Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
11
,
No.
1
,
Febr
uar
y
2021
, pp.
3
90
~
401
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v11
i
1
.
pp
3
90
-
401
390
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om
GPON
an
d V
-
b
an
d mmWave in
green b
ackh
aul
s
olution f
or 5G
u
ltra
-
d
en
se
n
etw
or
k
Ayod
e
ji Akee
m Aja
ni
1
,
Vi
tal
ic
e Ka
le
ch
a Odu
ol
2
,
Z
achaeus
K
ayod
e
Ad
e
yem
o
3
1
Depa
rt
m
ent
of
El
e
ct
ri
ca
l
Eng
in
ee
ring
,
PA
N Afric
an
Univer
sit
y
,
Instit
ute of
B
asi
c
Sci
ences,
Te
chno
log
y
a
nd
Innova
ti
on
(PA
UIS
TI)
,
Ken
y
a
2
Depa
rt
m
ent
of Electrical
Eng
in
ee
ring
,
Univ
ersity
of
Nai
robi, Ke
n
y
a
3
Depa
rt
m
ent
of
El
e
ct
ri
ca
l
and
E
l
ec
tron
ic
s E
ng
ineeri
ng,
L
adoke
Akintol
a
Univer
si
t
y
of
T
ec
hno
log
y
(
LAUTECH
)
,
Niger
ia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ma
r
15
, 20
20
Re
vised
Jun
1
0
,
20
20
Accepte
d
J
u
n
23
, 20
20
Ultra
-
dense
netw
ork
(UD
N)
is
cha
racte
ri
ze
d
b
y
m
assive
depl
o
ym
ent
of
sm
all
ce
l
ls
which
result
ed
int
o
comple
x
bac
khauling
of
the
ce
l
ls.
Thi
s
implie
s
that
for
5G
UD
N
to
be
en
erg
y
eff
ici
ent
,
appr
opri
ate
bac
khau
li
ng
sol
uti
ons
m
ust
be
provide
d
.
In
thi
s
pape
r,
we
have
evalua
t
ed
the
per
fo
rm
an
ce
of
gig
a
passive
opti
c
al
n
et
work
(
GPON
)
and
V
-
band
m
il
l
imetre
wave
(
m
m
W
ave
)
in
serving
as
gr
ee
n
bac
kh
aul
solu
tion
for
5G
UD
N.
Th
e
app
roa
c
h
was
to
first
rep
roduc
e
ex
isti
ng
bac
kh
aul
solu
ti
ons
in
ver
y
den
se
net
work
(
VDN
)
sce
nar
i
o
which
serve
d
as
benc
hm
a
rk
for
the
per
form
an
c
e
evalua
t
ion
for
the
UD
N
sce
nar
io
.
The
b
e
st
two
soluti
ons,
GP
ON
and
V
-
band
soluti
ons
from
the
VD
N
were
the
n
deplo
y
ed
in
5G
UD
N
sce
nar
io.
The
rese
ar
ch
was
done
b
y
sim
ula
ti
on
in
MA
TL
AB.
The
per
form
ance
m
et
ric
s
used
were
powe
r
consum
pti
on
an
d
ene
rg
y
eff
icie
nc
y
against
th
e
norm
al
iz
ed
ho
url
y
tra
ff
ic
profil
e
.
The
resu
lt
rev
e
al
ed
tha
t
GP
ON
and
V
-
b
and
m
mWave
outpe
rform
ed
othe
r
soluti
ons
i
n
VD
N
sce
nar
i
o.
How
eve
r,
this
per
form
anc
e
signifi
c
antl
y
droppe
d
in
th
e
UD
N
sce
nar
io
due
to
high
er
data
tra
ff
ic
r
equi
re
m
ent
of
UD
N
compare
d
to
VD
N.
Thus,
it
ca
n
be
conc
lud
e
d
tha
t
GP
ON
a
nd
V
-
band
m
mWave
are
no
t
best
suit
ed
to
s
erv
e
as
gr
ee
n
ba
ckha
ul
solu
ti
on
f
or
5G
UD
N
nec
essit
at
ing
fur
the
r investi
g
at
io
n
of
oth
er available
b
ac
kh
aul
tec
hnologi
es.
Ke
yw
or
d
s
:
5G
G
-
P
O
N
Gr
ee
n
b
ac
kh
a
ul
s
olu
ti
ons
Ultra
-
d
en
se
n
et
works
V
-
b
a
nd m
m
W
ave
This
is an
open
acc
ess arti
cl
e
un
der
the
CC
B
Y
-
SA
l
ic
ense
.
Corres
pond
in
g
Aut
h
or
:
Ayo
dej
i
A
kee
m
A
j
ani
,
Dep
a
rtm
ent o
f
Ele
ct
rical
En
gi
neer
i
ng,
PAN
Africa
n Un
i
ver
sit
y,
In
s
ti
tute of
Basi
c
Scie
nces,
Tec
hnol
og
y a
nd
Inn
ov
at
io
n (P
A
UISTI
)
,
PAUST
I
Com
plex
Bl
ock B,
JKUAT
m
ai
n
ca
m
pu
s,
JU
J
A,
P. O.
B
ox 62
000
–
00
200 Na
iro
bi, Kenya.
Em
a
il
:
ajan
i
aa@gm
ai
l.co
m
1.
INTROD
U
CTION
The
T
hird
a
nd f
ourt
h
ge
ner
at
i
on
netw
ork
(
3G and 4G)
ha
ve
sh
ow
n
that m
ob
il
e n
et
works
can
pro
vid
e
broa
db
a
nd
acce
ss
[1]
.
Howe
ver,
us
e
rs’
num
ber
s
and
ser
vice
ap
plica
ti
on
s
are
inc
reas
ing
a
nd
c
hangi
ng
.
This
acco
rd
i
ng
to
f
or
ecast
[2
,
3]
us
e
rs’
data
an
d
traf
f
ic
req
ui
rem
ent
s
will
ov
er
w
he
l
m
the
capaci
ty
of
the
existi
ng
m
ob
il
e
net
wor
ks.
Th
us
,
t
her
e
is
need
f
or
a
network
t
hat
can
handle
10
00x
m
or
e
data
traffi
c
an
d
this
netw
ork
is
known
as
the
fifth
g
ene
rati
on
net
work
(5G
)
[
4,
5]
.
Ca
pac
it
y
con
siderati
on
was
no
t
the
on
ly
adv
a
ncem
ent
exp
ect
e
d
of
5G
netw
ork
com
par
ed
to
e
xisti
ng
net
work
s
[1]
.
The
ot
her
e
xpect
ed
feat
ur
e
s
of
5G
netw
ork
is
a
s
sh
ow
n
i
n
Ta
bl
e
1.
I
n
orde
r
t
o
ac
hieve
it
s
ta
rg
et
e
d
perfor
m
ance,
[
1]
ide
ntifie
d
e
voluti
on
of
existi
ng
ra
dio
acce
ss
te
c
hnologies
(R
ATs
),
Hype
rd
e
ns
e
sm
all
‐
cell
de
plo
ym
ent
,
Self
‐
orga
nisin
g
ne
twor
k
(S
O
N
),
Ma
chi
ne
ty
pe
com
mu
nicat
io
n
(MT
C),
De
velo
ping
m
il
l
i
m
et
re
w
ave
(m
m
W
ave
)
RATs,
al
loca
ti
on
of
new
sp
ect
ru
m
for
5G,
S
pectr
um
sh
arin
g
,
R
adio
acce
ss
network
(R
AN)
vi
rtuali
sat
ion
,
E
nergy
ef
fici
enc
y
(EE)
and Redesi
gn
i
ng
bac
khaul li
nk
s
as t
he 10 p
il
la
rs
of 5G
net
work.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
GPO
N
an
d
V
-
band
mm
W
ave
in gree
n ba
ck
haul s
olu
ti
on f
or
5G u
lt
r
a
-
de
ns
e
network
(
Ayod
ej
i Ake
e
m Aj
an
i)
391
Table
1.
E
xpec
ta
ti
on
fro
m
the
5G fram
ewo
r
k
[
6]
Para
m
eters
Su
p
p
o
rt
Data Rates
10
-
1
0
0
x
m
o
r
e than
L
T
E
d
ata
rates
Mob
ility
Su
p
p
o
rt
f
o
r
h
ig
h
sp
eed u
sers (~50
0
Km
/h
)
Heterog
en
eo
u
s Netwo
rks
Mob
ility
su
p
p
o
rt
in
heterog
en
eo
u
s R
ad
io
Acce
ss
Techn
o
lo
g
ies alo
n
g
with
m
u
lti
-
co
n
n
e
ctiv
ity
capab
ilities
CAPEX/O
PEX
Su
stain
ab
le
New dep
lo
y
m
en
t c
ap
ab
ilities
Easy
W
ire
less
dev
ice de
n
sity
Su
p
p
o
rt
f
o
r
1
0
-
1
0
0
x
m
o
re
d
ev
ices
End
-
to
-
End
latenc
y
<1
m
s
Qu
ality
of
E
x
p
erie
n
ce
Co
n
tex
t bas
ed
(
f
lo
w,
m
o
b
ilit
y
pro
f
ile
,
etc.
