TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 16, No. 3, Dece
mbe
r
2
015, pp. 574
~ 582
DOI: 10.115
9
1
/telkomni
ka.
v
16i3.887
0
574
Re
cei
v
ed Au
gust 20, 20
15
; Revi
sed
No
vem
ber 1
2
, 2015; Accepte
d
No
vem
ber
28, 2015
Nodes Deployment Scheme of Heterogeneous Wireless
Sensor Network Based
on Organic Small Molecule
Model
Ju
w
e
i Zhan
g
1,2,3
, Y
u
Wang
1,2
, Yachuang Liu
1
, and Qiang
y
i
Li
1,2
1
Hena
n Univ
er
sit
y
of Scie
nce
and T
e
chno
log
y
, He
nan L
u
o
y
ang 4
7
1
023, C
h
in
a
2
Po
w
e
r Electro
n
ics Devic
e
an
d S
y
stem En
gi
neer
ing L
ab of
Hen
an, He
nan
Luo
ya
n
g
471
0
23, Chi
n
a
3
X
i
an
Ji
ao
to
ng
U
n
i
v
e
r
si
ty
, Shan
xi
Xi
an
7
1
0
0
4
9
,
C
h
i
na
Corresp
on
din
g
author, e-mai
l
:
A
b
st
r
a
ct
F
o
r nod
es d
e
p
l
oy
me
nt of het
erog
ene
ous s
e
nsor n
e
tw
ork, base
d
o
n
differ
ent pro
b
a
b
il
ity sensi
n
g
mo
de
ls of hete
r
oge
neo
us no
d
e
s, refer to organic s
m
al
l
mol
e
cul
e
mode
l,
‘
c
lass-
mol
e
cu
le
’
sensi
ng mod
e
l of
hetero
g
e
neo
us
nodes is
pro
pose
d
. Co
mb
i
ned w
i
th DS
mT
data fusion
mo
de
l, t
he ch
ang
es of netw
o
rk
covera
ge r
a
tio
after usi
ng th
e
new
se
nsin
g
mo
de
l a
nd
dat
a fusi
on
alg
o
rit
h
m is stu
d
i
ed.
Accordi
ng to
th
e
researc
h
result
s, the node-
de
ploy
m
ent strategy of heter
og
ene
ous se
nsor
netw
o
rks base
d
on or
gan
ic s
m
a
l
l
mo
lec
u
le
mo
d
e
l (NHOS) is prop
os
ed. Se
n
s
or netw
o
rk simu
lati
on
mod
e
l
is establis
he
d usin
g MAT
L
AB
softw
are, the results sh
ow
th
e effective
nes
s of t
he
alg
o
ri
thm, the
cov
e
rage
of n
e
tw
ork and
d
e
tectio
n
efficiency of n
o
des are i
m
prov
ed, t
he lifeti
m
e of netw
o
rk is prolo
nge
d,
ener
g
y
consu
m
pti
on
and the
nu
mb
e
r
of de
ploy
ment
nod
es is r
e
d
u
c
ed,
a
nd th
e sc
ope
of perc
e
iv
i
ng is
exp
a
n
ded
. As a resu
lt, N
H
OS can
impr
ove
the detecti
on p
e
rform
anc
e of the netw
o
rk.
Ke
y
w
ords
:
nod
e de
pl
oy
ment sche
m
e, h
e
terog
e
n
eous
w
i
reless se
nso
r
netw
o
rk, organic s
m
a
ll
mo
l
e
cul
e
mo
de
l
Copy
right
©
2015 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
With
the dev
elopme
n
t
of wirel
e
ss se
nsor
te
chn
o
logy
and man
u
facturing microel
ectro
n
ic
techn
o
logy, the wirele
ss
sensor net
wo
rk co
nsi
s
t of
a large nu
mbe
r
of micro
sen
s
or n
ode
s wh
ich
own
perce
ption, cal
c
ulatio
n and
comm
u
n
icatio
n abilit
y is applie
d to the military
or civilia
n are
a
s,
such as
environmental
moni
toring, i
ndustrial
control,
battlefie
ld
surveillance, detection of
high-
risk environm
ent, biologi
cal
medicin
e
, intelligent hou
sehold a
nd he
alth monitori
n
g
, etc [1].
Coverage
problem i
s
a
b
a
si
c p
r
oble
m
in wi
rele
ss sen
s
o
r
n
e
tworks depl
oyment
[2],
unde
r the co
ndition of the sen
s
o
r
no
de ene
rgy,
perception, co
mmuni
cation
and com
put
ing
ability limited, using
a ce
rtain
strategy
of
softwa
r
e
and h
a
rdware,
which
can en
su
re t
he
coverage a
r
e
a
and covera
ge time, reali
z
e effective a
w
arene
ss an
d monitori
ng,
is an impo
rt
ant
indicator of network pe
rce
p
tion se
rvice
quality,
and also is a ho
t problem in
sen
s
o
r
netwo
rk
research [3].
In many appli
c
ation
s
, there
are usually a
variet
y of monitorin
g
obje
c
ts in the mo
nitoring
area. S
u
ch a
s
, the
se
nsors n
ode
s
nee
d to mo
nitor
water temp
erature,
salinity
,
ph valu
e, et
c in
water envi
r
on
mental m
onit
o
ring,
and
th
e sen
s
ors
no
des ne
ed to
monitor a va
riety of ch
emi
c
al
diffusion in fa
ctory pollutio
n
early wa
rni
ng [4].
