I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
,
p
p
.
4
9
9
3
~
5
0
0
7
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
14
.i
6
.
p
p
4
9
9
3
-
5
0
0
7
4993
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Dy
na
mic serv
ice
-
a
wa
re net
wo
rk
se
lection f
ra
mewo
r
k f
o
r mul
ti
-
o
bject
iv
e opt
imiz
a
tion in 5
G
-
a
dv
a
nced he
terog
eneo
us wireless
networks
B
ha
v
a
na
Srini
v
a
s
1
,
Na
dig
Vi
j
a
y
endra
Um
a
Reddy
2
1
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s
a
n
d
C
o
m
mu
n
i
c
a
t
i
o
n
En
g
i
n
e
e
r
i
n
g
,
N
e
w
H
o
r
i
z
o
n
C
o
l
l
e
g
e
o
f
E
n
g
i
n
e
e
r
i
n
g
,
B
a
n
g
a
l
o
r
e
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
A
r
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
,
N
e
w
H
o
r
i
z
o
n
C
o
l
l
e
g
e
o
f
E
n
g
i
n
e
e
r
i
n
g
,
B
a
n
g
a
l
o
r
e
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
an
3
,
2
0
2
5
R
ev
is
ed
Au
g
2
,
2
0
2
5
Acc
ep
ted
Sep
7
,
2
0
2
5
Th
e
i
n
c
re
a
sin
g
c
o
m
p
lex
i
ty
o
f
h
e
t
e
ro
g
e
n
e
o
u
s
wire
les
s
n
e
tw
o
rk
s
(H
WNs)
a
n
d
th
e
d
iv
e
rse
re
q
u
irem
e
n
ts
o
f
m
o
b
i
li
ty
p
a
tt
e
r
n
s
a
n
d
se
rv
ice
c
las
se
s
n
e
c
e
ss
it
a
te
a
d
v
a
n
c
e
d
so
l
u
ti
o
n
s fo
r
n
e
two
r
k
s
e
lec
ti
o
n
a
n
d
re
so
u
rc
e
o
p
t
imiz
a
ti
o
n
.
Ex
isti
n
g
m
o
d
e
ls
o
f
ten
fa
ll
sh
o
rt
in
a
d
d
re
ss
in
g
d
y
n
a
m
ic
m
o
b
il
it
y
sc
e
n
a
rio
s
a
n
d
se
rv
ice
d
iffere
n
ti
a
ti
o
n
,
lea
d
i
n
g
t
o
in
e
ffic
ien
c
ies
in
re
so
u
rc
e
a
ll
o
c
a
ti
o
n
,
s
u
b
o
p
ti
m
a
l
th
ro
u
g
h
p
u
t,
a
n
d
i
n
c
re
a
se
d
late
n
c
y
.
To
o
v
e
rc
o
m
e
th
e
se
li
m
it
a
ti
o
n
s,
t
h
is
stu
d
y
p
ro
p
o
se
s a
d
y
n
a
m
ic se
rv
ice
-
a
wa
r
e
n
e
two
rk
se
lec
to
r
(D
S
AN
S
)
fra
m
e
wo
rk
fo
r
5G
-
a
d
v
a
n
c
e
d
e
n
v
ir
o
n
m
e
n
ts.
Th
e
fra
m
e
wo
rk
in
teg
ra
tes
a
n
a
d
a
p
ti
v
e
d
e
e
p
d
e
c
isio
n
n
e
tw
o
rk
(AD
DN
)
fo
r
m
u
lt
i
-
o
b
jec
ti
v
e
o
p
t
imiz
a
ti
o
n
,
a
d
d
re
ss
in
g
c
rit
ica
l
q
u
a
l
it
y
o
f
se
rv
ice
(Q
o
S
)
m
e
tri
c
s
su
c
h
a
s
t
h
ro
u
g
h
p
u
t,
d
e
lay
,
a
n
d
e
n
e
rg
y
e
fficie
n
c
y
w
h
il
e
e
n
h
a
n
c
in
g
q
u
a
li
t
y
o
f
e
x
p
e
rien
c
e
(
Qo
E)
f
o
r
a
p
p
li
c
a
ti
o
n
s
li
k
e
e
n
h
a
n
c
e
d
m
o
b
il
e
b
ro
a
d
b
a
n
d
(
e
M
BB
)
,
u
lt
ra
-
re
l
iab
le
lo
w
late
n
c
y
c
o
m
m
u
n
ica
ti
o
n
(
URL
LC
)
,
a
n
d
in
ter
n
e
t
o
f
th
i
n
g
s
(
I
o
T
).
Th
e
DSANS
fra
m
e
wo
rk
d
y
n
a
m
ica
ll
y
a
d
a
p
ts
t
o
m
o
b
il
it
y
p
a
tt
e
rn
s
a
n
d
v
a
r
y
in
g
n
e
two
r
k
c
o
n
d
i
ti
o
n
s,
e
n
s
u
rin
g
e
fficie
n
t
r
e
so
u
rc
e
e
stim
a
ti
o
n
a
n
d
o
p
ti
m
a
l
n
e
two
r
k
se
lec
ti
o
n
.
S
im
u
latio
n
re
su
l
ts
h
i
g
h
li
g
h
t
i
ts
su
p
e
rio
rit
y
,
a
c
h
iev
in
g
u
p
to
2
5
%
imp
ro
v
e
m
e
n
t
i
n
t
h
ro
u
g
h
p
u
t
a
n
d
a
1
5
%
re
d
u
c
ti
o
n
i
n
late
n
c
y
c
o
m
p
a
re
d
to
sta
te
-
of
-
th
e
-
a
rt
a
lg
o
rit
h
m
s.
T
h
e
se
fin
d
in
g
s
v
a
li
d
a
te
DSANS
a
s
a
ro
b
u
st
so
lu
ti
o
n
fo
r
m
it
ig
a
ti
n
g
th
e
li
m
it
a
ti
o
n
s
o
f
e
x
isti
n
g
m
o
d
e
ls,
o
p
ti
m
izi
n
g
n
e
two
rk
p
e
rfo
rm
a
n
c
e
,
a
n
d
m
e
e
ti
n
g
th
e
strin
g
e
n
t
d
e
m
a
n
d
s
o
f
n
e
x
t
-
g
e
n
e
ra
t
io
n
HWNs
.
K
ey
w
o
r
d
s
:
Dee
p
lear
n
in
g
D
y
n
a
m
i
c
s
e
r
v
ic
e
-
a
w
a
r
e
n
et
w
o
r
k
s
e
l
e
ct
o
r
H
e
t
e
r
o
g
e
n
e
o
u
s
w
i
r
e
l
e
s
s
n
e
t
w
o
r
k
s
Ma
ch
in
e
lear
n
in
g
Qu
ality
of
s
er
v
ice
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
B
h
av
an
a
Srin
iv
as
Dep
ar
tm
en
t o
f
E
lectr
o
n
ics an
d
C
o
m
m
u
n
icatio
n
E
n
g
in
ee
r
i
n
g
,
New
Ho
r
izo
n
C
o
lleg
e
o
f
E
n
g
i
n
ee
r
in
g
New
Ho
r
izo
n
Kn
o
wled
g
e
Par
k
,
B
ellan
d
u
r
Ma
in
R
o
a
d
,
Nea
r
Ma
r
ath
ah
alli,
B
en
g
alu
r
u
5
6
0
1
0
3
,
I
n
d
ia
E
m
ail:
b
h
av
an
as_
1
2
@
r
ed
if
f
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
ex
p
o
n
e
n
tial
g
r
o
wth
o
f
th
e
in
ter
n
et
o
f
t
h
in
g
s
(
I
o
T
)
an
d
th
e
in
c
r
ea
s
in
g
in
teg
r
atio
n
o
f
co
n
n
ec
te
d
d
ev
ices
in
to
d
aily
life
h
av
e
led
to
th
e
p
r
o
life
r
atio
n
o
f
h
e
ter
o
g
en
eo
u
s
wir
eless
n
etwo
r
k
s
(
HW
Ns).
T
h
ese
n
etwo
r
k
s
co
m
b
in
e
m
u
ltip
le
r
ad
io
ac
ce
s
s
tech
n
o
lo
g
ies
(
R
AT
s
)
s
u
ch
as
4
G/5
G,
Wi
-
Fi
,
a
n
d
lo
w
-
p
o
wer
wid
e
ar
ea
n
etwo
r
k
s
(
L
PW
ANs)
to
en
ab
le
s
ea
m
less
co
m
m
u
n
icati
o
n
.
W
ith
th
e
r
a
p
id
ev
o
lu
tio
n
t
o
war
d
5
G
-
a
d
v
an
ce
d
an
d
b
e
y
o
n
d
,
HW
Ns
h
av
e
em
er
g
ed
as
a
k
ey
en
a
b
ler
f
o
r
f
u
tu
r
e
s
m
ar
t
ap
p
licatio
n
s
,
in
cl
u
d
in
g
a
u
to
n
o
m
o
u
s
v
eh
icles,
s
m
ar
t
cities,
in
d
u
s
tr
ial
au
to
m
atio
n
,
a
n
d
c
r
itical
I
o
T
-
d
r
iv
en
in
f
r
astru
ctu
r
es
[
1
]
.
W
h
ile
HW
Ns
p
r
o
v
id
e
f
lex
ib
ilit
y
an
d
s
ca
lab
ilit
y
,
th
e
ir
h
eter
o
g
en
eity
also
i
n
tr
o
d
u
c
es
s
ig
n
if
ican
t
ch
allen
g
es
in
m
an
ag
in
g
n
etwo
r
k
r
eso
u
r
ce
s
an
d
e
n
s
u
r
in
g
s
ea
m
l
ess
h
an
d
o
v
er
o
p
er
atio
n
s
,
p
ar
ti
cu
lar
ly
f
o
r
I
o
T
d
ev
ices
ch
ar
a
cter
ized
b
y
d
iv
er
s
e
m
o
b
ilit
y
b
eh
av
io
r
s
an
d
s
er
v
i
ce
r
eq
u
ir
em
en
ts
.
Dev
ices
r
an
g
in
g
f
r
o
m
s
tatic
s
en
s
o
r
s
to
h
ig
h
ly
m
o
b
ile
n
o
d
es
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
4
9
9
3
-
5
0
0
7
4994
ex
h
ib
it
v
a
r
y
in
g
th
r
o
u
g
h
p
u
t,
laten
cy
,
an
d
e
n
er
g
y
c
o
n
s
tr
ain
ts
,
d
em
an
d
in
g
ad
ap
ti
v
e
s
o
l
u
tio
n
s
ca
p
a
b
le
o
f
ad
d
r
ess
in
g
th
ese
d
y
n
am
ic
s
ce
n
ar
io
s
.
T
r
a
d
itio
n
al
n
etwo
r
k
s
e
lectio
n
an
d
h
an
d
o
v
e
r
s
tr
ateg
i
es,
o
f
ten
r
elian
t
o
n
s
ig
n
al
s
tr
en
g
th
-
b
ased
m
etr
ics
s
u
ch
as
r
ec
eiv
ed
s
ig
n
al
s
tr
en
g
t
h
in
d
icato
r
(
R
SS
I
)
o
r
s
ig
n
al
-
to
-
in
ter
f
er
en
ce
-
p
lu
s
-
n
o
is
e
r
atio
(
SIN
R
)
,
f
ail
to
o
p
t
im
ize
s
y
s
tem
p
er
f
o
r
m
an
ce
u
n
d
er
s
u
ch
h
eter
o
g
e
n
eo
u
s
a
n
d
r
eso
u
r
ce
-
co
n
s
tr
ain
e
d
co
n
d
itio
n
s
[
2
]
,
[
3
]
.
T
o
m
ee
t
th
e
s
tr
in
g
en
t
q
u
ality
o
f
s
e
r
v
ice
(
Qo
S)
a
n
d
q
u
ality
o
f
ex
p
er
ien
ce
(
Qo
E
)
r
eq
u
ir
em
e
n
ts
,
n
ex
t
-
g
e
n
er
atio
n
HW
Ns m
u
s
t b
e
ac
co
m
m
o
d
ate
d
.
Div
er
s
e
m
o
b
ilit
y
p
atter
n
s
(
e
.
g
.
,
s
tatio
n
ar
y
,
lo
w
-
s
p
ee
d
,
a
n
d
h
ig
h
-
s
p
ee
d
u
s
er
s
)
th
at
s
ig
n
if
ican
tly
im
p
ac
t
h
an
d
o
v
er
f
r
e
q
u
en
c
y
,
r
eso
u
r
ce
allo
ca
tio
n
,
an
d
s
ig
n
a
lin
g
o
v
er
h
ea
d
.
Mu
ltip
le
s
er
v
ice
class
e
s
s
u
ch
as
r
ea
l
-
tim
e
v
id
eo
,
laten
cy
-
s
en
s
itiv
e
v
ir
tu
al
r
ea
lity
(
VR
)
,
an
d
lo
w
-
p
o
wer
in
d
u
s
tr
ial
in
ter
n
et
o
f
th
in
g
s
(
I
I
o
T
)
ap
p
licatio
n
s
,
ea
ch
im
p
o
s
in
g
u
n
iq
u
e
d
em
an
d
s
o
n
n
etwo
r
k
t
h
r
o
u
g
h
p
u
t,
d
elay
,
an
d
e
n
er
g
y
e
f
f
icien
cy
.
E
x
is
tin
g
h
an
d
o
v
er
an
d
n
etwo
r
k
s
elec
tio
n
m
eth
o
d
o
lo
g
ies
lack
th
e
n
e
ce
s
s
ar
y
in
tellig
en
ce
to
d
y
n
am
ically
ad
ap
t
to
th
ese
co
m
p
lex
ities
.