)
Energy
ef
f
icien
cy
Hig
h
Data Rates
10
-
1
0
0
x
m
o
r
e than
L
T
E
d
ata
rates
The
10
pill
ars
are
relat
ed
,
a
nd
m
os
t
tim
e
re
searche
s
com
ple
m
entaril
y
so
lve
prob
le
m
s
to
ad
dr
ess
t
w
o
or
m
or
e
of
the
se
key
featu
res
si
m
ultaneou
sl
y.
Energy
ef
fici
ency,
m
il
l
i
m
etr
e
wa
ve
(m
m
Wav
e
),
bac
kh
a
ul
an
d
Hype
rd
e
ns
e
S
m
al
l
‐
Ce
ll
Dep
loym
ent
in
oth
e
r
w
ords
,
ultra
-
dense
ne
tw
ork
s
(
UDN)
are
s
om
e
of
the
c
on
cept
on
fo
c
us
i
n
this
re
search
.
T
he
c
oncept
of
Ultra
-
Den
s
e
Netw
orks
i
n
5G
is
ab
ou
t
de
ns
ific
at
ion
of
sm
al
l
cells
a
nd
the
ex
plo
it
at
io
n
of
s
patia
l
r
e
us
e
of
sp
ect
r
um
to
m
eet
the
high
nu
m
ber
of
us
e
rs,
m
achine
an
d
se
rv
ic
es
[
7]
.
UDNs
boos
t
ca
pacit
y
and
e
nh
ance
c
ov
e
rag
e
with
lo
w
-
c
os
t
and
powe
r
-
e
ff
i
ci
ent
inf
rastr
uc
ture
in
5G
net
works.
5G
UDN
de
pl
oym
ents
are
e
nv
isa
ge
d
to
be
heteroge
neous
and
de
ns
e
,
pri
m
aril
y
throu
gh
the
prov
isi
on
i
ng
of
sm
a
ll
cells
su
ch
as
pic
ocell
s
and
fem
tocel
ls.
In
UDN,
us
ers
ca
n
be
within
the
vicinit
y
of
m
ulti
pl
e
cel
ls,
wh
ic
h
i
m
plies
high
interfe
re
nc
es
if
no
t
m
an
aged
[
8]
.
I
n
ad
diti
on
to
inter
f
eren
ce
iss
ue,
UDN
will
m
os
t
li
kely
face
the
pro
bl
e
m
of
redu
ndant
s
m
a
ll
cell
s,
SCs
ene
rg
y
con
s
um
ption
,
and
li
m
it
ed
backha
ul
capa
ci
ty
.
Ther
e
f
or
e,
t
he
re
is
need
to
do
e
xtensi
ve
r
esearch
i
nto
5G
U
DN
e
nerg
y
con
s
um
ption
and
it
backh
aulin
g
so
luti
on
[9]
.
Air
ports,
op
e
n
gathe
rin
g,
ca
m
pu
se
s,
apa
rtm
ents,
m
a
ll
,
rail
sta
ti
on
s
ar
e
exam
ples
of
place
s
wh
e
re
5G
U
D
N
is
ex
pecte
d
to
be
de
plo
y
ed
[
10]
.
Ta
ble
2
c
on
ta
in
s
th
e
pro
per
ti
es
of
U
D
N
com
par
ed
t
o
the pre
vious
ne
tworks
.
Table
2
. U
DN
pro
per
ti
es in
c
om
par
ison
to
ol
der
netw
orks
[
11
]
Tr
ad
itio
n
al
Netwo
rks
Den
ser Ne
t
wo
rks
Ver
y
Dens
e
Netwo
rks
VD
N
Ultra
-
Den
se
Netwo
rks
UD
N
Period
Bef
o
re
2
0
1
4
2015
–
2
0
1
7
2017
–
2
0
2
0
Beyon
d
20
2
0
Su
b
scrib
ed
data
1
GB /
m
o
n
th
2
–
5
GB /
m
o
n
th
5
–
1
0
GB /
m
o
n
th
20
-
5
0
GB /
m
o
n
th
Mini
m
u
m
us
e
r
th
r
o
u
g
h
p
u
t
4
M
b
p
s
8
M
b
p
s
1
0
M
b
p
s
10
–
2
0
M
b
p
s
Sp
ectru
m
2
x 1
0
0
M
Hz
2
x 1
2
0
M
Hz
2
x 1
4
0
MHz
2
x 1
6
0
M
Hz
Site /
K
m
2
7
sites
2
1
sites
2
6
sites
9
3
sites
Inter Site
Distan
ce
I
SD
3
9
5
m
2
3
7
m
2
0
9
m
1
1
2
m
Ther
e
a
re
nu
m
ero
us
bac
khaul
te
chnolo
gi
es
(w
ire
d
a
nd
wireless
)
a
vaila
ble
f
or
backha
uling
the
te
le
com
m
u
nicat
ion
netw
orks
.
Acc
ordin
g
to
Ron
y
et
al
.
[12]
,
bac
khaul
is
th
e
li
nk
betwee
n
one
Ba
se
Stat
ion
/e
N
ode
B
(BS/eNB)
to
ano
t
her.
Also
,
i
n
the
c
entrali
zed
a
ppro
ac
h,
i.e.
Ce
ntrali
zed
ra
dio
acce
ss
netw
ork
(CR
AN),
backha
ul
connects
bas
eband
un
it
(B
BU)
a
nd
the
cor
e
netw
ork.
Additi
on
al
ly
,
in
5G
netw
orks,
bac
khaul
will
carry
la
rg
e
am
ount
of
tr
af
fic
to/fr
om
the
cor
e
net
work
w
her
e
bo
th
distri
bu
te
d
r
adio
acce
ss
netw
ork
(
DR
AN)
(R
AN
processin
g
is
distribu
te
d
to
BSs),
an
d
CR
AN
co
-
exis
t.
Ba
ckh
a
ul
in
5G
is
exp
ect
e
d
to
ha
ve
low
la
te
ncy
and
lo
w
powe
r
consum
ption
[13]
.
Dif
fer
e
nt
te
chn
ol
og
ie
s
ha
ve
bee
n
co
ns
i
der
e
d
for
5G
bac
kh
a
ul
to
be
a
ble
to
m
eet
the
re
quirem
ents.
Eac
h
of
t
hese
te
ch
no
l
og
ie
s
ha
ve
their
ad
va
ntag
es
an
d
disad
va
ntages.
Broa
dly
bac
khaul
so
l
utio
ns
c
an
be
di
vid
e
d
into
wired
an
d
wireless
s
olu
t
ion
s
[
12
]
.
Wir
ed
s
olu
ti
on
s
hav
e
do
m
inate
d
t
he
backha
ul
netw
ork
in
t
he
past
w
hen
the
base
sta
ti
on
was
m
ajorly
m
acro
base
s
ta
ti
on
(MBS).
C
oppe
r
an
d
Fib
re
op
ti
cs
are
the
two
popula
r
opti
on
s
for
wi
red
backha
uling.
It
is
ho
we
ve
r
be
li
eved
that
copper
-
ba
sed
so
l
utio
n
will
no
t
be
a
bl
e
to
serv
e
as
backha
u
l
for
5G
a
nd
beyo
nd
due
to
it
s
lim
it
ed
capaci
ty
[14]
.
This
is
no
t
issue
with
fibre
op
ti
cs
as
it
has
high
cap
aci
ty
,
low
la
ten
cy
wh
ic
h
ca
n
m
ee
t
the
qu
al
it
y
of
serv
ic
e
(
Q
oS
)
requirem
ent
of
5G
an
d
bey
ond.
High
co
st
of
dep
l
oym
ent
and
scal
abili
ty
are
the
m
ajo
r
iss
ue
s
associat
e
d
with
fi
br
e
opti
cs
base
d
bac
khaul
so
l
utio
n.
I
n
order
to
a
ddr
ess
these
iss
ue
s,
the
re
has
been
e
volu
ti
on
of
passive
op
ti
cal
net
wor
k
(
PON)
te
c
hnology
to
im
pr
ove
the
perf
or
m
ance
of
fibr
e
ba
se
d
so
luti
ons
[
15, 16]
.
W
i
reless
so
l
ution
has
rece
ntly
gaine
d
at
te
ntion
as
m
any
research
es
[17
-
20]
reco
gniz
ed
it
as
po
s
sible
viable
op
ti
on
for
bac
kha
ulin
g
5G
a
nd
bey
ond.
Asi
de
th
e
pr
ese
nt
s
ub
6
G
Hz
a
nd
m
ic
rowav
e
fr
e
quency
pr
ese
ntly
in
use
,
othe
r
higher
fr
e
qu
e
ncies
w
irel
ess
opti
ons
wh
ic
h
prov
i
de
la
rg
e
r
li
nk
c
apacit
y
bu
t
a
re
ver
y
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
390
-
40
1
392
vu
l
ner
a
ble
to
env
i
ronm
ental
eff
ect
s
will
be
gr
eat
ly
exp
lore
d.
F
or
inst
ance,
m
m
W
av
e
op
erati
ng
in
three
diff
e
re
nt
bands
,
60
GH
z
(V
-
ba
nd),
70
/
80
G
Hz
(E
-
ba
nd)
ha
s
been
descr
i
bed
as
a
n
at
tract
ive
op
ti
on
f
or
fu
t
ure
wireless
backh
aul
an
d
ac
cess
netw
ork
te
ch
nolo
gy,
as
it
off
ers
ve
ry
la
r
ge
capaci
ty
(up
to
10G
bps)
c
ompar
e
d
to
ot
her
wi
reless
opti
ons
[19]
.
