
node
Simulation
scale
Channel
model
Energ
y
model
Obj
e
c
t
-
oriented
Application
OMNet+
+
[1
]
High
Lo
w
Larger
Simple
Simple
Y
e
s
Protocol and
debugging
OP
Net
[2
]
High
Lo
w
Larger
Simple
Simple
Y
e
s
Protocol
NS2
[3]
High
Lo
w
Large
Simple
Functional
Y
e
s
Protocol
SENSE
[4]
High
Lo
w
Large
Non
Functional
Y
e
s
Protocol
TO
SSIM
[5
]
Lo
w
High
Large
Non
Functional
No
Debugging
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 1, Janua
ry 2013 : 103 – 1
1
4
104
We mak
e
the following four c
ont
ributions
in this
paper:
1) We p
r
o
p
o
s
e a set of gene
ric
simu
lation ar
chite
c
ture to defi
ne nod
es an
d WSN. Lay
ed
stru
cture i
s
reali
z
ed
wi
th indep
end
ent
compo
n
ents
coo
p
e
r
ated with
e
a
ch
othe
r b
y
messagi
ng a
nd sh
arin
g da
ta in bulletin board.
2) Lin
k
relia
bility model from indu
stri
a
l
env
iorment
and energy model are i
n
tegre
a
ted in
to
s
i
mulation framework
to verify the effic
i
ency of WSN t
e
ch
nolo
g
ies
more
cre
d
ible
ly.
3) A mappin
g
sch
eme is
a
pplied to buil
d
up hardwar
e-in
-the-l
oop
emulation, in
whi
c
h re
al no
des
sen
d
op
erati
ng pa
ram
e
ters to
simulatio
n
platform
via data
colle
ction network
and
wo
rk
a
s
virtual com
p
o
nents in
simul
a
tion.
4) A parallel
architectu
re
deployed in
multi-co
re comp
uter with
cent
rali
zed
sch
eduli
ng is
prop
osed to simulate larger WSNs with l
a
yer topolo
g
y and more than 100
0 nod
e
s
.
Also, we u
s
e
the said
platform to
si
mula
te a typical in
dustri
a
l a
ppli
c
ation
compli
ant with
WIA-PA standard.
Our ev
aluation sh
ows the
correctness
and availability
of si
mulation platform,
and
p
a
rallel simulatio
n
a
nd
ha
rd
wa
re
-in-th
e-lo
op emualtion
im
prove perfo
rmance with less
runni
ng time
and large
r
scale WSNs ov
er traditio
nal
scheme.
The rem
a
ind
e
r of this pap
er is st
ru
ctured as follo
ws. Section 2 gives a introd
u
c
tion to
our
simul
a
tio
n
fram
e work a
s
solutio
n
of less
ca
p
a
city and
mo
re
runni
ng ti
me of tra
d
itional
simulatio
n
te
chni
que
s
and
detaile
d
desi
gn a
n
d
re
alization of
archit
ecture, mo
del
s, ha
rd
wa
re
-i
n-
the-loo
p
emu
a
ltion sy
stem
and pa
rall
el simulati
o
n
sy
stem. Sectio
n 3 analy
z
e
s
synchro
n
ization
scheme
in te
rms
of the
gran
ula
r
ity of the pa
ralleli
sm a
nd
simu
lation time, verify the mo
del
accuracy a
n
d
evaluate
s
the e
f
f
e
ct
iv
ene
ss
by
a
n
sim
u
lat
i
on
instan
ce of
indu
strial WSN
appli
c
ation. F
i
nally, we co
n
c
lud
e
in Secti
on 4.
2. Rese
arch
Metho
d
This sectio
n introdu
ce
s th
e fundame
n
tals of our
si
mulation fra
m
ewo
r
k PIWSNSim and
its unde
rlyin
g
model re
alizatio
n, function co
mpo
nent mappi
n
g
tactics an
d parall
e
lization
scheme. PIWSNSim en
able
s
a par
a
llel execution
of network simulatio
n
model
s to red
u
ce
runni
ng time
by means of
three p
r
op
erti
es: 1) It
intro
duces a laye
d stru
cture to easily de
scri
be
node, net
work and e
n
viorment. 2) It de
fi
ne
s a mappin
g
sche
me from re
a
l
node to vitual
comp
one
nt in
simulatio
n
to
reali
z
e a h
a
rdwa
re
-in-th
e-l
oop em
ualtio
n
, even co
de
debu
gging. 3
)
It
employs a p
a
ralleli
zatio
n
scheme to e
x
ecute
ind
e
p
ende
nt simul
a
tion events
simultan
eou
sl
y,
control by ce
ntralized sch
edule
r
. PIWSNSim enabl
e
s
a credibl
e simulation by building u
p
more
accurate link
realibility mod
e
ls an
d ene
rg
y models of n
ode an
d network.
2.1. Time-v
ar
y
i
ng Link Reliabilit
y
Mod
e
l in PIWSNSim
Many in
dust
r
i
a
l WS
Ns d
e
p
loyment i
s
unde
rway no
wad
a
ys to
al
low th
e e
ngi
neers to
reliably
acqui
re a
nd
co
ntro
l the r
eal
-time
data
of WS
Ns in
the fa
ct
ory at any time
anywh
ere.
Th
e
noise and i
n
terferen
ce
s are
signifi
cant due to
the wid
e
op
erating te
mp
eratu
r
e
s
, strong
vibration
s
, airborn
e
conta
m
inants, ex
cessive ele
c
tromagn
etic n
o
i
se
cau
s
e
d
b
y
large m
o
tors etc.
Therefore, it
is im
porta
nt for a
si
mulati
on fra
m
ewo
r
k to
playba
ck
ch
ara
c
teri
stics of
comm
uni
cati
on ch
ann
el in orde
r to verify the te
chn
o
logy and o
p
t
imize the de
sign of ind
u
st
rial
WSNs. IEEE
802.15.4 based WSNs are designed to
support com
m
unication over short ranges
with low data rate and redu
ced
energy consumpt
ion. So
the IEEE 802.15.