Since th
e h
a
r
dware
co
st
of cu
rrent
se
nso
r
n
ode
is high
er, in
m
u
lti-obje
c
t m
o
nitoring
appli
c
ation
s
,
each no
de a
s
sembly a v
a
riety of di
fferent type
s
of sen
s
o
r
s, these no
de
s
are
hetero
gen
eo
us. When
the
node
ene
rgy
is limited, th
e more
sen
s
ors are a
s
se
mbled
on a
n
ode;
the life of th
e
nod
e i
s
sho
r
ter. Two imp
o
rtant p
r
o
b
le
ms
sho
u
ld
be
co
nsi
dered i
n
multiple
obj
ect
monitori
ng
n
e
twork
cove
rage, n
a
mely
ho
w to
us
e
less
co
st of
network to
obtain th
e id
eal
netwo
rk cove
rage
pe
rform
ance, and
h
o
w to
wei
g
h
the impo
rta
n
ce
of net
wo
rk m
onito
ring
of
different obje
c
ts a
c
cording
to different objec
t
s
[5].
Heteroge
neo
us
ch
aracte
ristics of
he
ter
oge
neo
us wi
rele
ss
sensor netwo
rk give
expre
ssi
on t
o
three a
s
p
e
cts in
clu
d
in
g node
hete
r
oge
neity, link hete
r
og
en
eity and net
work
proto
c
ol hete
r
oge
neity [6-10]. Node het
erog
eneity in
clud
es pe
rce
p
tion ability,
cal
c
ulatio
n ability,
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Nod
e
s
Deplo
y
m
ent Schem
e of Hetero
ge
neou
s Wi
rele
ss Sen
s
o
r
Ne
twork… (Ju
w
ei Zhang
)
575
comm
uni
cati
on ability, energy, etc. And the comm
u
n
icatio
n ability, perceptio
n ability, and energy
make th
e bi
gge
st influen
ce o
n
coverage. T
he
co
verage
probl
em re
se
arch
in the exist
i
ng
literature abo
ut rando
mly deployment of hetero
gen
eo
us wi
rele
ss sensor net
wo
rk is le
ss.
A wirele
ss se
nso
r
n
e
two
r
k
coverage
opti
m
izat
ion
alg
o
r
ithm i
s
p
r
e
s
e
n
ted to
pe
rceive the
radiu
s
of the hete
r
og
en
eou
s, whi
c
h
can
e
ffecti
v
ely improve
the hete
r
o
gene
ou
s net
work
coverage
in
[4]; Based
o
n
integ
e
r ve
ctor pla
nning
, A multi-o
b
j
e
ctive
cove
rage
algo
rith
m is
prop
osed in [5], which
can
effectively sol
v
e the probl
e
m
of multiple target monito
ring. Both of the
two pe
rcepti
on mod
e
l wh
ich u
s
e
d
in [
4
] and [5] is relatively si
mple, and
can not m
eet
the
deman
d of the true conditi
on.
A routing p
r
otocol b
a
sed
evolutionary
al
gorithm i
s
prop
osed fo
r the hete
r
o
gene
ou
s
clu
s
ter no
de of WSN in literatu
r
e [11], Whi
c
h ca
n
effectively redu
ce the errors
of cluste
r nod
es
handli
ng data
combin
ation
and sepa
ratio
n
, prolon
g th
e
su
rvival time of netwo
rk,
but did not gi
ve
the algo
rithm
of non-clu
s
t
e
r no
de. An
efficient
dyna
mic cl
uste
rin
g
strat
egy (E
DCS
) which
can
effectively sol
v
e the selecti
on pro
b
lem o
f
the
multi-level heterog
e
neou
s network clu
s
ter n
o
d
e
,
effectively improve the p
e
rforman
c
e of
netwo
rk
and
prolo
ng the
survival time of network is
put
forwa
r
d in lite
r
ature [12]; howeve
r
, EDCS did
not give a spe
c
ific no
de depl
oyme
nt algorithm.
A new ide
a
i
s
propo
se
d b
y
incre
a
si
ng
hetero
gen
eo
us no
de
s to
prolo
ng the
survival o
f
WSN i
n
litera
t
ure [13], b
u
t
the hete
r
og
e
neou
s
node
p
e
rception
pro
b
lem
wa
s n
o
t co
nsi
dered.
To
solve the
se
nsin
g ra
diu
s
of heteroge
neou
s n
e
two
r
k
node
depl
oyment probl
em, Expandi
ng-
virtual force
algorith
m
(E
X-VFA) al
go
ri
thm is put
fo
rward in
[14],
but the
co
nn
ectivity and
d
a
ta
fusion p
r
obl
e
m
is little con
s
ide
r
ed in EX
-VFA.
For
hete
r
oge
neou
s Se
nsor n
e
two
r
k
node
s
cove
rage
blind
area p
r
obl
em,
a n
ode
deployme
nt coverage
alg
o
rithm ba
sed
on org
ani
c small
mol
e
cule
s m
odel
NHOS
(No
des
Depl
oyment Strategy of Heteroge
neo
us Sen
s
o
r
Networks-ba
s
e
d
on Organi
c Small Molecule
Model, NHO
S
) is pro
p
o
s
e
d
in this pap
e
r
. Based o
n
the org
ani
c small mole
cul
e
stru
cture m
odel,
hetero
gen
eo
us
n
ode
s det
ection mod
e
l is
e
s
tabli
s
h
e
d
.
In orde
r to i
m
prove th
e n
e
twork
cove
rage,
extend the lif
e of the
network an
d a
c
hie
v
e the optim
a
l
deploym
ent
of se
nsor
nod
es to
monito
red
area,
A d
a
ta
fusion
mo
del
wa
s propo
sed, combi
n
in
g with
the
DSmT data
fu
sion
an
d
deci
s
ion
rule
s. Ra
ndo
mly deploym
ent mobile
n
ode
s will m
o
ve acco
rding
to the pri
n
ciple of "virtu
al
potential field
"
. The effe
ctiveness
of NHOS i
s
ve
rified
by simul
a
tion
expe
rim
ents, whi
c
h can
effectively re
duce the
nu
mber
of de
ployment
no
de
and th
e no
d
e
ene
rgy
con
s
umptio
n, im
prove
the efficiency
of network coverag
e
, prol
ong the
network lifetime, expand
the d
e
tection rang
e,
and imp
r
ove the dete
c
tion
perfo
rman
ce
of the network.