A
pa
rt
f
ro
m
env
i
ronm
ental
l
os
ses,
m
os
t
m
m
W
aves
ba
nd
al
so
re
quires
li
ne
of
sigh
t
(L
OS
)
[
21]
.
Find
i
ng
an
e
ne
rg
y
e
ff
ic
ie
nt
and
en
vir
on
m
ental
ly
fr
ie
ndly
(gree
n)
s
olu
ti
on
s
f
or
prese
nt
and
f
utu
re
te
le
com
m
un
ic
a
ti
on
net
wor
ks
is
on
e
of
t
he
m
os
t
i
m
po
rtant
con
ce
pt
to
r
esearch
on
no
w
acco
r
ding
to
[
22
]
.
The
pr
ese
nt
ge
ner
at
io
n,
fift
h
ge
ne
rati
on
(
5G)
is
bu
il
t
on
gree
n
co
nce
pt.
T
her
e
a
re
effor
ts
[23
-
29]
from
researc
hers
to
m
ake
5G
greener
th
an
the
pr
e
vious
net
works
des
pite
increase
in
de
ns
ific
at
io
n
an
d
dat
a
rates
[
30
]
.
For
5G
an
d
beyo
nd
to
be
m
eet
i
ts
energy
ef
fici
en
cy
exp
ect
at
io
ns,
both
the
acc
ess
an
d
the
bac
kh
a
ul
par
ts
of
the
netw
orks
m
us
t
be
carefu
ll
y
desig
ned
a
nd
op
e
ra
te
d
[
31]
.
Partic
ularly
,
choosin
g
the
righ
t
te
chnolo
gy
to
achieve
the
best
res
ult
in
te
rm
s
of
energy
eff
ic
ie
ncy
and
powe
r
co
ns
um
ption
is
an
op
e
n
pro
blem
wh
ic
h
needs
m
or
e
ef
fort
m
os
t
especial
ly
wh
en
it
com
es
to
5G
UDNs
whose
backha
ul
will
be
ve
r
y
com
plex
due t
o hig
h nu
m
ber
of sm
all cell
s
[32]
.
In
res
pons
e
to
the
e
nv
i
ronm
ental
chall
en
ge
t
hat
com
es
from
the
e
m
issio
ns
of
t
he
m
ob
il
e
netw
ork
,
m
any researc
he
s
[30,
33
-
35]
hav
e
bee
n
do
ne
in
orde
r
to m
ake th
e
netw
ork
m
or
e env
i
ron
m
ental
ly
f
riendl
y and
energy
eff
ic
ie
nt.
Wh
il
e
ther
e
are
diff
e
re
nt
ways
to
achieve
gr
ee
n
ne
twork
,
Tom
baz
[36]
was
one
of
the
le
adin
g
au
thors
w
ho
a
r
gued
t
hat
bac
khaul
power
c
onsu
m
ption
c
ontribu
te
s
sig
nifi
cantl
y
to
the
ov
e
rall
powe
r
co
nsum
ption
a
nd
ene
r
gy
eff
ic
ie
ncy
of
the
m
ob
il
e
netw
ork.
Si
nc
e,
there
has
been
m
any
back
ha
ul
so
luti
ons
[
36
-
38]
for
diff
e
ren
t
ge
ner
at
io
ns
of
the
m
ob
il
e
network
us
i
ng
dif
f
eren
t
te
c
hnol
og
ie
s
.
Tom
baz
et
al
.
[36]
pro
ved
th
at
backhau
l
ha
s
sign
ific
a
nt
im
pact
in
the
ov
erall
powe
r
c
on
s
um
ption
an
d
ene
rgy
eff
ic
ie
ncy
of
the
net
work.
T
hey
pro
po
s
ed
Op
ti
cal
Po
i
nt
-
to
-
po
i
nt
Ether
net
bac
kh
a
ul
so
luti
on
f
or
t
he
3G
UMTS
net
wor
ks
.
T
heir
so
l
ution
s
howe
d
i
m
pro
ved
gr
ee
n
perform
ance
of
the
netw
ork.
Their
s
olu
ti
ons
were
base
d
on
3G
ne
twork
a
nd
aft
er
the
em
erg
en
ce
of
4G
ne
tw
ork
the
re
was
need
t
o
get
bet
te
r
bac
kh
a
ul
s
ol
ution
s
to cope
with t
he
co
m
plexity
o
f
the
n
e
w netw
ork.
Tom
baz
et
al
.
[39]
f
ur
t
he
r
as
sessed
the
im
pact
of
bac
khau
l
on
the
overal
l
energy
c
onsum
pt
ion
of
the
cel
lular
ne
tworks
us
in
g
di
ff
ere
nt
data
traff
ic
re
qu
irem
ents
from
20
14
to
2020.
T
he
y
pr
opose
d
th
r
ee
(3
)
backha
ul
arc
hi
te
ct
ur
es
us
in
g
three
(
3)
dif
f
eren
t
te
c
hnologies
(c
oppe
r,
fibr
e
an
d
m
ic
ro
wa
ve
).
T
w
o
(2)
of
the
pr
opos
e
d
arc
hitec
tures
are
hybri
d
there
by
m
a
king
us
e
of
two
te
ch
no
l
og
ie
s
sim
ultan
eo
us
ly
.
Fibr
e
-
to
-
the
-
no
de
(
FTT
N)
us
i
ng
V
DS
L
2,
M
ic
rowav
e
O
nly
and
Fib
re
-
to
-
t
he
-
B
uildin
g
(
F
TTB)
+
Mi
cr
owave
are
their
th
ree
(3)
arc
hitec
tur
al
i
m
ple
m
entation
s
.
T
heir
res
ult
showe
d
t
ha
t
backha
ul
can
be
re
spo
ns
ible
for
up
to
50%
of
t
he
powe
r
co
nsum
ption
on
the
ne
twork
.
T
he
hy
br
id
bac
kh
a
ul
m
ade
up
of
F
ibre
-
to
-
t
he
-
B
uildin
g
(F
TTB
) op
ti
on
and m
ic
ro
wa
ve
li
nk
s
h
a
d bett
er
perform
ance co
m
par
ed
to
t
he othe
r
a
rc
hit
ect
ur
es.
Su
a
rez
et
al
.,
[
37
]
st
ud
ie
d
e
ne
rg
y
e
ff
ic
ie
ncy
of
bac
khaul
i
n
heter
og
e
nous
netw
ork
.
T
he
y
propose
d
two
(
2)
hy
bri
d
bac
kh
a
ul
ar
chite
ct
ur
es:
Fi
br
e
-
to
-
the
-
buil
ding
(F
T
TB)
+
10Gbps
pas
sive
opti
cal
network
(10GPO
N
)
te
chnolo
gies
an
d
FTTB
+
m
ic
rowav
e
.
Thei
r
a
naly
ses
wer
e
ba
sed
on
Ar
ea
Power
C
on
s
um
pt
ion
[
W
/
km
2].
Their
resu
lt
showe
d
that
th
e
two
pro
po
sed
arc
hitec
tu
res
ha
d
bette
r
perf
or
m
ance
than
the conv
e
ntio
na
l backhaul in t
erm
s o
f
area pow
e
r
co
nsum
ption
[
W/
km
2].
Mowla et al
.
[
38]
pr
ese
nte
d
en
erg
y
-
eff
ic
ie
nt
com
m
un
ic
at
io
n
m
od
el
fo
r
5G
heter
og
e
ne
ous
netw
orks
(
HetNets)
.
They
consi
de
red
both
acce
s
s
an
d
backha
ul.
T
he
y
inv
est
igate
d
powe
r
co
nsum
ption
of
va
rio
us
bac
kha
ul
de
sign
s
,
the
n
pr
opos
e
d
tw
o
ba
ckh
a
ul
so
luti
ons.
W
i
r
ed
passive
opti
cal
netw
ork
(
PON
)
a
nd
wireless
V
-
ba
nd
m
i
ll
i
m
et
r
e
wa
ve
(m
mWave
)
.
The
pe
rfor
m
ance
evaluati
on
was
done
th
r
ough
sim
ulati
on
us
ing
netw
ork
si
m
ulator
2
(N
S
-
2).
They
s
howe
d
how
t
o
c
onnec
t
passive
o
ptica
l
netw
ork
un
i
ts/
te
r
m
inals
with
5G
acce
s
s
un
it
s
t
o
re
duc
e
the
ov
e
rall
powe
r
consum
ption
in
the f
irst s
olu
t
ion
. T
he
sec
ond
so
l
ution
inte
gr
at
e
d
m
m
W
ave b
ac
kh
a
ul un
it
s w
it
h
5G
SC
N
unit
s
to
reduce
po
wer
co
nsum
ption
.
Ba
c
kh
a
ul
power
co
nsu
m
pt
ion
in
W
a
tt
and
Ba
ckh
a
ul
Energy
Ef
f
ic
ie
ncy
Mbp
s/
Watt
w
ere
the
perfor
m
ance
m
e
tric
s
us
e
d.
T
he
re
s
ult
rev
eal
e
d
th
at
their
so
l
utio
n
can
save
up
to
48%
powe
r
co
nsum
ption.
