TELKOM
NIKA
, Vol.11, No
.3, March 2
0
1
3
, pp. 1652 ~ 1664
ISSN: 2302-4
046
1652
Re
cei
v
ed
No
vem
ber 5, 20
12; Re
vised Janua
ry 2
6
, 20
13; Accepted
February 8, 2
013
A Novel QoS Routing Algorithm in Wireless Mesh
Networks
Liu Hui
Jian
g
x
i Un
ivers
i
t
y
of Scie
nce a
nd T
e
chnol
og
y
Ganzho
u, Chi
n
a
e-mail: l
x
yl
iu
hu
i@16
3.com
A
b
st
r
a
ct
W
i
th the rapid deve
l
op
ment o
f
informati
on techno
logy, pe
op
le
’
s
da
ily life be
comes more a
nd mor
e
dep
en
dent on
w
i
reless technol
ogi
es. W
i
reless mes
h
net
w
o
rk consists
of a numb
e
r of characteristi
c
s
associ
ated w
i
th the return path,
w
i
th
a strong fault toleranc
e,
stability, w
i
d
e
ly used by
the light to the ci
ty
netw
o
rk constr
uction,
mi
litary
app
licati
ons,
and k
e
y servic
e prov
ider
s
an
d other fi
elds.
Co
mp
ared w
i
t
h
traditio
nal w
i
re
d commu
n
ic
ati
on techno
lo
gie
s
, how to
provide qu
alifi
ed QoS routing ser
v
ice is a prima
r
y
p
r
o
b
l
e
m
fo
r wire
l
e
ss
me
sh
ne
two
r
ks wa
i
t
i
n
g
fo
r e
ffe
cti
v
e so
l
u
tio
n
.
Re
ga
rd
in
g
th
i
s
p
r
ob
l
e
m
,
th
i
s
p
aper
app
lys the theory of Evolution
a
ry Game to QoS
routing algorit
hm fo
r
w
i
reless me
sh netw
o
rks a
n
d
prop
osed
a n
o
v
el a
l
gor
ith
m
c
a
lle
d EGW
R
A
w
h
ich can
not
only
improv
e the p
e
rfor
ma
nce of traditi
on
al
QoS
routin
g protoc
o
l
s but also b
e
a
b
le to re
d
u
ce t
he cost of the routin
g al
gorith
m
.
Ke
y
w
ords
: w
i
reless tech
no
lo
gy; QoS routin
g; me
s
h
netw
o
rks; evolutio
nar
y game theory
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Wirel
e
ss loca
l area netwo
rks are con
s
id
ered a
s
one of the
viable solutio
n
s for providin
g
Internet conn
ectivity to mobile users. Th
e comm
ercial
wirele
ss LA
Ns tod
a
y, e.g., IEEE 802.11 ,
provide si
ngl
e hop wirel
e
ss con
n
e
c
tion to the infrastructure. All th
e mobile use
r
s need to re
si
de
within the a
c
cess point’
s
re
ach, i.e., within the
basi
c
service
set wit
h
a radiu
s
of
100
~30
0
met
e
rs.
In order to provide full coverag
e
in relatively
large size area
s we n
eed to deploy a large number
of access poi
nts (AP). Thu
s
, the conve
n
t
ional infr
a
s
tructure-ba
s
ed
wirele
ss LA
Ns d
o
not scale
well with the t
a
rget coverag
e
area a
nd n
u
mbe
r
of nod
es.
We ca
n cla
s
sify wirele
ss networkin
g architec
tu
re
as follows 1
)
point-to-m
u
ltipoint
infras
t
r
uc
ture-aided approac
h
lik
e
wireless
s
i
ngle hop net
work
s
(e.g., IEEE 8
02.11 wireless
LAN) a
nd 2)
peer-to-p
e
e
r
multihop ap
proac
h e.g., mobile ad h
o
c
netwo
rks (MA
N
ETs).
The differen
c
e between m
e
sh
netwo
rks and conv
ent
ional infrastru
c
ture
wi
rele
ss LANs
is the fa
ct that mesh
net
works result in a
multiho
p
topology
which
req
u
ire
s
decentrali
ze
d
coo
r
din
a
tion. The differen
c
e between m
e
sh net
wo
rks and the mobile ad hoc networks re
side
s in
the existence
of the infrastructu
re conn
ection.
The a
c
cess point
s deploye
d
can
act both as a
peer of the internal wi
rele
ss ad ho
c ne
twork and th
e bridge to the wire
d network. To pro
v
ide
s
u
ffic
i
ent infras
truc
ture acc
e
s
s
b
and
wi
dth, multiple acce
ss p
o
in
t
s
ca
n be de
ployed withi
n
th
e
netwo
rk. T
r
af
fic balan
cin
g
can b
e
achie
v
ed by
the underlyin
g me
sh ro
uting p
r
otocol. The
mesh
netwo
rk featu
r
es p
r
e
s
e
n
ted
above lend t
hemselves to
easy scala
b
il
ity.
Wirel
e
ss me
sh networks
seaml
e
ssly integrat
e the
s
e two network architectu
res. This
integratio
n is
obtaine
d by the propo
sed
WM
R prot
o
c
o
l
, implemente
d
in ea
ch wi
reless no
de. T
he
con
n
e
c
tivity to the wi
red
backb
one i
s
provide
d
by the wirele
ss
infrast
r
u
c
ture
acce
ss p
o
in
ts.