4 standard
has
become a
re
cog
n
ized ind
u
stry sta
nda
rd and
well a
c
cepted by i
ndu
strial u
s
e
r
s. [6] mea
s
u
r
ed
channel characteri
stics
of sixt
een IEEE802.15.4
radi
o ch
annel
s in typical i
ndustrial environm
ent.
It is found tha
t
chann
el reli
ability is possibly and dr
a
m
atically different with ea
ch
other an
d no
ne
of the
channels can
al
ways pr
ovide good
reli
ability required
by i
ndustry appli
c
ation.
Furthe
rmo
r
e,
reliability of each
cha
nnel
varies
with time and spa
c
e.
Defini
tion 1
. PDS, Packet Drop Seq
u
ence, is the log of the succe
ss o
r
failure in one
time-sl
o
t com
m
unication
b
a
se
d on
ACK
re
ceived
or
not, sh
own a
s
Eq. 1,
wh
erein i d
enote
s
the
ith transmi
ssion. PDI, Packet dro
p
interval, is the time gap betwee
n
two adja
c
e
n
t lost packet
s
to
define the PDS distributio
n.
(1)
0
()
1
recei
ved
A
C
K
PDS
i
ot
h
e
r
s
Evaluation Warning : The document was created with Spire.PDF for Python.
TEL
K
H
a
A
CK
j+N-
1
and
N
perio
se
ct
i
o
PDI i
Figu
r
(
a
sho
w
large
trans
K
OM
NIKA
a
rdware-
i
n-
t
h
Defini
tio
identified b
y
1
, sh
own
as
N
umACK
is
r
From te
s
ds of (ti,t
j
]
w
o
n R1
~R25
w
s defin
e
d
in
r
e 1(a
)
. Also
a
) u a
nd
σ
v
a
PIWSNS
w
n
with a
n
i
n
ly implemen
t
MiXim fr
a
mission. Th
e
N
PD
R
h
e-lo
op an
d
P
n 2.
PDR,
P
y
the send
er
Eq. 2, wher
e
r
e
c
eived A
C
K
s
t-be
d mea
s
w
ith diffe
re
nt
w
ith val
u
e [1
Eq. 3, whe
r
the PDI di
st
r
a
lue from m
e
Figure
1
im as a sim
u
n
dep
ende
nt
s
t
ed in “pu
r
e”
Figure 2
a
mework pr
o
e
Analogu
e
M
N
um
Packet
NumP
a
c
I
S
P
arallel Sim
u
P
acket Drop
in the log wi
n
e
in Number
p
K
numb
e
rs.
s
urem
ent,
P
p
r
op
er
ty p
a
-4%], [5-8%
]
r
e u
an
d
σ
h
a
r
ibutio
n cu
rv
e
e
asu
r
em
en
t
1
. Time-vary
i
(
e
xpress
e
d
u
lation fram
e
s
truc
ture na
OMNe
T
+
+
;
. Time-varyi
n
o
v
i
de
s ch
an
n
M
odel
s are r
e
1
Num
A
CK
c
ket
S
SN: 2302-4
0
u
lation Archi
t
Ratio is
th
e
n
do
w
o
f
T
in
p
acket is the
P
DI follows
ram
e
ter
s
b
a
]
, … , [97-1
0
a
ve differen
t
e
is shown
a
(b) P
D
i
n
g
mo
del in
d
in
PDR, P
D
e
work
for
W
S
med a
s
EIF
but also
su
c
n
g ink
mode
n
el mo
del to
e
spon
sibl
e f
o
1
00%
0
46
t
ecture for
W
e
indicato
r o
f
clu
d
ing N tr
a
total sendin
g
sa
me d
i
st
ri
b
sed
on P
D
R
0
0%], and p
r
t
value in di
f
a
s Figu
re 1(
b
D
I probabilit
y
indu
stri
al a
p
D
I, and PDS)
S
Ns buil
d
s t
h
s
h
ow
n
as
F
c
cessfully in
c
l of EIF in PI
filter and p
r
o
r sim
u
latin
g
W
i
r
el
ess Sen
s
f
los
t
packet
s
a
nsmi
ssion
n
g
numbe
r
o
f
b
utio
n of
lo
g
R
value.
PD
R
obability de
n
f
ferent PDR
b
).
y
density
dis
t
p
pli
c
ation
h
e sai
d
time-
v
F
igure 2. EI
F
c
orpo
rate
d in
WSNSim
r
oce
s
s si
gn
a
g
the
attenu
a
s
or …
(Shih
o
s
numb
ers
w
n
umbe
red fr
o
f
WSN
in T
p
g
normal i
n
R
i
s
divided i
n
n
sit
y
dist
rib
u
t
s
e
ct
ion
s
sh
o
t
ribution in
R
v
arying link
F
can be n
o
to the MiXi
M
a
l in phy
si
ca
l
a
tio
n
of a re
c
o
ng D)
105
w
ithout
o
m j to
p
erio
d,
(2)
som
e
n
to 25
t
io
n of
o
wn in
(3)
R
1-R7
m
od
el
o
t only
M
[7]
.
l
layer
c
eiv
ed
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 1, Janua
ry 2013 : 103 – 1
1
4
106
sign
al, su
ch
as
shad
owi
n
g, fading an
d
path loss
. T
he De
cid
e
rs
are respon
si
ble for
classif
y
in
g
the sign
al as noise o
r
si
g
nal, and calculating bi
t error of the re
ceived messa
ges. Reliabilit
y of
cha
nnel
in
80
2.15.4 b
a
sed
indu
strial
WSNs varie
s
wi
th time d
ue t
o
all
kin
d
s of
factors,
su
ch
as
multi-path
int
e
rferen
ce,
wi
rele
ss
radio
i
n
terferen
ce a
nd so on.