T
heir
s
ol
ution
s
outpe
rfor
m
ed
the
pre
vious
s
olu
ti
on
in
te
rm
s
of
po
wer
c
onsu
m
ption
a
nd
energy ef
fici
en
cy
.
I
n
t
his
pap
e
r,
we
hav
e
eval
ua
te
d
al
l
this
s
olu
ti
on
in
t
he
pre
5G
netw
ork
scenari
o
to
det
erm
ine
wh
ic
h
is
best
of
the
existi
ng
gr
ee
n
backha
ul
s
ol
ution
s
.
F
ur
t
he
rm
or
e,
we
a
na
ly
sed
the
pe
rfor
m
ance
of
t
he
best
so
luti
ons
in
5G
U
DN
sce
na
r
io
thr
ough
sim
ulati
on
in
M
A
TLAB
us
i
ng
an
im
pr
ov
e
d
5G
m
od
el
and
al
gorithm
.
This
is
to
dete
rm
ine
if
the
be
st
two
s
olu
ti
on
s
(
gi
ga
passive
op
ti
cal
net
work
(GPO
N
)
an
d
V
-
band
m
il
l
i
m
et
re
wav
e
(m
m
W
ave)
)
are
well
su
it
ed
to
ser
ve
as
green
ba
ckh
a
ul
s
olu
ti
on
f
or
5G
U
D
N
net
wor
k.
W
e
hav
e
pr
ese
nted
the
m
od
el
and
i
m
pro
ved
al
gorit
hm
in
s
ect
ion
2
and
the
sim
ula
ti
on
desc
riptio
n
in
s
ect
ion
3.
Sect
ion
4
c
ont
ai
ns
t
he
r
es
ult
and d
isc
us
si
on
wh
il
e s
ect
io
n 5
h
as
the c
oncl
usi
on
s
fro
m
the
analy
sis.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
GPO
N
an
d
V
-
band
mm
W
ave
in gree
n ba
ck
haul s
olu
ti
on f
or
5G u
lt
r
a
-
de
ns
e
network
(
Ayod
ej
i Ake
e
m Aj
an
i)
393
2.
PROP
OSE
D MO
DEL
This
sect
io
n
presents
the
a
ppr
oac
h
us
e
d
f
or
this
researc
h.
It
incl
udes
the
re
quirem
e
nt
an
al
ysi
s
,
def
i
niti
on
of c
on
ce
pts a
nd to
ols, per
f
or
m
ance p
a
ram
et
ers
and m
et
rics as w
el
l as
def
init
i
on of sce
na
rios.
2
.
1.
5G
UDN
mod
el
This
sect
io
n
co
ntains
the
der
i
vations
of
the
m
at
he
m
at
ic
a
l
op
t
im
iz
ation
pro
blem
to
m
ak
e
UDNs
m
os
t
energy
ef
fici
ent
with
out
losi
ng
t
he
e
xp
ect
e
d
qual
it
y
of
s
erv
ic
e
(QoS
)
Re
qu
irem
ent.
The
5G
U
DN
m
od
el
pro
po
se
d
he
re
consi
st
of
a
central
m
acro
base
sta
ti
on
(MBS)
an
d
seve
ral
ran
dom
ly
distrib
uted
Sm
al
l
cel
ls.
The
c
hoic
e of
t
his
pap
e
r for s
m
al
l cel
l i
s p
ico
cel
ls beca
us
e
we
a
re investi
ga
ti
ng
a
n ou
t
do
or scena
rio
.
The follo
wing
set
o
f
notat
io
ns has
b
ee
n
a
dopt
ed
in t
his
pap
e
r:
Q
as
set
of
Q f
or a
5G m
ulti
-
t
ie
r
Heter
ogen
ous
Netw
ork, i
ndex
q.
J as set
of J S
C
N base stat
io
ns, in
dex
j.
T as set
of T T
raffic
class, i
ndex
t.
U
as
set
of
U Users
, inde
x u.
The
e
ne
rg
y
e
f
fici
ency
of
t
he
ultra
de
ns
e
ne
twork
(
+
)
with
res
pect
to
the
acce
ss
a
nd
ba
ck
ha
ul n
et
w
ork
ca
n be
giv
e
n
in
(
1)
,
,
1
1
A
N
+
B
H
H
e
t
N
e
t
U
D
N
A
N
+
B
H
H
e
t
N
e
t
U
D
N
E
E
=
QJ
qj
qj
d
P
(1)
wh
e
re
,
,
1
1
,
,
Q
J
qj
qj
qQ
d
jJ
is
the
total
data
rate
by
eve
ry
base
sta
ti
on
.
A
N
+
B
H
H
e
t
N
e
t
U
D
N
P
is
the
total
power
consum
ption
of
the
netw
ork
includi
ng
po
w
er
co
nsum
ed
by
both
acce
s
s
an
d
ba
ck
ha
ul
netw
ork.
T
he
total
powe
r
co
nsum
ption
(
A
N
+
B
H
H
e
t
N
e
t
U
D
N
P
)
is t
he
s
um
o
f
powe
r
c
on
s
um
ption
s
of MB
S (
P
MBSTotal)
and
SCs
(
P
S
CNTotal
)
and
this is as
def
i
ne
d
in
(2)
M
B
S
T
o
t
a
l
S
C
N
T
o
t
a
l
A
N
+
B
H
H
e
t
N
e
t
U
D
N
P
PP
(2)
Since UDN
is heter
og
e
nous
i
n
natu
re,
q
> 1
,
the MBS exist
at
q
= 1
an
d
the SCN
is at
q
=
2
,
∀
q
∈
Q
.
Fo
r
q
=
1 (MB
S)
.
P
MBSTotal
=
P
MB
SAN
+
P
MBS
BH
(3)
wh
e
re
P
MBSAN
is
the
po
wer
consum
ption
of
t
he
acce
ss
netw
ork
f
or
MB
S
(as
def
i
ned
i
n
(
4),
P
MBSBH
is
the po
wer co
nsum
ption
of t
he back
ha
ul n
et
w
ork for MB
S
(a
s d
e
fine
d
in
(
5
)
).
P
MBSAN
has
both
fixe
d
a
nd
l
oad
de
pend
e
nt
par
am
et
ers.
T
he
fi
xed
powe
r
co
nsum
ption
is
the
total
powe
r
co
nsum
ed
by
the
MB
S
irres
pecti
ve
of
w
hethe
r
the
re
is
data
t
raffic
or
not.
On
th
e
oth
e
r
hand,
the
lo
a
d
dep
e
ndent
c
on
su
m
ption
de
pe
nd
s
on
the
para
m
et
ers
of
the
traff
ic
bein
g
pa
ssed
a
nd
t
his
is
us
u
al
ly
in
a
dd
it
io
n
to the fi
xe
d.
f
i
x
e
d
M
B
S
A
N
M
d
y
n
a
BS
m
i
c
M
B
S
t
x
PP
P
(4)
wh
e
re
f
i
x
e
d
MBS
P
is
the
fi
xed
po
wer
co
nsum
ption
,
M
B
S
is
t
he
loa
d
de
pe
ndent
pa
ram
et
er
,
d
y
n
a
m
i
c
tx
P
is
the
dynam
ic
powe
r of
base
sta
ti
on
, it i
s
de
fine
d
as
Fo
r
the
MB
S
ba
ckh
a
ul
wh
ic
h usin
g
P
O
N,
t
he
pow
e
r
c
onsu
m
ption o
f
t
he b
a
ckh
a
ul is as
g
i
ven is
(
5).
M
B
S
u
l
S
F
P
+
M
B
S
B
H
o
g
g
P
N
P
N
P
N
P
(5)
wh
e
re
MBS
N
is
th
e
n
um
ber
of
MB
S,
P
o
is
the
po
wer
co
nsum
pti
on
of
a
n
O
N
U
,
N
g
is
t
he
num
ber
of
G
PON
port
in
an
OL
T,
P
g
is
the
po
we
r
co
nsum
ption
of
the
G
PON
port,
N
ul
is
the
nu
m
be
r
of
upli
nk
interface,
and
P
SFP+
is t
he
pow
e
r
c
onsum
pt
ion
of SF
P
+ m
od
ule.
Fo
r
q
=
2,
(S
C
N)
,
we
a
dopte
d
P
SCNAN
as
t
he
pow
er
co
nsu
m
pt
ion
of
the
acce
ss
netw
ork
f
or
SCN
a
nd
P
SCNBH
as
the pow
e
r
co
nsum
ption
of the back
hau
l
netw
ork
f
or
SC
N.