Each no
de in
the network is both a se
rvice pr
ovid
er and a servi
c
e con
s
um
er, i.e., each no
de
has forwa
r
di
n
g
ability similar to the nod
es’ fun
c
tionali
t
y in MANETs.
As a new type of wirele
ss netwo
rk, wi
rele
ss me
sh
netwo
rk [1, 2, 19, 31-34]
con
n
e
c
t
mesh n
ode
s throug
h wirel
e
ss links to construc
t a dynamic topol
o
g
y, self-org
a
n
izin
g and m
u
lti-
hop wi
rele
ss interco
nne
cted netwo
rk.
Compa
r
ed
with the tra
d
itional sin
g
l
e
-ho
p
wirele
ss
netwo
rks, it can extend co
verage,
enh
a
n
ce robu
stn
e
ss, redu
ce deployme
nt cost and incre
a
se
cap
a
city. Co
mpared to the traditional
switch
ed
n
e
tworks, wireless Me
sh netwo
rk
cabl
ing
betwe
en nod
es rem
o
ved need
s, but still has a di
strib
u
ted netwo
rk
provide
s
red
u
ndan
cy and re-
routing. In the
wirele
ss Mesh netwo
rk, if you w
ant to add a new d
e
vice, simply pl
ug in the power
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Novel
QoS Routin
g Algorithm
in Wirele
ss M
e
sh Net
w
orks (Li
u
Hu
i)
1653
on it,
it can automatically config
ure itse
lf,
and determine the best multi-hop transmi
ssion p
a
th.
Add or mobil
e
device, the network top
o
logy
cha
nge
s can auto
m
atically find and automatically
adju
s
t the tr
affic routing in orde
r to
obtain t
he most efficient transmissio
n
path. A typ
i
cal
Wirel
e
ss me
sh netwo
rk
ca
n be de
scribe
d like Figu
re
1
.
Wirel
e
ss m
e
sh network’s
b
u
sin
e
ss i
s
usually
gathe
re
d in the Me
sh Ro
uter o
r
Gateway,
easily lead to local network conge
stion,
makin
g
it difficult to maintain network
globally opti
m
al
routing. Th
us routing p
r
ot
ocol
s mu
st b
e
able to
ad
apt to this sit
uation so a
s
to provide b
e
tter
QoS for the
use
r
s. So, the rese
arch o
f
wireless
m
e
sh net
work routing p
r
oto
c
ol is of gre
a
t
signifi
can
c
e.
Figure 1. Typical st
ru
cture
of wirele
ss m
e
sh n
e
two
r
k
2. Related Works
Reg
a
rdi
ng th
e Qo
s proble
m
of Routin
g
Algorit
hm for Wireless
Mes
h
Networks
, lots
of
schola
r
s in thi
s
filed have
made some a
c
hievem
ents.
Some of them with rep
r
e
s
entatives can
be
listed a
s
follo
ws. Vo, Hu
ng
Quoc [3, 4, 16, 17, 18]
pro
pose a novel
approa
ch to control the traf
fic
to avoid cong
estion
by de
couplin
g forwa
r
ding
proc
ess from the
rout
ing p
r
o
c
ess.
The p
r
op
osal
is
intere
sted in
provisi
onin
g
quality of se
rvice gu
ara
n
tees fo
r pla
n
n
ed Wi
rele
ss
Mesh
Net
w
o
r
ks
whi
c
h are bei
ng used wi
de
ly as a bro
a
d
band
wirel
e
ss acce
ss net
work. Zho
u
, Hao [5, 6, 12,
13]
focu
se
s on
QoS ro
uting
with ba
nd
wi
dth co
nst
r
ai
nt in multi-radio multi-channel WMN, and
prop
oses
a n
e
w multimet
ri
c and
a QoS
routing p
r
ot
ocol MM
QR.
The ro
uting
metric h
a
s t
w
o
advantag
es.
First, it repla
c
es the tra
n
smissi
on rate
of ETT with available ban
d
w
idth so that the
node
s
with light load a
r
e
more li
kely to
be sel
e
cte
d
. Secon
d
, it take
s the
chan
nel diversity into
accou
n
t and assign
s a wei
ght to each link acco
rdin
g to the channe
ls of links within the range
o
f
three ho
ps. Sun, Xuemei
[7, 8, 14, 1
5
] propo
se
s a QoS routin
g algorithm b
a
se
d on cult
ure
-
particl
e swa
r
m optimizatio
n algo
rithm
s
. The alg
o
rith
m use
s
th
e
dual-evolutio
n
mechani
sm
of
culture algo
ri
thms and a
c
hieves furth
e
r
impr
ovem
e
n
t on global
optimum location mutati
on
particl
e swa
r
m optimizatio
n algorith
m
s
(MPSO) by
i
n
trodu
cin
g
the con
c
e
p
t of inertia weig
ht.