2.
Mathem
atica
l
Model and Assump
tion
We a
s
sume
that the mon
i
toring a
r
ea
H by
re
ctan
gular, N
kin
d
s of hete
r
o
gene
ou
s
perception ability of wirele
ss sensor nodes randomly distribut
ed
within the rectan
gle H, and al
so
assume that t
he wirele
ss
sensor net
wo
rk
(WSN) ha
s the followin
g
prop
ertie
s
:
(1)
Hete
rog
e
neou
s p
e
rce
p
tion ability sensor. Se
n
s
o
r
s h
a
ve different perce
ptio
n ability,
the pe
rceptio
n of
radiu
s
i
s
diffe
rent, a
nd th
e
no
de
model
pe
rce
p
tion i
s
p
r
ob
ability perce
p
t
ion
model. Th
e p
o
sition
of sen
s
or no
de i i
s
(,
)
ii
x
y
, perceptio
n di
stan
ce i
s
i
r
and
perceive
d
p
r
obability
is
i
P
whi
c
h mee
t
the function
()
ii
Pr
. The typical p
e
rception p
r
o
bability model
[13] is as follows:
()
0,
,
(,)
,
,
.
1,
ei
p
ir
ie
i
p
e
i
ei
p
rr
d
E
P
SQ
e
r
r
d
r
r
E
rr
d
(1)
Whe
r
e
(,)
i
PS
Q
is th
e perce
ption
prob
ability of the sen
s
o
r
node
i
S
to target node
Q,
ip
d
is
Euclide
an
distance
bet
wee
n
sen
s
or no
d
e
i
S
and th
e ta
rget n
ode
Q,
()
ee
rr
r
is a
un
cert
ainty
perceptio
n measure of the
sen
s
ors,
i
E
is the initial energy of senso
r
node
s,
ir
E
is the remaining
energy,
()
ip
e
dr
r
,
and
is the
perceptio
n of
sen
s
o
r
n
o
d
e
s
within the
scope
of
monitori
ng an
d thing
s
the a
ttenuat
ion
co
efficient of qu
ality.
is rand
om num
ber
meeting n
o
rm
al
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 16, No. 3, Dece
mb
er 201
5 : 574 – 582
576
distrib
u
tion, whi
c
h sho
w
s the reality of various
inte
rfere
n
ce on the influen
ce
of the percei
v
ed
probability;
(2) All sen
s
or node
s have
the ability of
dat
a fusio
n
, unde
rwater
communi
catio
n
s have
enou
gh en
ergy to finish se
lf-positio
ning
and c
an mov
e
to the spe
c
i
f
ied locatio
n
freely;
(3) Before th
e de
ploymen
t
algo
rithm e
x
ecute, all
n
ode
s h
a
ve fi
nish
ed
self
-p
ositioni
ng
accurately, and the nod
es’
coordinate
s
are kno
w
n;
(4) T
here a
r
e thre
e kin
d
s of wo
rki
ng
state in
node life: dorm
a
n
c
y, detection,
comm
uni
cati
on. The
nod
e ene
rgy
con
s
umptio
n in
comm
uni
cati
on state
is t
he la
rge
s
t, a
nd is
smalle
st in
a
do
rmant
stat
e. The
rel
a
tio
n
shi
p
of
ene
rgy co
nsumpti
on a
nd th
e ti
me of
nod
es
in
comm
uni
cati
on
state a
r
e
positive. T
h
e
larg
er of
co
mmuni
cation
distan
ce, th
e
more e
nerg
y
is
con
s
um
ed.
DSmT (De
z
e
r
t - Smarand
ach
e
found
) i
s
put fo
rwa
r
d
by the Fren
ch
schola
r
Deze
rt in
2002 [14], an
d later i
s
dev
elope
d jointly by Dezert an
d Smara
nda
che schol
ar a
n
d
so o
n
. DS
mT
is an exten
s
i
on of the cl
a
ssi
cal th
eory
of evi
dence, but different
from ba
sical
l
y D-S theo
ry.
DSmT can combi
ne wit
h
any type of trust
function of inde
pend
ent so
u
r
ce, but mai
n
ly
con
c
e
n
trate i
n
the co
mbi
ned un
ce
rtai
nty, high
co
nflict, and in
accurate sou
r
ce
of evide
n
ce.
Espe
cially when the
conf
lict between
sou
r
ce b
e
co
me big
ger or eleme
n
t is
vague, relati
vely
impre
c
i
s
e, DSmT ca
n be
yond the limi
t
ations of
D-S theoreti
c
al
frame
w
o
r
k t
o
solve
com
p
lex
fusion p
r
obl
e
m
of static or
dynamic [15
-
20].
On the ba
si
s
of the above
assumptio
n
s
and DSm
T
re
lated theo
ry, we give the f
o
llowin
g
definition:
Defini
tion 1:
p-reliability co
verage: Assu
me that point
Q exists in th
e monitorin
g
area
H,
in sensor net
work the
credibility of the ta
rget node poi
nt Q test
results P(Q) meet:
()
(
0
1
)
PQ
p
p
(2)
Defini
tion 2
:
Effective co
verage
p
r
op
o
r
tion: If the
area
H n
eed
ed to
be
mo
nitored
is
two-dimension area, the acreag
e of H is S(H), the acreage m
eeting p-reli
ability coverage is
(H
)
p
S
, and the ratio
of
(H
)
p
S
and S(H)
is:
(H)
100%
(H)
p
S
S
(3)
Whe
r
e
is the effective coverag
e
propo
rti
on of two-di
mensi
on mo
n
i
toring a
r
ea
H.