T
he
to
ta
l
po
we
r
co
nsum
ption
s of SC
N
is
def
i
ned in
(6) while
P
SCNAN
and
P
SCNBH
are d
efi
ned in
(
7
)
a
nd
(
8
)
res
pecti
vely
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
390
-
40
1
394
S
C
N
T
o
t
a
l
S
C
N
A
N
S
C
N
B
H
P
P
P
(6)
j
,
f
i
x
e
d
j
,
d
y
n
a
m
i
c
S
C
N
S
C
N
S
C
N
S
C
N
A
N
,tx
s
PP
P
H
(7)
P
SCNBH
=
{
∑
(
ℎ
,
+
ℎ
,
)
,
∈
−
,
(8)
wh
e
re
j
,
f
i
x
e
d
S
C
N
P
is
the
fi
xed
po
wer
co
nsum
ption
of
S
CN,
∀
j
∈
J
,
H
s
is
the
traf
fic
l
oad
f
or
SC
N
(this
is
e
qual
to
zero
in
sle
e
p
m
od
e),
S
C
N
is
the
slop
e
of
loa
d
-
dep
e
ndent
po
wer
co
nsum
pti
on,
j
,
d
y
n
a
m
i
c
S
C
N
,
t
x
P
is
the
tra
nsm
issi
on
powe
r
f
or
the
SCN
(
picocell
=
21
dBm
),
b
h
,
f
i
x
e
d
k
m
P
is
the
fixe
d
po
wer
c
onsu
m
ption
of
eac
h
SC
N
m
m
W
ave
backha
ul
li
nk
,
bh,t
x
k
m
P
is
the
loa
d
de
pe
nd
e
nt
ra
dio
f
r
equ
e
ncy
tra
nsm
it
po
wer
co
nsum
ption
of
th
e
SCN
m
m
W
a
ve
backha
ul
li
nk
,
P
o
is
the
po
wer
co
ns
um
ption
o
f
a
n
O
ptica
l
Netw
ork
Un
it
(
ONU)
an
d
N
SC
N
-
ONU
is
the
nu
m
ber
of
li
nk
in
g betwee
n
the
SC
N
a
nd
the ON
U.
Eq
uation
(1)
c
an
be
m
od
ifie
d
to
m
ini
m
iz
a
t
ion
of
powe
r
consum
ption
i
ns
te
ad
of
m
axi
m
iz
ation
of
the
energy
ef
f
ic
ie
ncy.
This
is
to
av
oid
f
racti
on
al
it
y
in
the
optim
iz
at
ion
prob
le
m
.
This
is
po
s
sible
acco
r
ding
to
equ
at
io
n
wh
e
r
e
it
can
be
in
ferred
that
if
the
ex
pected
data
rate
is
be
ing
delivere
d,
m
ini
m
izing
powe
r
consum
ption
will
resu
lt
in
m
axi
m
iz
ing
EE.
Th
us,
pro
bl
e
m
fo
rm
ulati
on
to
m
ini
m
iz
e
powe
r
co
nsum
ption
is
as
giv
e
n
in
(9)
Mi
ni
m
iz
e
+
=
P
MBS
f
ixe
d
+ Δ
MBS
.
P
tx
dy
n
a
mic
+
+
+
+
+
,
+
.
∆
.
,
,
+ *P
SCNBH
(9)
*P
SCNBH
can
as
su
m
e
ta
ke
two
def
init
io
ns
de
pendin
g
on
w
hethe
r
m
m
W
ave
or
PON
is
bein
g
us
e
d.
This
(
9
)
ca
n be
fur
the
r
sta
te
d
as
(
9a) w
hich
m
ake u
se
of m
m
W
ave a
nd
(
9b)
whe
n
P
ON is use
d.
Mi
ni
m
iz
e
+
=
P
MBS
f
ixe
d
+ Δ
MBS
.
P
tx
dy
n
a
mic
+
+
+
+
+
,
+
.
∆
.
,
,
+
∑
(
ℎ
,
+
ℎ
,
)
∈
(9
a
)
Mi
ni
m
iz
e
+
=
P
MBS
f
ixe
d
+ Δ
MBS
.
P
tx
dy
n
a
mic
+
+
+
+
+
,
+
.
∆
.
,
,
+
−
(
9b)
Subj
ect
t
o
m
a
x
t
x
,
q
j
t
x
,
,
(
,
)
uU
P
u
P
q
Q
j
J
(10)
qj
0
,1
,
(
,
,
,
)
t
u
U
t
T
q
Q
Su
jJ
(11)
qj
11
1
,
,
,
TJ
t
tj
u
U
q
j
S
QJ
(12)
th
qj
1
,
,
,
T
t
t
u
U
t
T
qQ
u
u
S
u
jJ
(13)
q
j
q
j
11
,
,
q
j
<
q
j
,
TT
tt
tt
S
u
S
u
U
q
J
u
Q
j
(14)
qos
q
j
q
j
()
,
)
(,
,
u
U
Q
u
qj
u
J
(15)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
GPO
N
an
d
V
-
band
mm
W
ave
in gree
n ba
ck
haul s
olu
ti
on f
or
5G u
lt
r
a
-
de
ns
e
network
(
Ayod
ej
i Ake
e
m Aj
an
i)
395
S
C
N
S
C
N
1
1
,
,
,
J
j
j
r
mm
j
W
N
W
W
q
q
Q
j
J
(16)
Eq
uation
(
10)
ens
ur
e
s
the
m
axi
m
u
m
transm
issi
on
po
wer
is
with
the
m
axi
m
u
m
value
(
m
a
x
tx
P
).
The
m
ax
i
m
u
m
val
ues
f
or
M
BS
an
d
pic
oce
ll
s
are
41
dBm
an
d
21
dBm
r
especti
vely
.
(
u
)
re
pr
ese
nts
us
e
r’
s
associat
ion
to a
p
arti
cular bas
e stat
ion
j
at ti
e
r
q
in (11).
T
his takes b
ina
ry v
al
ues
‘
0’ o
r
‘
1’
w
it
h
‘1’
w
he
n
us
e
r
u
is ass
ociat
ed
with
par
ti
cula
r base stat
io
n q,j
w
it
h
t
raffic
ty
pe
t
or 0 ot
herwise.
Eq
uation
(
12)
ens
ur
es
that
a u
ser
is
c
onnect
e
d
to
a
si
ng
le
ba
se
sta
ti
on
at
a
n
in
sta
nce.
E
quat
ion
(
13)
is
respo
ns
ible
f
or
m
aking
su
re
us
ers
are
c
onnect
wit
h
a
par
ti
cula
r
Sig
nal
to
In
t
e
rf
e
r
ence
to
N
oise
Ra
ti
o
(S
I
NR,
(
)
)
val
ue
w
hich
great
er
tha
n
the
th
res
ho
l
d
S
I
NR
(
ℎ
)
.
Eq
uation
(
14)
is
for
us
e
r’
s
as
so
ci
at
ion
to
the
near
est
ba
s
e
sta
ti
on
.
Eq
ua
ti
on
(
15),
is
to
ensure
the
Q
oS
requirem
ent
is
m
et
by
the
a
chieva
ble
dow
nlink
thr
oughput
f
or
a
us
er
(
(
)
)
is
gr
eat
er
tha
n
or
eq
ua
l
to
the
le
ast
achieva
ble
dow
nl
ink
th
rou
ghpu
t
fo
r
a
us
e
r
wh
ic
h
m
eet
s the QoS
re
qu
ire
m
ent o
f
the
us
e
r
(
η
(
)
)
.
E
quat
ion
(16
)
ens
ur
es
that
th
e
assigne
d
bandw
i
dth
(
)
to
N
um
ber
of
SC
N
s
need
e
d
(
)
is
gr
eat
er
t
han
or
at
le
ast
equ
al
to
t
he
e
xcess
band
width
require
d.
The
e
xcess
ba
ndwidt
h
re
qu
ired
is
the
dif
f
ere
nce
betwee
n
the
t
ot
al
bandw
i
dth
require
d
(
)
a
nd
the
ba
ndwi
dth
of
t
he
MB
S
(
).
This
m
od
el
has
in
cl
uded
bac
kh
a
ul
powe
r c
onsu
m
ption
f
or
ea
ch o
f
the
s
olu
ti
ons i
n
the
pro
blem
f
or
m
ulati
on
unli
ke
t
he
one
adopted
by
[
38]
.
This
is
part
ic
ularly
i
m
po
r
ta
nt
in
U
DN
s
cenari
o.
T
o
s
ol
ve
the
prob
le
m
in
(9
)
,
we
use
d
a
n
i
m
pr
oved
5G
UDN
e
nerg
y e
ff
ic
ie
nt
al
gorithm
.
2
.
2
.
Impr
ove
d 5G
UDN
e
n
ergy ef
fici
enc
y alg
orithm
An
al
gorithm
to
s
olv
e
t
he
prob
le
m
in
(
9)
i
s
prese
nted
he
re.
T
his
im
pr
oved
al
gorithm
is
to
al
lo
w
m
at
he
m
at
ic
a
l
s
olu
ti
on
pro
gra
m
m
ing
in
MATLAB.
T
his
al
gorithm
al
so
e
xp
at
ia
te
s
on
how
the
a
ddit
ion
s
a
nd
pu
tt
in
g
to
slee
p of SCs a
re
be
ing
done
.
Algori
th
m
:
Im
pro
ved 5G
UDN
E
nergy E
ff
ic
ie
ncy
1:
Input
:
Se
t
of
us
er
s,
uU
;
se
t
of
tr
af
fi
c
cl
as
s,
tT
;
Se
ts
of
5G
mu
lt
i
-
ti
er
Heterogenous Network,
q ϵ Q;
Set of Small cell base stations,
j ϵ J
.
