Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
13
,
No.
1
,
Jan
uar
y
201
9
,
pp.
249~
257
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
3
.i
1
.pp
249
-
257
249
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
A
ne
w look
a
t ene
rgy harv
esting ha
lf
-
du
plex
D
F
p
ower spli
tti
ng
proto
co
l
re
l
ay netwo
rk
o
ver ri
cian chann
el
i
n case
o
f
maximi
zin
g capacity
Phu
Tr
an
Ti
n
1
, M
in
h Tra
n
2
, Ta
n
N.
Nguye
n
3
,
Th
an
h
-
L
ong
Nguy
e
n
4
1
Facul
t
y
of Elect
ronic
s T
ec
hno
lo
g
y
,
Industri
al Unive
rsit
y
of
Ho C
hi
Minh
C
ity
,
H
o
Chi
Minh
Cit
y,
Vie
tna
m
2
Optoel
e
ct
roni
cs
Resea
r
ch
Group
,
Facu
lty
of Elec
tri
c
al
and El
e
ct
r
onic
s E
ng
ine
e
rin
g,
Ton
Duc Thang
Univer
sit
y
,
Ho
Chi
Minh C
ity
,
Vi
et
nam
3
W
ire
le
ss
Com
muni
cations
Rese
a
rch
Group,
Facult
y
of El
ec
tr
ic
a
l a
nd
Elec
tron
ic
s
E
ngine
er
ing,
Ton
Duc Thang
Univer
sit
y
,
Ho
Chi
Minh C
ity
,
Viet
nam
4
Cent
er
for
Infor
m
at
ion
T
ec
hnolo
g
y
,
Ho Chi
Min
h
Cit
y
Un
ive
rsit
y
of
Food Indust
r
y
,
Ho Ch
i
Minh
Cit
y
,
Vi
et
n
am
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
A
ug
2
9
, 201
8
Re
vised N
ov 1
9
, 2
018
Accepte
d
Oct
1
, 2
018
In
thi
s
pape
r
,
th
e
s
y
st
em
per
for
m
anc
e
in
t
erm
of
the
e
rgodic
c
apa
c
ity
of
a
hal
f
-
dupl
ex
decode
-
and
-
forward
rel
a
y
ing
ne
twork
over
Ric
ian
Fading
Channe
l
is
inves
ti
gat
ed
.
The
p
ower
split
ti
ng
p
rotoc
ol
is
proposed
for
the
s
y
stem
m
odel
.
For
thi
s
purpose,
the
anal
y
tical
m
at
hemat
ic
a
l
expr
essions
of
the
erg
odi
c
ca
p
a
ci
t
y
in
c
ase
s
of
m
axi
m
iz
e
and
no
-
m
axi
m
iz
e
e
rg
odic
ca
p
ac
i
t
y
are
d
eri
ved
an
d
discussed.
F
urthe
rm
ore
,
the
eff
e
ct
of
var
i
ous
s
y
stem
par
amete
rs
on
t
he
s
y
st
em
per
fo
rm
anc
e
is
rigor
ousl
y
studi
ed.
Final
l
y
,
the
ana
l
y
t
ical
r
esults
are
al
so
de
m
onstrat
ed
b
y
Monte
-
Carl
o
si
m
ula
ti
on
i
n
compari
son
wit
h
the
anal
y
ti
c
al
expr
essions.
Th
e
rese
arc
h
results
show
that
the
anal
y
tical
m
at
hemati
c
al
a
nd
sim
ula
te
d
result
s
m
at
ch
for
al
l
poss
ibl
e
par
amete
r
va
lue
s
for
both
sche
m
e
s.
Ke
yw
or
ds:
Decode
-
an
d
-
f
orwa
rd (DF
)
Energy
harvest
ing
(E
H)
Ergo
dic cap
aci
ty
Power spli
tt
ing p
ro
t
oco
l
Re
la
y network
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed.
Corres
pond
in
g
Aut
h
or
:
Tan N.
Ng
uyen,
W
i
reless C
omm
un
ic
at
ion
s R
esearch
Gr
oup,
Faculty
of Elec
tric
al
an
d El
ect
ronics E
nginee
rin
g,
To
n Du
c
Th
a
ng
Un
i
ver
sit
y,
Ho Chi Mi
nh
Ci
ty
, V
ie
tnam
.
Em
a
il
: ng
uyen
nh
at
ta
n@t
dt
u.
e
du.vn
1.
INTROD
U
CTION
I
n
t
he
la
st
de
cades,
the
pro
li
fer
at
ion
of
c
el
lular
net
wor
ks
a
nd
wi
reless
sen
sor
networks,
ha
ve
dem
and
ed
the
higher
qual
it
y
of
po
wer
s
uppl
y
fr
om
the
so
ur
ce
node
to
wi
reless
dev
ic
es
.
In
the
co
nv
e
nt
ion
al
m
et
ho
d,
wirele
ss
de
vices
are
gen
e
rall
y
pow
ered
by
batte
ri
es,
w
hich
ha
ve
to
be
r
eplace
d/rec
harged
on
tim
e
m
anu
al
ly
.
In
the
com
m
un
ic
at
ion
net
wor
k,
the
proces
s
of
batte
ry
re
pl
ace
m
ent
can
be
li
nk
e
d
to
so
m
e
disad
va
ntages
su
c
h
as
inc
onve
nient,
in
feasib
le
fo
r
s
om
e
app
li
cat
ion
s
[
1
-
4]
.
In
the
la
st
de
cade
,
o
ne
of
t
he
best
ways
to
trans
f
er
energy
in
the
com
m
un
ic
at
ion
netw
ork
i
s
us
in
g
RF
sign
al
s
.