Zhou, Ha
o [21, 22, 23] propo
se
s a ne
w multim
etri
c and a QoS routing p
r
oto
c
ol MMQ
R. The
routing
metri
c
has t
w
o a
d
vantage
s. First
,
it repl
aces t
he tran
smi
s
si
on rate
of ETT with availa
ble
band
width so
that the nodes with light l
oad are mo
re
likely to be selecte
d
. In additional, it takes
the chan
nel diversity into accou
n
t and assign
s a
we
ight to each link acco
rdin
g to the chann
els
of links withi
n
the range
of three hop
s. Agarwal,
Anjali [24, 25,
26] propo
se
s an Ants-in-Mesh
(AIM) ro
uting
proto
c
ol for
wirel
e
ss me
sh netwo
rks,
whi
c
h is b
a
sed on ide
a
s
from the nat
ure
-
inspi
r
ed Ant Colo
ny Optimization
(ACO) frame
w
o
r
k. AIM agent distribut
e
s
forward ants
on
deman
d to search for the route
s
to the
desti
n
a
tion and then activates co
rrespo
nding ba
ckward
ants to
confirm the ro
utes and
up
date
the phe
rom
one. AIM en
able
s
only th
e de
stination
to
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 2302-4
046
TELKOM
NIKA
Vol. 11, No
. 3, March 20
13 : 1652 – 1
664
1654
cho
o
se k mu
ltiple paths b
a
se
d on Ants Phero
m
on
e
,
which is ba
sed on
several link-releva
nt
Quality of Se
rvice
(QoS
)
metrics. Ron
g
Bo [
27, 2
8
,
29, 30] p
r
o
pose a
novel
netwo
rk g
r
a
p
h
prep
ro
ce
ssin
g app
roa
c
h t
o
enabl
e traf
fic engin
e
e
r
i
ng and
enh
a
n
ce th
e pe
rforma
nce of QoS
multica
s
t ro
uting alg
o
rithm
s
. In this
app
roa
c
h,
we e
m
ploy pri
o
riti
zed
admi
s
sio
n
co
ntrol
sch
e
me
and devel
op
a utility-con
s
traine
d optima
l
priority gain
policy.
Traditio
nal Q
o
S routing al
gorithm
s like
the
achievem
ents listed ab
ove only conside
r
ed
singl
e obje
c
ti
ve perfo
rman
ce p
a
ra
meters, su
ch a
s
d
e
lay boun
d
or ba
nd
width
limitations,
or
static multi-o
b
jective con
s
train
ed
situ
ati
on. WM
N can not m
eet the nee
ds of Ge
ne
ral
Req
u
ire
m
ent
s for
som
e
busi
n
e
ss li
ke
dynamic m
u
lti-obje
c
tive
perfo
rman
ce,
su
ch a
s
d
e
l
a
y,
band
width a
n
d
Freq
uen
cy con
g
e
s
tion.
This
pape
r
applie
s the t
heory of Ev
olutiona
ry G
a
me to Q
o
S routing
algo
rithm for
wirel
e
ss m
e
sh networks which
ca
n not
only improv
e
the pe
rform
a
nce
of traditio
nal QoS
routi
ng
proto
c
ol
s but also b
e
able t
o
redu
ce
the
co
st of the ro
uting algo
rith
m.
We a
r
e g
o
ing
to extract a
stable core i
n
MANET in terms of mobilit
y in the goal t
o
se
rve
QoS-aware
routing to ta
ke network m
obility to
fi
nd
the sp
eci
fi
c resou
r
ces.
S
u
ch an
extra
c
tion
can de
fi
n
e
a sub
s
et of MNsin the netwo
rk whe
r
e hi
s
mobility is low and the lin
ks b
e
twe
en them
are reliable i
n
time. Therefore, the sel
e
cted p
a
th
throug
h this co
re is mo
re st
able in term
s of
mobility, and con
s
e
que
ntly the require
d
QoS are mo
re guaranteed
. Indeed, we can
fi
nd a pa
th
that reque
sts our requi
rem
ent in QoS such a
s
band
width, but the mobility of some interme
d
iate
MNs
con
s
i
s
ting of this path is high. Taking in to acco
unt the coope
ration bet
wee
n
MNs in terms
of mobility is
a promisi
ng i
dea to
fi
n
d
th
e ade
quate
stable core an
d co
nsequ
ent
ly a stabl
e pa
th.
In this asse
ssment so a
s
to extract this co
re, we use a co
ope
rative game theory. For this
obje
c
tive, we consi
der M
N
s in MANE
T as players, and we try to
fi
nd a coo
perative co
ali
t
ion
(partition) of players that minimizes
the mobility criteria in the net
work.
3. Summar
y
of EGWRA
3.1. The fra
m
e
w
o
r
k o
f
EGWRA
EGWRA, our propo
se
d archite
c
ture, is
desig
ned to
take full ad
vantage of WM
N
architectu
re.
EGWRA mixes p
r
oa
ctive route co
mputa
t
ion for route
r
s and o
n
-de
m
and path
se
tup
for clients. This design
eases the
management of
client m
obility and reduces routing table
si
ze.
Proa
ctive rou
t
e comp
utatio
n is pe
rform
e
d usin
g a Di
stance Ve
ctor
(DV)
app
roa
c
h. On-d
eman
d
path setup i
s
perform
ed b
y
new extensions introdu
c
ed in the rou
t
ing proto
c
ol in orde
r to take
advantag
e of the architectu
re of WM
Ns.