2:
Output
:
Se
t
of
ac
ti
ve
SC
Ns
me
et
s
al
l
th
e
co
ns
tr
a
in
ts
in
Eq
ua
ti
on
s
(1
0
to
16
)
an
d
fulfils the optimum value for Equation (9):
S
C
N
;
(
)
j
J
N
n
J
3:
for
each hour of the day,
ℎ
∈
01, 02, …, 24
do
4:
while
the set of users is not empty:
U
do
5:
User
u
ca
lc
u
la
te
s
th
e
SI
NR
qj
()
u
fr
om
ea
c
h
SC
N
or
MB
S
th
en
as
so
ci
at
es
to
th
e
SC
N
or
MBS q,j that satisfies
th
qj
(
)
(
)
uu
based on QoS
requirements.
6:
end while
7: Calculate
r
m
W
(total required bandwidth)
8:
if
r
mm
WW
(MBS available bandwidth/spectrum)
then
9:
for
each SCN
jJ
do
10:
SC
N
0
j
N
(all SCNs are turned to sleep mode)
11:
end for
12:
Handover all users to the MBS
13:
else
14:
Calculate
r
e
m
m
W
W
W
15: Set of SCNs is empty:
J
;
S
C
N
()
N
n
J
;
16:
Com
pute total power consumption model
P
SCN
for all SCNs.
17:
Add the SCN q,j that gives the minimum
P
SCN
to the set:
J
J
q
j
;
S
C
N
()
N
n
J
;
18:
Us
in
g
th
e
cu
rr
en
t
se
t
of
SC
Ns
,
c
om
pu
te
th
e
cu
rr
en
t
ne
tw
or
k
po
we
r
co
ns
um
pt
io
n
A
N
-
B
H
H
e
t
N
e
t
UDN
P
subject to the constraints
19:
while
S
C
N
S
C
N
1
J
jj
e
j
W
W
N
do
20:
Temporarily add next SCN q,j with the lowest
P
SCN
to the set:
J
J
q
j
;
()
S
C
N
N
n
J
;
21:
Us
in
g
th
e
c
ur
re
nt
se
t
of
SC
Ns
J
,
co
mp
ut
e
te
mp
or
ar
y
ne
tw
or
k
po
we
r
c
on
su
mp
ti
on
A
N
-
B
H
H
e
t
N
e
t
UDN
(
t
e
m
p
)
P
subject to the constraints.
22:
if
A
N
-
B
H
A
N
-
B
H
H
e
t
N
e
t
U
D
N
H
e
t
N
e
t
UDN
(
t
e
m
p
)
PP
then
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
390
-
40
1
396
23:
Ac
ce
pt
th
e
SC
N
as
be
ca
us
e
it
ha
s
no
t
in
c
re
as
ed
th
e
po
we
r
co
ns
um
pt
io
n
of
the network
A
N
-
B
H
H
e
t
N
e
t
UDN
P
A
N
-
B
H
H
e
t
N
e
t
UDN
(
t
e
m
p
)
P
24:
else
25:
Reject the S
CN because i
t has increa
sed the powe
r consumed b
y the networ
k
\
J
J
q
j
;
S
C
N
()
N
n
J
;
26:
end if
27:
end while
28:
end if
29:
end for
This
al
gorithm
ens
ur
e
s
the
a
dd
it
io
n
of
sm
a
ll
cel
ls
wh
e
re
need
e
d
with
ou
t
com
pr
om
isi
ng
the
power
consum
ption
of
the
netw
ork
and
m
akes
othe
r
SCs
sle
e
p
w
hen
not
re
quire
d.
T
he
sim
ulatio
n
of
the
so
l
ution
i
s
done
i
n
M
ATL
AB.
3.
RESEA
R
CH MET
HO
D
Si
m
ulati
on
des
cripti
ons
are
presente
d
he
re,
wh
ic
h
inclu
des
the
ch
oice
of
par
am
et
ers,
pe
rfor
m
ance
m
et
rics and
t
he
d
esc
riptio
n of
the sce
nar
io
s.
3
.
1
.
Sim
ulat
i
on
of 5G
UDN
The
stu
dy
involve
d
dif
fer
e
nt
backhau
l
s
olu
ti
ons
us
in
g
a
vaila
ble
te
chnolo
gies
in
5G
ver
y
dense
netw
ork
,
VDN
and
UDN
sce
nar
i
os
.
Th
ere
f
or
e
,
the
resea
r
ch
re
qu
i
red
de
sign
of
5G
VDN
an
d
U
D
N
sc
enar
i
os
and
this
was
s
i
m
ulate
d
in
MATL
AB.
T
he
si
m
ulati
on
dec
isi
on
is
in
sp
ir
ed
by
w
ork
re
viewe
d
[36
-
39]
an
d
recent stu
dy in
5G
UDN
[
32, 40
-
43]
.
The
c
hoic
e o
f
sim
ulatio
n
is becau
se 5
G UDN is sti
l
l evo
lvin
g
an
d do
i
ng
a li
ve
ex
pe
rim
ent w
a
s
no
t a
va
il
able.
The
sim
ulati
on
pa
ram
et
ers
use
d
fo
ll
ows
the
work
of
[
38
]
.
The
norm
al
ised
hourl
y
traf
fic
pro
file
f
ro
m
ci
sco
was
use
d
to
this
eff
ect
as
pr
ese
nted
in
Figure
1.
The
perform
ance
m
et
rics
us
ed
a
re
powe
r
co
ns
um
ptio
n
in
W
at
ts a
nd e
nergy ef
fici
enc
y i
n
Mb
ps
/
W
at
t.
The
full
sim
ulati
on
pa
ram
e
te
rs
are
as
giv
e
n
in
Ta
ble 3.
Table
3
.
Sim
ul
at
ion
p
aram
et
e
rs
Para
m
eters
Para
m
eters
Desc
ri
p
tio
n
Valu
e
f
c
Access Net
wo
rk C
arr
ie
r
Fr
eq
u
en
cy
2
.0 GHz
W
m
Macr
o
-
ti
er
sp
ect
rum
1
0
M
Hz
Dy
n
a
m
i
c po
wer
of
m
ic
rocell b
ase sta
tio
n
4
3
dB
m
,
,
Dy
n
a
m
i
c tr
an
s
m
is
sio
n
po
wer
f
o
r
f
e
m
to
cell
1
7
dB
m
Fix
ed
po
wer
co
n
su
m
p
tio
n
of
m
ic
rocel
l bas
e statio
n
1
3
0
W
,
Fix
ed
po
wer
co
n
su
m
p
tio
n
of
f
e
m
to
cel
l
4
.8 W
∆
m
Load
-
d
ep
en
d
en
t Par
a
m
e
ter
4
.7
∆s
Slo
p
e of
load
-
d
ep
e
n
d
en
t po
wer
co
n
sum
p
tio
n
8
Sp
ectru
m
Allo
cati
o
n
Partition
ed
Tr
af
f
ic M
o
d
el Cap
acity
Fu
ll Bu
f
f
er
Env
iron
m
en
t
Su
b
-
u
rban
Maxi
m
u
m
nu
m
b
er
of
do
wn
lin
k
interfac
e
24
Α
W
eig
h
tin
g
para
m
e
ter
0
.9
/
/
Maxi
m
u
m
po
wer c
o
n
su
m
p
tio
n
o
f
switch
/
Po
wer
con
su
m
p
tio
n
of
on
e
d
o
wn
lin
k
interf
ace
/
Po
wer
con
su
m
p
t
io
n
of
o
n
e up
lin
k
i
n
terface
3
0
0
/
1
/
1
W
P
m
o
d
e
m
/ P
DS
L
A
M
/
Po
wer
co
n
su
m
p
t
io
n
of
VDSL
2
m
o
d
e
m
/
D
SLA
M
/
Fibr
e switch
5
/
8
5
/
3
0
0
W
P
SFP
/ P
S
FP
+
/ P
lo
w
-
c
/
P
h
i
g
h
-
c
Po
wer
co
n
su
m
p
t
io
n
of
SFP
m
o
d
u
le /
SFP+
m
o
d
u
le /
L
o
w /
Hig
h
p
o
wer
co
n
su
m
p
tio
n
r
eg
io
n
of
a
m
icro
wav
e anten
n
a
1
/
1
/
3
7
/
9
2
.5
W
/
/
g
/
P
o
Maxi
m
u
m
po
wer c
o
n
su
m
p
tio
n
of
GE
S /
Po
we
r
co
n
su
m
p
tio
n
o
f
m
i
crowav
e
switch
/ Power
co
n
su
m
p
ti
o
n
of
OLT
/
Po
we
r
co
n
su
m
p
tio
n
of
ONU
5
0
/
5
3
/
2
.9 / 5
W
/
/
/
/
Nu
m
b
e
r
o
f
po
rts per DS
LA
M
switch
/ Fibre swit
ch
/
G
E
S /
sp
litter
/
m
i
crowav
e
switch
1
6
/
2
4
/
1
2
/
2
4
/
16
Ag
m
ax
/
/
Maxi
m
u
m
tr
af
f
ic s
witch
can han
d
le /
Maxi
m
u
m
tr
an
s
m
i
ss
io
n
r
ate of
th
e interf
ace
/
Max
i
m
u
m
capacit
y
of
m
i
crowav
e
switch
2
4
/
1
0
/
3
6
Gb
p
s
ℎ
/
W
m
k
Back
h
au
l f
requ
en
c
y
us
in
g
m
m
W
av
e
/
Back
h
au
l link
bandwid
th
6
0
/
1
.76
GHz
D
Leng
th
of
eac
h
back
h
au
l link
1
0
0
m
L
P
/ L
Po
/ L
i /
L
sh
/
L
a
Tolerable p
ath
lo
ss
/
Path
los
s at 1
m
d
istan
ce /
i
m
p
l
e
m
en
tatio
n
los
s /
Sh
ad
o
win
g
los
s /
Atten
u
atio
n
los
s
1
0
8
/
6
8
/4 / 1 /
3
.2 d
B
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
GPO
N
an
d
V
-
band
mm
W
ave
in gree
n ba
ck
haul s
olu
ti
on f
or
5G u
lt
r
a
-
de
ns
e
network
(
Ayod
ej
i Ake
e
m Aj
an
i)
397
The
ch
oice
of
adoptin
g
norm
al
ise
d
hourl
y
l
oad
tra
ff
ic
is
to
m
ake
the
stud
y
easi
ly
adap
ta
ble
to
oth
e
r
sit
uations
.