Ma
n
y
co
n
v
en
t
io
n
al
ap
p
r
o
ac
h
es
s
u
f
f
er
f
r
o
m
h
ig
h
s
ig
n
alin
g
o
v
er
h
ea
d
,
f
r
eq
u
en
t
h
a
n
d
o
v
er
f
ailu
r
es,
an
d
in
ad
eq
u
ate
r
eso
u
r
ce
o
p
tim
izatio
n
,
p
a
r
ticu
lar
l
y
in
en
v
ir
o
n
m
en
ts
with
d
en
s
e
u
s
er
tr
af
f
ic
an
d
lim
ited
b
an
d
wid
th
av
ailab
ilit
y
.
C
en
tr
alize
d
alg
o
r
ith
m
s
,
w
h
ile
ef
f
ec
tiv
e
in
s
m
all
-
s
ca
le
s
ce
n
ar
io
s
,
b
ec
o
m
e
co
m
p
u
tatio
n
ally
p
r
o
h
ib
itiv
e
a
n
d
im
p
r
ac
tical
f
o
r
lar
g
e
-
s
ca
le
HW
Ns
d
u
e
to
s
ca
lab
ilit
y
an
d
p
r
iv
ac
y
c
o
n
ce
r
n
s
.
Me
an
wh
ile,
h
eu
r
is
tic
-
b
ased
m
eth
o
d
s
f
ail
to
d
eliv
e
r
o
p
tim
al
s
o
lu
tio
n
s
in
r
ea
l
-
tim
e
f
o
r
d
y
n
am
ic
a
n
d
m
u
lti
-
s
er
v
ice
en
v
ir
o
n
m
e
n
ts
[
4
]
.
R
ec
en
t
ad
v
an
ce
m
en
ts
in
m
a
ch
in
e
lear
n
i
n
g
(
ML
)
a
n
d
d
is
tr
ib
u
ted
o
p
tim
izatio
n
tech
n
iq
u
es
h
av
e
u
n
lo
ck
ed
n
ew
p
o
s
s
ib
ilit
ie
s
f
o
r
ad
d
r
ess
in
g
th
e
co
m
p
le
x
ch
allen
g
es
f
ac
ed
in
HW
Ns
.
T
h
ese
n
etwo
r
k
s
,
ch
ar
ac
te
r
ized
b
y
d
iv
e
r
s
e
co
n
n
ec
tiv
ity
o
p
tio
n
s
an
d
an
in
c
r
ea
s
in
g
n
u
m
b
er
o
f
I
o
T
d
e
v
ices,
m
u
s
t
m
ee
t
th
e
s
tr
in
g
e
n
t
r
e
q
u
ir
em
en
ts
o
f
Q
o
S
an
d
Qo
E
u
n
d
e
r
d
y
n
am
ic
a
n
d
r
eso
u
r
ce
-
co
n
s
tr
ain
ed
co
n
d
itio
n
s
.
T
h
e
n
ee
d
f
o
r
ef
f
icien
t
r
eso
u
r
c
e
m
an
ag
em
en
t
an
d
ad
a
p
tiv
e
d
ec
is
io
n
-
m
ak
in
g
is
p
ar
ticu
lar
ly
cr
itical,
as
HW
N
s
s
er
v
e
d
ev
ices
ex
h
ib
itin
g
v
ar
y
in
g
m
o
b
ilit
y
b
eh
av
io
r
s
an
d
s
er
v
ice
d
em
a
n
d
s
.
T
r
ad
itio
n
a
l
n
etwo
r
k
s
elec
tio
n
an
d
r
eso
u
r
ce
allo
ca
tio
n
m
eth
o
d
s
,
o
f
ten
r
elian
t
o
n
s
tatic
th
r
esh
o
ld
s
o
r
ce
n
t
r
alize
d
ar
ch
it
ec
tu
r
es,
s
tr
u
g
g
le
to
h
an
d
le
th
e
r
ap
id
ly
ch
a
n
g
in
g
n
etwo
r
k
co
n
d
itio
n
s
[
5
]
–
[
7
]
.
C
en
tr
alize
d
ap
p
r
o
ac
h
es
f
ac
e
ch
allen
g
es
s
u
ch
as
s
ca
lab
ilit
y
is
s
u
e
s
,
in
cr
ea
s
ed
s
ig
n
alin
g
o
v
er
h
ea
d
,
an
d
co
n
ce
r
n
s
o
v
er
p
r
iv
ac
y
w
h
en
s
h
ar
in
g
r
aw
d
ata
ac
r
o
s
s
n
o
d
es.
I
n
a
d
d
itio
n
,
th
e
h
ete
r
o
g
en
eity
o
f
HW
Ns
in
tr
o
d
u
ce
s
s
ig
n
if
ican
t
ch
allen
g
es
in
m
an
ag
i
n
g
d
ev
ices
o
p
er
atin
g
u
n
d
er
d
i
f
f
er
en
t
m
o
b
ilit
y
p
atter
n
s
r
an
g
in
g
f
r
o
m
s
tatio
n
ar
y
n
o
d
es
to
h
ig
h
-
s
p
ee
d
m
o
b
ile
d
ev
ices
an
d
ca
ter
in
g
to
m
u
ltip
le
s
er
v
ice
class
e
s
,
in
clu
d
in
g
laten
cy
-
s
en
s
itiv
e,
th
r
o
u
g
h
p
u
t
-
in
ten
s
iv
e,
an
d
en
er
g
y
-
co
n
s
tr
ain
ed
ap
p
licatio
n
s
.
W
h
ile
d
is
tr
ib
u
ted
lear
n
i
n
g
tech
n
iq
u
es
h
av
e
em
er
g
e
d
as
p
r
o
m
is
in
g
s
o
lu
tio
n
s
to
im
p
r
o
v
e
s
ca
lab
ilit
y
an
d
ad
ap
tab
ilit
y
,
th
e
in
teg
r
atio
n
o
f
s
u
ch
m
eth
o
d
s
f
o
r
ac
h
i
ev
in
g
m
u
lti
-
o
b
jectiv
e
p
er
f
o
r
m
an
ce
o
p
tim
izatio
n
r
em
ain
s
lim
ited
.
E
x
is
tin
g
ap
p
r
o
ac
h
es
o
f
ten
f
ail
to
ac
co
u
n
t
f
o
r
th
e
co
m
b
in
e
d
im
p
ac
t
o
f
m
o
b
ilit
y
b
eh
av
io
r
s
an
d
s
er
v
ice
-
s
p
ec
if
ic
co
n
s
tr
ain
ts
o
n
n
etwo
r
k
p
er
f
o
r
m
an
ce
,
lea
d
in
g
to
f
r
eq
u
e
n
t
h
an
d
o
v
e
r
f
ailu
r
es,
in
ef
f
icien
t
r
eso
u
r
ce
u
tili
za
tio
n
,
an
d
d
e
g
r
ad
ed
u
s
er
s
atis
f
ac
tio
n
.
T
h
e
a
b
s
en
ce
o
f
d
y
n
am
ic
lo
a
d
b
ala
n
cin
g
m
ec
h
an
is
m
s
f
u
r
th
er
e
x
ac
er
b
ates th
ese
is
s
u
es,
p
ar
ticu
lar
ly
d
u
r
in
g
co
n
g
esti
o
n
an
d
h
ig
h
-
lo
ad
c
o
n
d
itio
n
s
[
8
]
.
T
h
e
r
ap
i
d
p
r
o
life
r
atio
n
o
f
I
o
T
d
ev
ices
in
HW
Ns
h
as
in
tr
o
d
u
ce
d
s
ig
n
if
ica
n
t
ch
allen
g
es
in
en
s
u
r
in
g
s
ea
m
less
co
n
n
ec
tiv
ity
an
d
o
p
tim
al
p
er
f
o
r
m
an
ce
.
Dev
ices
with
d
iv
e
r
s
e
m
o
b
ilit
y
b
eh
a
v
io
r
s
r
an
g
in
g
f
r
o
m
s
tatio
n
ar
y
to
h
ig
h
-
s
p
ee
d
u
s
er
s
ex
p
er
ien
ce
f
r
e
q
u
en
t
h
an
d
o
v
er
s
,
lead
in
g
to
s
ig
n
alin
g
o
v
e
r
h
ea
d
an
d
d
eg
r
a
d
ed
p
er
f
o
r
m
an
ce
.
Ad
d
itio
n
ally
,
c
ater
in
g
to
m
u
ltip
le
s
er
v
ice
class
es,
s
u
ch
as
laten
cy
-
s
en
s
itiv
e
ap
p
licatio
n
s
(
e.
g
.
,
VR
)
an
d
en
er
g
y
-
c
o
n
s
tr
ain
ed
I
o
T
s
en
s
o
r
s
,
d
em
an
d
s
ad
ap
tiv
e
r
eso
u
r
ce
allo
ca
tio
n
t
o
b
alan
ce
Qo
S
a
n
d
Qo
E
.
E
x
is
tin
g
m
eth
o
d
s
lack
th
e
ca
p
ab
ilit
y
to
d
y
n
am
ically
o
p
tim
ize
n
etwo
r
k
s
elec
tio
n
an
d
r
eso
u
r
ce
esti
m
atio
n
wh
ile
ad
d
r
ess
in
g
th
ese
co
m
p
lex
ities
[
9
]
,
[
1
0
]
.
T
h
e
r
ef
o
r
e,
t
h
er
e
is
a
p
r
ess
in
g
n
ee
d
f
o
r
i
n
tellig
en
t,
m
u
lti
-
o
b
jectiv
e
o
p
tim
izatio
n
s
tr
ateg
ies
th
at
ca
n
ad
ap
t
to
v
ar
y
in
g
m
o
b
ilit
y
p
atter
n
s
an
d
s
er
v
ice
-
s
p
ec
if
ic
r
eq
u
ir
em
e
n
ts
,
en
s
u
r
in
g
ef
f
ici
en
t
r
eso
u
r
ce
u
tili
za
tio
n
,
r
ed
u
ce
d
laten
cy
,
an
d
im
p
r
o
v
e
d
u
s
er
s
atis
f
ac
t
io
n
in
HW
Ns.
Hi
s
s
tu
d
y
ad
d
r
ess
es
th
e
lim
itatio
n
s
o
f
ex
is
tin
g
n
etwo
r
k
s
elec
tio
n
an
d
r
eso
u
r
ce
o
p
tim
izatio
n
m
o
d
els
in
5G
-
ad
v
an
ce
d
HW
Ns.
T
h
e
k
ey
co
n
tr
ib
u
tio
n
s
ar
e
as f
o
llo
ws
:
‒
Pro
p
o
s
ed
f
r
am
ew
o
r
k
:
a
n
o
v
el
d
y
n
am
ic
s
er
v
ice
-
awa
r
e
n
etw
o
r
k
s
elec
to
r
(
DSANS)
f
r
am
ewo
r
k
d
esig
n
ed
to
o
p
tim
ize
n
etwo
r
k
p
e
r
f
o
r
m
an
ce
b
y
d
y
n
am
ically
ad
ap
tin
g
to
m
o
b
ilit
y
p
atter
n
s
a
n
d
s
er
v
ice
class
v
ar
iatio
n
s
.
‒
Mu
lti
-
o
b
jectiv
e
o
p
tim
izatio
n
:
in
teg
r
atio
n
o
f
an
ad
a
p
tiv
e
d
ee
p
d
ec
is
io
n
n
etwo
r
k
(
ADDN
)
to
b
alan
c
e
cr
itical
Qo
S
m
etr
ics
s
u
ch
as
t
h
r
o
u
g
h
p
u
t,
d
elay
,
an
d
en
er
g
y
ef
f
icien
cy
wh
ile
en
h
a
n
cin
g
Q
o
E
f
o
r
d
i
v
er
s
e
ap
p
licatio
n
s
lik
e
e
n
h
an
ce
d
m
o
b
ile
b
r
o
ad
b
an
d
(
eM
B
B
)
,
u
ltra
-
r
eliab
le
lo
w
laten
cy
co
m
m
u
n
icatio
n
(
UR
L
L
C
)
,
an
d
I
o
T
.
‒
Scalab
ilit
y
an
d
ef
f
icien
cy
:
d
ev
elo
p
m
en
t
o
f
alg
o
r
ith
m
s
f
o
r
ef
f
icien
t
r
eso
u
r
ce
esti
m
atio
n
an
d
ad
ap
tiv
e
n
etwo
r
k
s
elec
tio
n
,
e
n
s
u
r
in
g
s
c
alab
ilit
y
ac
r
o
s
s
v
ar
y
in
g
u
s
er
d
em
an
d
s
an
d
n
etwo
r
k
co
n
d
itio
n
s
.
‒
Per
f
o
r
m
an
ce
v
alid
atio
n
:
d
em
o
n
s
tr
ated
s
u
p
er
io
r
ity
o
f
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
th
r
o
u
g
h
s
im
u
latio
n
s
,
ac
h
iev
in
g
u
p
to
2
5
%
im
p
r
o
v
em
en
t
in
th
r
o
u
g
h
p
u
t
an
d
a
1
5
%
r
ed
u
ctio
n
in
laten
cy
co
m
p
ar
ed
to
s
tate
-
of
-
th
e
-
ar
t a
lg
o
r
ith
m
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Dyn
a
mic
s
ervice
-
a
w
a
r
e
n
etw
o
r
k
s
elec
tio
n
fr
a
mewo
r
k
fo
r
mu
lti
-
o
b
jective
o
p
timiz
a
tio
n
…
(
B
h
a
va
n
a
S
r
in
iva
s
)
4995
2.
RE
L
AT
E
D
WO
RK
Q
o
S
a
n
d
i
n
t
e
ll
i
g
e
n
t
f
l
o
w
c
o
n
tr
o
l
a
r
e
k
e
y
a
s
p
e
ct
s
i
n
n
et
w
o
r
k
o
p
t
i
m
i
z
at
i
o
n
.