In EGWRA, the ba
ckbon
e
beco
m
e
s
totally tr
anspare
n
t to the clie
nts, whi
c
h do
not
need to emb
ed any new feature. Thi
s
mean
s that
if two clients a
s
soci
ated to different WM
Rs
wish to com
m
unicate, the set of WMRs will forward the traf
fi
c at
the IP level and not at the
link
level. To obtain su
ch a b
ehavior, WM
Rs h
a
ve to be more tha
n
a simple fo
rwarde
r and
more
than a simpl
e
router. In particul
a
r, on the client
su
b
netwo
rk inte
rface, whi
c
h can be descri
bed
like figure 2.
Figure 2.The
frame
w
ork of
EGWRA
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Novel
QoS Routin
g Algorithm
in Wirele
ss M
e
sh Net
w
orks (Li
u
Hu
i)
1655
The
WMR acts in
such a
way to let local c
lie
nts thin
kthat remote
client
s a
r
e in
the
local WLAN. Then, it is up to the WMRs to
fi
nd out to which WM
R the
client is asso
ciated
and
route the pa
ckets a
c
co
rdin
gly.
The ba
sic ta
sk of the Enha
nce
d
DV mo
dule is to mai
n
tain pro
a
ctiv
ely a route toward
any other WMR pre
s
e
n
t in the backbo
ne sub
net
wo
rk. Thi
s
Enh
anced DV m
odule is a
n
IPv6
impleme
n
tation of DSDV
, based
on
existing
IPv4 code. Som
e
enha
ncem
ents have
b
een
introdu
ce
d in order to co
o
perate
with the Client Ma
nage
r modul
e. This modu
le is the only one
that accesse
s
the kernel ro
uting table.
The
NDP-P
roxy module
provide
s
the
WM
R with t
heability to a
c
t like
a Nei
ghbo
r
Discove
r
y Protocol
proxy.This
allo
ws t
o
tran
sp
are
n
tly forwa
r
d
p
a
ckets towa
rd the
WM
R
the
destin
a
tion cli
ent is associ
a
t
ed with. In
this
way, no particula
r mech
anism
s are nece
s
sary on the
client side, in order to communi
cate with remo
te cli
ents. The NDP-proxy module will correctly
answe
r to the ICMPv6 requ
est se
nt by the local
client
s.
The IPv6 Forwarde
r modul
e enca
p
sulates all
IPv6 pa
ckets in anot
her IPv6 packet, in
orde
r to ship
packets toward the de
stination c
lie
nt.
This solution,
though intro
duci
ng a ce
rtain
amount of overhe
ad, avoid
s
kee
p
ing sta
t
e in t
he WMRs alo
ng the path bet
we
en
clients. Inde
ed,
only the WMRs at the edges of the path must be
aware that the two client
s are communi
ca
ting.
Another sig
n
i
fi
cant advantage is the robustn
ess to
mobilit
y. If a
client chang
es the WMR to
whi
c
h it is asso
ciated, onl
y the
WMRs at the
edges of the path must updat
e the information
(3
WM
Rs in tot
a
l). WM
Rs
al
ong the path
do not nee
d
to make an
y update, wh
ile contin
uing
to
relay pa
ckets.
The Cli
ent M
anag
er m
odu
le is respon
si
ble
for di
scov
ering
whi
c
h
client is a
s
so
ciated
to which WMR in an
on
-dem
and fa
shion. Wh
en
a local
client
wish
es to
communi
cate
to a
remote
clie
nt (a
ssoci
a
ted
to anothe
r
WMR), it
sen
d
s
a m
u
ltica
s
t
req
u
e
s
t in o
r
de
r to di
sco
v
er
whe
r
e the
cli
ent is. Th
e result of thi
s
query i
s
st
ored in the
Foreign
Client T
able. The
Cli
ent
Manag
er mo
dule also ma
nage
s the reply to
reque
st
s sent by other WM
Rs. Fu
rtherm
o
re, since
Mesh
DV mu
st be a
w
a
r
e
of the clie
nts that are
a
ssociate
d
to th
e WM
R, the
Client Ma
nag
er
module m
onit
o
rs the
set of client
s asso
ci
ated to wl
anI
and sto
r
e
s
them in the Local Client Tabl
e.
3.2. Summary
of Ev
o
l
utionar
y
Game T
h
eor
y
yy
Non
-
coop
erative gam
e
theory is the deci
s
io
n-m
a
kin
g
in a distribute
d
envi
r
onm
ent,
the analysi
s
of individual utility maximi
zation Playe
r
for the optimal policy ch
oice. Evolutio
nary
game, no
n-coope
rative g
a
me, a bran
ch of a g
a
m
e strategy
for furthe
r a
nalysi
s
of g
a
me
popul
ations i
n
a lon
g
-te
r
m
stability. Evolution of
the
Na
sh e
quilib
rium (all Pl
ayer of the
opti
m
al
strategy
) with groups
of st
ability, which is executed
when
the other Player
bal
anced strategy,
any Player can not be balanced by a u
n
ilateral de
pa
rture from the strategy
for more effec
t
ive;
Mean
while, the implem
en
tation of a balan
ced
strat
egy can
reve
al the individ
ual pro
p
o
r
tio
n
of
total populati
on.