In
the
l
ast
tim
e,
the
s
yst
e
m
perform
ance
of
the
e
nergy
harvesti
ng
rela
y
networ
k
has
been
c
onside
red
i
n
m
any
s
tud
ie
s
.
Su
c
h
a
s,
[
5
]
inv
est
igate
d
th
e
fu
ll
-
duple
x
e
nergy
ha
rvest
ing
relay
netw
ork
with
the
s
i
m
ultaneou
s
e
nergy
ha
rvest
ing
a
nd
inf
or
m
at
ion
tr
ansm
issi
on
.
M
or
e
over
,
the
de
velo
pm
ent
of
co
op
e
rati
ve
prot
oco
ls
f
or
en
erg
y
harvesti
ng
relay
netw
ork
is
de
eply
stud
ie
d
i
n
[
6
-
7
]
.
Furth
erm
or
e
,
[
8
-
9
]
pro
po
se
d
a
“h
arv
est
-
the
n
-
tra
ns
m
it
”
pr
oto
c
ol
f
or
a
m
ul
ti
-
us
er
rela
y
netw
ork.
I
n
al
l
pap
e
rs
a
bo
ve,
t
he
e
nergy
so
urce
of
the
whole
netw
ork
is
only
the
source
or
the
acce
ss
point.
I
n
t
he
tre
nd
s
to
im
pr
ovin
g
the
e
nergy
ha
r
vestin
g
a
nd
i
nfor
m
at
ion
tran
s
m
issi
on
pro
ces
ses
in
the
wireless
re
la
y
netwo
r
k,
s
om
e
research
e
rs
pro
posed
th
e
idea
of
de
pl
oying
ded
ic
at
e
d
powe
r
beac
on
node
(P
B).
In
the
wireless
relay
network
with
us
i
ng
the
PB
node
,
the
destinat
io
n
(D)
can
ha
r
ve
st
wireless
en
erg
y
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
249
–
257
250
no
t
only
from
the
so
urce
(S)
node
bu
t
al
so
from
the
dep
loye
d
PB
node
[
10
-
1
1
]
.
For
this
new
m
od
el
,
th
e
qu
e
sti
on
a
bout
syst
e
m
per
f
orm
ance
is
sti
ll
op
en
an
d
is
nec
essary
to
i
nv
es
ti
gate.
H
ow
e
ve
r,
the
re
a
re
no
t
m
any
works
c
on
ce
nt
rate
on
t
h
e
Ri
ci
an
fa
ding
c
ha
nn
el
ye
t,
a
nd
t
he
ou
ta
ge
perf
or
m
ance
analy
sis
of
s
uch
cha
nn
el
s
is
fun
dam
ental
to
pract
ic
e. T
he
r
e
m
ai
nin
g ga
p
c
an fil
l by this
work.
In
t
his
w
ork
,
we
propose
a
nd
dem
on
strat
e
the
syst
em
perform
ance
of
the
EH
DF
HF
relay
ing
netw
ork
in
the
c
ase
of
m
axim
iz
ing
an
d
non
-
m
axi
m
i
zi
ng
ergo
dic
capaci
ty
.
Fo
r
detai
ls
on
this
a
naly
sis,
the
energy
,
a
nd
i
nfo
rm
ation
are
t
ran
s
ferre
d
f
rom
th
e
so
urce
t
o
the
relay
nodes
by
the
po
wer
s
plit
ti
ng
prot
oco
l
,
and
al
l
chann
el
s
are
con
sid
ered
as
the
Ri
ci
an
fad
i
ng
ch
ann
el
s
.
The
m
ai
n
con
tri
buti
ons
of
the
paper
are
su
m
m
arized as foll
ow
s:
1)
T
he
syst
e
m
m
od
el
of
the
ha
lf
-
du
plex
rela
yi
ng
netw
ork
ov
e
r
the
Ri
ci
an
fa
ding
cha
nnel
s
is
pr
opos
e
d
with
the PS
pr
oto
c
ol
.
2)
T
he
m
at
hem
at
ic
al
exp
res
sion
s
of
t
he
e
rgo
dic
capaci
ty
in
cases
m
a
xim
iz
e
,
and
non
-
m
axi
m
iz
e
of
the
syst
e
m
m
od
el
are
der
ive
d.
3)
Th
e
in
flue
nc
e
of
th
e
m
ai
n
pa
ram
et
ers
on
the
syst
em
per
f
or
m
ance
is
dem
on
strat
ed
e
nt
irel
y
by
the
Mon
te
Ca
rlo
sim
ulatio
n
.
The
st
ru
ct
ur
e
of
this
pa
per
is
pro
pose
d
as
fo
ll
o
ws.
Sect
i
on
s
I
I
present
s
the
syst
e
m
m
od
el
of
t
he
relay
ing
netw
ork
.
Sect
io
ns
I
I
I
der
ive
the
a
naly
ti
cal
exp
re
ssion
of
e
rgo
di
c
capaci
ty
in
cases
m
axi
m
ize
an
d
non
-
m
axi
m
iz
e
of
the
m
od
el
syst
e
m
.
Sect
ion
IV
pro
vid
e
s
th
e
nu
m
erical
resu
lt
s
and
s
om
e
discuss
i
on
s
.
F
inall
y,
Sect
ion
V
c
onc
lud
es
the
pa
per.
2.
SY
STE
M M
O
DEL
In
this
pa
per
,
t
he
syst
em
m
od
el
is
the
e
nerg
y
harvesti
ng
ha
lf
-
du
plex
(
HF)
decode
-
an
d
-
f
orward
(
DF
)
powe
r
sp
li
tt
ing
(PS)
protoc
ol
relay
ing
ne
twork
,
a
s
sho
wn
in
Fig
ure
1
.