A
cc
o
r
d
i
n
g
t
o
B
a
[
1
1
]
,
a
t
h
r
e
e
-
s
t
e
p
Q
o
S
-
f
o
r
e
c
a
s
t
i
n
g
s
ch
e
m
e
i
s
p
r
o
p
o
s
e
d
t
o
e
n
a
b
l
e
u
s
e
r
e
q
u
i
p
m
e
n
t
(
U
E
)
t
o
i
n
t
e
l
li
g
e
n
t
l
y
d
e
t
e
r
m
i
n
e
t
h
e
d
a
t
a
f
l
o
w
d
i
r
ec
t
i
o
n
b
as
e
d
o
n
n
e
t
w
o
r
k
c
h
a
r
a
c
t
e
r
i
s
ti
cs
a
n
d
l
o
a
d
d
i
s
t
r
i
b
u
t
i
o
n
a
c
r
o
s
s
n
o
d
e
s
.
T
h
i
s
a
p
p
r
o
a
c
h
p
r
i
o
r
i
t
i
z
es
l
o
a
d
b
a
l
a
n
c
i
n
g
a
n
d
u
s
e
r
f
ai
r
n
e
s
s
w
h
i
l
e
ass
i
g
n
i
n
g
d
a
t
a
t
o
s
e
c
o
n
d
a
r
y
n
o
d
e
s
(
SN
s
)
b
as
e
d
o
n
Q
o
S
r
e
q
u
i
r
e
m
e
n
t
s
a
n
d
t
h
e
a
v
e
r
a
g
e
t
r
a
n
s
m
is
s
i
o
n
c
a
p
a
b
il
i
ti
e
s
o
f
th
e
U
E
.
S
u
c
h
s
c
h
e
m
es
a
i
m
t
o
e
n
h
a
n
c
e
n
e
t
w
o
r
k
e
f
f
i
c
i
e
n
c
y
w
h
i
l
e
c
a
te
r
i
n
g
t
o
u
s
e
r
-
s
p
e
c
i
f
i
c
p
r
i
o
r
i
ti
es
.
I
n
t
h
e
c
o
n
t
e
x
t
o
f
n
e
t
w
o
r
k
s
e
l
e
c
t
i
o
n
s
t
r
ate
g
i
e
s
,
Ma
e
t
a
l
.
[
1
2
]
in
tr
o
d
u
ce
s
a
m
u
lti
-
ag
en
t
Q
-
l
ea
r
n
in
g
-
b
ased
ap
p
r
o
ac
h
,
r
ef
e
r
r
ed
to
as
th
e
m
u
ltiag
en
t
Q
-
lear
n
in
g
n
etwo
r
k
s
elec
tio
n
(
MA
QNS
)
alg
o
r
ith
m
.
T
h
is
m
eth
o
d
em
p
l
o
y
s
Nash
Q
-
lear
n
in
g
to
ac
h
ie
v
e
a
jo
in
t
o
p
tim
al
s
elec
tio
n
s
tr
ateg
y
th
at
im
p
r
o
v
es
s
y
s
tem
th
r
o
u
g
h
p
u
t
an
d
r
e
d
u
ce
s
u
s
er
b
lo
ck
in
g
wh
ile
m
ee
tin
g
th
e
s
tr
in
g
en
t r
eq
u
ir
e
m
en
ts
o
f
I
o
T
s
er
v
ices.
T
h
e
u
s
e
o
f
d
is
cr
ete
-
tim
e
Ma
r
k
o
v
ch
ain
s
f
o
r
m
o
d
elin
g
n
etwo
r
k
s
ele
ctio
n
,
co
u
p
le
d
with
tech
n
iq
u
es
lik
e
th
e
an
aly
tic
h
i
er
ar
ch
y
p
r
o
ce
s
s
(
AHP)
an
d
g
r
ay
r
elatio
n
al
an
aly
s
is
(
GR
A)
,
h
elp
s
ca
p
tu
r
e
u
s
er
p
r
ef
er
en
ce
s
an
d
en
h
a
n
ce
d
ec
is
io
n
-
m
ak
in
g
ac
cu
r
ac
y
.
T
h
e
ap
p
licatio
n
o
f
g
en
er
ativ
e
ad
v
er
s
ar
ial
n
etwo
r
k
s
(
GANs)
in
[
1
3
]
f
o
r
ce
ll
lo
ad
esti
m
atio
n
d
em
o
n
s
tr
ates
s
ig
n
i
f
ican
t
p
r
o
g
r
ess
in
5
G
c
o
m
m
u
n
icatio
n
n
etwo
r
k
s
.
T
h
is
ap
p
r
o
ac
h
esti
m
ates
ce
ll
lo
ad
b
ased
o
n
ter
m
in
al
-
m
e
asu
r
ed
wir
eless
in
f
o
r
m
atio
n
,
ad
d
r
ess
in
g
cr
itical
ch
allen
g
es
s
u
ch
as
lo
w
d
ata
tr
an
s
m
is
s
io
n
r
ates,
h
ig
h
s
ig
n
ali
n
g
c
o
s
ts
,
an
d
d
elay
s
in
tr
ad
iti
o
n
al
lo
a
d
esti
m
atio
n
m
eth
o
d
s
.
B
y
en
ab
lin
g
r
ea
l
-
ti
m
e,
ter
m
in
al
-
s
id
e
d
ec
is
io
n
-
m
ak
in
g
,
th
is
m
eth
o
d
im
p
r
o
v
es
o
v
er
all
n
etwo
r
k
p
er
f
o
r
m
an
ce
an
d
en
h
an
ce
s
u
s
e
r
ex
p
er
ie
n
ce
.
Fu
r
th
er
ad
v
a
n
cin
g
r
eso
u
r
ce
o
p
tim
izatio
n
,
C
ab
r
er
a
et
a
l.
[
1
4
]
p
r
esen
ts
SVOR
A,
a
n
o
v
el
ap
p
r
o
ac
h
in
teg
r
atin
g
v
ir
tu
alize
d
/o
p
en
r
ad
io
ac
ce
s
s
n
etwo
r
k
(
V/O
-
R
AN)
co
n
ce
p
ts
with
a
s
er
v
ice
-
b
ased
ar
ch
itectu
r
e
(
SB
A)
.
T
h
e
p
r
o
p
o
s
ed
d
elay
-
awa
r
e
en
er
g
y
e
f
f
icien
cy
-
b
ased
R
AN
s
elec
tio
n
alg
o
r
ith
m
(
DE
E
R
)
lev
er
ag
es
u
tili
ty
f
u
n
ctio
n
s
to
ac
co
u
n
t
f
o
r
Qo
S
m
etr
ics
s
u
ch
as
th
r
o
u
g
h
p
u
t,
r
eso
u
r
ce
b
l
o
ck
u
tili
za
tio
n
,
d
elay
,
an
d
en
er
g
y
co
n
s
u
m
p
tio
n
.
DE
E
R
’
s
f
lex
ib
ilit
y
allo
ws
it
to
ad
ap
t
to
v
ar
y
in
g
p
r
i
o
r
ities
,
in
clu
d
in
g
c
r
itical
d
elay
-
s
en
s
itiv
e
s
er
v
ices,
en
er
g
y
ef
f
icien
cy
,
o
r
a
b
alan
ce
d
o
p
tim
izatio
n
ap
p
r
o
ac
h
,
m
ak
in
g
it
h
ig
h
ly
a
p
p
licab
le
in
d
y
n
am
ic
n
etwo
r
k
s
ce
n
ar
i
o
s
.
I
n
th
e
s
a
m
e
v
ein
,
C
ab
r
er
a
et
a
l
[
1
4
]
e
m
p
h
asizes
SVOR
A’
s
u
tili
ty
in
o
p
tim
izin
g
en
er
g
y
co
n
s
u
m
p
tio
n
a
n
d
m
i
n
im
izin
g
co
m
m
u
n
icatio
n
d
elay
s
in
h
e
ter
o
g
en
eo
u
s
n
etwo
r
k
s
.
B
y
i
n
t
e
g
r
a
t
i
n
g
SB
A
w
i
t
h
R
A
N
s
e
le
c
t
i
o
n
m
e
c
h
a
n
is
m
s
,
t
h
i
s
a
p
p
r
o
a
c
h
d
e
m
o
n
s
t
r
at
e
s
t
h
e
a
b
i
l
i
t
y
t
o
a
d
d
r
es
s
t
h
e
d
e
m
a
n
d
s
o
f
m
o
d
e
r
n
c
o
m
m
u
n
i
c
a
t
i
o
n
s
y
s
t
e
m
s
t
h
r
o
u
g
h
r
o
b
u
s
t
a
n
d
a
d
a
p
t
a
b
l
e
s
o
l
u
ti
o
n
s
.
Z
h
u
e
t
a
l
.
[
1
5
]
p
r
o
p
o
s
es
R
E
M
NS
,
a
n
o
v
e
l
a
c
c
e
s
s
s
e
l
e
ct
i
o
n
m
e
c
h
a
n
is
m
t
a
il
o
r
e
d
f
o
r
I
o
T
s
e
r
v
i
c
es
i
n
5
G
h
et
e
r
o
g
e
n
e
o
u
s
n
e
t
w
o
r
k
s
.
T
h
is
m
ec
h
a
n
i
s
m
i
n
t
r
o
d
u
c
e
s
a
f
u
z
z
y
l
o
g
i
c
-
b
a
s
e
d
p
r
e
-
a
s
s
es
s
m
e
n
t
f
r
a
m
e
w
o
r
k
t
o
f
il
t
e
r
p
o
t
e
n
t
i
a
l
n
et
w
o
r
k
s
,
e
n
s
u
r
i
n
g
t
h
a
t
o
n
l
y
t
h
e
m
o
s
t
s
u
i
t
a
b
le
o
p
t
i
o
n
s
a
r
e
c
o
n
s
i
d
e
r
e
d
.
F
u
r
t
h
e
r
m
o
r
e
,
i
t
i
n
c
o
r
p
o
r
a
t
e
s
a
d
u
al
-
e
v
a
l
u
a
t
i
o
n
f
r
a
m
ew
o
r
k
c
o
m
b
i
n
i
n
g
s
u
b
j
e
ct
i
v
e
-
o
r
i
e
n
t
e
d
A
H
P
a
n
d
o
b
j
e
c
t
i
v
e
-
o
r
i
e
n
t
e
d
en
t
r
o
p
y
w
e
i
g
h
t
m
e
t
h
o
d
(
E
W
M)
t
o
a
s
s
es
s
p
r
e
f
e
r
e
n
ce
d
e
g
r
e
e
s
o
f
I
o
T
s
e
r
v
i
c
es
f
o
r
v
a
r
i
o
u
s
n
e
t
w
o
r
k
a
t
t
r
i
b
u
t
e
s
.
R
E
M
N
S
e
f
f
e
ct
i
v
e
l
y
b
a
la
n
c
e
s
u
s
e
r
-
c
e
n
t
r
i
c
Q
o
E
o
p
t
i
m
i
z
a
t
i
o
n
wi
th
e
f
f
i
c
i
e
n
t
n
e
t
w
o
r
k
u
t
i
l
i
z
at
i
o
n
.
E
f
f
i
c
i
e
n
t
h
a
n
d
o
v
e
r
(
H
O
)
m
a
n
a
g
e
m
e
n
t
a
n
d
r
e
s
o
u
r
c
e
o
p
t
i
m
i
z
a
ti
o
n
i
n
h
e
t
e
r
o
g
e
n
e
o
u
s
n
e
t
w
o
r
k
s
(
H
e
t
-
N
e
t
s
)
a
r
e
c
r
i
ti
c
a
l
f
o
r
e
n
s
u
r
i
n
g
s
e
a
m
l
e
s
s
c
o
n
n
e
c
ti
v
i
t
y
a
n
d
Q
o
S
.
A
c
c
o
r
d
i
n
g
t
o
T
a
s
h
a
n
e
t
a
l
.
[
1
6
]
,
a
v
e
l
o
c
i
t
y
-
a
w
a
r
e
f
u
z
z
y
l
o
g
i
c
c
o
n
t
r
o
l
l
e
r
w
i
t
h
w
e
i
g
h
te
d
f
u
n
c
t
i
o
n
(
V
A
W
-
F
L
C
-
W
F
)
alg
o
r
ith
m
is
p
r
o
p
o
s
ed
to
en
h
a
n
ce
th
e
HO
s
elf
-
o
p
tim
izatio
n
p
r
o
ce
s
s
.
T
h
is
alg
o
r
ith
m
in
co
r
p
o
r
ates
a
tr
ig
g
er
tim
er
to
r
ed
u
ce
h
a
n
d
o
v
er
p
in
g
-
p
o
n
g
(
HOPP)
ef
f
ec
ts
an
d
aim
s
to
ad
d
r
ess
is
s
u
es
lik
e
HOPP,
r
ad
io
lin
k
f
ailu
r
e
(
R
L
F),
an
d
r
ec
eiv
e
d
s
ig
n
al
r
ef
e
r
en
ce
p
o
wer
(
R
SR
P).
Ad
d
itio
n
ally
,
ca
te
g
o
r
izin
g
s
p
ee
d
s
ce
n
ar
io
s
is
h
ig
h
lig
h
t
ed
as
a
s
ig
n
if
ican
t
f
ac
to
r
i
n
m
itig
atin
g
m
o
b
ilit
y
-
r
elate
d
ch
allen
g
es,
with
co
m
p
ar
ativ
e
r
esu
lts
d
em
o
n
s
tr
atin
g
its
ef
f
ec
tiv
en
ess
o
v
er
n
o
n
-
ca
teg
o
r
ized
s
ce
n
ar
io
s
.