As the n
o
vel
achi
evement
in the resea
r
ch
field of n
on-coo
p
e
r
ative gam
e theo
ry [9,
10, 11, 20],
the rese
arch
on evolutionary game t
heory attract
s
great attentions in not only
aca
demy but
also in
du
stry field. Integration
of ev
olutiona
ry ga
me theo
ry, eco
nomi
cs
a
n
d
evolutiona
ry biology of rati
onal thought,
no longer h
u
man mod
e
l into the game supe
r-ratio
nal
side, that the human is usually achieve
d
th
rough tria
l and erro
r method of game balance, and
biologi
cal evolution
i
s
co
mmon,
the choice
Th
e
ba
lance is the
balan
cing
pro
c
e
ss to
a
c
hie
v
e a
balan
ce
d function, and th
us the histo
r
i
c
al, inst
itutio
nal factors a
nd the balan
cing p
r
o
c
e
s
s are
some of the details of the game will a
ffect the choice of multiple equilibria.
Set the evolution game lo
cated in an N-node
MA
NET
,
any node wi
th M, that except
the node i ot
her than the
colle
ction.
i
N
-
denotes the d
a
ta packet
s
ge
nerate
d
by source no
de i
whi
c
h is
calle
d i's group. Data betwe
en
sou
r
ce
and d
e
stinatio
n no
des transmit a compl
e
te d
a
ta
servi
c
e is cal
l
ed a sessio
n; the node mobility w
ill l
ead to chan
ges in netwo
rk topolo
g
y, the
compl
e
tion of
a sessio
n re
quire
s multipl
e
routing p
a
th
s to create dif
f
erent group
s.
The othe
r p
a
ram
e
ter M
trust set up
in all the
main co
mpo
nents of the
set of
strategi
es for the real number field o
n
an in
terval
X, strategy
by the probability densi
t
y
distrib
u
tion f (x) to ch
ara
c
te
rize
and
de
si
gn the fitne
s
s function
(,
)
ux
l
is a
contin
uou
s fu
nction i
n
domain
X
´L
, which is the envi
r
onm
ental pa
ramete
rs,
a
s
Environme
n
t, the probabili
ty density.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 2302-4
046
TELKOM
NIKA
Vol. 11, No
. 3, March 20
13 : 1652 – 1
664
1656
Note the pro
bability densi
t
y
function of the com
posi
t
ion of all
th
e set. The evolution of th
e
netwo
rk
confi
guratio
n soft
ware, game
evolutiona
ry
sele
ction i
s
linke
d with fitness. Whe
n
gi
ven
the definition
of the condi
tions of envir
onmental p
a
rameters sele
ction op
erato
r
:
X
X
TD
D
l
®
and the ave
r
age
sele
ctio
n operator
:
XX
TD
D
l
®
, the environ
ment, the average fitne
s
s
function.
Acco
rdi
ng to the statements listed ab
ov
e. The evolution game
for WNN can be
defined. So t
hat G =
(I, P, U).
Whe
r
e: I = {
N
i, N-i
}
for
the Player
col
l
ection,
Ni sai
d
nod
e i, and
Ni
indicates that
the netwo
rk
node
s in ad
di
tion to a
ll the node
s outsi
d
e
i; P is the strategy set P =
{pi, p-i}, 1 p
a
c
ket tran
smission, refused
to transmi
t a messag
e to 0
.
U is the utili
ty function, the
utility
function with Ui (si,
s-
i) to represent. It proceeds B (si,
s-i)> 0 from the i (the correct
transmissio
n) and cost
s C (si, s-
i) <0 (energy con
s
umption
)
of two part
s
. No
de i can be the
watchdo
g or
other mo
nitori
ng and fee
d
b
a
ck mechani
sms, that messag
e tran
smi
ssi
on to the n
e
xt
node if there
are Byzanti
ne errors
. Whether the b
enefits or
co
st
s i have a
strategy with
all
relevant pa
rticipa
n
ts, this is a strategy g
a
me
or actio
n
nodes are the embodim
e
n
t
of interactio
n,
about the Byzantin
e fault-tolera
nt
WMN classic pri
s
o
ner'
s
dilemm
a problem
s a
nd to link the
two
adja
c
ent Betwee
n nod
es
Ni and
Nj pri
s
oner's dil
e
m
m
a sho
w
n in
Table1.
Table 1. The
game Matrix
betwe
en lab
o
r node
s in WNN
successfu
l
failur
e
The packet reac
hs node Nj
( B ,C )
( 0 , C )
The packet not r
eachs node Nj
( B, 0)
( 0, 0 )
3.3. Summary
of Ev
o
l
utionar
y
Game T
h
eor
y
Assu
ming th
e
initial time in the game, th
e us
e of tech
nical
coo
p
e
r
a
t
ion in su
pply
-
sid
e
strategy
of p
opulatio
n rati
o of p(
01
p
££
). The
pro
portio
n
o
f
non-co
ope
rative strate
gi
es u
s
in
g
(1-p).Te
ch
nol
ogy used i
n
co
ope
rati
on de
man
d
-side
strateg
y
of popul
ation ratio
of
q(
01
q
££
).The prop
ortion of non
-co
ope
rative strategi
es u
s
i
ng 1-q. The t
e
ch
nical coo
peratio
n
strategy for the sele
ction
of
side popul
ation expecte
d to pay
for can be expre
s
sed with eq
u
a
tion
1.