I
n
this
m
od
el
,
the
inf
or
m
at
ion
is
trans
ferred
f
rom
the
sou
rce
(
S)
t
o
the
destin
at
ion
(D),
thr
ough
e
ne
rg
y
c
onstrai
ne
d
inter
m
ediat
e
relay
(
R).
T
he
energy ha
rv
e
sti
ng
a
nd in
form
a
ti
on
pr
ocessin
g
of
t
he
syst
em
m
od
el
with
PS
protoc
ol
are
pro
posed
i
n
Fig
u
re
2.
In
t
his
sc
hem
e,
T
is
the
blo
c
k
tim
e
in
wh
ic
h
the
s
o
urce
fu
ll
y
transm
it
the
i
nfor
m
at
ion
dat
a
to
the
destina
ti
on.
In
the
first
interval
tim
e
(T/2)
,
the
relay
ha
rv
est
s
e
nergy
(ρT)
an
d
re
c
e
ives
inf
or
m
at
i
on
((1
-
ρ
)T)
from
the
so
urce
sig
nal
,
wh
e
re
ρ
is
the
powe
r
sp
li
tt
ing
factor
ρ
∈
(0,
1)
.
I
n
the
rem
ai
nin
g
half
-
ti
m
e
T/2,
the
relay
node
trans
fer
s
i
nfo
r
m
at
ion
to
the
de
sti
nation
node
.
All
the
fa
ding
cha
nn
el
s
f
ro
m
S
to
R
an
d
R
t
o
D
are
pr
opose
d
as
the
Ri
ci
an
fad
i
ng
cha
nnel
s.
More
detai
ls
of
the
analy
ti
cal
m
a
them
a
ti
cal
m
od
el
of
t
he
e
rgo
dic
capaci
t
y
of
th
e
syst
e
m
m
od
el
is
prese
nted
a
nd a
naly
zed
i
n
t
he follo
wing s
ect
ion
s
[
12
-
16
]
, [
18
]
.
S
ou
r
c
e
De
s
t
i
n
at
i
on
Re
l
ay
h
g
1
d
2
d
I
nf
or
m
a
ti
on tr
a
ns
m
is
s
ion
E
ne
r
gy
ha
r
ve
s
t
i
ng
(
EH
)
Figure
1. Syst
em
m
od
el
EH
at
R
(
ρ
T
)
R
to
D
T
/
2
S
to
R
((
1
-
ρ
)
T
)
T
/
2
T
Figure
2. The
powe
r
s
plit
ti
ng
pro
t
oco
l
3.
THE
SYSTE
M
PE
RFO
R
MAN
CE
In
this
s
ect
ion,
we
analy
zed
a
nd
de
m
on
strat
ed
t
he
syst
em
perform
ance
a
naly
sis
of
t
he
m
od
el
syst
e
m
in
the
po
wer
s
plit
ti
ng
protoc
ol
[
1
2
-
1
4
].
I
n
the
first
tra
ns
m
i
ssion
phase
ba
sed
on
t
he
a
bove
syst
em
m
od
el
,
th
e
receive
d
sig
nal
at the
re
la
y ca
n be
giv
e
n by (
1)
:
1
1
(
1
)
r
s
r
m
y
h
x
n
d
(1)
wh
e
re
h
is
t
he source t
o rela
y chan
nel
gain
, d
1
is t
he
s
ource t
o rela
y dist
anc
e, and
m
is
the
path
l
os
s e
xponent.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
A
new l
oo
k
a
t
ener
gy ha
rve
sti
ng
half
-
du
plex
D
F
powe
r s
plit
ti
ng
prot
oco
l r
el
ay
…
(
Ph
u
Tr
an Tin
)
251
Her
e
,
s
x
is
the
t
ran
sm
it
te
d
sign
al
at
the
so
urce,
n
r
is
the
add
it
ive
w
hite
Gau
s
sia
n
nois
e
(AWGN)
with
var
ia
nce
N
0
a
nd
01
is
powe
r
s
plit
ti
ng
rati
o
at
t
he
relay
.
M
or
e
over
,
2
ss
xP
,
:
exp
ect
at
io
n
op
erator,
a
nd P
s
i
s av
e
rag
e
tra
nsm
it
p
ow
e
r
at
th
e sour
ce
.
22
1
1
1
(
/
2
)
(
/
2
)
(
/
2
)
ss
h
r
m
m
m
P
h
T
P
h
E
P
d
T
d
T
d
(2)
In (2)
,
01
is
an
e
ne
rg
y
c
onve
rsion e
ff
ic
ie
ncy c
oe
ff
ic
ie
nt.
In
t
his
syst
em
m
od
el
,
we
pro
po
s
ed
t
hat
the
m
od
el
is
work
i
ng
in
t
he
de
co
de
-
a
nd
-
forw
a
r
d
m
od
e.
T
he
n
the sig
nal to
no
ise
r
at
io
(SNR)
at the in
f
or
m
ation
proces
sin
g ph
a
se
of
t
he
relay
can
be
cal
c
ul
at
ed
by:
2
,
10
(
1
)
s
sr
m
Ph
SNR
dN
(3)
Fr
om
the e
qu
at
ion
(
3)
,
the ca
pa
ci
ty
o
f
S
-
R l
in
k
ca
n
be
cal
c
ulate
d
as:
2
,
2
,
2
10
(
1
)
11
l
o
g
(
1
)
l
o
g
1
22
s
s
r
s
r
m
Ph
C
S
NR
dN
(4)
In the sam
e w
a
y, the
receive
d si
gn
al
at t
he de
sti
nation
i
n
the
seco
nd
ph
ase
can
be
e
xpress
ed
as:
2
1
d
r
d
m
y
g
x
n
d
(5)
Wh
e
re
g
is
t
he
so
urce
to r
el
ay
channel
gai
n,
d
2
is
the r
el
ay
to
destinat
io
n
distance,
a
nd
2
rr
xP
By
co
m
bin
ing
with the
equati
on (2), the
S
N
R of the
destin
at
ion
ca
n
be
e
xpress
ed
as:
2
2
2
,
2
0
1
2
0
rs
rd
m
m
m
P
g
P
h
g
SNR
d
N
d
d
N
(
6)
The
ca
pacit
y o
f
R
-
D
li
nk ca
n
be
cal
culat
e
d
a
s:
22
,2
1
2
0
1
l
og
1
2
s
rd
mm
P
h
g
C
d
d
N
(
7)
In
this
a
naly
sis,
we
c
onsider
h,
g
gain
fac
tor
a
s
Ri
ci
an
fad
i
ng
cha
nnel
,
a
nd
h,g
ha
ve
a
r
an
dom
distrib
ution.