If the monitoring a
r
ea
by three-dime
n
s
ion a
r
e
a
H,
its volume i
s
V(H), the
volume
meeting p-reli
ability coverage is
(H
)
p
V
, and the ratio of
(H
)
p
V
and V(H) is
:
(H)
100%
(H)
p
V
V
(4)
Whe
r
e
is the effective coverag
e
propo
rtion of 3-D m
o
nitoring a
r
e
a
H.
Defini
tion 3
:
Effective detection
rate: If there
are
(H)
N
target n
ode
s in
the monito
ri
ng
area
H, in whi
c
h
(H)
A
N
target nod
es are effecti
v
ely monitore
d by sen
s
or n
e
twork, then:
(H)
100%
(H)
A
N
N
(5)
Whe
r
e
is sen
s
or n
e
two
r
k for monito
ring
area
H effecti
v
e detection rate.
Defini
tion 4:
Free DSmT
model: Set U a recogn
ition frame
w
ork,
12
U={
}
n
，，
，
con
s
i
s
t of n
d
e
tailed
eleme
n
ts
set
(the e
l
ement in
the
set
ca
n ove
r
lap), a
nd
oth
e
r a
s
sum
p
tio
n
s
and
con
s
trai
n
t
s of the ele
m
ent (o
r the
s
is) a
r
e
not
co
nsid
ere
d
, and
then the mo
del of the
DSmT
model
(U)
f
M
is free
dom mod
e
l.
Give a gene
ral re
cognitio
n
framework U, defi
ne a ba
sic p
r
ob
ability assi
gnme
n
t functio
n
m:
[0
,
1
]
U
D
is associ
a
t
ed with a given so
urce of eviden
ce, na
mely:
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Nod
e
s
Deplo
y
m
ent Schem
e of Hetero
ge
neou
s Wi
rele
ss Sen
s
o
r
Ne
twork… (Ju
w
ei Zhang
)
577
()
0
m
(
A
)
1
U
AD
m
，
(6)
Whe
r
e m
(
A) i
s
gen
erali
z
e
d
the basi
c
probability assi
gnment fun
c
tion of A, its trust functio
n
a
n
d
likelihood function is as foll
ows:
(A
)
(
)
U
BA
BD
B
EL
m
B
(7)
()
()
U
BA
BD
PL
A
m
B
(8)
Whe
n
freed
o
m
DSmT mo
del wo
rk, we have:
,(
)
1
U
AD
P
L
A
(9)
Whe
n
2
n
in Equation
(6)
，
th
e data of
ea
ch p
e
rce
p
tion atomi
c
whi
c
h
com
b
ined
into sig
nal
sou
r
ce is ind
i
vidual sou
r
ce, mixed DSmT unde
r
the mixed free
DSmT mod
e
l of combin
ation
rule
s are as f
o
llows:
()
1
2
3
,
(
)
[
()
()
()
]
def
U
MU
A
Dm
A
S
A
S
A
S
A
(10)
12
12
1
()
1
,,
,
=
()
(
)
f
U
n
n
n
def
ii
MU
i
XX
X
D
XX
X
A
SA
m
A
m
X
（）
(11)
12
11
2
,,
,
1
[
(
)
(
)
]
[(
(
)
(
)
)
(
)]
()
(
)
n
kn
t
n
de
f
ii
XX
X
i
uX
uX
A
u
X
u
X
A
I
SA
m
X
(12)
12
12
12
3
1
,,
,
()
()
(
)
U
n
n
n
n
def
ii
i
XX
X
D
XX
X
A
XX
X
SA
m
X
(13)
3.
Percep
tual Element Mode
l
Orga
nic smal
l mole
cule
s i
s
the
organi
c com
pou
nd
who
s
e
mole
cular
wei
ght i
s
u
nde
r
1000, such a
s
methan
e (CH4), etha
ne (C2
H
6), etha
n
o
l (C2
H
5
O
H),
benzene (C6
H
6), et
c.
The main
ato
m
s of small o
r
gani
c mol
e
cules
a
r
e
ca
rb
on atom
s, hydrog
en atom
s, oxygen
atoms, o
r
nit
r
ogen
atom
s e
t
al. They g
r
o
uped to
gethe
r a
c
cordi
ng to
ce
rtain
stru
ct
ure i
s
sho
w
n
in
Figure 1, Inspired
by this, we combin
e
d
differ
ent st
ructure sen
s
o
r
nod
es into
one sen
s
ing
unit
according
to certai
n
rule
s. The comp
oun
d
nod
es
havi
ng different complem
entary sen
s
ing
no
des
can
effectivel
y expand th
e sco
pe of
perceptio
n,
i
m
prove
the
efficien
cy of
a si
ngle
nod
e's
perceptio
n, incre
a
se effecti
v
e covera
ge
of sen
s
or n
e
tworks by app
rop
r
iate data
fusion mo
del.
Figure 1. The
stru
cture m
o
del of methan
e
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 16, No. 3, Dece
mb
er 201
5 : 574 – 582
578
Defini
tion 5:
Perceive ato
m
s: Assumin
g
there a
r
e N se
nsor no
des in the m
onitorin
g
area
H,
in
clu
d
ing M
n
ode
s whi
c
h
have
ability
to pe
rceiving
target
nod
e alone,
and we
call
t
h
e
node
s that
can monito
r t
he targ
et an
d tran
smit
d
a
ta to the cl
uster
hea
d n
ode
s individu
ally
perceptio
n atoms.
Defini
tion 6
:
Perceptio
n
molecule
s: Assu
med th
at
n kind
s
of sen
s
o
r
n
ode
s are
deploye
d
in
monitori
ng
area, an
d the
d
i
fferent
ki
nd
s of
se
nsor
no
d
e
s are combi
ned
i
n
to sen
s
ing
units i
n
a
c
co
rdan
ce
with th
e ap
pro
p
ri
ate
data fu
sio
n
model. We
call
the se
nsi
n
g
unit pe
rcep
tion
molecule
s. Perception m
o
l
e
cul
e
s a
r
e th
e smalle
st pe
rce
p
tion unit
and commu
ni
cation u
n
it.