R
es
o
u
r
ce
allo
ca
tio
n
in
h
eter
o
g
en
e
o
u
s
m
o
b
ile
e
d
g
e
co
m
p
u
tin
g
(
Het
-
ME
C
)
n
et
wo
r
k
s
is
f
u
r
th
er
ex
p
lo
r
ed
i
n
[
1
7
]
,
wh
er
e
th
e
en
er
g
y
-
ef
f
icien
t
r
eso
u
r
ce
all
o
ca
tio
n
p
r
o
b
lem
is
f
o
r
m
u
lat
ed
as
a
tim
e
-
v
ar
ia
n
t
m
ix
e
d
-
in
teg
er
n
o
n
lin
ea
r
p
r
o
g
r
a
m
m
in
g
(
MI
NL
P)
p
r
o
b
lem
.
T
o
ad
d
r
ess
th
is
,
a
m
u
lti
-
ag
en
t
d
ee
p
r
ein
f
o
r
ce
m
en
t
le
ar
n
in
g
(
MA
DR
L
)
-
b
ased
alg
o
r
ith
m
is
p
r
o
p
o
s
ed
,
f
ea
tu
r
in
g
th
e
m
u
lti
-
ac
to
r
s
h
a
r
ed
-
cr
itic
(
MA
SC
)
ar
ch
itectu
r
e
an
d
th
e
r
eg
io
n
al
tr
ain
in
g
d
is
tr
ib
u
te
d
ex
ec
u
tio
n
(
R
T
DE
)
f
r
am
ewo
r
k
.
T
h
is
in
n
o
v
ativ
e
a
p
p
r
o
ac
h
s
tab
ilizes
m
o
d
el
tr
ain
in
g
an
d
r
ed
u
ce
s
in
f
o
r
m
atio
n
ex
ch
a
n
g
e,
en
s
u
r
in
g
ef
f
icien
t
an
d
s
ca
lab
le
r
eso
u
r
ce
m
an
ag
em
e
n
t.
B
u
ild
in
g
o
n
th
e
ad
v
an
tag
es
o
f
d
ec
e
n
tr
alize
d
s
o
lu
tio
n
s
,
Xiao
et
a
l.
[
1
8
]
i
n
tr
o
d
u
ce
s
a
d
e
ce
n
tr
alize
d
MA
D
R
L
-
b
ased
r
eso
u
r
ce
allo
ca
tio
n
alg
o
r
ith
m
.
B
y
e
m
p
lo
y
in
g
a
d
ec
e
n
tr
alize
d
p
ar
tially
o
b
s
er
v
ab
le
Ma
r
k
o
v
d
ec
is
io
n
p
r
o
ce
s
s
(
d
ec
-
POMDP)
an
d
a
m
ix
e
d
-
c
en
tr
alize
d
-
d
ec
en
t
r
alize
d
(
MC
D)
f
r
am
ewo
r
k
,
th
is
ap
p
r
o
ac
h
ef
f
ec
tiv
ely
a
d
d
r
ess
es
p
ar
tial
o
b
s
er
v
a
b
ilit
y
an
d
s
ca
l
ab
ilit
y
ch
allen
g
es.
Fu
r
th
er
m
o
r
e,
a
r
ewa
r
d
f
u
n
ctio
n
is
d
esig
n
ed
with
o
b
jectiv
e
d
ec
o
m
p
o
s
itio
n
,
b
aselin
e
-
g
u
id
ed
s
ca
lin
g
,
an
d
Qo
S
v
i
o
latio
n
p
e
n
alties,
en
ab
lin
g
ag
en
t
s
to
o
p
er
ate
i
n
a
co
o
r
d
in
ate
d
m
a
n
n
er
.
T
o
en
h
a
n
ce
in
tellig
en
t
r
eso
u
r
ce
m
an
a
g
em
en
t,
Ya
n
g
et
a
l.
[
1
9
]
p
r
o
p
o
s
es
a
m
u
lti
-
a
g
en
t
d
u
elin
g
d
ee
p
-
Q
n
etwo
r
k
-
b
as
ed
alg
o
r
ith
m
th
at
lev
er
a
g
es
d
is
tr
ib
u
ted
co
o
r
d
in
ate
d
lear
n
in
g
.
T
h
is
ap
p
r
o
ac
h
u
tili
ze
s
a
d
u
elin
g
ar
ch
itectu
r
e
to
esti
m
ate
b
o
th
s
tate
-
v
alu
e
an
d
ac
tio
n
ad
v
an
tag
e
f
u
n
c
tio
n
s
,
en
ab
lin
g
th
e
alg
o
r
ith
m
to
r
ap
id
l
y
co
n
v
er
g
e
to
an
o
p
tim
ized
p
o
licy
.
T
h
e
d
is
tr
ib
u
ted
lear
n
in
g
f
r
am
ewo
r
k
en
s
u
r
es
ef
f
icien
t
p
o
licy
d
e
v
elo
p
m
e
n
t f
o
r
in
tellig
en
t r
eso
u
r
ce
m
an
ag
e
m
en
t.
Fin
ally
,
Xu
et
a
l.
[
2
0
]
f
o
r
m
u
l
ates
th
e
r
eso
u
r
ce
allo
ca
tio
n
p
r
o
b
lem
as
a
co
m
b
in
atio
n
o
f
m
u
lti
-
ar
m
ed
b
an
d
it
a
n
d
o
p
tim
izatio
n
p
r
o
b
l
em
s
,
in
tr
o
d
u
cin
g
th
e
n
etwo
r
k
co
o
r
d
in
atio
n
s
elec
tio
n
alg
o
r
ith
m
(
NC
SA)
an
d
th
e
n
etwo
r
k
s
elec
tio
n
alg
o
r
ith
m
(
NSA)
.
Ad
d
itio
n
ally
,
th
e
m
u
lti
-
tr
af
f
ic
n
etwo
r
k
s
elec
tio
n
alg
o
r
ith
m
(
MT
-
NSA
)
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
4
9
9
3
-
5
0
0
7
4996
d
ev
elo
p
e
d
to
ad
d
r
ess
th
e
d
iv
er
s
e
tr
af
f
ic
r
eq
u
ir
em
en
ts
o
f
d
ev
ices,
o
f
f
er
in
g
a
tailo
r
ed
ap
p
r
o
ac
h
to
n
etwo
r
k
s
elec
tio
n
th
at
ac
co
u
n
ts
f
o
r
v
a
r
y
in
g
tr
af
f
ic
ty
p
es
an
d
d
ev
ice
-
s
p
ec
if
ic
n
ee
d
s
.
A
d
is
tin
g
u
is
h
in
g
f
ea
tu
r
e
o
f
5
G
n
etwo
r
k
s
is
th
eir
ca
p
ab
ilit
y
t
o
m
ee
t
d
iv
er
s
e
Qo
S
r
eq
u
ir
em
en
ts
ac
r
o
s
s
d
if
f
er
e
n
t
s
er
v
ices
wh
ile
m
ain
tain
in
g
s
lice
in
d
ep
en
d
e
n
ce
a
n
d
en
ab
lin
g
f
lex
i
b
le
r
eso
u
r
ce
s
h
a
r
in
g
.
Als
en
wi
et
a
l.
[
2
1
]
p
r
o
p
o
s
e
a
m
u
lti
-
s
er
v
ice
p
ar
titi
o
n
in
g
s
tr
ateg
y
d
esig
n
ed
to
b
alan
ce
f
u
n
ctio
n
al
is
o
latio
n
with
th
e
o
p
tim
al
s
h
ar
in
g
o
f
r
eso
u
r
c
es.
T
h
is
s
y
s
tem
lev
er
ag
es
a
POMDP
to
en
h
a
n
ce
n
etwo
r
k
s
lice
s
er
v
i
ce
lev
el
ag
r
ee
m
en
t
(
SLA
)
c
o
m
p
lian
ce
,
s
p
ec
tr
u
m
u
tili
za
tio
n
ef
f
icien
cy
,
an
d
f
ai
r
n
ess
.
Similar
ly
,
Pap
a
et
a
l.
[
2
2
]
in
tr
o
d
u
ce
s
a
two
-
lev
el
d
y
n
a
m
ic
n
etwo
r
k
s
licin
g
f
r
am
ewo
r
k
th
at
in
co
r
p
o
r
ates
ten
an
ts
’
p
r
io
r
ities
,
b
aseb
a
n
d
r
eso
u
r
ce
s
,
in
ter
f
e
r
en
ce
,
a
n
d
t
h
r
o
u
g
h
p
u
t
in
to
its
d
esig
n
.
T
h
e
f
r
am
ew
o
r
k
o
p
er
a
tes
with
an
u
p
p
er
lev
el
d
ed
ic
ated
to
ad
m
is
s
io
n
co
n
tr
o
l
an
d
u
s
er
ass
o
ciatio
n
,
wh
ile
th
e
lo
wer
lev
el
en
s
u
r
es f
air
r
ad
io
r
eso
u
r
ce
allo
ca
tio
n
a
m
o
n
g
u
s
er
s
.
T
h
e
s
o
lu
tio
n
is
i
m
p
lem
en
ted
u
s
in
g
a
g
r
ad
ien
t
-
b
ased
o
p
tim
izatio
n
a
p
p
r
o
ac
h
,
w
h
ich
c
o
n
s
id
er
s
h
is
to
r
ical
r
eso
u
r
ce
allo
ca
tio
n
s
an
d
u
s
er
s
’
a
v
er
ag
e
d
ata
r
ates.
Mo
r
eo
v
er
,
wo
r
k
s
p
r
esen
ted
in
[
2
3
]
,
[
2
4
]
em
p
h
asize
n
e
two
r
k
th
r
o
u
g
h
p
u
t o
p
tim
izatio
n
wh
ile
m
ain
tain
in
g
f
air
n
ess
ac
r
o
s
s
n
etwo
r
k
s
lices.
3.
P
RO
P
O
SE
D
M
E
T
H
O
DO
L
O
G
Y
T
h
e
p
r
o
p
o
s
ed
s
tu
d
y
ad
d
r
ess
es
th
e
ch
allen
g
es in
v
o
lv
e
d
in
ef
f
i
cien
t selec
tio
n
o
f
a
n
etwo
r
k
co
n
s
id
er
in
g
ad
v
an
ce
d
5
G
en
v
ir
o
n
m
en
t.
T
h
er
ef
o
r
e,
an
in
tellig
en
t
an
d
p
r
iv
ac
y
awa
r
e
r
eso
lu
tio
n
is
p
r
o
p
o
s
ed
f
o
r
th
e
s
elec
tio
n
o
f
n
etwo
r
k
wh
ile
allo
win
g
ef
f
icien
t
an
d
p
r
o
p
er
u
s
e
o
f
th
e
n
etwo
r
k
a
n
d
im
p
r
o
v
is
in
g
u
s
er
ex
p
er
ie
n
ce
.
W
e
co
n
s
id
er
th
at
a
n
etwo
r
k
i
s
u
s
ed
as
a
s
et
o
f
u
s
er
r
eso
u
r
ce
th
at
is
d
is
tr
ib
u
ted
at
r
an
d
o
m
ex
p
r
ess
ed
as
h
av
in
g
s
u
b
-
in
d
e
x
{
1
,
2
,
…
,
}
.
Acc
o
r
d
i
n
g
to
th
e
SLAs
o
f
p
r
io
r
s
tu
d
ies,
th
e
u
s
er
s
c
o
u
ld
b
elo
n
g
to
eith
er
o
n
e
o
f
t
h
e
th
r
ee
s
tag
es
o
f
p
r
io
r
ity
=
{
1
,
2
}
.
Fo
r
=
{
1
}
,
wh
ich
im
p
lies
th
at
th
e
p
r
io
r
ity
is
h
ig
h
r
esu
ltin
g
in
h
ig
h
lev
el
clien
ts
th
at
p
ay
h
ig
h
e
r
f
o
r
b
est
Qo
S
lev
el.
On
th
e
co
n
tr
ar
y
,
=
{
2
}
im
p
lies
th
e
p
r
io
r
ity
is
lo
w
in
v
o
lv
in
g
n
o
r
m
al
u
s
er
s
th
at
d
o
n
o
t
p
ay
m
u
ch
a
n
d
ar
e
a
p
p
ea
s
ed
with
th
e
q
u
ality
o
f
n
ec
ess
ar
y
s
er
v
ices
b
ein
g
m
i
n
im
u
m
.
W
e
ca
teg
o
r
ize
th
e
two
ty
p
es
o
f
u
s
er
s
,
s
en
s
o
r
s
h
av
in
g
f
ix
ed
p
o
s
itio
n
a
n
d
h
ig
h
p
r
io
r
i
ty
as
well
as
ce
llu
la
r
u
s
er
r
es
o
u
r
ce
h
av
in
g
f
ix
ed
o
r
r
a
n
d
o
m
way
p
o
in
t
m
o
tio
n
an
d
o
n
e
o
f
t
h
e
s
tag
es o
f
p
r
io
r
it
y
th
at
is
p
o
s
s
ib
le.
C
o
n
s
id
er
a
s
eq
u
en
ce
b
ase
s
ta
tio
n
s
ex
p
r
ess
ed
as
h
av
in
g
s
u
b
in
d
ex
{
1
,
2
,
…
,
}
,
in
wh
ich
=
⋃
ℚ
.
Her
e,
th
e
co
llectiv
e
g
r
o
u
p
o
f
b
ase
s
tatio
n
s
is
g
i
v
en
as
f
o
r
ter
r
estrial
n
etwo
r
k
b
ase
s
tatio
n
s
.
T
h
e
s
et
o
f
n
etwo
r
k
s
elec
tio
n
s
is
g
iv
en
as
ℙ
,
h
av
in
g
s
u
b
in
d
ex
{
1
,
2
,
…
,
}
th
at
co
u
ld
b
e
o
b
tain
ab
le
o
r
n
o
t
f
o
r
v
a
r
io
u
s
b
ase
s
tatio
n
s
.
T
h
e
p
a
r
ticu
lar
s
er
v
ice
t
h
at
is
r
e
q
u
ested
is
r
ep
r
esen
ted
as
ℛ
.