Ti
me-varyin
g
chara
c
te
risti
c
o
f
cha
nnel
ca
n
be exp
r
e
s
sed a
s
o
scill
a
t
ions in
sh
ort perio
d an
d
mutation
when
some
e
v
en
t
happ
en
s in i
ndu
strial field
s
. By fit test usin
g dfittool
in Matlab, o
s
cillation i
n
small win
d
o
w
and
interval b
e
tween t
w
o
eve
n
ts all
follo
w the lo
g-lo
gi
stic dist
ributi
on with
diffe
rent paramet
ers;
Proba
bility density dist
rib
u
tion of affected cha
nnel
numbe
rs and
affected con
t
inuou
s freq
u
ency
cha
nnel nu
m
ber a
r
e all ke
ep to the log norm
a
l distri
b
u
tion with different pa
ram
e
ter values. EIF is
respon
sibl
e for sim
u
lating
the sign
al loss in ty
pical m
e
tallurgy p
r
o
c
ess indu
strial
appli
c
ation
s
.
EIF provide
s
a se
parate si
mple mo
dule
nam
ed
as In
dustri
a
lChan
nelControl to
descri
b
e
the loss rate
of 16 freq
ue
ncy chan
nels. Indus
trial
C
h
annel
Cont
rol
posse
sse
s
th
e scala
b
le a
nd
extensibl
e
mech
ani
sm.
1)
The
op
erat
ion p
a
ramete
rs can
be
con
f
igur
ed. The
para
m
eters a
r
e provid
ed in
the xml
file, and the
xml file nam
e is spe
c
ified
in the
sim
u
l
a
tion
configu
r
ation. So
u
s
ers a
r
e
easy
to
adju
s
t para
m
eters a
c
co
rdi
ng to cha
nnel
characte
ri
stic in diverse p
r
acti
cal ap
plications.
2) T
h
e
mod
e
l can
be
re
placeable.
Currently,
test
from
thre
e
typical in
du
stry field
s
introdu
ce
s th
e idust
r
ialCha
nnel
Control which
will be e
x
pande
d, imp
r
oved o
r
sub
s
tituted for mo
re
indu
stry environm
ent. Industrial
C
h
ann
elCo
ntrol
in
essen
c
e i
s
a
c++cl
a
ss; other mo
del
s ca
n
derive from it.
Industrial lin
k model n
a
m
e
is han
ded o
v
er
to the ph
ysical laye
r, so it is easy
for
physi
cal layer can make decisi
on which link model it calls.
2.2. La
y
e
d A
r
chitec
tur
e
o
f
PIWSNSim
PIWSNSim is desig
ned to
descri
be el
e
m
ents
in
WS
N, inclu
d
ing
n
ode
s, enviorment, and
manag
eme
n
t units. Layed
simulation a
r
chite
c
tu
re is employed to
simulate
WSN in detail
and
reality by cle
a
rly definin
g
element
s in
WSN
with
mo
dular mode. Shown as Fig
u
re
3,
PIWS
NSim
based o
n
O
MMet++ platf
o
rm e
m
pha
si
ze
s the im
po
rtance of e
n
vironm
ental m
odelin
g and
node
modelin
g. Fo
ur laye
r
stru
cture
com
b
ine
s
n
ode
s i
n
re
sou
r
ce laye
r
with
wirel
e
ss
medium
defin
ed
in media lay
e
r un
der th
e
regul
atory u
n
its in man
a
ger laye
r to
simulate th
e
WSN a
ppli
c
a
t
io
n
meticulo
usly,
and the n
e
twork o
p
e
r
ation and
pe
rf
o
r
man
c
e
anal
ysis a
r
e exhi
bited to user by
friendly G
U
I l
a
yer. PIWSNSim use
s
discrete
event h
andlin
g me
ch
anism
provid
ed by OM
Net
++,
manag
es eve
n
t queue p
u
shed by com
p
onent
s of WS
N and di
strib
u
tes event
s to CPU
kernel
.
PIWSNSim i
n
this p
ape
r
is re
alized a
s
a comp
osit
e
model defi
ned
in OMNet++ and
calle
d by Sim kernel. Th
e comp
one
nts in lay
ed st
ructu
r
e a
r
e i
ndep
ent and
can be e
a
sily
repla
c
e
d
an
d
updated to
help researchers e
s
tabli
s
h WSN
simul
a
tion for
spe
c
ific a
pplications,
and a
nalyze
techn
o
logy a
nd alg
o
rithm
for their con
c
ern. Th
e lay
ed a
r
chitectu
re em
ployed
in
PIWSNSim a
r
e na
rrate
d a
s
follows.
1)
Re
sou
r
ce l
a
yer in
clu
d
e
s
all n
ode
s i
n
t
he n
e
two
r
k.
Nod
e
s are d
e
f
ined a
s
hiera
r
chi
c
al
st
ru
ct
ure
to supp
ort de
scription of isom
orphi
sm a
nd heteroge
n
eou
s nod
es.
2) Me
dia lay
e
r in
clu
des
modelin
g wi
reless tr
an
smi
ssi
on cha
nne
l
and sam
p
lin
g
ch
aracte
rist
ics.
Link reli
ability model
de
scri
bed i
n
se
ctio
n 2.1 i
s
reali
z
ed a
s
wirele
ss lin
k m
odel
i
n
medi
a laye
r
to
accu
rately analyze WSN
tech
nology.S
ensi
ng
m
odel
with
sam
p
lin
g envio
rment
model
are
use
d
to simul
a
te sen
s
in
g p
r
ocess an
d chara
c
te
risti
c
s of system on
test.
3) Ma
nage
m
ent layer ta
ke
s re
sp
on
sibilit
y of setting th
e network pro
pertie
s
an
d b
u
ilding
WSN to
sup
port self-formation an
d
self-maint
enan
ce
of
WSN, in
clu
d
i
ng ma
nag
e
m
ent unit
s
of
topology, syn
c
hroni
zation,
con
n
e
c
ti
on, secu
rity sche
me, informati
on logg
er to
make it ea
sie
r
to do re
sea
r
ch on relate
d tech
nolo
g
ies
of improving
WSN p
e
rfo
r
m
ance.
4)
GUI l
a
yer provides u
s
ers
with frie
n
d
ly gr
aphi
cal
interfa
c
e
an
d repla
c
e
or
extend
wind
o
w
s
interface p
r
o
v
ided by O
M
Net++.