From
that,
the
prob
a
bili
ty
den
si
ty
functi
on
(
P
DF
)
of
a
ra
ndom
var
ia
ble
(R
V)
i
wh
e
re
i=
1
,
2
is
form
ulate
d
as in
[1
4
].
Wh
e
re
22
12
,
hg
2
0
()
()
(
!
)
i
l
l
bx
l
bK
f
x
a
x
e
l
(8)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
249
–
257
252
Wh
e
re
we de
note
(
1
)
1
,
K
ii
K
e
K
ab
In
(
8)
,
i
is
the
m
ean
value
of
RV
i
w
hich
i
=1,2
re
sp
ect
iv
el
y
and
22
12
12
,
mm
hg
dd
.
More
ov
e
r,
K
is
the
Ri
ci
an
K
-
f
act
or
de
fine
d
a
s
the
rati
o
of
t
he
powe
r
of
th
e
li
ne
-
of
-
sig
ht
(LOS)
com
pone
nt
to
the
scat
te
re
d
c
om
po
ne
nts a
nd
0
I
is t
he
ze
ro
-
th
order m
od
ifie
d B
essel
fun
ct
io
n
of the
first
kin
d.
Her
e
,
the
cum
ulati
ve
de
ns
it
y
functi
on(CD
F
)
of
R
V
i
wh
e
r
e
i=
1,
2
ca
n
be
com
pu
te
d
as
in
[1
4
]
.
W
e
assum
e that
12
, t
hen w
e
h
a
ve:
00
0
(
)
(
)
1
!!
ii
ln
l
nb
ln
a
K
b
F
f
x
d
x
e
b
l
n
(8)
3.1.
M
ax
im
iz
e capacit
y C
as
e
Fo
r
dec
ode
-
a
nd
-
forwar
d
rela
yi
ng
the
ov
e
ra
ll
Erg
odic
cap
aci
ty
and
SN
R
from
so
ur
ce
t
o
desti
nation
can
be writt
en a
s foll
ows:
2
12
,
,
2
2
1
0
1
2
0
(
1
)
11
m
i
n
,
m
i
n
l
o
g
1
,
l
o
g
1
22
s
s
D
F
s
r
r
d
m
m
m
Ph
P
C
C
C
d
N
d
d
N
(
9)
2
2
2
,,
1
0
1
2
0
(
1
)
m
in(
,
)
m
in
,
ss
DF
s
r
r
d
m
m
m
P
h
P
h
g
SNR
SNR
SNR
d
N
d
d
N
(
10)
In
e
quat
io
ns
(
9)
an
d
(10
),
we
can
ob
ta
in
the
fo
ll
owin
g
valu
e
of
ρ
m
axi
m
i
zes
the
DF
C
and
the
sam
e
for
DF
S
N
R
as the
foll
owin
g:
*
2
2
1
1
DF
m
d
(
11)
Pr
oo
f: See t
he Appe
nd
i
x A.
3.2.
N
on
-
ma
xi
mi
z
e cap
acit
y case
In the case
no
n
-
m
axi
m
iz
e capacit
y, we
ca
n
c
al
culat
e the
er
godic ca
pacit
y a
s:
,
,
0
1
(
)
1
l
n
2
1
sr
S
N
R
sr
F
Cd
(12)
,
,
0
1
(
)
1
l
n
2
1
rd
S
N
R
rd
F
Cd
(13)
Fr
om
these e
quat
ions,
we ha
ve:
,
1
2
,
10
11
1
00
(
1
)
(
)
Pr
Pr
Pr
(
1
)
(
1
)
sr
s
S
NR
s
r
m
mm
Ph
F
S
NR
dN
dd
F
(14)
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
A
new l
oo
k
a
t
ener
gy ha
rve
sti
ng
half
-
du
plex
D
F
powe
r s
plit
ti
ng
prot
oco
l r
el
ay
…
(
Ph
u
Tr
an Tin
)
253
Fr
om
the e
qu
at
ion
(
8)
,
w
e
h
a
ve
:
1
0
1
(
1
)
11
00
00
1
(
1
)
!
!
(
1
)
m
n
bd
mm
ln
l
ln
dd
a
K
b
Fe
b
l
n
(15)
,
22
12
,1
02
1
2
0
(
)
P
r
P
r
P
r
rd
mm
s
S
NR
r
d
mm
P
h
g
dd
F
SNR
d
d
N
(16)
,
1
2
12
2
2
2
02
0
(
)
|
(
)
rd
mm
S
N
R
dd
F
F
f
d
(17)
Si
m
il
arity
,
12
02
2
,
1
2
12
22
2
0
0
0
02
0
(
)
1
!
!
(
!
)
mm
rd
n
b
d
d
mm
l
k
n
k
l
b
k
S
N
R
l
k
n
dd
Kb
F
a
e
e
d
l
n
k
(18)
12
02
2
,
1
2
12
22
2
0
0
0
0
0
(
)
1
!