T
h
e
s
eq
u
e
n
ce
o
f
n
etwo
r
k
s
elec
tio
n
s
to
a
d
ju
s
t
th
e
u
s
er
r
eso
u
r
ce
is
ex
p
r
ess
ed
as
ℙ
,
in
wh
ich
ℙ
⊆
ℙ
.
E
v
er
y
h
as
its
ca
p
ac
ity
ex
p
r
ess
ed
co
n
s
id
er
in
g
r
eso
u
r
ce
b
lo
ck
s
o
f
a
co
n
s
tan
t
b
an
d
wid
th
ℎ
.
T
h
er
e
ex
is
ts
a
f
ix
ed
co
u
n
t
o
f
r
eso
u
r
ce
s
as
well
as
ch
an
g
in
g
v
a
r
iatio
n
s
f
o
r
s
er
v
ice
r
eq
u
ests
,
th
at
p
r
i
o
r
d
ef
in
e
th
e
r
eso
u
r
ce
b
lo
ck
s
f
o
r
e
v
er
y
n
etwo
r
k
s
elec
tio
n
p
r
o
ce
s
s
th
at
co
u
ld
r
esu
lt
in
in
ef
f
ec
tiv
e
u
s
ag
e
o
f
r
eso
u
r
ce
s
h
av
in
g
a
b
ad
ef
f
ec
t
o
f
th
e
Qo
S
.
Fu
r
th
e
r
,
we
tak
e
in
to
ac
co
u
n
t
t
h
e
s
licin
g
s
ch
em
e
wh
ile
o
m
itti
n
g
th
e
r
e
s
o
u
r
ce
s
b
ein
g
f
ix
e
d
f
o
r
e
v
er
y
n
etwo
r
k
s
elec
tio
n
.
T
h
e
r
eso
u
r
ce
b
lo
c
k
s
th
at
ar
e
av
ailab
le
r
ep
r
esen
ted
as
f
o
r
th
e
b
ase
s
tatio
n
is
allo
ca
ted
d
y
n
am
ically
wh
ile
lo
o
k
in
g
at
th
e
co
u
n
t
o
f
r
eq
u
ests
,
p
r
io
r
ity
o
f
th
e
u
s
er
s
,
th
e
ch
allen
g
es
o
n
th
e
Q
o
S
an
d
t
h
e
m
o
b
ilit
y
ch
ar
ac
ter
is
tics
.
T
h
e
p
ar
am
eter
{
0
,
1
}
wh
ich
ex
p
r
ess
es
th
e
r
e
s
o
u
r
ce
s
th
at
ar
e
av
ailab
le
f
o
r
th
e
in
th
e
,
im
p
ly
in
g
th
at
ze
r
o
in
d
icate
s
th
e
co
u
n
t
o
f
r
eso
u
r
ce
s
av
ailab
le
ar
e
in
ad
eq
u
ate
to
b
e
allo
ca
t
ed
f
o
r
m
in
im
u
m
th
r
o
u
g
h
p
u
t
(
)
n
ee
d
ed
b
y
th
e
u
s
er
f
o
r
th
e
(
)
.
T
h
e
th
at
ca
n
b
e
ac
ce
s
s
ed
th
r
o
u
g
h
t
h
e
is
ex
p
r
ess
ed
as
{
0
,
1
}
,
in
th
is
ca
s
e
ze
r
o
co
u
ld
im
p
l
y
th
at
th
e
is
n
o
t
av
ailab
le
f
r
o
m
th
e
b
ec
au
s
e
o
f
th
e
ac
ce
s
s
b
ein
g
im
p
o
s
s
ib
le
f
o
r
s
er
v
ice
o
r
lac
k
o
f
f
u
n
ctio
n
ality
.
T
h
e
s
et
o
f
s
er
v
ices
p
o
s
s
ib
le
d
en
o
ted
as
d
ef
in
ed
as
ℕ
h
av
in
g
s
u
b
-
in
d
ex
{
1
,
2
,
…
,
}
.
Fo
r
th
e
p
r
o
p
o
s
ed
wo
r
k
,
we
ta
k
e
in
to
ac
co
u
n
t
th
r
ee
d
if
f
er
en
t
s
er
v
ices,
n
am
ely
VR
,
v
i
d
eo
(
v
i
d
)
,
a
n
d
I
I
o
T
Ap
p
s
,
th
is
is
ch
ar
ac
ter
ized
u
s
in
g
v
ar
i
o
u
s
n
ee
d
c
o
n
s
id
er
in
g
th
e
,
d
ela
y
(
)
,
an
d
co
n
s
u
m
p
tio
n
o
f
e
n
er
g
y
(
)
.
Hen
ce
,
ev
er
y
s
er
v
ice
is
m
ap
p
ed
o
n
to
v
ar
io
u
s
n
etwo
r
k
s
elec
tio
n
s
.
T
h
e
s
ig
n
if
ican
ce
o
f
ev
e
r
y
s
er
v
ice
p
r
o
v
id
es
th
e
p
ar
am
eter
s
o
f
‘
Qo
S’
ex
p
r
ess
ed
u
s
in
g
weig
h
ts
ℎ
,
ℎ
an
d
ℎ
with
ℎ
+
ℎ
+
ℎ
.
T
h
e
th
at
a
u
s
er
r
eso
u
r
ce
ca
n
g
ain
co
n
s
id
er
i
n
g
th
e
(
,
)
d
ep
en
d
in
g
o
f
th
e
r
ec
e
p
tio
n
co
n
s
tr
ain
ts
o
f
th
e
u
s
er
an
d
th
e
r
eso
u
r
ce
b
l
o
ck
s
,
f
r
o
m
.
,
=
ℎ
ℎ
×
,
×
ℎ
(
1
)
I
n
th
is
ca
s
e,
,
ℎ
.
T
h
e
ℎ
ℎ
is
th
e
ef
f
icien
cy
r
elatin
g
to
th
e
m
o
d
u
latio
n
en
c
o
d
in
g
m
ec
h
a
n
is
m
g
ain
ed
b
y
r
eg
ar
d
in
g
.
,
is
r
ep
r
esen
ted
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Dyn
a
mic
s
ervice
-
a
w
a
r
e
n
etw
o
r
k
s
elec
tio
n
fr
a
mewo
r
k
fo
r
mu
lti
-
o
b
jective
o
p
timiz
a
tio
n
…
(
B
h
a
va
n
a
S
r
in
iva
s
)
4997
in
s
ec
o
n
d
,
d
ef
in
es
th
e
d
elay
o
f
th
e
u
s
er
f
o
r
ac
ce
s
s
in
g
th
e
s
er
v
ice
th
at
is
n
ee
d
ed
u
s
in
g
t
h
e
.
Fo
r
th
e
(
2
)
,
t
h
e
d
ela
y
in
tr
an
s
m
is
s
io
n
is
g
iv
en
as
,
,
an
d
th
e
d
elay
in
q
u
eu
in
g
is
g
i
v
en
as
,
,
.
T
h
e
co
n
s
u
m
p
tio
n
o
f
e
n
er
g
y
,
m
ea
s
u
r
ed
in
jo
u
les
is
d
en
o
ted
as
,
is
ev
alu
ated
b
y
th
e
ac
ce
s
s
o
f
u
s
er
to
th
e
s
er
v
ice
th
at
is
r
eq
u
ested
th
r
o
u
g
h
th
e
,
th
is
is
f
o
r
m
u
lated
a
s
(
3
)
.
,
=
,
,
+
,
,
(
2
)
,
=
,
∗
(
3
)
H
e
r
e
,
i
s
t
h
e
c
o
n
s
u
m
e
d
p
o
w
e
r
t
h
a
t
i
s
g
i
v
e
n
a
s
w
a
t
t
r
e
p
r
e
s
e
n
t
i
n
g
t
h
e
u
s
e
r
r
e
s
o
u
r
c
e
f
o
r
t
h
e
r
e
c
e
p
t
i
o
n
o
f
p
a
r
t
i
c
u
l
a
r
s
e
r
v
i
c
e
.
T
h
e
v
a
l
u
e
t
h
a
t
i
s
n
o
r
m
a
l
i
z
e
d
g
i
v
e
n
a
s
,
,
i
s
e
v
a
l
u
a
t
e
d
t
h
r
o
u
g
h
t
h
e
u
p
w
a
r
d
c
r
i
t
e
r
i
a
a
s
i
n
(
4
)
.
H
o
w
e
v
e
r
,
t
o
g
a
t
h
e
r
t
h
e
d
e
l
a
y
a
n
d
e
n
e
r
g
y
c
o
n
s
u
m
p
t
i
o
n
v
a
l
u
e
s
t
h
a
t
a
r
e
n
o
r
m
a
l
i
z
e
d
b
e
i
n
g
d
e
n
o
t
e
d
a
s
,
,
a
n
d
,
,
,
r
e
s
p
e
c
t
i
v
e
l
y
.
T
h
e
d
o
w
n
w
a
r
d
c
r
i
t
e
r
i
a
e
q
u
a
t
i
o
n
i
s
g
i
v
e
n
a
s
(
5
)
.
=
{
0
,
ℎ
1
−
(
(
−
)
(
′
×
(
−
)
)
−
1
)
,
ℎ
1
ℎ
(
4
)
=
{
1
,
ℎ
1
−
(
(
−
)
(
′
×
(
−
)
)
−
1
)
,
ℎ
0
ℎ
(
5
)
I
n
th
is
ca
s
e,
′
ℎ
2
is
th
e
r
ef
i
n
ed
s
te
ep
n
ess
v
ar
iab
le
an
d
,
ar
e
th
e
m
ax
an
d
m
in
im
u
m
s
co
r
es
r
elatin
g
to
ev
er
y
s
er
v
ice
ch
allen
g
e
an
d
th
e
p
ar
ticu
la
r
u
tili
ty
f
u
n
ctio
n
s
.
Fo
r
ca
lcu
latin
g
th
e
co
n
d
itio
n
s
o
f
th
e
n
etwo
r
k
to
b
e
s
atis
f
ied
f
o
r
ev
er
y
s
er
v
ice
r
eq
u
est,
th
e
d
im
en
s
io
n
less
v
alu
e
f
u
n
ctio
n
is
g
iv
en
as
,
[
0
,
1
]
.
T
h
is
u
n
it
is
a
co
m
b
in
atio
n
o
f
v
ar
io
u
s
n
o
r
m
alize
d
f
ea
tu
r
es
co
n
s
id
er
in
g
th
e
ac
ce
s
s
ib
ilit
y
o
f
n
etwo
r
k
s
elec
tio
n
,
av
ailab
ilit
y
o
f
r
eso
u
r
ce
,
u
s
er
as
well
as
ap
p
p
r
o
f
iles
.
W
h
en
,
,
it
im
p
lies
th
at
ca
n
s
atis
f
y
r
eq
u
ests
th
at
m
ax
im
ize
th
e
Qo
S
.
T
h
e
,
is
f
o
r
m
u
lated
as
(
6
)
.
,
=
{
,
,
1
,
,
0
0
=
0
∨
,
,
=
0
∨
,
,
=
0
(
6
)
T
h
e
v
alu
e
t
o
atten
d
th
e
r
e
q
u
est
o
f
th
e
u
s
er
b
y
th
e
is
d
en
o
ted
as
,
wh
en
th
e
n
etwo
r
k
h
as
s
u
f
f
icien
t
r
eso
u
r
ce
s
an
d
is
ex
p
r
ess
ed
in
(
7
)
.
T
h
e
v
alu
e
at
an
o
v
er
lo
a
d
is
g
iv
en
as
,
wh
ich
is
also
ev
alu
ated
in
th
e
(
7
)
g
iv
en
.
Fo
r
th
e
(
7
)
,
′′
ℎ
is
a
v
alu
e
f
ac
to
r
th
at
ad
ap
ts
to
,
s
co
r
e
to
b
en
ef
it th
e
b
ase
s
tatio
n
s
o
m
itti
n
g
th
e
o
v
er
l
o
ad
.
,
=
×
(
ℎ
×
,
,
+
ℎ
×
,
,
+
ℎ
×
,
,
)
,
=
(
(
(
′′
)
−
1
)
×
ℎ
×
,
+
ℎ
×
)
(
7
)
r
e
p
r
e
s
e
n
ts
t
h
e
p
o
s
s
i
b
l
e
r
es
o
u
r
c
e
b
l
o
c
k
t
h
a
t
t
h
e
h
a
s
u
n
t
il
al
l
th
e
s
e
u
s
e
r
s
h
a
v
e
t
h
e
l
e
as
t
p
o
s
s
i
b
l
e
c
o
n
s
i
d
e
r
i
n
g
th
e
r
e
s
t
r
i
c
t
i
o
n
s
o
n
t
h
e
s
e
r
v
ic
e
.
T
h
e
n
o
r
m
a
l
i
z
e
d
p
o
w
e
r
c
o
n
s
u
m
p
t
i
o
n
d
e
n
o
t
e
d
a
s
w
h
i
c
h
i
s
e
v
a
l
u
a
t
e
d
as
g
i
v
e
n
i
n
(
8
)
.
I
n
t
h
i
s
c
a
s
e
,
t
h
e
m
a
x
i
m
u
m
a
n
d
m
i
n
i
m
u
m
c
o
u
n
t
o
f
r
e
s
o
u
r
c
e
b
l
o
c
k
s
t
o
a
t
t
ai
n
t
h
e
a
n
d
a
r
e
e
x
p
r
e
s
s
ed
a
s
,
,
a
n
d
,
,
,
r
e
s
p
e
c
t
i
v
el
y
.
T
h
e
s
a
t
is
f
a
ct
o
r
y
t
h
r
o
u
g
h
p
u
t
,
f
o
r
t
h
e
d
i
m
e
n
s
i
o
n
l
ess
s
c
o
r
e
h
a
v
e
a
r
a
n
g
e
b
e
t
w
e
e
n
0
a
n
d
1
.