Net
w
ork
co
nfigu
r
ation, o
n
line
or
offline a
nalysi
s
an
d
forecastin
g o
f
simulation
data are re
al
ized in
GUI l
a
yer. GUI a
p
p
licatio
n ca
n
be not only
calle
d by sim
kernel, but a
l
so can run a
l
one to
re
ad/
write
configu
r
ation file and
sele
ct data
from datab
ase to give performa
n
ce enh
ancement rep
o
rt.
Nod
e
s in PIWSNSim plat
form sh
ould
have sa
m
e
functio
n
with real no
de to sampl
e
,
comp
ute and
commu
nicati
on. So simul
a
tion node
s
ar
e supp
osed
to have a clear fram
ework to
sup
port chan
ge or control of the network topolo
g
y and desgin algo
rithm. A stack structu
r
e fo
r a
node i
s
pro
p
o
se
d in this
pape
r sh
own
in Figure 4
(
a)
. Nod
e
wit
h
stack st
ructure co
nsi
s
ts of
a
numbe
r
of dif
f
erent fu
nctio
n
sta
c
ks that
are
de
ri
ved from the same template s
t
ac
k
.
Func
tions of
comm
uni
cati
ng, sam
p
ling,
energy man
aging a
nd p
o
sitioni
ng are
respe
c
tively implemente
d
in
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Hardware-in
-
t
he-lo
op an
d Parallel Sim
u
lation Architecture for Wi
rel
e
ss Sensor
… (Shiho
ng
D)
107
spe
c
ific
stack. The highe
st
level of each stack i
s
a share
d
ap
plica
t
ion layer. Stack structu
r
e
in
this pa
pe
r b
r
i
ngs th
e b
ene
fits of the n
e
twork
pr
oto
c
ol
sta
ck i
n
to d
e
scribi
ng
rela
ted key ta
sks,
su
ch a
s
co
de
reuse, stand
ardi
zation e
n
han
cment an
d interchan
ge
able protocol desi
gn.
Figure 3. The Layed archi
t
ecture
of PIWSNSim
(a) Ba
se tem
p
late sta
ck
(b) Thre
e-tier-a
rchitecture of virtual node
Figure 4. Virtual nod
e defi
ned in PIWS
NSim
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 1, Janua
ry 2013 : 103 – 1
1
4
108
The virtual n
ode in PIWSNSim is a co
mposit
e
com
pone
nt class
written in c++ code to
s
i
mu
la
te
w
o
rk
pr
oc
es
s
o
f
r
e
a
l
s
e
ns
or
no
d
e
s
w
i
th th
e
embed
ded
software. Al
so
the virtual n
ode
is a
compl
e
x module i
n
OMNet
+
+ environment. In
order to facilitate the im
plementation of
different
sim
u
lation
sen
a
ri
o and
switch
betwe
en
different fu
nctio
n
s, the vi
rtua
l node
ad
opt
s a
three
-
tier-architecture a
r
chitecture pa
rt
itioned
by h
a
rd
wa
re reso
uce
s
a
nd
so
ftware fun
c
ti
on,
sho
w
n in Fig
u
re 4
(
b), in
clu
d
ing layer of functio
n
, man
ageme
n
t and
resou
r
ce.
Re
sou
r
ce lay
e
r i
s
respon
si
ble for esta
bli
s
hin
g
mo
del
s of nod
e ha
rd
ware resou
r
ces a
nd
descri
b
ing th
e state ch
an
ge of hard
w
are res
ource
s and en
erg
y
consumptio
n. The com
m
on
hard
w
a
r
e
co
mpone
nts include
a p
o
wer su
pply m
o
d
u
le
whi
c
h i
s
norm
a
lly a
b
a
ttery mod
u
l
e
, a
stora
ge u
n
it, a CPU, seve
ral sen
s
ing in
terface
s
a
nd
several RF
module
s
. Th
e battery mo
dule
provide
s
dischargi
ng m
o
d
e
l of
some
ba
ttery and
re
cord
s the
cha
nge
of the
wo
rkin
g
state of
the
battery by p
o
we
ring to
o
t
her resource
membe
r
s.
RF mo
dule i
m
pleme
n
ts
si
gnal tra
n
scei
ver
function. T
r
a
n
smi
ssi
on
sig
nal attenu
ate
s
in a
c
co
rdan
ce
with the ti
me-varyin
g
m
odel d
e
fined
by
radio
ch
anne
l and be
sen
t
to the RF module
of
target node via
a wirel
e
ss chann
el. Sensing
interface
sen
d
sa
mple
m
e
ssag
es to
sensi
ng p
h
ysi
c
al
cha
nnel.
It simulate
s the colle
ction
of
physi
cal information ba
se
d on the model defined
b
y
sensin
g ph
ysical
chan
n
e
l and gen
erates
the cha
r
a
c
teri
stical
sampl
e
data.
Manag
eme
n
t layer take
s re
spo
n
si
bil
i
ty to coord
i
nate the work
of the variou
s
comp
one
nts i
n
a virtual no
de for the sp
ecified a
ppli
c
ation tasks.
Manag
eme
n
t module
s
re
al
ized
in PIWSNSi
m
incl
ude
s
energy man
ageme
n
t, mobile ma
nag
ement, data
mana
geme
n
t an
d
colla
boration
mana
geme
n
t. Collab
o
ra
tion mana
ge
ment provide
s
a b
u
lletin
board to
whi
c
h
variou
s co
mp
onent
s in no
de co
uld re
gi
ster me
ss
ag
e type respo
nded by theirsleve
s. Speci
f
ic
messag
e will
be issu
ed b
y
source co
mpone
nt to bulltin board, and then be
transfe
re
d to the
comp
one
nts that sub
s
cribe the
me
ssag
e, fi
nally
the
comp
o
nents exe
c
ut
e the
re
spo
n
se
operation. Bu
lletin board m
odule p
r
ovide
s
a unified in
t
e
rface for different tasks to
use re
so
urce
s.