T
hi
s
m
eans
diff
er
ent
stud
y
can
us
e
this
with
out
necessa
rily
us
in
g
the
am
o
un
t
of
traf
fic
use
d
he
re
or eve
n wit
h di
ff
e
ren
t
hourl
y prof
il
e.
Figure
1. N
orm
al
ise
d
t
raf
fic
l
oad
hourl
y p
r
of
il
e
[
2]
3
.
2
.
Des
cri
pt
i
on
of the
simu
lation
scen
ari
os
Ther
e
are
2
se
ts
of
sce
nar
i
os
i
m
ple
m
ented
her
e
.
T
he
first
set
is
the
re
pro
duct
ion
of
t
he
s
olu
t
io
n
rev
ie
wed
i
n
V
ery
De
ns
e
Net
work
scena
rio
us
in
g
eac
h
of
t
he
s
olu
ti
ons
hi
gh
li
ghte
d
in
S
ect
ion
1.
T
his
is
for
ver
ific
at
io
n
an
d
validat
io
n
pur
pose.
T
he
ot
her
set
is
to
te
st
the
best
2
so
luti
ons
in
the
U
DN
sce
na
rio
to
evaluate
their
perform
ance.
The
sce
nar
io
is
m
od
el
le
d
after
an
outd
oor
sit
uatio
n
li
ke
a
pu
blic
con
ce
rt
,
sp
ort
sta
diu
m
,
transpo
rtat
ion
hub
or
a
ny
oth
e
r
la
rg
e
ga
therin
g
w
her
e
the
tho
usa
nd
us
ers
re
quire
high
band
width si
m
ultaneo
us
ly
.
3
.
3
.
Perf
orm
ance
metric
s
The
perf
or
m
ance
m
e
tric
s
us
e
d
a
re
power
c
on
s
um
ption
i
n
W
at
ts
per
nor
m
al
iz
ed
hour
l
y
traff
ic
for
each
of the s
ol
utions
a
nd en
e
r
gy ef
fici
ency i
n
Mb
ps
/
Watt
.
4.
RESU
LT
S
A
ND
DI
SCUS
S
ION
S
The
re
su
lt
s
of
t
he
sim
ulati
on
are
bei
ng
pr
e
se
nted
in
t
his
sec
ti
on
as
well
as
the
detai
le
d
discuss
io
n
of
the r
es
ults.
4
.
1.
Res
ult
of
existin
g
s
olut
i
on
s
in
5G
VDN
The
re
viewe
d
so
luti
ons
we
re
i
m
ple
m
ented
i
n
VDN
net
wor
k
for
ve
rificat
ion
a
nd
vali
dation
purpose
.
The res
ult of th
e sim
ulatio
n
is
as prese
nted
i
n Fi
gure
2 an
d
F
igure
3
Figure
2. G
raph
of powe
r
c
on
su
m
ption
of
t
he
b
ac
kh
a
ul s
ol
utions
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
390
-
40
1
398
V
-
ba
nd
m
m
W
ave
an
d
PON
so
luti
on
ha
d
the
le
ast
power
con
s
um
ed
com
par
ed
to
othe
r
so
luti
ons
.
This
is
beca
use
there
ar
e
com
po
ne
nts
suc
h
as
Ethe
r
ne
t
switc
h,
fi
br
e
switc
h,
DS
L
acce
ss
m
ult
i
plexe
r
,
m
ic
ro
wa
ve
sw
it
ch,
Giga
bit
Ether
net
switc
h
in
the
pr
e
vi
ou
s
s
olu
ti
ons
respo
ns
ible
f
or
thei
r
high
-
powe
r
consum
ption
.
Con
si
der
i
ng
e
nergy
eff
ic
ie
nc
y,
wh
ic
h
is
the
data
rate
per
power
c
onsu
m
ed,
hi
gh
power
consum
ption
of
bac
khaul
s
olu
ti
on
will
resul
t
in
low
ene
r
gy
e
ff
ic
ie
ncy
an
d
lo
w
data
power
c
onsu
m
ption
will
m
ean h
ig
h
e
ne
rg
y e
ff
ic
ie
ncy
as sho
wn in Fi
gure
3.
Figure
3
.
Grap
h of
e
ne
rg
y e
ffi
ci
ency of the
backha
ul s
olu
ti
on
s
Si
m
il
ar
to
power
c
onsu
m
ption
gr
a
ph,
V
-
ba
nd
m
m
W
ave
an
d
PON
so
luti
on
had
the
bette
r
perform
ance
in
te
rm
s
of
e
nerg
y
eff
ic
ie
ncy
c
om
par
e
to
ot
hers.
T
his
m
ean
the
tw
o
so
l
utio
ns
a
re
well
s
uited
for
the
scena
rio.
T
he
te
ch
no
l
og
ie
s
e
m
plo
ye
d
ha
d
sim
il
ar
tim
eline
with
the
ne
twork
dep
l
oyed
in.
T
her
e
fore
,
it
is
i
m
per
at
ive
to
te
st
the
best
s
olu
ti
on
in
t
he
ne
w
netw
ork
scenari
o,
UDN
to
e
valuate
it
per
f
or
m
ance
for
su
it
abili
ty
. Th
e
r
es
ult o
f
this e
valuati
on is a
s
pr
ese
nted
in Se
ct
ion
4.2
.
4
.
2.
Res
ult
of
th
e b
est tw
o s
olut
i
on
s
(GP
O
N and
V
-
Ba
n
d mmW
ave)
in 5G
VDN an
d UD
N
scen
ar
ios
The
P
ON
a
nd
m
m
W
ave
so
lu
ti
on
was
s
im
ul
at
ed
in
U
D
N
s
cenari
o
an
d
w
as
com
par
ed
t
o
the
VDN
scenari
o
perf
orm
ance
pr
ese
nt
ed
in
4.1
.
Fi
gure
4
an
d
Figure
5
sho
w
t
he
powe
r
c
on
su
m
ption
a
nd
energy
eff
ic
ie
ncy
gr
a
phs
of the t
wo s
olu
ti
ons i
n bo
t
h VDN a
nd
U
DN scen
ario
s respect
ively
.
Figure
4. G
raph
of powe
r
c
on
su
m
ption
of
P
ON an
d
m
m
W
ave in
V
DN an
d U
DN
Figure
5. G
raph
of powe
r
c
on
su
m
ption
of
P
ON an
d
m
m
W
ave in
V
DN an
d U
DN
As
rev
e
al
ed
i
n
Table
2,
the
data
rate,
nu
m
ber
of
us
e
rs
a
nd
sit
es
in
U
DN
are
great
e
r
tha
n
t
hat
of
the
VDN.
T
hus,
there
is
an
in
crease
in
the
powe
r
co
nsum
e
d
by
the
so
luti
on
s
to
deliver
t
he
increase
d
da
ta
rate
there
by
le
adin
g
to
dro
p
in
en
erg
y
ef
fici
ency
as
sh
own
in
F
igure
5.
H
owe
ver,
UDN
net
work
is
ex
pect
ed
to
be
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
GPO
N
an
d
V
-
band
mm
W
ave
in gree
n ba
ck
haul s
olu
ti
on f
or
5G u
lt
r
a
-
de
ns
e
network
(
Ayod
ej
i Ake
e
m Aj
an
i)
399
gr
ee
ner
tha
n
VDN
[44]
tha
t
is
le
ss
powe
r
co
nsum
ption
an
d
hi
gher
e
nergy
ef
fici
en
cy
.
Th
us
,
t
his
open
a
resear
ch
a
rea
to
in
vestigat
e
ot
he
r
te
ch
nolo
gies
li
ke
NG
-
P
ON
2
a
nd
E
-
B
and
m
m
W
ave
in
U
D
N
sce
nari
o
to
know if
they c
an deli
ver the
desire
d per
for
m
ance.
5.
CONCL
US
I
O
N
GPON
a
nd
V
-
band
m
m
W
ave
backha
ul
so
l
ut
ion
s
wer
e
in
ve
sti
gated
in
5G
U
D
N
scena
r
io
.
T
he
res
ult
rev
eal
e
d
that
GPON
an
d
V
-
ba
nd
m
m
W
ave
s
olu
ti
ons
ou
t
perform
ed
oth
e
r
e
xisti
ng
so
l
ution
s
in
V
D
N
scenari
os
,
but
the
pe
rfo
rm
ance
dro
pp
e
d
in
U
DN
sce
na
rio
t
her
e
by
le
adin
g
to
a
sug
gestio
n
of
in
vestigat
i
on
of
oth
e
r
te
ch
nolo
gies
in
order
to
achie
ve
gr
ee
ne
r
perform
ance
in
5G
UDN
sc
enar
i
o.