!
(
!
)
mm
rd
n
b
d
d
mm
l
k
n
k
l
b
kn
S
N
R
l
k
n
dd
Kb
F
a
e
e
d
l
n
k
(19)
Apply [
3.4
71,9]
o
f
ta
ble of i
ntegr
al
[
17
]
, we
hav
e:
,
1
1
2
2
1
2
1
2
1
2
0
0
0
00
(
)
1
2
2
!
!
(
!
)
rd
kn
m
m
m
m
l
k
n
k
l
S
N
R
k
n
l
k
n
d
d
d
d
Kb
F
a
K
b
l
n
k
(20)
Re
place (
15)
i
nto
(
12),
(
20)
i
nto
(
13),
finall
y we
hav
e:
1
0
1
(
1
)
1
,
00
0
0
1
l
n
2
(
1
)
!
!
(
1
)
m
n
bd
m
ln
l
sr
ln
d
a
K
b
C
e
d
ln
(21)
1
21
2
1
2
1
2
,1
2
0
0
0
00
0
21
2
l
n
2
(
1
)
!
!
(
!
)
kn
m
m
m
m
l
k
n
k
l
r
d
k
n
l
k
n
d
d
d
d
a
K
b
C
K
b
d
l
n
k
(22)
,,
m
in
(
,
)
D
F
s
r
r
d
C
C
C
(23)
4.
N
U
MER
IC
A
L RES
ULTS
AND DIS
C
USSION
In
t
his
sect
io
n,
the
Mo
nte
Ca
rlo
sim
ulati
on
is
us
e
d
f
or
va
l
i
d
at
ing
t
he
a
na
ly
ti
cal
exp
ress
ion
i
n
the
above
sect
io
n
.
The
syst
em
perform
ance
in
te
rm
of
ergod
ic
ca
pacit
y
is
analy
zed
a
nd
dem
on
strat
ed
in
connecti
on
wi
th
η
,
d
1
=d
2
,
P
s
/N
0
an
d
K
.
We
c
onside
r
a
netw
ork
wit
h
on
e
s
ource,
one
relay
,
a
nd
one
destinat
io
n,
w
he
re source
-
rela
y and relay
-
des
ti
nation dist
anc
es are
both
n
orm
al
iz
ed
to unit
v
al
ue
.
T
he
ef
fect
of
P
s
/N
0
for
the
powe
r
s
plit
ti
ng
protoc
ol
on
th
e
ergo
dic
capa
ci
ty
of
the
pro
po
s
ed
relay
netw
ork
syst
e
m
is
pr
ese
nted
in
Fig
ure
3
.
Fig
ur
e
3
plot
s
the
er
godic
capaci
ty
of
the
syst
em
m
o
del
f
or
m
axi
m
iz
e
case
a
nd f
or
a
non
-
m
axi
m
iz
e
case
with
t
he
powe
r
s
plit
ti
ng
fact
or o
f
0.2
a
nd 0
.
5.
I
n
Fi
gure
3, we
set
the
syst
e
m
par
a
m
et
ers
as
d
1
=0.65,
d
2
=
0.8
5,
η=
0.
8
an
d
K=2.
From
t
he
res
ults,
t
he
analy
ti
cal
and
the
si
m
ulati
on
res
ults
m
at
ch
fo
r
al
l
possible
values
P
s
/N
0
.
Figure
4
s
ho
ws
that
t
he
e
r
godic
ca
pacit
y
ha
s
a
consi
der
a
bl
e
in
crease
w
hile
P
s
/N
0
increases
f
ro
m
0
to
30
.
M
or
e
over,
th
e
er
godic
capaci
ty
in
m
axi
m
iz
e
ca
se
in
bette
r
tha
n oth
ers
cases
.
Fu
rt
her
m
or
e,
t
he
er
godic
ca
pa
ci
ty
of
the
pr
opos
e
d
syst
em
ver
s
us
d
1
=d
2
i
s
il
lustrate
d
in
the
F
ig
ur
e
4
with
P
s
/
N
0
=2
0
and
K=
2.
F
rom
the
resu
lt
s,
we
show
that
the
ergo
dic
capaci
ty
decr
e
a
se
s
sign
ific
a
ntly
with
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
249
–
257
254
increasin
g
t
he
distances
d
1
=d
2
from
0.
5
to
1.5.
It
can
be
pro
ve
by
t
he
opinio
n
t
hat
the
stren
gth
of
th
e
RF
sign
al
de
creas
es
in
the
di
recti
on
of
inc
reasi
ng
tr
ansm
issi
on
dista
nce
bet
ween
S,
R
an
d
D.
All
the
an
al
yt
ic
al
and
sim
ulati
on
res
ults
are
a
gree
well
with
each
oth
e
r
.
I
n
the
sam
e
way,
the
i
nf
l
uen
ce
of
K
on
the
er
godic
capaci
ty
of
the
syst
e
m
m
od
el
in
bo
t
h
m
axi
m
iz
e
and
non
-
m
axim
iz
e
cases
i
s
propose
d
in
Fig
u
re
5.
In
Fig
u
re
5,
we
set
P
s
/
N
0
=1
0,
ρ=
0.3
a
nd
η
=0.8.
From
the
res
ults,
the
er
godic
capaci
ty
in
al
l
cases
ha
s
sli
gh
t
inc
rease
in
th
e
increasin
g
tre
nd
of
the
K
f
r
om
0
to
4.
Final
ly
,
Fig
u
re
6
il
lustrate
s
the
ef
f
ect
of
η
ver
s
us
ergod
ic
ca
pacit
y
with
the
m
a
in
par
a
m
et
ers
of
P
s
/
N
0
=2
0,
K=2
,
and
η=
0.8
.