Energy ma
na
gement u
n
it monitor
and a
nalyze e
n
e
r
g
y
to suppo
rt the sim
u
lation
of energy-a
ware
strategy. M
o
bility manag
e
m
ent defin
es a large
num
ber
of mobil
e
model, virutal nod
e movi
ng
according to
spe
c
ified m
o
bility model regula
r
ly
and
notify conne
ction mana
ge
ment modul
e
to re-
establi
s
h the
netwo
rk
co
nn
ectivity and update the
top
o
logy ba
sed
on the cove
rage attrib
ute of
the node ra
dio frequ
en
cy module. Data manag
e
m
ent establi
s
he
s data st
orag
e and f
u
sion
model, provid
ing mana
gem
ent strate
gi
es for related a
pplication
s
.
Nod
e
fun
c
tio
n
is
re
spo
n
si
ble to d
e
fine
node
ope
ratio
n
and
sim
u
lat
e
wo
rkflow
of physi
cal
node
s. The
task
of vitual node
co
nsi
s
ts of tr
an
smissi
on, sen
s
e a
nd
com
putation. Wh
en
transmitting data, the applicatio
n layer will send the data to r
outing layer,
MAC layer and
physi
cal laye
r, and finally n
o
tify the RF
module to
co
mplete the
si
gnal tra
n
smission. Sign
al from
radio
ch
ann
el
will be
re
cei
v
ed by the RF modul
e of the no
de
s in t
he covera
ge
and u
p
strea
m
ed
to the a
ppli
c
ation laye
r
seque
ntially. Whe
n
sam
p
ling d
a
ta, ap
pl
ication
layer
will n
o
tify se
nso
r
driver, then t
o
sen
s
in
g int
e
rface to a
c
quire
phy
si
ca
l information.
At last, sa
mpling d
a
ta are
colle
cted a
n
d
issued to
the appli
c
ati
on layer.
Be
cau
s
e
com
p
utation task
is to do vari
ous
operation
s
of data, so co
mputation mo
del is de
sc
rib
ed as
CPU
state tran
sition
diagram u
s
ed
in
data mana
ge
ment and en
e
r
gy model.
The
ba
se
cl
as
s of
t
e
mp
lat
e
st
a
c
k i
s
st
r
a
tified m
u
lti function layer. Each layer,
indep
ende
nt from oth
e
r lay
e
rs,
contai
ns
enou
gh ma
ni
pulation ta
sks and coo
r
dina
tes thro
ugh th
e
approp
riate in
terface. L
e
vels of the stack are define
d
a
s
follows:
1) Inte
rfa
c
e
Layer is resp
onsi
b
le fo
r
communi
catio
n
with
the
ap
prop
riate
ha
rdwa
re
an
d m
a
ke
compl
e
xity of implementin
g the hardwa
r
e fun
c
ti
on
s invisible to up
per mo
dule.
For exam
ple,
in comm
uni
caton tasks of
the node, inte
rface laye
r in
deed i
s
physi
cal layer.
2) The mi
ddl
e layer provides lo
w-level
function of e
rro
r che
cki
ng,
queue
sche
duling a
nd et
c to
wo
k a
s
rel
a
y
of data
flow a
nd
cont
rol flo
w
. Th
e MA
C l
a
yer i
n
comm
unication ta
sk i
s
a typical
middle laye
r.
3) Mana
gem
ent layer pro
v
ides high
-le
v
el func
tion of event detection, netwo
rk form
ation and
self-m
aintain
ence to wo
rk as the task
contro
le
r. Routing laye
r in comm
uni
ca
tion task i
s
a
typical mana
gement laye
r.
4) Shared ap
plicatio
n layer coordinate
s
the coll
abo
rati
on between d
i
fferent sta
c
ks.
The three lo
wer layers
of the sta
c
k
structu
r
e
re
sp
e
c
tively wo
rks as l
ogi
cal in
terface,
pro
c
e
ssi
ng u
n
it and co
ntro
lling unit. Architecture wi
th
shared ap
pli
c
ation laye
r take
s into a
c
count
the cha
r
a
c
teri
stics of the m
odern WS
N communi
catio
n
proto
c
ol an
d gene
ral pa
rallellizaito
n.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Hardware-in
-
t
he-lo
op an
d Parallel Sim
u
lation Architecture for Wi
rel
e
ss Sensor
… (Shiho
ng
D)
109
2.3. Design
of Mapping
Scheme Be
tw
e
e
n
Real Node With Vir
t
ual compon
ent
In ord
e
r to re
duce runni
ng
time of a
lar
ge scale WS
N simulatio
n
and also
de
b
ug
the
embed
ded
co
de in se
nsor
node
s, ha
rdware
-in-th
e-l
o
o
p
emulation
p
l
atform is ext
ende
d from t
h
e
la
ye
r
a
r
c
h
itectu
r
e
p
r
op
os
ed
in
s
e
c
t
io
n 2.2
b
y
us
in
g a ma
pp
in
g sc
he
me
b
e
t
we
e
n
r
e
a
l
no
d
e
w
i
th
virtual com
p
o
nent in PIWSNSim. The
mappin
g
sch
e
me integ
r
at
es real no
de
s into sim
u
la
tion
platform
as
embed
ded
compon
ents o
f
virtual n
ode
s.
Runni
ng
p
a
ram
e
ters
of re
al n
ode
s
are
transmitted to PIWSNSim
; relative enviorment
mo
dels defin
ed i
n
PIWSNSim
provide phy
sica
l
appli
c
ation
e
n
viorme
nt si
mlation, tran
sfer
data fro
m
so
me real
node
to ot
her
nod
es,
and
emulate p
r
a
c
t
i
cal
WSN.
Ru
nning i
n
form
a
t
ion from
real
node
s to PI
WSNSim
su
p
ports de
buggi
ng
cod
e
easily. Real no
de
s are u
s
ed in
hard
w
a
r
e
-
in-t
he-lo
op emu
a
ltion to cre
d
ibly verify the
effectivene
ss of some o
p
timization t
e
ch
nolo
g
y,
provide reliabl
e data ba
sis for the act
ual
deployme
nt of the network,
and form di
stributed
simul
a
tion to redu
ce runni
ng tim
e
.