Furthe
r
res
earch
is
on
go
i
ng
on
in
vestigat
io
n
of
NG
-
P
ON
2
and
E
-
Ba
nd
m
m
W
ave
sol
ution
s
in
ser
vi
ng
as
gr
e
en
ba
ckh
a
ul
so
l
ution
f
or
5G UD
N.
ACKN
OWLE
DGE
MENTS
Sp
eci
al
apprec
ia
ti
on
to
Africa
n
U
nion
Com
m
issi
on
for
spon
s
ori
ng
the
re
search
,
Japa
n
I
nter
nationa
l
Cooperati
on
A
gen
cy
(J
ICA
)
as
well
as
t
he
sta
ff
of
P
A
N
Africa
n
U
nive
rsity
In
sti
tute
f
or
Ba
sic
Scie
nces
Tech
no
l
og
y
a
nd
Inn
ovat
ion,
PAUI
STI,
Ken
ya
.
We
t
hank
Dr
Kibe
t
Lang
et
,
c
oor
din
at
or
,
Ele
ct
rical
En
gin
eeri
ng,
P
AUIST
I
,
Dr
A
kande
Oluwol
e
(Lado
ke
A
ki
nto
la
U
niv
e
rsity
of
Te
c
hnology,
Ni
ger
ia
)
,
En
gr.
Olum
ide
Aj
ay
i
(Adeleke
U
ni
ver
sit
y,
Ni
ger
i
a)
,
As
so
ci
at
e
P
rofess
or
Md
M
unjure
M
ow
la
(Rajs
ha
hi
Un
i
ve
rsity
of Enginee
rin
g an
d
Tec
hnol
ogy, Ba
ng
la
des
h)
,
our fam
i
li
es
and
fr
ie
nds
for t
heir
s
uppo
rt.
REFERE
NCE
S
[1]
R.
Vannit
h
amb
y
and
S.
Ta
l
w
ar,
“
Towa
rds
5G
Applic
a
ti
ons,
R
equi
rement
s
and
Candi
da
e
Tech
nologi
es
,”
John
Wil
e
y
&
Sons,
Lt
d
,
201
6
.
[2]
Cisco
S
y
stems
I
ncor
pora
t
ion,
“
C
isco
Annual
In
ternet
Repor
t
(201
8
–
2023)
W
hit
e P
ape
r,
”
2020.
[O
nli
ne]
.
Avai
la
bl
e
:
htt
ps://
ww
w.cis
c
o.
com/c/en/us/so
lut
ions/c
o
ll
a
te
r
al/execut
iv
e
-
per
sp
ec
t
ive
s/annu
al
-
i
nte
rne
t
-
rep
ort
/whit
e
-
p
ape
r
-
c11
-
7414
90.
html.
[3]
J.
Saunder
s
an
d
N.
Marshal
l
,
“
Mobile
bac
k
haul
opt
ions
-
Spect
rum
ana
l
ysis
and
rec
om
m
enda
ti
ons
,
”
G
SM
Associat
ion
,
201
8.
[4]
R.
Cháve
z
-
San
ti
ago
,
et
al.
,
“
5G:
The
Con
ver
gence
of
W
ire
le
ss
Comm
unic
at
ions,
”
Wirel
ess
Pe
rs
onal
Comm
un
ic
ati
ons
,
vol
.
83
,
no
.
3
,
p
p.
1617
-
1642
,
2
015.
[5]
M.
H.
Alsharif
and
R.
Nordin
,
“
Evol
uti
on
towa
rds
fift
h
gen
era
t
ion
(5G)
wire
l
e
ss
net
works
:
Curre
nt
tr
ends
an
d
c
hallenge
s
in
th
e
depl
o
y
m
ent
of
m
il
li
m
et
re
wav
e,
m
assive
MIM
O,
and
sm
al
l
cells,
”
Tel
ec
omm
u
n
ic
aton
Syst
ems
,
vol.
64
,
no
.
4
,
pp
.
617
-
637
,
2017
.
[6]
M.
Agiwal,
et
a
l.
,
“
Next
Gene
r
at
ion
5GW
ire
le
s
s
Networks:
A
Com
pre
hensive
Surve
y
,
”
IEEE
Comm
un
ic
ati
ons
Surv
ey
and
T
u
to
rials
,
vol
.
18
,
no
.
3
,
pp
.
1617
-
16
54,
2016
.
[7]
T.
Q.
Duong,
et
al
.
,
“
Ultra
-
Den
se
Networks
f
or
5
G
a
nd
Be
y
on
d
Modell
ing
,
Anal
y
s
is,
a
nd
Ap
pli
c
at
ions
,”
Joh
n
Wil
e
y
&
Sons Lt
d
,
2019
.
[8]
V.
Poirot,
e
t
a
l.
,
“
Ene
rg
y
eff
ic
i
en
t
m
ult
i
-
conne
c
tivit
y
al
gori
thms
for
ult
r
a
-
dense
5
G
net
works
,
”
Wi
rel
ess
Net
works
,
vol.
26
,
no
.
3
,
pp
.
2207
-
2222
,
20
20.
[9]
C.
Dai
,
et
al.
,
“
A
Cluste
r
-
Based
Sm
al
l
Cell
On/
Off
Scheme
for
Ene
rg
y
Eff
i
cie
nc
y
Optimiz
at
io
n
in
Ultr
a
-
Dens
e
Networks
,
”
In
te
r
nati
onal
Con
fe
re
nce
on
Comm
uni
cat
ions a
nd
Networ
ki
ng
in
Chin
a,
pp
.
385
-
401
,
2019
.
[10]
M.
J.
Daa
s
,
“
Ene
rg
y
Man
age
m
e
nt
Fram
ework
f
or
5G
Ultra
-
Dense
Networks
Us
i
ng
Graph
The
or
y
,
”
IEEE
Acce
ss
,
vol.
7
,
pp
.
17531
3
-
175323,
20
19
.
[11]
Nokia,
“
Ultr
a
D
ense
Network
(
UD
N)
W
hit
e
Pa
per
,
”
Nok
ia
,
201
6.
[12]
R.
I.
Ron
y
,
et
al
.
,
“
Joint
ac
ce
s
s
-
bac
khaul
per
s
pec
t
ive
on
m
ob
il
ity
m
ana
g
ement
in
5G
net
works
,
”
2017
IEEE
Conf
ere
nc
e
on
S
tand
ards
for
Co
mm
un
ic
ati
ons a
nd
Net
work
ing
,
pp.
115
-
120
,
20
17.
[13]
K.
M.
S.
Huq
an
d
J.
Rodriguez,
“
Bac
khau
li
ng/Fro
ntha
uli
ng
for
Fut
ure
W
ire
l
ess
S
y
stems
,”
John
W
iley
&
Son
s,
Ltd
,
2017
.
[14]
T.
and
I.
Ajani,
et
al.
,
“
A
Revi
ew
of
Gree
n
Bac
khaul
Soluti
ons
for
5G,”
in
8th
Inte
rnational
Co
nfe
renc
e
on
Mobil
e
e
-
Serv
i
ce
s (
ICoMe
S)
,
pp.
49
-
52
,
2019
.
[15]
A.
L.
Rezaa
ba
d,
et
al.
,
“
Ultr
a
-
Dense
5G
S
m
al
l
Cel
l
Depl
o
y
m
ent
for
Fiber
and
W
ire
le
s
s
Bac
khaul
-
Aw
ar
e
Infra
struct
ur
es,
”
IEE
E
Tr
ans
act
io
ns on
Veh
ic
u
lar
Technol
ogy
,
vo
l. 67, no. 12, pp. 1
2231
-
12243,
20
18.
[16]
The
Bro
adba
nd
Forum
,
“
The
Fut
ure
of
Pass
ive
O
pti
c
al
Ne
tworkin
g
is He
r
e
NG
-
PO
N2,”
2018.
[17]
J.
G.
Andrews
,
et
al
.
,
“
Modeli
ng
and
Anal
y
z
i
ng
Mill
imete
r
W
ave
Cel
lular
S
y
stems
,
”
IEEE
Tr
an
s
act
ions
on
Comm
un
ic
ati
ons
,
vol
.
65
,
no
.
1
,
p
p.
403
-
430
,
201
7.
[18]
Y.
Niu
,
et
a
l.
,
“
Ene
rg
y
-
Eff
ic
i
ent
Scheduling
for
m
mWave
Bac
kh
aul
ing
of
Sm
al
l
Cel
ls
in
Het
ero
g
ene
ous
Ce
ll
ul
ar
Networks,”
I
EEE
Tr
ans
act
ions
on
Veh
ic
ular
Te
chnol
ogy
,
vo
l. 6
6,
no
.
3
,
pp
.
267
4
-
2687,
2017
.
[19]
P.
Huang
and
K.
Ps
ounis,
“
Optimal
b
ac
khau
l
ing
for
d
ense
s
m
al
l
-
cell
d
ep
lo
y
m
ent
s
using
m
mWave
li
nks,
”
Comput
er
Comm
un
ic
ati
on
s
,
vol
.
138
,
pp
.
32
-
44
,
2019.
Evaluation Warning : The document was created with Spire.PDF for Python.