T
he
ergo
dic
cap
aci
ty
increase
s
rem
ark
ab
ly
wh
il
e
η
increases
from
0
to
1.
T
he
re
su
lt
s
show
that
al
l
si
m
ula
ti
on
and
a
nal
yti
cal
resu
lt
s
are
m
atch
ed
well
with
each
oth
e
r
.
M
or
e
ov
er
, t
he
er
godic
capaci
ty
in
cas
e m
axi
m
iz
e is bett
er th
a
n
t
he
rem
a
in
ing
case
s
.
Figure
3
.
Er
go
dic capa
ci
ty
v
e
rsu
s
P
s
/N
0
Figure
4
.
Er
go
dic capa
ci
ty
v
e
rsu
s
d
1
=d
2
Figure
5
.
Er
go
dic
capa
ci
ty
v
e
rsu
s
K
Figure
6
.
Er
go
dic capa
ci
ty
v
e
rsu
s
η
5.
CONCL
US
I
O
N
In
this
pa
per
,
we
in
vestigat
e
the
syst
e
m
per
f
or
m
ance
of
t
he
EH
half
-
du
plex
DF
relay
i
ng
netw
ork
ov
e
r
the
Ri
ci
a
n
fa
ding
c
hann
el
in
powe
r
spl
it
t
ing
pr
oto
c
ol
.
For
pe
rfo
rm
a
nce
analy
sis
s
yst
e
m
,
we
der
i
ve
the
analy
ti
cal
exp
r
essions
of
er
godic
ca
p
aci
t
y
for
bot
h
m
axi
m
iz
e
and
non
-
m
axi
m
i
ze
case
s.
Th
e
Mo
nte
Ca
rl
o
si
m
ulati
on
is
us
ed
f
or
va
l
i
d
at
ing
th
e
analy
ti
cal
m
et
ho
d
.
T
he
resu
lt
s
sho
w
that
the
analy
ti
cal
m
at
he
m
a
ti
ca
l
and
si
m
ulate
d
resul
ts
m
a
tc
h
fo
r
a
ll
po
ssible
para
m
et
er
values
for
both
sche
m
es
and
the
m
axim
i
ze
case
pr
ovide
s
the
bette
r
res
ul
ts
in
co
m
par
ison
with
the
non
-
m
axi
m
iz
e
case
of
the
ergo
dic
capaci
ty
.
The
res
ults
cou
l
d
pro
vid
e
t
he pr
ospect
ive
rec
omm
end
at
ion
f
or
the co
m
m
un
ic
at
ion
netw
ork
i
n t
he near
futu
re
.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
A
new l
oo
k
a
t
ener
gy ha
rve
sti
ng
half
-
du
plex
D
F
powe
r s
plit
ti
ng
prot
oco
l r
el
ay
…
(
Ph
u
Tr
an Tin
)
255
APPE
N
D
IX
A.
We ca
n
see
tha
t
DF
C
is m
axi
m
iz
ed
w
hen all
of its ar
gu
m
ent b
ec
om
es equ
al
. H
ence,
t
he value
of
*
DF
can
be
ob
ta
ine
d by e
quat
ing C
s,r
to C
r,d
. Replace
(12
)
int
o
(
11)
we
hav
e:
1
2
0
1
2
2
1
1
2
2
1
2
0
2
1
DF
s
m
m
m
mm
m
P
SN
R
d
d
d
d
d
N
d
(A1)
Wh
e
re
we de
note
0
0
s
P
N
.
The
n
the
Erg
odic
capa
ci
ty
can be calc
ulate
d as:
2
00
1
(
)
1
(
)
l
og
(
1
)
l
n
2
1
DF
DF
SNR
DF
SNR
F
C
f
d
d
(A2)
12
0
1
2
2
1
1
2
2
1
1
2
2
1
1
2
1
2
2
2
0
2
0
2
0
(
)
P
r
P
r
(
)
(
)
P
r
|
(
)
DF
DF
m
m
m
S
N
R
m
m
m
m
m
m
F
S
N
R
d
d
d
d
d
d
d
d
d
F
f
d
(A3)
Wh
e
re
is
a
th
re
sh
ol
d
of the
syst
e
m
.
Fu
rt
her
m
or
e,
fro
m
the equati
on
s
(3),
(4) we
hav
e:
2
2
2
1
1
2
00
02
0
2
1
1
2
22
2
0
02
()
(
)
1
!!
()
()
e
xp
(
!
)
DF
n
m
m
m
ln
l
S
N
R
ln
m
m
m
l
b
l
l
d
d
d
a
K
b
F
b
l
n
b
d
d
d
bK
ed
l
(A4)
2
1
2
12
2
0
0
0
02
0
2
1
1
2
22
02
(
)
1
1
!
!
(
!
)
()
e
xp
DF
n
n
mm
l
k
n
k
l
S
N
R
l
k
n
m
m
m
b
k
dd
Kb
Fa
l
n
k
b
d
d
d
ed
(A5)
In this a
naly
sis, b
y a
pply
ing
t
he
e
qu
at
io
n
0
()
n
n
n
t
t
t
n
x
y
x
y
t
to
(20),
t
he ou
ta
ge
proba
bili
ty
c
an be
dem
on
strat
ed
a
s foll
ow
:
1
1
2
0
0
2
2
1
2
12
22
2
0
0
0
0
02
0
(
)
1
!
!
(
!
)
m
m
m
DF
n
t
b
d
b
d
d
mm
l
k
n
k
ln
b
k
S
N
R
l
k
n
t
n
dd
Kb
F
a
e
e
e
d
t
l
n
k
(A6)
1
1
2
0
0
2
2
1
2
12
22
2
0
0
0
0
0
0
(
)
1
!