Assu
med tha
t
there are n
perception
a
t
oms
in the
molecule, ea
ch pe
rceptio
n atomic
data is
12
n
、、
、
respe
c
tively, and then perce
ption
molecul
a
r fu
sion d
a
ta so
u
r
ce i
s
:
12
U={
}
n
，，
，
(14)
4.
Node
Deplo
y
ment Stra
teg
y
Based on Organic Sm
all Molecular Model
With con
s
ide
r
ation of the p
e
rceiving u
n
it
model, an
d
data fusi
on m
odel, we put forward
a sen
s
o
r
net
work no
de de
ployment alg
o
rithm ba
se
d on small m
o
l
e
cul
e
model
as follo
ws:
Step 1:
Build
app
ro
priate
perceptio
n m
odel, d
e
termi
ne the
unit
(perceptio
n m
o
lecule)
perceptio
n
range
an
d th
e po
sition
of
the
atom
i
n
eve
r
y pe
rception
unit a
c
cordi
n
g
to t
he
monitori
ng a
r
ea of H m
onitorin
g
re
q
u
irem
ent
s. Compute the
optimal distance b
e
twe
en
molecule
s, and the numb
e
r
of all kind
s of sen
s
ing n
o
des.
Step 2:
Ran
domly deploy
sen
s
ing no
d
e
s in the mo
nitoring a
r
ea
H according
to the
numbe
r of all kind
s of se
nsi
ng nod
es in
step 1.
Step 3:
Sel
e
ct a
startin
g
no
de,
whi
c
h
usually i
s
a n
ode
s o
n
vertex
po
sitions
o
f
monitori
ng a
r
ea
H, then
select th
e corresp
ondi
ng n
o
de
clo
s
est to
the
starting
node
in o
r
de
r to
build a sen
s
o
r
mole
cule
s, and then its I
D
is set to 1.
Step 4:
Cho
o
se
a p
e
rcep
tion nod
e
clo
s
e
s
t to pe
rce
p
tion mol
e
cul
e
s
1, and
bu
ild the
se
con
d
perce
ption mole
cul
e
according to step 3, its
ID is set to 2. Acco
rdi
ng to "molecular force"
prin
ciple, the
force
betwee
n
mole
cule
s i
s
a
s
so
ci
ated
with the b
e
tween the m
o
le
cule
s. Assum
e
that the dista
n
ce b
e
twe
en
the two pe
rce
p
tion mole
cul
e
s is
'
D
, the best distance for
molecule
s is
i
D
, t is the threshold value;
Whe
n
the dist
ance
'
D
betwee
n
perceptio
ns molecul
e
s m
eet:
'
i
DD
t
(15)
The interm
ole
c
ula
r
force is
repul
sio
n
, no
de 2 move far away a units.
Whe
n
the dist
ance
'
D
betwee
n
perceptio
ns molecul
e
s m
eet:
'
i
DD
t
(16)
The interm
ole
c
ula
r
force is
attraction, no
de 2 move cl
osely a unit
s
.
Until the perception mole
cular di
stan
ce
meet:
'
ii
Dt
D
D
t
(17)
Node 2 moves
c
o
mpletely.
Step 5:
By the greedy al
gorithm, r
epe
at step
s 3 an
d 4, adju
s
t the mole
cula
r
positio
n,
until the entire molecula
r a
r
e adju
s
ted.
Step 6:
Ea
ch
perce
ption m
o
lecule
s com
p
lete initiali
zation for data
f
u
sio
n
t
a
s
k
all
o
cat
i
o
n
and data fu
si
on nod
e rota
tion se
quen
ce, and the p
e
rception
mol
e
cul
a
r
sen
s
o
r
s no
de
s act
as
data fusio
n
n
ode
s and
co
mmuni
cation
node
s in turn.
Step 7:
T
h
e
data fusion
node
s collect
the data of
other se
nso
r
s for d
a
ta fusio
n
at
regul
ar inte
rvals.
Step 8:
Th
e
node
s fo
r dat
a fusio
n
send
the data to
si
nk n
ode
s, an
d then
spe
c
if
y a data
fusion
node
and b
r
oa
dca
s
t the me
ssa
ges to
other
node
s a
c
cording to the d
a
ta fusio
n
no
de
rotation o
r
de
r.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Nod
e
s
Deplo
y
m
ent Schem
e of Hetero
ge
neou
s Wi
rele
ss Sen
s
o
r
Ne
twork… (Ju
w
ei Zhang
)
579
Step 9:
Sen
s
ors n
ode
s repeat ste
p
7
and ste
p
8
after re
ceivi
ng the data
fusion
comm
and.
Pseud
o co
de
of the algorith
m
is as follo
ws:
Functio
n_n
od
e_initi(N,1): I
n
itialise th
e
m
o
lecule
nod
e wh
ose ID is N, turn the fl
ag of the
node to 1, det
erm
i
ne rotatio
n
orde
r of ato
m
nodes that fuse data.
Functio
n_find
_
m
i
n(N,0):Fi
n
d the
ne
are
s
t
se
nsi
n
g
nod
e(N+1
)
wh
ose flag
is 0,turn the
ID
to N+
!;
Functio
n_find
_
m
i
n(N,n):Fi
n
d n n
e
a
r
e
s
t sen
s
in
g no
de
s
who
s
e fla
g
is 0, tu
rn th
e ID to
N+!;
Functio
n_b
uil
d_m
olecul
e(N):Build th
e
node
s
who
s
e
ID are
N int
o
m
o
lecule
n
ode, turn
the ID to N, t
u
rn the flag to 1;
Functio
n_m
o
v
e_fa
r
(N):M
o
ve th
e m
o
lecule n
ode
N fa
r a
w
a
y
from
the m
o
lecule
node
N-
1;
Functio
n_m
o
v
e_
clo
s
e
(
N):
M
ove the m
o
l
e
cul
e
nod
e N close to the m
o
lecule no
d
e
N-1;
Opt_di
st: The
optim
al distance b
e
twe
en
two m
o
lecule
node
s;
Thr_
dist; The
thresh
old di
stance
valu
e o
f
the optim
al
distan
ce.