Shown i
n
Fig
u
re 5, ma
ppi
ng nod
es a
r
e
deployed in
resou
c
e laye
r of PIWSNSim with
two tier a
r
e
respon
sibl
e for tra
c
king t
he a
c
t
ual op
eration
of the se
nsor n
o
des
and
passin
g
runni
ng i
n
formation to
wi
rele
ss lin
k
model
und
er the
som
e
manag
eme
n
t strategy to
form
clu
s
terin
g
wireless
se
nsor network
whi
c
h t
r
an
smits
data p
a
kca
g
e
from
some
se
nde
r n
o
d
e
to
other re
ceive
r
n
ode
s. Sen
s
or n
ode
s
se
nd
run
n
ing
st
ate to m
appi
ng n
ode
via
wirel
e
ss met
hod,
su
ch as usi
n
g
net
work protocol
sta
c
k to
com
m
uni
cate with
sim
u
lation virtu
a
l
nod
e, or wi
red
method, such
as the se
rial
interface or
USB in
terface.
Mappin
g
nod
e repla
c
e the
some fun
c
tio
n
of virtual no
de an
d b
u
ild up th
e m
appin
g
bet
ween real no
des and
virtual
no
de
s. Multi
comm
uni
cati
ons mo
de
s use
d
betwe
e
n
physi
ca
l n
ode
s and virtual node
s in
PIWSNSim and
multi-core
an
d multi-threa
d
mechani
sm
provide
d
by PIWSNSim platform sup
ports l
a
rg
e-scale
netwo
rk
simul
a
tion with mo
re than 10
00
node
s.
Figure 5. Mapping
schem
e and dat
a d
e
f
ition in PIWSNSim
Real
nod
es
sen
d
net
wo
rk relate
d dat
a
to map
p
ing
node
with t
w
o tier
and
work a
s
a
comp
one
nt in PIWSNSim
by su
ch m
a
pping
schem
e to re
pla
c
e
task l
a
yer
an
d re
so
urce la
yer
defined in
si
mulation no
d
e
s. The m
a
n
ageme
n
t layer are
ke
pt in virtual node t
o
simul
a
te so
me
monitori
ng
or co
ope
ration
strategy. Da
ta str
eam
be
tween
real
node
s
and
virtual n
ode
s
are
divided into two cl
asse
s. One is n
e
two
r
k related
d
a
ta sen
d
amon
g real no
de
s, and the othe
r i
s
runni
ng state
data for code
debu
gging. T
he network
re
lated data is
a quintupl
e of (node ID, )
The inte
rfaces of diffe
rent layers,
node
s a
nd
model
s a
r
e
defined
uni
fied. The
comp
one
nts
in different l
a
yers
ca
n b
e
ea
s
ily assembled
and
disma
n
tled
in the syste
m
.
Simulation st
eps to imple
m
ent hard-in
-the-lo
o
p
emul
ation are d
e
scibe
d
as follo
ws:
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 1, Janua
ry 2013 : 103 – 1
1
4
110
1) Buidi
ng a
new
co
mpon
ent of map
p
i
ng layer in m
appin
g
no
de
to repl
ace
th
e functio
n
la
yer
and re
so
urce
layer define
d
in virtual node
sh
o
w
n
as Figure 5
.
The interfa
c
e
s
between
mappin
g
laye
r and m
anag
ement layer
or wi
rele
ss chann
el are
u
n
ified to use the simul
a
tio
n
model
s ea
sily.
2) Ma
ppin
g
l
a
yer in m
app
ing no
de ta
ke re
sp
onsi
b
ili
ty for con
n
e
c
ting re
al no
d
e
s via
wirele
ss
or
wire
mode, a
nd catchi
ng the re
al no
de
runni
ng
statu
s
, trackin
g
/de
buggi
ng no
de
s emb
edd
ed
function
s to
map
real
no
d
e
to th
e virtu
a
l no
de. Al
so
sim
u
lation
m
odel
s em
ploy
ed in
ma
ppin
g
node
s integ
r
a
t
e real nod
e into PIWSNSi
m
virtual network.
2.4. Paralleli
z
a
tion E
v
ent Scheme on Multicore
Computer
Multi-processor comp
uters
[8
]
con
s
titute t
he de
-fa
c
to d
e
fault ha
rd
wa
re pl
atform e
v
en for
desktop
syst
ems to p
r
o
v
ide che
a
p
yet powe
r
fu
l “private
comp
uting
cl
usters”. So the
parall
e
lization
of discrete e
v
ent simulati
ons
signi
fi
can
t
ly gained im
portan
c
e
and
can b
e
put i
n
to
earth to redu
ce runni
ng time of larg
e-scale WS
N
simulatio
n
. This pa
pe
r prese
n
ts a
sim
p
le
dynamic cent
ralized p
a
rall
el sche
dulin
g
method to
su
pport th
e sim
u
lation of la
rg
e-scal
e
WSN
by
sched
uling i
n
depe
ndent
e
v
ent pro
d
u
c
e
d
in
simulati
o
n
progress.
Queu
e of tim
e
-sta
mpe
d
e
v
en
t
with time module
s
is crea
ted and mai
n
tained dy
na
mically by rel
a
ted com
pon
ents. A paral
lel
event sch
ed
uler propo
se
d in this paper takes re
pon
sibility to manage ev
ent queue a
nd
dis
t
ribute event to s
o
me core of c
o
mputer.
Parallel
simul
a
tion in PIWSNSim in
clud
es m
odelin
g
executio
n tim
e
of ea
ch
event, and
dynamical ce
ntralized eve
n
t sch
edul
i
n
g
sch
eme. Th
e step
s of pa
rallel
simulati
on is
sho
w
n
as
follows
:
1) Emb
edd
e
d
software of
WSN
nod
es are
de
si
gn
e
d
and
develo
ped a
c
co
rdin
g to ap
plication
requi
rem
ents of
WSN.
2) Event execution time m
odelin
g meth
od is em
pl
oy
ed to define
static minim
u
m time of event,
and the
dyna
mic
run
-
time
corre
c
tion. K
e
y value of t
he time m
o
d
e
l are record
ed in th
e xml
data files.