=
{
0
,
ℎ
,
,
(
(
)
(
,
,
)
−
1
)
,
,
,
≤
≤
,
,
1
,
ℎ
(
8
)
3
.
1
.
Dy
na
m
ic
s
er
v
ice
-
a
w
a
re
net
wo
rk
s
elec
t
o
r
T
h
e
p
r
o
p
o
s
ed
s
tu
d
y
f
o
r
th
e
DSANS
m
ain
ly
in
clu
d
es
th
e
s
elec
tio
n
o
f
n
etwo
r
k
as
well
as
r
eso
u
r
ce
esti
m
atio
n
.
At
ev
er
y
tr
an
s
m
is
s
io
n
p
er
io
d
,
ev
er
y
lo
ca
l
ev
alu
atio
n
m
o
d
el
g
ath
e
r
s
th
e
s
er
v
ice
r
eq
u
ests
f
r
o
m
th
e
u
s
er
s
th
at
ar
e
n
ew,
s
er
v
ice
u
p
d
ates
f
r
o
m
u
s
er
s
th
at
ar
e
ex
is
tin
g
,
o
r
th
e
ch
a
n
n
el
in
d
ica
tio
n
o
f
th
e
ex
is
tin
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
4
9
9
3
-
5
0
0
7
4998
u
s
er
co
n
s
id
er
i
n
g
t
h
e
th
r
esh
o
ld
.
T
h
is
is
ex
p
lain
ed
in
t
h
e
Fig
u
r
e
1
.
T
h
e
ev
e
n
ts
th
at
a
r
e
ca
p
tu
r
ed
s
ep
ar
atel
y
b
y
th
e
ap
p
licatio
n
ar
e
g
ath
er
e
d
b
y
th
e
g
en
er
al
ev
alu
atio
n
m
o
d
e
l
f
o
r
a
to
tally
ev
en
ts
.
W
h
en
e
v
er
y
e
v
en
t
en
d
s
,
th
e
alg
o
r
ith
m
s
r
u
n
f
o
r
th
e
n
e
x
t
+
1
ev
en
t
f
r
o
m
t
h
e
co
u
n
t
o
f
ev
en
t
th
at
ar
e
i
d
en
tifie
d
f
o
r
t
h
e
p
ar
ticu
lar
tr
an
s
m
is
s
io
n
p
er
io
d
.
T
h
e
ev
e
n
t
r
eq
u
ests
th
at
ar
e
r
ejec
ted
a
r
e
s
to
r
ed
in
th
e
s
tar
tin
g
o
f
th
e
q
u
eu
e
f
o
r
+
1
b
ein
g
th
e
n
ex
t tr
an
s
m
is
s
io
n
p
e
r
io
d
.
Fig
u
r
e
1
.
D
y
n
am
ic
s
er
v
ice
-
aw
ar
e
n
etwo
r
k
s
elec
to
r
Af
ter
ev
er
y
lo
ca
l
m
o
d
el
m
a
k
es
a
s
ep
ar
ate
d
ec
is
io
n
co
n
s
id
er
in
g
th
e
p
a
r
ticu
lar
in
f
o
r
m
ati
o
n
f
o
r
th
e
f
o
r
,
th
e
s
elec
tio
n
m
o
d
el
th
en
ch
o
o
s
es
th
e
b
ase
s
tatio
n
h
av
in
g
th
e
h
ig
h
est
,
,
am
o
n
g
all
th
e
m
o
d
els
th
at
ten
d
to
th
e
r
eq
u
est.
T
h
is
s
elec
ti
o
n
m
o
d
el
is
u
n
awa
r
e
o
f
th
e
l
o
ca
l
in
f
o
r
m
atio
n
,
as
it
o
n
ly
g
r
asp
s
th
e
,
,
,
u
p
h
o
ld
i
n
g
t
h
e
p
r
iv
ac
y
an
d
d
ec
r
ea
s
in
g
o
v
er
h
ea
d
co
m
m
u
n
icatio
n
.
T
h
e
p
r
o
p
o
s
ed
im
p
r
o
v
is
ed
d
ec
en
tr
alize
d
r
a
d
io
ac
ce
s
s
n
etwo
r
k
h
as
a
d
y
n
am
ic
b
itra
te
tr
af
f
ic
f
o
r
n
etwo
r
k
s
en
v
ir
o
n
m
en
t,
p
r
i
o
r
ity
o
f
t
h
e
u
s
er
an
d
s
er
v
ice
d
i
f
f
icu
lties
.
W
h
en
th
er
e
is
s
u
f
f
ic
ien
t c
ap
ac
ity
o
f
t
h
e
n
etwo
r
k
,
th
e
alg
o
r
ith
m
allo
ca
tes
th
e
co
u
n
t
o
f
r
eso
u
r
ce
b
lo
c
k
s
f
o
r
wh
ile
tak
in
g
in
to
ac
c
o
u
n
t
th
e
s
er
v
ice
n
ee
d
s
an
d
n
eg
lects
th
e
p
r
io
r
ity
o
f
th
e
u
s
er
.
On
th
e
co
n
tr
ar
y
,
if
th
e
b
ase
s
tati
o
n
lack
s
s
u
f
f
icien
t
r
eso
u
r
ce
,
th
e
DSANS
im
p
lem
en
ts
a
r
eso
u
r
ce
esti
m
atio
n
s
ch
em
e
co
n
s
id
er
in
g
th
e
p
er
f
o
r
m
a
n
ce
o
f
s
er
v
ice
an
d
th
e
s
er
v
ice
lev
el
ag
r
ee
m
en
t
p
r
o
v
es
to
b
e
ad
v
an
tag
e
o
u
s
to
u
s
er
s
an
d
n
eg
lects
th
e
s
u
d
d
en
d
eg
r
ad
atio
n
o
f
th
e
Qo
S
.
T
h
e
r
eso
u
r
ce
esti
m
atio
n
f
o
cu
s
es
o
n
r
elea
s
in
g
r
eso
u
r
ce
s
as
we
ll
as
ten
d
in
g
to
n
ew
u
s
er
s
co
n
s
id
er
in
g
th
e
r
eq
u
ir
em
e
n
ts
.
Alth
o
u
g
h
,
if
t
h
e
r
eq
u
est
f
o
r
s
er
v
ice
r
elate
s
to
th
e
n
etwo
r
k
h
a
v
in
g
a
n
ew
u
s
e
r
,
th
e
ch
o
s
en
b
ase
s
tatio
n
p
r
o
ce
ed
s
to
th
e
allo
ca
t
io
n
o
f
r
eso
u
r
ce
s
f
o
r
th
e
n
etwo
r
k
s
elec
tio
n
p
r
o
ce
s
s
.
Ass
u
m
e
th
e
n
o
b
ase
s
tatio
n
is
ch
o
s
en
im
p
ly
i
n
g
=
0
,
as
th
er
e
ar
e
in
a
d
eq
u
ate
p
o
s
s
ib
le
r
eso
u
r
ce
s
f
o
r
r
elea
s
in
g
as
well
a
s
s
atis
f
y
in
g
th
e
least
co
n
s
tr
ain
ts
.
T
h
en
,
th
e
m
o
d
el
g
ets
to
a
ter
m
in
al
s
tag
e
(
=
1
)
an
d
n
ew
ev
en
ts
ar
e
n
o
t a
tten
d
ed
to
.
T
h
is
co
n
d
itio
n
p
r
ev
ails
u
n
til r
elea
s
e
o
f
t
h
e
r
e
s
o
u
r
ce
s
.
l
im
→
∞
∑
∑
,
,
=
1
=
1
.
ℎ
,
ℎ
,
,
ℎ
,
,
ℎ
,
ℎ
(
9
)
T
h
e
DSANS
f
o
cu
s
es
o
n
th
e
b
est
b
ase
s
tat
io
n
f
o
r
s
atis
f
y
in
g
th
e
r
eq
u
ests
an
d
o
p
tim
izatio
n
o
f
s
licin
g
th
e
u
s
ag
e
o
f
r
eso
u
r
ce
s
f
o
r
ev
er
y
tr
an
s
m
is
s
io
n
p
er
io
d
.
Hen
ce
,
th
e
alg
o
r
ith
m
is
f
o
r
m
u
lated
as
g
iv
en
.
Her
e,
ex
p
r
ess
es
th
e
to
tal
co
u
n
t
o
f
t
r
an
s
m
is
s
io
n
p
er
io
d
s
.
T
h
e
Qo
S
an
d
p
er
ce
p
tio
n
o
f
th
e
u
s
er
is
d
ir
ec
tly
af
f
ec
ted
by
,
,
,
wh
ile
co
n
s
id
er
in
g
th
e
d
iv
e
r
s
ity
o
f
n
etwo
r
k
co
n
d
itio
n
s
,
t
y
p
es
o
f
u
s
er
s
,
co
n
s
tr
ain
ts
o
n
s
er
v
ice
as we
ll a
s
ac
ce
s
s
to
s
licin
g
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Dyn
a
mic
s
ervice
-
a
w
a
r
e
n
etw
o
r
k
s
elec
tio
n
fr
a
mewo
r
k
fo
r
mu
lti
-
o
b
jective
o
p
timiz
a
tio
n
…
(
B
h
a
va
n
a
S
r
in
iva
s
)
4999
3
.
2
.
P
r
o
po
s
ed
a
da
ptiv
e
deep
decisi
o
n net
wo
rk
T
h
e
p
r
o
p
o
s
ed
DSANS
f
o
r
ch
o
o
s
in
g
th
e
n
etwo
r
k
is
b
ased
o
n
th
e
ADDN
im
p
lem
en
tin
g
co
l
lab
o
r
ativ
e
k
n
o
wled
g
e
ag
g
r
eg
atio
n
f
r
a
m
ewo
r
k
(
C
KAF)
ap
p
r
o
ac
h
to
d
ev
elo
p
a
ML
g
lo
b
al
m
o
d
u
le.
L
o
ca
l
ag
e
n
ts
co
o
p
er
ate
to
ch
o
o
s
e
th
e
p
o
lic
y
∗
th
at
in
cr
ea
s
es
th
e
Qo
S
f
o
r
e
v
er
y
u
s
er
o
f
th
e
n
etwo
r
k
as
well
as
m
ax
im
izes
th
e
o
p
tim
izatio
n
o
f
u
s
in
g
r
eso
u
r
ce
s
th
at
is
ex
p
o
s
ed
t
o
d
iv
e
r
s
e
d
em
an
d
s
o
f
th
e
u
s
er
as
we
ll
as
co
n
s
tr
ain
ts
o
n
th
e
s
er
v
ice.
C
o
n
s
id
er
′
as
a
p
ar
t
icu
lar
d
ec
is
io
n
p
e
r
io
d
,
wh
er
e
′
{
1
,
2
,
.
.
,
′
}
an
d
′
=
×
×
.
T
h
e
co
u
n
t
o
f
ev
e
n
ts
at
tr
an
s
m
is
s
io
n
p
er
io
d
is
g
iv
en
as
.
th
e
co
u
n
t
o
f
tr
an
s
m
is
s
io
n
p
er
io
d
s
at
p
er
io
d
is
d
en
o
ted
as
an
d
is
th
e
co
u
n
t
o
f
ep
is
o
d
es
at
th
e
tr
ai
n
in
g
p
h
ase.
I
n
itially
,
th
e
g
lo
b
al
m
o
d
u
le
in
itializes
th
e
g
lo
b
al
p
ar
am
eter
at
r
an
d
o
m
an
d
is
s
h
ar
ed
with
th
e
lo
ca
l
m
o
d
u
le.
T
h
e
r
e
is
a
lo
ca
l
ag
en
t
f
o
r
ev
er
y
b
ase
s
tatio
n
th
at
ex
ec
u
tes
th
e
tr
ain
i
n
g
p
h
ase
f
o
r
ea
ch
ev
en
t
in
,
g
ath
er
in
g
n
ew
p
a
r
am
eter
s
.
Fu
r
th
er
,
to
o
m
it
o
v
er
h
ea
d
co
m
m
u
n
icatio
n
,
f
o
r
ea
ch
d
ec
is
io
n
p
er
io
d
ℎ′
,
lo
ca
l
p
ar
am
eter
s
ar
e
tr
a
n
s
f
er
r
ed
to
th
e
g
lo
b
al
m
o
d
u
le
ag
g
r
eg
atin
g
th
em
th
r
o
u
g
h
a
f
ed
er
ated
av
er
a
g
e
s
ch
em
e
d
en
o
ted
as
r
esu
ltin
g
in
th
e
n
ew
g
lo
b
al
m
o
d
u
le
.
,
′
+
1
=
−
1
×
∑
,
′
=
1
(
10
)
Her
e,
th
e
c
o
u
n
t
o
f
lo
ca
l
ag
en
ts
th
at
ar
e
in
v
o
l
v
ed
i
n
th
e
tr
ain
in
g
p
h
ase
is
r
ep
r
esen
ted
as
.
Fu
r
th
er
m
o
r
e
,
th
e
weig
h
t
o
f
th
e
m
o
d
el
is
tr
an
s
f
er
r
ed
b
ac
k
to
th
e
lo
ca
l
m
o
d
u
le.
T
h
is
iter
atio
n
co
n
tin
u
es
till
th
e
alg
o
r
ith
m
r
ea
ch
es
t
h
e
id
ea
l
o
p
tim
ized
g
lo
b
al
m
o
d
u
le
∗
.
T
h
ese
p
ar
am
eter
s
1
=
⋯
=
=
∗
ar
e
u
s
ed
b
y
th
e
lo
ca
l
a
g
en
ts
wh
ile
p
r
ev
en
t
in
g
tr
an
s
f
er
o
f
s
en
s
itiv
e
in
f
o
r
m
atio
n
with
in
th
em
.