1
1
11
11
11
1
[(
1
)
]
(
1
)
(
)
uq
a
v
c
q
a
v
c
av
q
a
v
c
p
p
=+
-
+
-
-
=
+
-
(1)
Tech
nolo
g
y sup
p
liers ex
pect to pay
for the average p
opul
ation, whi
c
h can be
descri
bed with
equatio
n2.
11
12
11
1
1
11
1
(1
)
()
(
1
)
q
Up
u
p
u
p
av
q
a
v
c
p
a
v
av
q
p
c
p
av
p
p
=+
-
=
+-
+
-
=
-+
(2)
Similarly you can cal
c
ula
t
e the demand-side tech
nology coo
p
e
ration with
non-
coo
perative populatio
n exp
e
cted to pay
21
u
,
22
u
, and the average expe
cted
payoff .
21
2
1
2
22
2
[(
1
)
]
(
1
)
(
)
u
p
av
c
p
av
c
av
p
a
v
c
p
p
=+
-
+
-
-
=
+
-
(3)
22
2
2
2
(1
)
up
a
v
p
a
v
a
v
=+
-
=
(4)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Novel
QoS Routin
g Algorithm
in Wirele
ss M
e
sh Net
w
orks (Li
u
Hu
i)
1657
21
1
1
2
11
1
1
22
2
(1
)
()
(
1
)
Uq
u
q
u
q
a
v
q
av
c
q
av
av
p
q
c
q
a
v
p
p
=+
-
=
+-
+
-
=
-
+
(5)
Therefore, the network eq
uation
s
for the
dynamic re
plicatio
n can be descri
bed
with
equatio
n (6
).
11
1
1
1
1
11
1
11
21
2
2
2
22
2
2
22
()
(
)
(
1
)(
),
()
(
)
(1
)
(
)
dp
pu
U
p
tv
q
a
v
c
dt
tv
q
p
c
p
a
v
pp
a
v
q
c
dq
qu
U
q
t
v
p
a
v
dt
ct
v
p
q
c
q
a
v
qp
a
v
p
c
p
p
p
p
p
p
ì
ï
ï
=-
=
+
-
-
ï
ï
ï
ï
+
-
=
ï
ï
ï
ï
-
-
ï
ï
í
ï
ï
=-
=
+
-
ï
ï
ï
ï
ï
-
+
-
=
ï
ï
ï
-
-
ï
ï
î
(6)
Equation (6) descri
bed
the grou
p
dynamic,
to solve thi
s
euq
ation
s
. Set
0
dp
dt
=
,
0
dq
dt
=
,five
steady points can be rea
c
he
d, whi
c
h ca
n be listed as follows.
1
(0
,
0
)
E
=
,
2
(0
,
1
)
E
=
,
3
(1
,
0
)
E
=
,
4
(1
,
1
)
E
=
,
21
5
21
(,
)
cc
E
av
a
v
pp
=
.
With
0,
1
pq
££
, when
2
2
1
c
av
p
>
a
nd
1
1
1
c
av
p
>
,
5
E
does n
o
t exist. The Jaco
bi Matrix
can
be ca
cul
a
ted
accordingly.
11
1
22
2
(1
2
)
(
)
(1
)
(1
)
(
1
2
)
(
)
pa
v
q
c
p
p
a
v
J
qq
a
v
q
a
v
p
c
pp
pp
é
ù
--
-
ê
ú
=
ê
ú
--
-
ê
ú
ë
û
(7)
Whe
n
de
t
0
J
¹
, The matrix is non
sing
ular, the
r
e is a uniq
ue
solutio
n
.
(1) In ste
a
d
y
point
1
(0
,
0
)
E
=
,
1
2
0
0
c
J
c
éù
-
êú
=
êú
-
êú
ëû
,
12
de
t
0
,
0
Jc
c
t
r
J
=>
<
,
and the poi
nt
1
(0
,
0
)
E
=
is proven at
steady state.
(2) In th
e st
eady point
2
(0
,
1
)
E
=
,
11
2
0
0
av
c
J
c
p
éù
-
êú
=
êú
êú
ëû
,
12
de
t
0
,
0
Jc
c
t
r
J
=>
<
, and the p
o
in
t
2
(0
,
1
)
E
=
is proven at
steady state.
(3) In the
steady poi
n
t
3
(1
,
0
)
E
=
,
1
22
0
0
c
J
av
c
p
é
ù
ê
ú
=
ê
ú
-
ê
ú
ë
û
,
12
2
de
t
(
)
J
ca
v
c
p
=-
, when
2
2
01
c
av
p
<<
,
3
(1
,
0
)
E
=
is not the steady point
of the system,
whe
n
2
2
1
c
av
p
>
,
3
(1
,
0
)
E
=
is the steady
point of syste
m
.