3) Simul
a
tion
mode
is co
nfirmed. If
si
mulation i
s
need
ed, virtu
a
l no
de
with
three
tiers i
s
desi
gne
d; if emulation is n
eede
d, mappi
ng nod
e with
two tier is buil
t
up by mapping schem
e.
4) Simul
a
tion
WSN is e
s
tablished foll
o
w
ing
the a
p
p
licatio
n ne
e
d
s, in
clu
d
ing
to defin
e th
e
netwo
rk top
o
l
ogy, the chan
nel model, se
nso
r
mod
e
l a
nd so o
n
.
5) Event
schedul
er
use
the the
dy
namic
centra
lized
pa
rallel
event
sche
duling
metho
d
to
simulate the
operatio
n of large
-
scal
e WSN
a
nd a
nalyze p
e
rfo
r
mance. Event sche
dulin
g
method in thi
s
pap
er i
s
pa
rallel exe
c
uti
on strategy
to initiatively issue event to
variou
s co
re
and
optimi
z
e
d
sch
edule,
which
is propo
sed
for
co
mp
uter
of multi-core processo
r
with sh
are
d
memory.
The eve
n
t sche
dule
r
i
s
core u
n
it to
achieve
ce
ntralized m
a
nagem
ent, d
y
namic
optimizatio
n. Event in simu
lation su
cce
s
sively
be put
into pendi
ng
executio
n event queu
e; each
CPU
core ha
s a worke
r
threa
d
an
d m
a
intain
s a
st
orag
e unit ho
lding event p
r
ocessin
g
tasks.
Memory unit of
executing event
of
t
a
sk
is set
a spi
n
lo
ck
t
o
p
r
ov
ide
saf
e
sh
arin
g serv
i
c
e
i
n
mu
lt
i-
core p
r
ocessor sy
stem. F
unctio
n
co
mp
onent
s execu
t
e task a
nd
manag
eme
n
t, prod
uce time-
stampe
d events
a
nd pu
sh
event
to
pendi
ng eve
n
t que
ue,
which
is p
r
ote
c
ted
with
m
u
te
x
sema
pho
re o
r
co
nditional
variable
s
. Fig
u
re 6 d
e
scri
b
e
s the sch
e
d
u
lling process co
ntrole
d b
y
centralized
sche
dule
r
.
Th
e solid line is event strea
m
, and
the dash
ed line is the synchro
nou
s
data stream.
Key elements in paralleli
za
tion are defin
ed with
sets, variable
s
an
d
function
s. E={e| e is
event prod
uced in simul
a
tion model
s}
，
F={
e
|e is so
me pendi
ng
event}, D={e|
e is issued e
v
ent
to CPU core
by sche
dule
r
}
,
W
i
={t | t is worker thre
ad i
n
ith
CPU
co
re}, an
d F
⊆
E. Each
event h
a
s
two prope
rty value {T
start
, T
end
}, wherein
T
start
is initial
event executi
on time and
T
end
is event end
t
i
me.
Twin is
s sy
nch
r
o
n
o
u
s wi
ndo
w of
cur
r
ent even
t. Task = NULL mean
s m
e
meo
r
y unit of
executin
g event is empty. Functi
on of
waitFo
rOn
e
T
h
rea
d
is defi
ned to que
ry if there is idle
thread, if not
the executio
n is blo
c
ked
until som
e
th
read i
s
idl
e
whi
c
h is
processed
as
ret
u
rn
value. The scheduli
ng algo
rithm is de
sig
ned a
s
follows.
1) Du
rin
g
initializatio
n, function comp
one
nts pu
sh stati
c
events into
F, and T
wi
n
=
∞
。
2) Event sch
edule
r
extract the first ev
ent in F as curre
n
t event as. Then T
win
is updated
by
comp
ari
ng T
st
art
of E
current
and T
wi
n
. If T
sta
r
t
≤
T
wi
n
,
e
can be
exe
c
ute
d
parall
e
l with E
current
, else
e
will be wait
until the value of T
wi
n
is chan
ged. CheckPa
r
alli
za
tion function
judges the
simultan
eity of event.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Hardware-in
-
t
he-lo
op an
d Parallel Sim
u
lation Architecture for Wi
rel
e
ss Sensor
… (Shiho
ng
D)
111
Figure 6. Sch
edule
r
processing in PIWS
NSim
checkParallization(){
do{
e=m
i
n{T
starte
’
|e’
∈
F}
if (T
starte
≤
T
wi
n
)
retur
n
(e
)
else
waitFo
rSom
eThre
ad(w|
Tw
in
.job=NULL
)
T
wi
n
=
m
in{Ten
d(e'
)|e'
∈
D}
}whil
e
(true)
}
3) S
c
he
dule
r
queri
e
s the
st
atus
of worke
r
th
read
to fin
d
the
availabl
e on
e, move
s E
current
to s
e
t
of
D, di
stribute
s
E
current
to availabe
worker thre
ad
, and finall
y
cha
nge t
he value
of
T
wi
n
=min{T
end
(e)|e
∈。
D}
4)
Worker thread
checks the loca
l ev
ent status, that means t
he
status of
memory uni
t
o
f
executin
g ev
ent of ta
sk. If task if free, t
he eve
n
t w
ill
b
e
pr
oc
es
se
d; e
l
s
e
ta
sk is
s
p
in
lo
ck
ed
to
synchro
n
ize threa
d
s.
5) Wo
rker thread executes ev
ent proce
ss p
r
og
ram t
o
gene
rate n
e
w events m
a
rked with time
stamp a
n
d
pushed in
to F. Once
the wo
rk
e
r
thre
ad is finished, t
he value o
f
T
wi
n
=min{T
end
(e)|e
∈
F}.
Centralized sche
duling architecture
with
m
anag
emen
t of dynamic
events contro
led by
event sche
du
ler can b
a
lan
c
e lo
ad
s effe
ctively.