T
h
e
i
n
ter
ac
tio
n
s
o
f
t
h
e
lo
ca
l
m
o
d
u
le
ar
e
f
o
r
m
alize
d
co
n
s
id
er
in
g
s
tates,
r
ewa
r
d
s
(
,
,
)
as
well
as
ac
tio
n
s
th
at
ar
e
d
e
s
cr
ib
ed
in
d
etail
as
f
o
llo
ws
:
i)
State:
is
d
ef
i
n
ed
f
o
r
ev
e
r
y
l
o
ca
l
m
o
d
u
le
as
an
d
h
as
u
s
er
,
n
etwo
r
k
in
f
o
r
m
atio
n
a
n
d
ap
p
licatio
n
r
elatin
g
to
th
e
b
a
s
e
s
tate=
io
n
.
T
h
e
s
tate
is
n
o
ticed
b
y
e
v
er
y
lo
ca
l
m
o
d
u
le
r
elate
d
to
th
e
b
ase
s
tatio
n
at
th
e
tim
e
o
f
an
ev
e
n
t
d
u
r
in
g
;
ii)
R
e
war
d
:
ev
er
y
lo
ca
l
m
o
d
u
le
attain
s
a
r
ewa
r
d
d
en
o
te
d
as
,
,
b
ec
au
s
e
o
f
th
e
ac
ti
o
n
th
at
co
n
tr
ib
u
tes
to
th
e
p
h
ase
o
f
lear
n
in
g
.
T
h
e
co
m
b
in
ed
r
ewa
r
d
is
m
ax
im
ized
b
y
tr
ain
i
n
g
th
e
l
o
ca
l
ag
en
ts
g
iv
en
as
=
∑
∑
×
,
,
;
an
d
iii)
Actio
n
:
th
e
p
o
s
s
ib
le
s
et
o
f
ac
tio
n
s
th
at
ar
e
to
b
e
p
er
f
o
r
m
e
d
b
y
ev
e
r
y
l
o
ca
l
ag
en
t
is
d
en
o
ted
as
=
{
0
,
1
,
2
}
.
T
h
e
ac
tio
n
p
er
f
o
r
m
ed
b
y
lo
ca
l
ag
en
t
at
th
e
ev
en
t
at
,
r
elatin
g
to
r
eq
u
est
is
g
iv
en
as
,
,
.
W
h
en
,
,
=
1
,
it
im
p
lies
th
at
th
e
b
ase
s
tatio
n
ca
n
ten
d
to
th
e
r
eq
u
est
with
am
p
le
r
eso
u
r
ce
s
,
if
,
,
=
2
,
th
e
b
ase
s
tatio
n
ca
n
also
ten
d
to
th
e
r
e
q
u
est
b
u
t
h
as
to
u
n
d
er
g
o
th
e
p
r
o
ce
s
s
o
f
r
eso
u
r
ce
esti
m
atio
n
b
ec
a
u
s
e
o
f
o
v
er
lo
ad
i
n
g
.
W
h
er
ea
s
,
if
,
,
=
0
,
th
e
b
ase
s
tat
io
n
is
u
n
ab
le
to
ten
d
to
th
e
r
eq
u
est
an
d
,
=
0
.
At
th
is
s
tag
e,
th
e
b
ase
s
tate
is
n
o
t
s
u
itab
le
f
o
r
th
e
s
elec
to
r
m
o
d
el.
H
o
wev
er
,
if
th
e
b
a
s
e
s
tatio
n
s
atis
f
ies
th
e
d
em
an
d
s
b
u
t
ev
e
r
y
ac
tio
n
is
ze
r
o
,
th
e
n
th
e
r
eq
u
est
o
f
th
e
u
s
er
is
r
ejec
ted
wh
ich
th
er
ef
o
r
e
af
f
ec
ts
th
e
Qo
S
.
T
h
e
ADDN
en
h
an
ce
s
th
e
le
ar
n
in
g
a
b
ilit
y
o
f
t
h
e
m
o
d
el
an
d
ev
a
d
es
th
e
to
o
o
p
tim
is
tic
r
ewa
r
d
ev
alu
atio
n
u
s
in
g
a
p
p
r
o
x
im
ati
n
g
f
u
n
ctio
n
v
ia
two
n
eu
r
al
n
etwo
r
k
s
f
o
r
v
al
u
e
f
u
n
ctio
n
.
T
h
e
in
itial
Q
-
v
al
u
e
f
u
n
ctio
n
(
,
,
)
,
wh
er
e
th
e
v
ec
to
r
f
o
r
th
e
n
eu
r
al
n
etwo
r
k
weig
h
t
s
is
g
iv
en
as
,
th
is
is
u
s
ed
is
s
elec
tin
g
th
e
ac
tio
n
.
T
h
e
s
ec
o
n
d
Q
-
v
alu
e
f
u
n
ctio
n
̂
(
,
−
,
)
f
o
r
ev
alu
atio
n
o
f
th
e
r
ewa
r
d
.
W
e
ass
u
m
e
th
at
−
.
Fu
r
th
er
,
th
e
p
ar
am
eter
s
o
f
̂
ar
e
u
p
d
ated
c
o
n
s
id
er
in
g
th
e
r
ate
o
f
u
p
d
atin
g
d
e
n
o
ted
as
o
f
th
e
d
esti
n
ed
n
etwo
r
k
.
T
h
e
ag
en
ts
u
s
e
a
g
r
ee
d
y
m
ec
h
a
n
is
m
b
ased
o
n
ep
s
ilo
n
f
o
r
ch
o
o
s
in
g
ac
tio
n
s
an
d
av
o
id
in
g
s
tallin
g
.
E
v
e
r
y
ag
en
t
co
n
s
id
er
s
th
e
b
est
ac
tio
n
(
,
(
,
,
)
)
b
ased
o
n
p
r
io
r
e
x
p
er
ien
c
es
h
av
in
g
1
−
as
a
p
r
o
b
a
b
ilit
y
.
E
v
e
r
y
a
g
en
t
u
tili
z
es
th
e
ex
p
er
ie
n
ce
to
en
h
a
n
ce
e
f
f
icien
cy
,
t
h
e
ex
p
er
ien
ce
s
ℶ
,
=
,
,
,
,
,
,
,
,
,
+
1
is
s
to
r
ed
in
th
e
r
ep
lay
b
u
f
f
e
r
d
en
o
ted
a
s
.
W
h
en
th
e
ex
p
er
ie
n
ce
s
th
at
ar
e
s
to
r
ed
|
|
is
ad
eq
u
ate
to
b
e
s
am
p
led
at
r
an
d
o
m
at
a
s
m
aller
b
atc
h
d
im
en
s
io
n
|
ℳ
|
,
th
is
ac
tio
n
is
p
er
f
o
r
m
ed
b
y
th
e
ag
en
ts
.
T
h
e
d
esti
n
ed
s
co
r
e
o
f
ev
er
y
lo
ca
l
m
o
d
u
le
at
th
e
p
h
ase
o
f
tr
ain
in
g
is
f
o
r
m
u
lated
as
(1
1
)
.
,
=
,
,
+
×
̂
(
,
+
1
,
(
,
+
1
,
,
;
)
;
−
)
)
(1
1
)
I
n
th
is
ca
s
e,
if
th
e
m
o
d
el
attain
s
th
e
ter
m
in
al
s
tate,
th
e
d
esti
n
atio
n
s
co
r
e
is
eq
u
iv
alen
t
to
,
,
.
T
h
e
Q
-
s
co
r
e
th
at
is
u
p
d
ated
is
g
iv
en
as
(
1
2
)
.
Fo
r
th
e
(1
2
)
,
[
0
,
1
]
wh
ich
d
ef
in
es
th
e
r
ate
o
f
lear
n
in
g
co
n
t
r
o
llin
g
th
e
s
p
ee
d
at
wh
ich
th
e
alg
o
r
ith
m
lear
n
s
.
E
v
er
y
lo
ca
l
m
o
d
u
le
ev
alu
ate
s
th
e
lo
s
s
f
u
n
ctio
n
an
d
u
tili
ze
s
it
to
m
in
im
ize
th
e
er
r
o
r
th
at
o
cc
u
r
s
d
u
r
in
g
tr
ain
i
n
g
.
T
h
is
is
r
elatin
g
to
th
e
m
ea
n
s
q
u
ar
ed
er
r
o
r
an
d
if
g
iv
en
as
(
1
3
)
.
Her
e,
th
e
s
u
b
in
d
ex
f
o
r
iter
atio
n
f
o
r
all
th
e
elem
en
ts
in
th
e
s
m
all
b
atch
is
g
iv
en
as
.
I
n
co
n
clu
s
io
n
,
t
h
e
lo
s
s
f
u
n
ctio
n
g
lo
b
ally
is
f
o
r
m
u
lated
as
g
iv
e
n
as
(
1
4
).
T
h
e
en
tire
p
r
o
ce
s
s
o
f
o
p
tim
ize
d
ADDN
d
escr
ib
ed
in
Alg
o
r
ith
m
1
.
(
,
+
1
,
,
+
1
;
)
=
(
1
−
)
×
(
,
,
,
;
)
+
×
,
(1
2
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
4
9
9
3
-
5
0
0
7
5000
(
)
=
(
|
ℳ
|
)
−
1
∑
(
−
(
,
,
)
)
2
ℳ
(1
3
)
(
)
=
(
∑
|
ℳ
|
=
1
=
1
)
−
1
∑
|
ℳ
|
×
(
)
=
1
(1
4
)
Alg
o
r
ith
m
1
.
Op
tim
ized
DDN
at
th
e
tr
ain
in
g
p
h
ase
Input:
,
,
,
,
ℎ
′
,
|
ℳ
|
,
,
,
,
Initialization:
′
,
′
,
ℎ
:
,
ℎ
:
=
−
=
,
|
|
←
∅
Output:
∗
Step 1:
For every episode
=
1
,
…
,
do
Initialization
,
=
0
Step 2:
While
=
0
do
For every transmission period
do
Step 3:
For every event
do
′
+
+
Local Modules
Step 4:
For every ML model
do
On observing the present state
,
The agent takes action
,
,
relating to
Step 5:
Agent receives
,
,
Situation varies to
,
+
1
and
stores
ℶ
,
Step 6:
If
|
|
ℎ
|
ℳ
|
then
Small batch of
|
ℳ
|
is sampled from
Step 7:
,
and
are evaluated using (10)
and
(12)
is used to update
Updating after back propagation;
−
←
+
(
1
−
)
−
Step 8:
End
End
Step 9:
If
(
′
,
ℎ
′
)
0
then
′
+
+
;
Global Module
Gather
,
′
,
′
−
to every
machine learning
model
Step 10:
End
End
End
End
Step 11:
If
ℎ
0
,
1
then
Apply
−
scheme
Step 12:
End
End
3
.
3
.
Ada
ptiv
e
re
s
o
urce
m
a
n
a
g
em
ent
m
o
du
le
T
h
e
ad
ap
tiv
e
r
eso
u
r
ce
m
an
a
g
em
en
t
m
o
d
u
le
(
AR
MM
)
h
as
t
o
b
e
im
p
lem
en
ted
w
h
en
th
e
c
h
o
s
en
b
ase
s
tatio
n
is
o
v
er
lo
a
d
ed
.
I
t
e
n
h
an
ce
s
th
e
ex
p
e
r
ien
ce
o
f
th
e
u
s
er
b
y
d
ec
r
ea
s
in
g
laten
c
y
a
n
d
r
ed
u
ce
s
th
r
o
u
g
h
p
u
t
b
y
p
r
ev
en
tin
g
co
n
g
esti
o
n
o
f
n
et
wo
r
k
.
I
t
also
e
n
s
u
r
es
all
th
e
n
etwo
r
k
co
n
tr
ib
u
tes
to
th
e
co
m
p
lete
p
er
f
o
r
m
an
ce
,
lead
in
g
to
ef
f
icien
t
u
s
ag
e
o
f
r
eso
u
r
ce
s
.
AR
MM
p
lay
s
an
ess
en
tial
r
o
le
in
en
ab
lin
g
a
s
ea
m
less
an
d
ef
f
icien
t
n
etwo
r
k
o
p
e
r
atio
n
s
b
y
d
y
n
a
m
ic
d
is
tr
ib
u
tio
n
o
f
tr
af
f
ic
an
d
o
p
tim
izatio
n
o
f
r
eso
u
r
ce
u
s
ag
e
v
ia
C
KAF.
T
h
e
p
r
o
ce
s
s
o
f
AR
MM
is
ex
p
lain
ed
in
d
etail
u
s
in
g
th
e
Alg
o
r
ith
m
2.
Alg
o
r
ith
m
2
.
AR
MM
Input:
ℝ
,
,
,
,
Initialization:
=
{},
=
0
,
(
)
=
0
Output:
,
,
,
Step 1:
Priority base scheduling is used
→
∗
Step 2:
For each E
⊂
∗
If
∑
+
|
|
=
1
then
Evaluate affected users
,
Step 3:
Append
(
)
to
End
Step 4:
For
do
Evaluate
(
)
Step 5:
If
(
)
ℎ
(
)
then
(
)
=
(
)
=
End
Step 6:
=
∑
|
|
=
1
+
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Dyn
a
mic
s
ervice
-
a
w
a
r
e
n
etw
o
r
k
s
elec
tio
n
fr
a
mewo
r
k
fo
r
mu
lti
-
o
b
jective
o
p
timiz
a
tio
n
…
(
B
h
a
va
n
a
S
r
in
iva
s
)
5001
4.
RE
SU
L
T
S
AND
D
I
SCU
SS
I
O
N
W
e
co
m
p
ar
e
o
u
r
p
r
o
p
o
s
ed
al
g
o
r
ith
m
ag
ain
s
t
f
o
u
r
s
tate
-
of
-
t
h
e
-
ar
t
alg
o
r
ith
m
s
.