(4) In th
e p
o
i
nt
4
(1
,
1
)
E
=
,
11
22
0
0
ca
v
J
ca
v
p
p
é
ù
-
ê
ú
=
ê
ú
-
ê
ú
ë
û
, simila
rl
y, when
2
2
0
c
av
p
>
and
1
1
1
c
av
p
<
,
4
(1
,
1
)
E
=
is the stead
y point of
the system. when
2
2
1
c
av
p
>
and
1
1
1
c
av
p
>
then
4
(1
,
1
)
E
=
is
not the stead
y state of the
system.
(5) fo
r the
21
5
21
(,
)
cc
E
av
a
v
pp
=
, o
n
ly when
2
2
0
c
av
p
>
and
1
1
1
c
av
p
<
,
5
E
can
be con
s
ide
r
ed th
e
steady poi
nt.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 2302-4
046
TELKOM
NIKA
Vol. 11, No
. 3, March 20
13 : 1652 – 1
664
1658
Evolutionary stable strategy (ESS) is a
descr
i
p
tion of the evolution of
the game i
n
the
evolution of the con
c
e
p
t of steady
state, dynamic system to de
scri
be the local stability analysis
of the above balan
ce.
(1) When
1
1
1
c
av
p
>
and
2
2
1
c
av
p
>
, this system has four ste
ady points
1
234
,,
,
E
EE
E
. W
h
er
e
1
E
is stabl
e no
d
e
, (coope
rate
, coop
erate
)
a syste
m
of
evolutiona
rily st
able
strate
gy, the other
three eq
uilibri
um points a
r
e unstabl
e. At this
point,
military and civilian technol
ogy, supply-side
and de
mand
-side p
opul
ations th
rou
gh l
ong-te
rm ev
o
l
ution of the popul
ation wi
ll stabilize in
the
(non
-coo
pera
t
ion, non-coo
perative
)
stra
t
egy, which do
es not form a
netwo
rk.
(2)Wh
en
1
1
0
c
av
p
>
and
2
2
1
c
av
p
<
, this system has four ste
ady points
12
3
4
,,
,
E
EE
E
. W
h
er
e
E1 and E4
for the stabl
e node, (co
operate, co
o
perate
)
an
d
(co,
co) i
s
the system
of
evolutiona
rily stable
strate
gy, E2 and E
3
for the
un
st
able n
ode, E
5
for the
sad
d
le poi
nt. At this
point, network techn
o
logy,
network syst
ems and
de
mand
-si
de su
pply-si
de pop
ulation
s
throu
gh
long-te
rm ev
olution of the
populatio
n to E5 acco
rdi
ng to the initial value for the sa
ddle p
o
i
nt
were stable
at different (non-co
ope
rati
on, non-c
o
o
peratio
n) an
d (co
ope
rati
on, coop
eration)
strategy.
Tech
nolo
g
y
sup
p
ly-si
de
and d
e
man
d
-
sid
e
spe
c
ie
s po
pulatio
n
s
un
de
r different
techn
o
logie
s
to get their investment
co
st and te
chn
o
logy ben
efits of te
ch
nolo
g
y, the ratio of
benefit-co
s
t ratio can b
e
descri
bed a
s
technol
ogy,w
hi
ch is u
n
d
e
r certain te
chni
cal b
ene
fits
comm
uni
cat
i
on co
st
s.
Figure 3 Describe
d in the plane S system
techn
o
logy for military and civilian
popul
ations side an
d dem
and-sid
e
dyn
a
mics of ga
m
e
pop
ulation
s
. The sy
stem
has five p
a
rt
ial
equilib
rium p
o
int, the two
unsta
ble e
q
u
ilibriu
m
poin
t
E2 and E3
and E5
con
necte
d into t
he
sad
d
le-point line for the different
states the system
conv
e
r
ge
s to the critical line, that line
the
uppe
r right (E2, E5, E3, E4 part)
system
converg
e
n
c
e
in partnershi
p
, in
the line
of the lower left
(El, E2, E5,
E3 part) the
system will converge
to the co-operati
ve relationshi
p. Because the
system'
s
evol
ution is a long proc
ess, probably in a very long time
to maintain a
coo
perative and
comp
etitive coexisten
c
e.
2
(0
,
1
)
E
4
(1
,
1
)
E
3
(1
,
0
)
E
1
(0
,
0
)
E
21
5
21
(,
)
cc
E
av
a
v
Figure 3. The
evolutionary
model for me
sh net
work g
a
me
4. Work
flo
w
of EGW
R
A
We co
nsi
d
e
r
the dynamic nat
ure of MANET as a game with N players, and
we call
about coalitio
n all non-emp
t
y party S of
N.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Novel
QoS Routin
g Algorithm
in Wirele
ss M
e
sh Net
w
orks (Li
u
Hu
i)
1659
4.1. The Co
mmunicatin
g Step of EG
WRA
Traf
fi
c
fl
owi
n
g in a mesh n
e
twork
can b
e
cla
ssi
fi
ed into three types:
a) traf
fi
c bet
ween two
client
s asso
ciate
d
with the sam
e
WM
R.
b) traf
fi
c bet
ween two
client
s asso
ciate
d
with different
WM
Rs.
c
)
traf
fi
c b
e
tween a cli
ent a
nd a gate
w
ay
. All types of traf
fi
c are deta
iled belo
w
.