In WSN sim
u
latio
n
s, mobility
of node
s, time-
varying characteri
stics of
chan
nel can
bring u
npre
d
ica
b
le chag
en of wo
rk l
oad. Ce
ntrali
zed
manag
eme
n
t scheme
will
be well
ada
pted to su
dde
n
alteration du
ring runin
g
wi
thout
addition
al
balan
ce
strat
egie
s
.Sche
d
u
l
er takes
re
spon
sibility to query eve
n
t queu
e in se
q
uen
ce, and fi
nd
out the cu
rre
nt event, and distrib
u
te ev
e
n
t after che
cking the paralli
zation.
3. Results a
nd Analy
s
is
This
se
ction
pre
s
ent
s eval
uation of co
rrectn
e
s
s the link reliabiility model, effecti
v
ess of
parall
e
lization
sche
me, a
n
d
at la
st
use
s
PIWS
NSim
platform to
s
i
mulate
a practical industrail
WSN
with 40
0 node
s to an
alyze the extensi
b
ilty
of our simul
a
tion a
nd emulatio
n frame
w
ork.
3.1. Link Reliabilit
y
Model (EIF) Anal
y
s
is
Two
simulati
ons a
r
e d
e
si
g
ned to test th
e co
rre
ctne
ss and validity of EIF. First one is to
prod
uce PDS
acco
rdin
g to
model
ba
se
d on
offline t
e
st
sampl
e
s
in ou
r lab
by
usin
g EIF a
n
d
OMNet
+
+. M
ean
while
a p
r
actical te
st is
execut
e
d
bet
wee
n
two
no
des whi
c
h
re
corded
their
own
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 1, Janua
ry 2013 : 103 – 1
1
4
112
PDS and
ano
ther
simulatio
n
ba
se o
n
T
w
o-Ray Groun
d ch
ann
el Mo
del with
erro
r ratio of 1
0
%
is
impleme
n
ted
by using
NS2
.
The analy
s
is shown
in F
i
gure
7 prove
s
the lin
k reli
ability mode i
n
PIWSNSim is more correc
t.
Figure 7. Correctn
ess resu
lt of link reliab
ility model co
mpared with
NS2 and p
r
a
c
tical test
Secon
d
sim
u
l
a
tion is to te
st validity and scal
ability of EIF by implementing the
spe
c
ific
hoppi
ng
sche
me. Ch
annel
redu
nda
ncy
is on
e of the
techni
que
s t
o
increa
se t
h
e relia
bility and
determi
nism
of data comm
unication be
cause none of
the chan
nel
s can p
r
ovide a
good relia
bili
ty
requi
re
d by in
dustry
appli
c
ation. Slotted
hoppi
ng a
nd
slo
w
ho
ppin
g
are
always u
s
ed i
n
ind
u
st
rial
appli
c
ation. I
n
slotted
ho
pping,
cha
n
n
e
l-ho
ppin
g
timeslot
s o
ccupy equ
al d
u
ration
s. Ta
king
accou
n
t of link
cha
r
a
c
teri
stics to incre
a
se
the
relia
bility, some
adaptive fre
q
uen
cy hoppi
ng
method
s h
a
ve be
en
pro
p
o
s
ed
to
satisfy
the hig
h
relia
bility requi
re
ments. T
hat i
s
u
s
in
g the
st
able
and reliable
cha
nnel
with
out hoppi
ng
until the ch
a
nnel go
es to
bad, and th
en ch
oo
se a
new
cha
nnel
with
high
relia
bili
ty to rea
c
h
a high
reli
ab
le network
communi
catio
n
. This ki
nd
of
hoppi
ng is
cal
l
ed ada
ptive frequ
en
cy hop
ping
[9]
.
Table 2. Reli
ability of AFH and time-slot
hoppin
g
in si
ngle cl
uste
r start WSN
Nodes
1
2
3 4 5 6
7
8
AFH
(
%
)
98.1
96.9
94.5 98.0 95.4
100
94.5
96
Time-slot
hoppin
g
(%)
80
81
79 80 80 80
81
80
EIF provid
es the
both
ho
pping
st
rateg
y
to in
cre
a
se the
reli
abil
i
ty and T
D
M
A
MAC
proto
c
ol to
a
v
oid interfe
r
e
n
ce
s
amon
g
node
s
and
verify the tran
smissio
n
d
e
la
y. Hoppi
ng m
ode
is exp
r
e
s
sed
as a
configu
r
ed
pa
ramete
r in
EIF for u
s
ers to
cho
o
s
e
whi
c
h
ho
p
p
ing
method
is
calle
d by TDMA MAC
[10]
module in th
eir sim
u
latio
n
s. The co
nclu
sio
n
is that the adap
tive
freque
ncy h
o
pping m
a
inta
ins the hi
gh
reliability du
e to swit
chin
g to the cha
nnel with hi
g
h
reliability com
m
unication shown in Tabl
e 2.
3.2. Time effectiv
eness of parallel sch
eme in PIWSNSim
An instan
ce
simulation i
s
impleme
n
ted to
verify the time effectivene
ss
of parallel
scheme i
n
tro
duced in Se
ction 2.4. The
simulatio
n
pl
aygrou
nd i
s
1000m
*100
0,
and the
r
e o
ne
sin
k
no
de lo
cated in
coo
r
di
nates
of (0,0
), other n
ode
s are
ran
doml
y
place
d
on
some lo
cation
in
the playg
r
ou
n
d
. No
de
s
con
s
ults with
nei
ghbo
rs in
a
ccorda
n
ce
with
the lo
cation
to self-o
rga
n
ized
into clu
s
ters with less than
64 membe
r
node
s.
Nodes are equipmented
with network
prot
ocol
com
p
liant with IEEE 802.15.4
standrad
to uploa
d dat
a re
gula
r
ly. Clu
s
ter
hea
d
s
a
r
e
sele
ct
e
d
acco
rdi
ng t
o
the di
stan
ce from
sin
k
n
ode
and
nod
e ID.
Topolo
g
y of a
clu
s
ter is sta
r
st
ru
ct
ure,
a
nd T
D
MA m
e
cha
n
sim
is e
m
ployed to
send
Evaluation Warning : The document was created with Spire.PDF for Python.