T
h
at
i
m
p
lem
en
t
in
tr
a
-
s
lice
s
ch
ed
u
lin
g
m
ec
h
an
is
m
s
.
T
h
ese
in
cl
u
d
e
t
h
e
u
s
er
-
o
r
ien
t
ed
q
u
ality
o
f
s
er
v
ice
(
UQo
S)
alg
o
r
ith
m
[
2
3
]
,
t
h
e
m
ax
b
it
r
ate
th
r
o
u
g
h
p
u
t
m
u
lt
i
-
co
n
n
ec
tiv
ity
(
T
MC)
g
r
ee
d
y
alg
o
r
ith
m
[
2
4
]
.
T
h
e
p
r
io
r
ity
-
b
ased
p
r
o
p
o
r
tio
n
al
f
air
n
ess
(
PP
F)
alg
o
r
ith
m
[
2
5
]
,
[
2
6
]
.
4
.
1
.
Resul
t
s
T
h
e
cu
m
u
lativ
e
d
is
tr
ib
u
tio
n
f
u
n
ctio
n
(
C
DF
)
p
lo
t
illu
s
tr
ates
th
e
d
is
tr
ib
u
tio
n
o
f
av
er
ag
e
s
er
v
ice
th
r
o
u
g
h
p
u
t
ac
r
o
s
s
f
o
u
r
n
etwo
r
k
s
lices,
h
ig
h
lig
h
tin
g
t
h
eir
p
er
f
o
r
m
a
n
ce
v
a
r
iab
ilit
y
an
d
Q
o
S
d
if
f
er
e
n
tiatio
n
.
Netwo
r
k
s
lice
1
ac
h
iev
es
th
e
h
ig
h
est
th
r
o
u
g
h
p
u
t,
c
ater
in
g
to
b
an
d
wid
th
-
in
ten
s
iv
e
s
er
v
i
ce
s
,
wh
ile
s
lice
4
d
em
o
n
s
tr
ates
lo
wer
p
er
f
o
r
m
an
ce
,
lik
ely
allo
ca
ted
f
o
r
l
o
w
-
th
r
o
u
g
h
p
u
t
ap
p
licatio
n
s
s
u
ch
as
I
o
T
.
T
h
is
d
is
tr
ib
u
tio
n
r
ef
lects
th
e
im
p
ac
t
o
f
m
o
b
ilit
y
p
atter
n
s
a
n
d
s
er
v
ice
class
es
o
n
h
eter
o
g
en
eo
u
s
n
etwo
r
k
p
er
f
o
r
m
an
ce
,
alig
n
in
g
with
th
e
o
b
jectiv
e
o
f
a
n
aly
zin
g
t
h
eir
in
f
lu
en
ce
.
Ad
d
itio
n
ally
,
th
e
r
a
n
g
e
o
f
t
h
r
o
u
g
h
p
u
t
ac
r
o
s
s
s
lices
u
n
d
er
s
co
r
es
t
h
e
r
elev
an
ce
o
f
d
ev
elo
p
i
n
g
m
u
lti
-
o
b
jectiv
e
r
eso
u
r
ce
allo
ca
tio
n
s
tr
ateg
ies
to
o
p
tim
ize
f
air
n
ess
,
th
r
o
u
g
h
p
u
t,
an
d
Qo
S,
en
s
u
r
in
g
ef
f
ici
en
t
r
eso
u
r
ce
u
tili
za
tio
n
ac
r
o
s
s
d
iv
er
s
e
s
er
v
ice
d
em
an
d
s
.
Fig
u
r
e
2
s
h
o
ws th
e
C
DF o
f
av
er
ag
e
s
er
v
ice
th
r
o
u
g
h
p
u
t
f
o
r
n
etwo
r
k
s
lices.
T
h
e
C
D
F
p
l
o
t
d
e
m
o
n
s
t
r
a
t
es
t
h
e
r
a
d
i
o
r
e
s
o
u
r
c
e
a
l
l
o
c
a
ti
o
n
d
i
s
t
r
i
b
u
t
i
o
n
f
o
r
f
o
u
r
n
e
t
w
o
r
k
s
l
i
c
e
s
,
s
h
o
w
c
as
i
n
g
t
h
e
s
y
s
te
m
'
s
a
b
i
l
ity
t
o
h
a
n
d
l
e
d
i
f
f
e
r
e
n
t
i
at
e
d
r
e
s
o
u
r
c
e
d
e
m
a
n
d
s
.
N
et
w
o
r
k
s
l
i
c
e
1
a
c
h
i
e
v
e
s
t
h
e
h
i
g
h
e
s
t
a
l
l
o
c
a
ti
o
n
,
s
u
g
g
e
s
t
i
n
g
p
r
i
o
r
i
ti
za
t
i
o
n
f
o
r
h
i
g
h
-
d
e
m
a
n
d
a
p
p
l
i
c
a
t
i
o
n
s
,
w
h
i
le
s
li
c
e
4
e
x
h
i
b
i
t
s
t
h
e
l
o
w
e
s
t
al
l
o
c
a
ti
o
n
,
l
i
k
e
l
y
r
es
e
r
v
e
d
f
o
r
l
o
w
-
r
es
o
u
r
c
e
s
e
r
v
i
ce
s
s
u
c
h
as
I
o
T
.
T
h
e
g
r
a
d
u
a
l
a
n
d
u
n
i
f
o
r
m
r
i
s
e
i
n
ea
c
h
c
u
r
v
e
i
n
d
i
c
a
t
es
f
a
i
r
n
e
s
s
w
i
t
h
i
n
s
l
i
c
es
,
t
h
o
u
g
h
i
n
t
e
r
-
s
l
i
ce
g
a
p
s
h
i
g
h
l
i
g
h
t
o
p
p
o
r
t
u
n
i
t
i
e
s
f
o
r
o
p
t
i
m
i
z
i
n
g
r
e
s
o
u
r
c
e
s
h
a
r
i
n
g
.
T
h
i
s
a
n
a
l
y
s
is
a
l
i
g
n
s
w
i
t
h
t
h
e
o
b
j
ec
t
i
v
e
o
f
d
e
s
i
g
n
i
n
g
r
e
s
o
u
r
c
e
es
t
i
m
at
i
o
n
a
l
g
o
r
i
t
h
m
s
t
o
b
a
l
a
n
c
e
Q
o
S
a
n
d
e
n
s
u
r
e
e
f
f
i
c
i
e
n
t
r
es
o
u
r
c
e
u
t
i
li
z
a
ti
o
n
a
cr
o
s
s
h
e
t
e
r
o
g
e
n
e
o
u
s
n
et
w
o
r
k
s
l
ic
e
s
.
Fi
g
u
r
e
3
s
h
o
ws
t
h
e
r
a
d
i
o
r
e
s
o
u
r
c
e
a
l
l
o
ca
t
i
o
n
.
Fig
u
r
e
2
.
C
DF o
f
av
e
r
ag
e
s
er
v
ice
th
r
o
u
g
h
p
u
t
f
o
r
n
etwo
r
k
s
lices
Fig
u
r
e
3
.
R
ad
io
r
eso
u
r
ce
allo
c
atio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
4
9
9
3
-
5
0
0
7
5002
Fig
u
r
e
4
d
em
o
n
s
tr
ates
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
p
atter
n
s
ea
r
ch
(
PS
)
alg
o
r
ith
m
a
cr
o
s
s
th
r
ee
co
n
f
ig
u
r
atio
n
s
(
N1
,
N2
,
a
n
d
N3
)
in
ter
m
s
o
f
s
lice
th
r
o
u
g
h
p
u
t
as
th
e
n
u
m
b
er
o
f
eM
B
B
n
etwo
r
k
s
er
v
ices
in
cr
ea
s
es.
N1
ex
h
ib
its
lim
ited
th
r
o
u
g
h
p
u
t
(
ca
p
p
e
d
at
8
0
Mb
p
s
)
,
wh
ile
N2
ac
h
iev
es
u
p
to
1
4
0
M
b
p
s
,
an
d
N3
o
u
tp
er
f
o
r
m
s
b
o
th
with
a
p
ea
k
o
f
1
6
5
Mb
p
s
,
s
h
o
wca
s
in
g
s
ig
n
if
ican
t
s
ca
lab
ilit
y
an
d
r
eso
u
r
ce
ef
f
icien
cy
.
T
h
e
tr
an
s
itio
n
f
r
o
m
N1
to
N3
h
ig
h
lig
h
ts
a
s
u
b
s
tan
tial
im
p
r
o
v
em
en
t
in
r
eso
u
r
ce
u
tili
za
tio
n
,
with
th
r
o
u
g
h
p
u
t
g
ain
s
o
f
7
5
a
n
d
1
7
.
8
%
f
o
r
N2
an
d
N3
,
r
esp
ec
tiv
ely
.
T
h
is
r
ef
lects
th
e
alg
o
r
ith
m
’
s
ad
a
p
tab
ilit
y
to
d
iv
er
s
e
s
er
v
ice
d
em
an
d
s
,
alig
n
in
g
wi
th
th
e
o
b
jectiv
es
o
f
o
p
tim
izin
g
r
eso
u
r
ce
allo
ca
tio
n
an
d
ad
d
r
ess
in
g
m
o
b
ilit
y
p
atter
n
s
in
h
eter
o
g
en
eo
u
s
n
et
wo
r
k
s
.
Fig
u
r
e
5
illu
s
tr
ates
th
e
s
er
v
in
g
p
r
o
b
a
b
ilit
y
o
f
th
e
PS
alg
o
r
ith
m
ac
r
o
s
s
th
r
ee
co
n
f
ig
u
r
atio
n
s
(
N1
,
N2
,
an
d
N3
)
as
th
e
n
u
m
b
er
o
f
eM
B
B
n
etwo
r
k
s
er
v
ices
in
cr
ea
s
es.
N1
s
h
o
ws
s
tead
y
g
r
o
wth
b
u
t
s
tab
ilizes
at
0
.
8
,
in
d
icatin
g
lo
wer
r
elia
b
ilit
y
u
n
d
e
r
h
ea
v
y
lo
ad
s
.
I
n
co
n
tr
ast,
N2
an
d
N3
o
u
tp
er
f
o
r
m
N1
,
with
N
3
ac
h
iev
in
g
n
ea
r
-
o
p
tim
al
s
er
v
in
g
p
r
o
b
ab
ilit
y
(
~
0
.
9
5
)
,
d
e
m
o
n
s
tr
atin
g
s
u
p
er
i
o
r
s
ca
lab
i
lity
an
d
r
eso
u
r
ce
allo
ca
tio
n
.
T
h
e
p
r
o
g
r
ess
io
n
h
i
g
h
lig
h
ts
th
e
ef
f
icien
c
y
o
f
th
e
PS
alg
o
r
ith
m
in
m
ee
tin
g
Qo
S r
eq
u
ir
em
e
n
ts
an
d
its
ad
ap
tab
ilit
y
to
in
cr
ea
s
ed
s
er
v
i
ce
d
em
an
d
s
,
alig
n
in
g
with
th
e
o
b
jectiv
es
o
f
o
p
tim
izin
g
r
eso
u
r
ce
allo
ca
tio
n
an
d
en
h
an
cin
g
u
s
er
s
atis
f
ac
tio
n
in
h
eter
o
g
en
e
o
u
s
n
etwo
r
k
s
.
Fig
u
r
e
4
.
Per
f
o
r
m
an
c
e
o
f
t
h
e
PS
alg
o
r
ith
m
ac
r
o
s
s
th
r
ee
co
n
f
ig
u
r
atio
n
s
Fig
u
r
e
5
.
Ser
v
i
n
g
p
r
o
b
a
b
ilit
y
o
f
th
e
PS
alg
o
r
ith
m
ac
r
o
s
s
th
r
ee
co
n
f
ig
u
r
atio
n
s
Fig
u
r
e
6
co
m
p
a
r
es
th
e
p
s
eu
d
o
r
an
d
o
m
ized
s
ea
r
ch
s
tr
ateg
y
(
PR
S
S),
T
MC,
an
d
th
e
p
r
o
p
o
s
ed
PS
alg
o
r
ith
m
ag
ain
s
t
th
e
to
tal
n
u
m
b
er
o
f
eM
B
B
n
etwo
r
k
s
er
v
i
ce
s
.
T
h
e
PS
alg
o
r
ith
m
co
n
s
is
ten
tly
o
u
tp
er
f
o
r
m
s
th
e
o
th
er
s
,
ac
h
iev
in
g
a
p
ea
k
th
r
o
u
g
h
p
u
t
o
f
ap
p
r
o
x
im
ately
1
6
5
Mb
p
s
,
s
h
o
wca
s
in
g
s
u
p
er
io
r
r
eso
u
r
ce
allo
ca
tio
n
an
d
s
ca
lab
ilit
y
.
T
h
e
T
MC
alg
o
r
ith
m
ex
h
ib
its
m
o
d
er
ate
p
e
r
f
o
r
m
an
ce
,
s
tab
ilizin
g
n
ea
r
1
4
0
Mb
p
s
,
wh
ile
th
e
PR
S
S
alg
o
r
ith
m
lag
s
b
e
h
in
d
,
p
ea
k
in
g
at
ar
o
u
n
d
1
2
0
Mb
p
s
.
T
h
e
r
esu
lts
h
ig
h
lig
h
t
th
e
e
f
f
icien
cy
o
f
th
e
PS
alg
o
r
ith
m
in
o
p
tim
izin
g
th
r
o
u
g
h
p
u
t,
esp
ec
ially
u
n
d
er
h
ig
h
s
er
v
ice
d
em
an
d
s
,
alig
n
in
g
with
o
b
jectiv
es
r
elate
d
to
en
h
an
ci
n
g
Qo
S a
n
d
ef
f
icien
t r
eso
u
r
ce
u
tili
za
tio
n
in
h
eter
o
g
en
eo
u
s
n
etwo
r
k
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.