!
(
)
!
(
!
)
m
m
m
DF
n
t
b
d
b
d
d
mm
l
k
n
k
ln
b
kt
S
N
R
l
k
n
t
dd
Kb
F
a
e
e
e
d
l
t
n
t
k
(A7
)
Apply t
he
e
qua
ti
on
[3.47
1,9] of
ta
ble
of
i
nte
gr
al
[],
we hav
e:
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
249
–
257
256
1
0
1
1
2
2
1
2
1
2
2
0
0
0
0
00
12
1
0
(
)
1
2
!
!
(
)
!
(
!
)
2
m
DF
kt
n
t
bd
m
m
m
m
l
k
n
k
ln
S
NR
l
k
n
t
mm
kt
d
d
d
d
Kb
F
a
e
l
t
n
t
k
dd
Kb
(A8)
1
0
21
1
2
2
1
2
0
0
0
0
0
1
2
2
1
2
1
0
(
)
1
2
!
!
(
)
!
(
!
)
2
m
DF
n
k
t
bd
m
l
k
n
k
ln
S
N
R
l
k
n
t
kt
m
m
m
kt
d
Kb
F
a
e
l
t
n
t
k
d
d
d
Kb
(A9)
Wh
e
re
()
v
K
is t
he
m
od
ifed
Be
sse
l functi
on
of
t
he
seco
nd
kind
and
v
th
orde
r.
Finall
y, we ha
ve:
1
0
1
21
2
2
2
0
0
0
0
21
2
1
1
2
1
00
0
2
l
n
2
!
!
(
)
!
(
!
)
2
1
m
kt
m
l
k
n
k
ln
DF
l
k
n
t
n
k
t
bd
m
m
m
kt
d
a
K
b
C
l
t
n
t
k
d
d
d
K
b
e
d
(A10)
En
d of
Pro
of
.
REFERE
NCE
S
[1]
Bi,
S.,
Ho
,
C.
K
.
,
&
Zha
ng
,
R.
(
2015,
04).
W
irel
ess
powere
d
co
m
m
unic
at
ion:
Opportunit
i
es
and
cha
llenges.
IEEE
Comm
unic
ati
ons Magazine
,
53
(4
),
117
-
125
.
doi
:1
0.
1109/mcom
.
2015.
7081084
.
[2]
Ni
y
a
to,
D.
,
Kim
,
D.
I.
,
Maso,
M.,
&
Han,
Z.
(2017).
W
ire
les
s
Po
were
d
Com
m
unic
at
ion
Networks:
Resea
r
ch
Dire
ctions a
nd
T
ec
hnolog
ic
a
l
Ap
proa
che
s.
IE
EE
Wirel
ess Comm
unic
ati
ons
,
2
-
1
1
.
doi:
10.
1109
/mw
c.
2017.
1600116
.
[3]
Nasir,
A.
A.,
Zhou,
X.,
Durrani
,
S.,
&
Kenne
d
y
,
R.
A.
(2013,
07).
Rel
a
y
i
n
g
Protocol
s
for
W
ire
le
ss
Ene
rg
y
Harve
sting
and
Inform
at
ion
Proce
ss
ing.
IE
EE
Tr
ansacti
ons
o
n
Wirel
ess
Comm
unic
ati
ons,
12
(7),
3622
-
363
6.
doi:
10.
1109
/t
wc.2013.062413.
12
2042
.
[4]
Bai
das,
M.
W
.
,
&
Alsus
a,
E
.
A.
(2016,
02)
.
Pow
er
al
lo
cation,
r
el
a
y
se
lecti
on
an
d
ene
rg
y
coope
r
at
ion
str
ategie
s
i
n
ene
rg
y
h
arv
esti
n
g
coope
r
at
iv
e
wi
rel
ess
net
works
.
Wirel
ess
Comm
unic
ati
ons
and
Mobil
e
Comput
i
ng,
16
(14),
2065
-
2082.
doi
:10.
100
2/wcm.2668.
[5]
Krikidi
s,
I.
(20
14
,
03).
Sim
ult
ane
ous
Inform
at
ion
and
Ene
rg
y
Tra
nsfer
in
Large
-
Scale
Networks
with/
without
Rel
a
y
ing
.
IEEE T
rans
act
ions o
n
Comm
unic
ati
ons,
62
(3)
,
900
-
912
.
doi:
10
.
1109/tco
m
m
.
2014.
020914.
130825.
[6]
Ding,
Z
.
,
Krikidis
,
I.,
Sharif
,
B
.
,
&
Poor,
H.
V.
(2
014,
08). W
ire
l
e
ss
Inform
at
ion
a
nd
Pow
er
Tra
nsf
er
in
Coop
erati
v
e
Networks
W
it
h
Spati
all
y
R
ando
m
Rel
a
y
s.
IE
EE
Tr
an
sacti
ons
on
Wirel
ess
Com
municat
ions,
13
(8),
4440
-
4453.
doi:
10.
1109
/t
wc.2014.2314114.
[7]
Mheic
h,
Z
.
,
&
Savin,
V.
(2017
,
03)
.
Cooper
at
i
ve
comm
unic
ati
on
protoc
o
ls
wit
h
ene
rg
y
har
v
esti
ng
r
el
a
y
s.
201
7
Wirel
ess Days
.
d
oi:
10.
1109
/wd.
2
017.
7918116.
[8]
Huang,
C.
,
Zh
a
ng,
R.
,
&
Cu
i,
S.
(2013,
08)
.
Throughput
Ma
ximiza
ti
on
for
t
he
Gauss
ia
n
Re
lay
Chann
el
wi
t
h
Ene
rg
y
Harve
st
ing
Constrai
nts.