1. Based on
influen
cing fa
ctors, divid
e
dict
ion a
r
ea t
o
I parts, cal
c
ulate the n
u
m
ber of
m
o
lecule no
d
e
s(Ni
),the red
unda
ncy rate
(R
r) and the o
p
tim
a
l distance betwe
en to node
s
in part I;
2. Deplo
y
N n
ode
s ran
dom
ly;
3. N
’
= N*Rr;
4. Wake N
’
n
ode
s ran
dom
ly, flag = 0;
5. The flag of node
0 = 1;
6. for
(
i
=0;i<N
i;i+
+)
7. Functio
n_find
_
m
i
n(i,0);
8.
Functio
n_find
_
m
i
n(i);
9. Functio
n_b
uil
d_m
olecul
e(i
+
1);
10. if(Dist
>
Opt_
dist+Thr_di
st)
11.
Functio
n_m
o
v
e_
clo
s
e
(
i+1);
12.
elseif(Di
s
t < Opt_di
st-Th
r
_
d
ist)
13.
Functio
n_m
o
v
e_fa
r(i
+1
);
14.
endif
15. endif
16. endfor
5. Simulation
Analy
s
is
We
simul
a
te
NHOS by
MATLAB, an
d assu
me
th
at the monit
o
ring
area
H is
squa
re
who
s
e
side le
ngth is a
=
30
0
0
m. The aim i
s
to obtai
n p
-
reliability cove
rage in th
e monitorin
g
area
,
and p
=
0.85.
The effect of target di
strib
u
tion and
oth
e
r
environ
menta
l
factors are n
o
t con
s
ide
r
ed
.
We a
s
sum
e
that two type
s sen
s
o
r
s a
r
e
use
d
in
unde
rwate
r
sen
s
o
r
net
works: p
a
ssive
son
a
r
se
nsor and the
gia
n
t
magneti
c
re
sista
n
ce se
n
s
or. Fo
r pa
ssi
v
e son
a
r th
e
effective ra
di
u
s
of perception
is
50
effic
R
m
unde
r the premi
s
e of probability cove
rage.
Acco
rdi
ng to the cha
r
a
c
ter of giant magnet
o re
sista
n
c
e sen
s
or a
n
d
magneti
c
propertie
s
of the target
node, the
p
e
rception
ra
n
ge of gi
ant m
agneto
re
si
stance sen
s
o
r
is irreg
u
la
r, a
n
d
can
be
assu
med to
be
elliptic, elliptical
long
ra
diu
s
60
a
R
m
and
sho
r
t rad
i
us
30
b
R
m
a un
de
r
the condition
of meetin
g t
he p
r
ob
abilit
y cover
age.
The n
u
mb
ers of two
kin
d
s
of
se
nsor
are
deploye
d
by ratio of 1:3.
Ran
dom de
ployment algorithm
and based on the ‘virtual forc
e’ d
eployment al
gorithm
simulatio
n
re
spe
c
tively.
These
two kin
d
s of
al
g
o
rit
h
m si
mulatio
n
re
sult
s
co
mpared
with
the
perfo
rman
ce
of the de
ploy
ment alg
o
rith
m ba
sed
NHOS. The hi
g
her
of the n
e
t
work
covera
g
e
rate and th
e greate
r
effective detection rate
of the target node sho
w
s that the b
e
tter
perfo
rman
ce
of network m
onitorin
g
.
The re
sult
s are respe
c
tively sho
w
n in Fig
u
re 2, Figu
re
3 and Figu
re
4.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 16, No. 3, Dece
mb
er 201
5 : 574 – 582
580
Figure 2.
Rel
a
tions of effe
ctive coverag
e
and pe
rceive node
s num
ber
Figure 3. Rel
a
tions of effe
ctive det
ectio
n
rate and ta
rget node
s nu
mber
Figure 4. Rel
a
tions of net
work resi
dual
energy
Figure 2 is
sho
w
n that
the cha
nge
of t
he effective covera
ge
rate (p-pro
bability
coverage
) in
the same te
st area with th
e increa
se of
the numbe
r of sen
s
or n
o
de by differe
n
t
algorith
m
s of
the
RDM,
VFA and
NHO
S
algo
rithm.
Comp
ared
wi
th RDM al
go
rithm a
nd VF
A,
NHOS algo
rithm ha
s more effective coverag
e
and
single n
ode'
s pe
rception
is more effici
ent.
NHOS
can
redu
ce th
e
pe
rce
p
tion
sco
p
e
of
se
nsor
node
ove
r
lap
p
ing
and
the
pe
rceived
bl
ind
area
by usin
g DSmT d
a
ta fusion
alg
o
rithm.
The
r
efore, NHOS
algorithm
o
u
tperfo
rms V
F
A
algorith
m
and
RDM alg
o
rith
m in terms of
coverage effi
cien
cy.
Figure 3 is
shown the rel
a
tionship bet
wee
n
e
ffective detectio
n
rate and the n
u
mbe
r
of
target nod
es.
As can b
e
see
n
from th
e figure 3
th
at effective detection
rate
of the target
is
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Nod
e
s
Deplo
y
m
ent Schem
e of Hetero
ge
neou
s Wi
rele
ss Sen
s
o
r
Ne
twork… (Ju
w
ei Zhang
)
581
highe
r in NHOS than in RDM an
d VFA. The rea
s
o
n
is that NHO
S
adopted
DSmT data fu
sion
algorithm
whi
c
h i
s
more
effective and
can im
prov
e the detection
probabilit
y of
the target node,
increa
se the
coverage effi
cien
cy.
Figure 4 is
shown that the
cha
nge of th
e node
re
sid
ual ene
rgy of
sen
s
o
r
net
works
with
the time go
n
e
at the
sam
e
dete
c
tion
a
r
ea
and th
e
same
num
be
r (5
00
gro
u
p
s
), a
nd the
same
coverage p
e
rforman
c
e, pe
rce
p
tion abilit
y of node.
As can be
seen
from the figure 4 that NHOS
co
st mo
re
en
ergy at
begi
n
n
ing
be
cau
s
e
of the
node
s’ move
ment.
VFA algo
rith
m and
NHO
S
co
st less en
ergy pe
r unit time than RDM algo
rith
m due to the redu
nda
ncy node dorm
ancy
mech
ani
sm o
f
VFA and
NHOS. NHOS
algorith
m
con
s
ume
s
l
e
ss e
nergy th
an th
e VFA algo
rit
h
m.
Therefore, NHOS algo
rith
m survive
s
m
o
re
time com
pare
d
with ot
her two al
go
ri
thms.
It can be se
e
n
that NHOS
can effective
l
y r
educe the
numbe
r of sensor no
de
s use
d
in
the heteroge
neou
s sen
s
o
r
network wit
h
the sam
e
perceptio
n a
b
ility, improve the perce
p
t
ion
efficien
cy of
node, re
du
ce
t
he perce
ption of blind area, and
expa
nd the detect
i
on ran
ge ba
sed
on the org
ani
c sm
all mole
cular mo
del of hetero
gen
eo
us sen
s
or n
e
twork.
6.
Conclu
sion and Futu
r
e
Work
In this paper,
a heteroge
n
eou
s sen
s
o
r
netwo
rk n
o
d
e
deployme
n
t
algorithm NHOS is
pre
s
ente
d
ba
sed o
n
the organi
c sm
all molecula
r mo
del. DSmT d
a
ta fusion i
s
use
d
in NHO
S
to
redu
ce th
e d
eployment n
o
de num
ber, i
m
prove th
e
e
fficiency of n
e
twork
cove
rage. We de
scrib
e
the detail
s
of
our
algo
rith
m and
co
mp
are it
with ot
her
pro
p
o
s
ed
schem
es.
Compa
r
ison a
nd
theory a
nalysis
sho
w
th
at the p
r
op
osed
scheme
outp
e
rform
s
mo
st of the
existin
g
sch
e
me
s. T
h
e
further
wo
rk i
s
to improve,
optimize the
model in
thi
s
pap
er a
nd
try to other fusio
n
strateg
y
to
optimize the
hetero
gen
eo
us sen
s
or n
e
twork no
de de
ployment alg
o
rithm ma
ke i
t
more efficie
n
t.
Ackn
o
w
l
e
dg
ements
The fina
ncial
sup
port
s
fro
m
the Nation
al Natu
ral S
c
i
ence Fo
unda
tion of Chi
n
a
(NSF
C)
unde
r the
grant No.6
090
4023,
No.61
0400
10, No
.6110
1167
a
nd the A
e
ro
nautical Sci
e
nce
Found
ation
of Chin
a (A
SFC)
und
er the
grant
No.20
110
1
4200
3, No.2
0115
1420
05
are
ackno
w
le
dge
d gratefully. The auth
o
rs
also g
r
atefull
y
ackn
owl
e
d
ge the hel
pfu
l
comme
nts
and
sug
g
e
s
tion
s of the editors
and revie
w
e
r
s,
whi
c
h have
improved the
presentation.
Referen
ces
[1]
David
More
no
-Salin
as, Anto
nio P
a
sco
al,
Joaq
uin
Aran
d
a
. Sens
or Ne
t
w
orks f
o
r Op
timal T
a
rget
Loca
lizati
o
n
w
i
t
h
B
eari
ngs
-Onl
y Me
asur
ements
in
C
onstrai
ned
T
h
ree-Dim
ensi
o
n
a
l Sc
enar
ios.
Sensors.
20
13;
13: 103
86-1
0
4
17.
[2]
Marjan
Mor
adi
, Javad
R
e
za
zade
h, Ab
dul
Sama
d Isma
il. A
Revers
e
Loc
aliz
atio
n
Scheme
fo
r
Und
e
r
w
ater Ac
oustic Sens
or Net
w
orks.
Sensors.
2012; 1
2
: 4352-
43
80.
[3]
Achint Ag
gar
w
a
l, F
r
ank Kirc
h
ner. Object R
e
cogn
it
ion an
d Loca
lizati
on:
T
he
R
o
le
of T
a
ctile Se
nsors.
Sensors.
20
14;
14: 322
7-32
66
.
[4]
Luo
Xu,
C
hai
Li,
Ya
ng
J
un. Multi-ob
jectiv
e Strate
g
y
of M
u
ltiple
C
o
vera
ge
in
Heter
oge
ne
ous S
enso
r
Net
w
orks.
Jour
nal of Electro
n
i
cs and Infor
m
a
t
ion T
e
chn
o
l
o
g
y
.
2014; 36(
3): 692-
695.
[5]
Du
Xia
o
-
y
u, S
un Li-
j
ua
n, Guo Jian, Ha
n Ch
ong.
Cov
e
ra
ge
Optimization
Algorit
hm for Heterog
e
n
e
o
u
s
WSNs.
Journal
of Electronics
and Infor
m
ati
o
n T
e
chno
lo
gy.
201
4; 36(3): 69
6-70
2.