IEE
E
Journal
on
Sel
e
ct
ed
Areas
in
Comm
uni
cat
ions,
31
(8),
1469
-
147
9.
doi:
10.
1109
/j
sac
.
2013.
130811
.
[9]
Ahm
ed,
I.
,
Ikhl
ef,
A.,
Schober
,
R.
,
&
Mall
ik
,
R.
K.
(2013,
04).
Joint
Pow
er
Alloc
a
ti
on
and
Rel
a
y
Select
ion
in
Ene
rg
y
Harve
sting
AF
Rel
a
y
S
y
st
ems
.
IEE
E
Wire
l
ess
Comm
unic
ati
ons
Lett
ers,
2
(2),
239
-
2
42.
doi:
10.
1109
/wcl
.
2013.
012513.
13
0007.
[10]
Ze
ng,
Y
.
,
Chen
,
H.,
&
Zha
ng,
R.
(2016,
05)
.
Bidi
re
ct
ion
al
W
ire
l
ess
Inform
at
ion
and
Pow
er
Tra
nsfer
W
it
h
a
Helpi
ng
R
el
a
y
.
I
EE
E
Comm
unications
Letters
,
20
(5),
862
-
865
.
d
oi:
10.
1109
/l
com
m
.
2016.
2549515.
[11]
Guraka
n,
B.
,
Oze
l
,
O.,
Yang
,
J.,
&
Ulukus,
S.
(2013,
06).
En
erg
y
coope
r
at
io
n
in
ene
rg
y
har
vesti
ng
two
-
w
a
y
comm
unic
at
ions
.
2013
IEEE
In
ter
nati
onal
Conf
ere
nce on
Comm
u
nic
ati
ons
(
ICC)
.
doi:
10.
1109
/i
c
c.2013.6655023.
[12]
Chaudhr
y
,
M.,
&
Zuba
ir
,
S.
(2002,
10).
Ext
en
ded
inc
om
ple
t
e
gamm
a
fu
nct
ions
with
appl
ic
ations.
Journal
of
Mathe
mati
cal
A
naly
sis and Appl
ic
ati
ons
,
274
(2)
,
725
-
745.
doi:
10
.
1016/s0022
-
247
x(02)00354
-
2~.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
A
new l
oo
k
a
t
ener
gy ha
rve
sti
ng
half
-
du
plex
D
F
powe
r s
plit
ti
ng
prot
oco
l r
el
ay
…
(
Ph
u
Tr
an Tin
)
257
[13]
J.,
J.
E.
(1979,
08).
The
H
-
Functi
on
with
Appli
ca
t
ions
in
Stat
isti
cs
and
Other
Discipl
ine
s
.
Tec
hnometric
s,
21
(
3),
392
-
393.
do
i
:10.1080/00401706.1979.10489802.
[14]
Ta
bl
e
of
Inte
gr
als
,
Serie
s
,
an
d
Pr
oduct
s.
(2015
). d
oi:
10.
1016
/c
201
0
-
0
-
64839
-
5.
[15]
Tra
n
Hoang
Quang
Minh.
"H
y
b
rid
Ti
m
e
-
Pow
er
Sw
it
chi
ng
Protocol
of
Ene
rg
y
H
arv
esti
ng
Bidi
r
e
ct
ion
al
Rela
y
ing
Network:
Thro
ughput
and
E
rgodic
C
apaci
t
y
An
aly
sis."
TEL
KOMNIKA
(
Tele
communic
ati
on
Computi
ng
El
e
ct
ronics
and
Control)
,
16,
no
.
5
(10
,
2018):
18
9.
ht
tp:
/
/dx.
doi
.
o
rg/10.
12928/
te
lk
om
nika
.
v16i5.
91
18
.
[16]
Ti
n,
Phu
Tra
n
,
T
ran
Hoang
Quan
g
Minh,
Ta
n
N
.
Ngu
y
en
,
and
Mi
roslav
Vozna
k
.
"
S
y
stem
Perform
anc
e
Anal
y
sis
of
Half
-
Duplex
Re
lay
Network
o
ver
Ric
i
an
Fad
ing
Channe
l
.
"
TEL
KOMNIKA
(
Tele
communic
ati
on
Computin
g
El
e
ct
ronics
and
Control)
.
16,
no
.
1
(02
,
2018):
18
9.
doi
:10.
12928
/
te
lkomnika
.
v16i
1.
7491.
[17]
Rashid,
T
ari
qu
e,
Sunil
Kum
ar,
Aks
ha
y
Verm
a,
Prate
ek
R
aj
G
au
ta
m
,
and
Arvind
Kum
ar.
"P
m
-
EE
MRP
:
Pos
tura
l
Movem
en
t
Based
Ene
rg
y
Eff
i
cient
Multi
-
hop
R
outi
ng
Protocol
for
Intra
W
ire
l
e
ss
Body
Sensor
Network
(Intr
a
-
W
BS
N)."
TEL
K
OMNIKA
Telec
omm
unic
ati
on
Computing
Elec
tronic
s
and
Co
ntrol)
,
16,
no
.
1
(02,
2018)
:
1
66.
doi:
10.
12928
/t
e
l
kom
nika
.
v16i1.
7
318.
[18]
A.
F.
Morabit
o
,
"P
ower
S
y
nthe
si
s of
Mask
-
Constrai
ned
Shaped
Bea
m
s T
hrough M
axi
m
al
l
y
-
Spar
se
Plana
r
Arra
y
s
,
"
T
ELKOMNIKA
(Tele
communic
a
t
ion
Computing
El
e
ct
ronics
and
Control)
,
vol. 14
,
n
.
4
,
pp
.
1217
-
1219,
2016
.
Evaluation Warning : The document was created with Spire.PDF for Python.