[6]
Hua
ng S
h
u
a
i,
Che
ng
Lia
ng-l
un. A L
o
w
Re
dun
da
nc
y Cov
e
rag
e
-en
h
a
n
ci
ng Al
gor
ithm for Dir
ection
a
l
Sensor
Net
w
or
k Based
on
F
i
ctitious F
o
rce.
Chin
ese J
our
n
a
l of S
ensors
and Act
uators
.
201
1; 24(
3)
:
418-
422.
[7]
Hon
g
Z
hen, Yu
Li, Z
hang Gui-
jun. Efficient a
nd
D
y
n
a
mic Cl
usterin
g
Sche
me for Heterog
ene
ous Multi-
level Wireless
Sensor Net
w
or
ks.
Acta Automatica Sinica.
20
13; 39(4): 4
54-
464.
[8]
Li Mi
ng. Stud
y
on C
o
ver
a
g
e
Alg
o
rithms
for
Hetero
ge
n
eous W
i
r
e
less
Sensor
Net
w
orks. Ph.D
.
dissertati
on. C
hon
gqi
ng U
n
iv
ersit
y
; 2
011.
[9]
Kumar D, Aseri T
C
, Patel RB. EEHC: Energy
Effi
cie
n
t Hete
roge
neo
us C
l
u
s
tered Sch
e
me
for W
i
reles
s
Sensor N
e
t
w
or
ks.
Comp
uter Co
mmun
icati
o
ns.
2009; 3
2
(4)
:
662-66
7.
[10]
Seng
upta S,
Das S, Nasir
MD,
et al. Mul
t
i-obj
ective N
o
de De
pl
o
y
men
t
in W
S
Ns: in Search
of an
Optimal T
r
ade
-off amon
g
Co
verag
e
, L
i
fetim
e
, Ener
g
y
Con
s
umptio
n, a
n
d
Co
nnectiv
i
t
y
.
Engi
neer
in
g
Appl
icatio
ns of Artificial Intel
lig
ence.
20
13; 26
(1): 405-4
16.
[11]
Bara’a A Attea, Enan A
Khal
il. A Ne
w
Ev
oluti
o
n
a
r
y
Bas
ed R
o
u
t
ing Protoc
ol
for Clustere
d
Hetero
gen
eo
u
s
W
i
reless Sen
s
or Net
w
orks.
Appl
oed S
o
ft Computi
n
g
. 20
1
2
; 12: 195
0-19
57.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 16, No. 3, Dece
mb
er 201
5 : 574 – 582
582
[12]
Hon
g
Z
hen, Yu
Li, Z
hang Gui-
jun. Efficient a
nd
D
y
n
a
mic Cl
usterin
g
Sche
me for Heterog
ene
ous Multi-
level Wireless
Sensor Net
w
or
ks.
Acta Automatica Sinica
. 20
13; 39(4): 4
54-
460.
[13]
Subir
Hal
der,
Sipra
Das B
i
t. Enha
ncem
ent
of W
i
reless
Sensor
Net
w
o
r
k Lifetime
b
y
Dep
l
o
y
i
n
g
Hetero
gen
eo
u
s
Nodes.
Jour
n
a
l of Netw
ork and Co
mputer A
pplic
atio
n.
201
4; 38: 106-
124.
[14]
Che
n
Ji
e, Du
Qing-
w
e
i,
Li
Xia
o
-
y
u, D
i
n
g
F
e
n
g
. Res
earch
on
the
De
plo
y
m
ent
Algorit
hm
o
f
Hetero
gen
eo
u
s
Sens
or N
e
t
w
orks Bas
ed
on
Proba
bil
i
t
y
Mo
del.
J
ourn
a
l of Chin
ese
Co
mp
uter
Syste
m
s.
201
2; 33(1): 50
-53.
[15]
Card
ei M, T
hai MT
, Ying-shu L, et al.
Energ
y
-efficient T
a
rg
et Covera
ge i
n
W
i
reless Sens
or Netw
orks
.
24th An
nu
al J
o
int Co
nfere
n
c
e
of the IEEE
Co
mput
er an
d Commu
nic
a
tions S
o
cieti
e
s
(INF
OCOM
200
5). Miami. 200
5: 197
6-19
84.
[16]
Kashi
SS, Sh
arifi M. C
o
ver
age
Rat
e
C
a
lc
ulati
on
in
W
i
re
less S
ensor
N
e
t
w
o
r
ks.
C
o
mputin
g.
2
012
;
94(1
1
): 833-
85
6.
[17]
Dezert J. F
o
u
ndati
ons
of a
Ne
w
T
heor
y
o
f
Plausi
b
l
e
a
n
d
Par
ado
xical
Reas
oni
ng.
In
fo
rma
ti
on
and
Security Journal.
200
2; 13(9):
90-95.
[18]
Dezert J, T
c
hamov
A, Smara
ndac
he F
,
et a
l
.
T
a
rget T
y
pe
T
r
acking w
i
th
PCR5
and
De
mpster
’
s
R
u
les
:
A Co
mp
arative
Analys
is
. Proc
eed
ings
of F
u
s
i
on
200
6 Intern
ation
a
l co
nfere
n
ce o
n
Informa
tion F
u
si
on
.
F
i
renze, Ital
y
.
200
6.
[19]
Smaran
dach
e
F
,
Dezert J.
Informatio
n
F
u
sion Bas
ed o
n
New
Proporti
ona
l Confl
i
ct Redistri
buti
o
n
Rules.
Proc
ee
din
g
s of F
u
sion
2005 C
onfer
e
n
ce, Phil
ad
elp
h
ia, 20
05.
[20]
Smaran
dach
e
F
,
Dezert J. A
pplic
atio
ns
an
d Adv
anc
es of
DSmT
for Informatio
n
F
u
si
o
n
. Re
hob
oth:
America
n
Res
earch Press. 2
006.
Evaluation Warning : The document was created with Spire.PDF for Python.