Client
s asso
ciated with the
same
WM
R. In this ca
se,
client
s do n
o
t need a
n
y pa
rticula
r
attention, sin
c
e the wl
anI interf
ace aut
omatically b
r
i
dge
s the traf
fi
c b
e
twe
en them. This i
s
an
embed
ded fe
ature of the HostAP mode
of wirele
ss ca
rd drive
r
s.
4.2. Working
Step of EG
WRA
The packa
ge
S may be an element or equal to N.
It is assume
d given on parts of N, a
function
whi
c
h com Acco
rding to the definition
of evolution gam
e
, we can de
sign the working
st
ep
s of
E
G
WRA
a
s
f
o
llo
ws.
/* Step 1
:
Finding Availab
l
e EGWRA-Neighb
or */
TCBT
C-
N(u
)
Φ
;
T
D
(u
)
Φ
;
p(
u)
p0
;
α
=
π
/3
while ( p
(
u
)
<
P and gap
-
α
(TD
(
u))
)
begin
bc
as
t (u, p(u), (Hello, p(u))
)
u rec
e
ives
Ack
(ack
, p(u)) mess
age from v
if v’s game score is m
o
re th
an the thre
sh
old
u cal
c
ulate
s
the dire
ction o
f
discove
red
neigh
bor v dir(u, v),
the tran
smitting po
wer a
n
d
the directio
n deter
mi
ne
s the neigh
bor v(
p(u
)
, dir(u, v) )
TCBT
C-
N(u
)
=
TCBT
C-
N
(
u
)
∪
{ v | disco
vered nei
ghb
or v }
TCBT
C-
D(u
)
=
TCBT
C-
D
(
u
)
∪
{dir(u, v) | discovered n
e
ighb
or v }
Pow(u) =
Pow(u)
∪
{ p(u, v) | discove
re
d neigh
bor v }
p(u
)
Incre
a
s
e(p(u))
end
/* Step 2
:
Finding Availab
l
e DT-Nei
ghb
or */
k is the up
per bound of no
d
e
degree, k = 6
Sort all qualifi
ed neig
hbo
rs
found in Step
1 in orde
r of increa
sing di
stance o
r
po
wer
Pow = { p
1
, p2, p3, ……, pm }
,
p1
≤
p2
≤
p3
≤
……
≤
pm
while (payoff value>th
re
sh
old)
begin
for( i=
1; i
≤
m;
i++ )
begin
u transmits
with power pi, 1
≤
i
≤
m
dra
w
a pe
rpe
ndicular bi
se
ctor betwe
en u
and the nod
e
corre
s
po
ndin
g
to the powe
r
pi
end
end
T
D
T-
N(
u
)
=
TD
T
-
N
(
u
)
∪∈
{ v | v
TCBTC-N(u) a
nd v has corre
s
po
n
d
ing Voronoi
Edge }
T
D
T-
D(
u
)
=
TD
T
-
N
(
u
)
∪∈
{ v | v
TCBTC-N(u) a
nd v has corre
s
po
n
d
ing Voronoi
Edge }
/* Step 3
:
Filling the
α
-gap
*/
Sort TCBTC-N(u
)
an
d TDT
-
N(u) in o
r
d
e
r of increa
sin
g
directio
n
if gap-
α
( TDT-N(u
)
)
then
α
the
smalle
r direct
ion of the gap
β
the larg
e
r
dire
ction of the gap
if dp, dq, dr
∈
TCBT
C-N(u
)
and
dp
≤
dq
≤
dr and dp
=
α
, dr =
β
then d
q
is dro
ppe
d in Step
2 and can fill the
α
-g
ap
TN(u) =
TDT
-
N(u
)
∪
{ v | th
e dire
ction of v is dq }
TD(u) =
TDT
-
D(u
)
∪
{ dir(u,
v) | the directi
on of v is dq }
Pow(u) =
Pow(u)
∪
{ p(u, v) | the directi
on of v is dq }
/* Step 4
:
Edge Removal
*/
Suppo
se no
d
e
v, w
∈
N(u
)
s
e
nd(u, p(u, v), relation(v, w), v)
recv
(u, relatio
n
(v
, w), v
)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 2302-4
046
TELKOM
NIKA
Vol. 11, No
. 3, March 20
13 : 1652 – 1
664
1660
if relation
(v
, w) i
s
“Y” a
n
d
| TN(
u
) |
- 1
≥
3 th
en T
N
(u
)
=T
N(u
)
- {v} P
r
o
c
e
dure
ga
p-
α
(T
D(
u)
)
Suppo
se T
D
(u) = {
d1,d2,d
3
,…,dn} Sort
dire
ction
s
in TD(u) in in
cre
a
sin
g
ord
e
r
TD(u)
=
{dk
1
,
d
k
2
,dk
3
,…,dkn}, dk
1
≤
dk
2
≤
dk3
≤
…
≤
dk
n
,
1
≤
ki
≤
n, 1
≤
i
≤
n
for( i=1; i
≤
n;
i++
)
begin
if dki+1
- dki
≤
2
π
/3 then contin
ue
end
The aim of the genetic o
perato
r
s is to
generate a secon
d
gen
eration po
pul
ation of
solutio
n
s fro
m
those sel
e
cted thro
ugh
geneti
c
ope
rators: cro
s
so
ver, and/or m
u
tation. For e
a
ch
new
sol
u
tion
to be p
r
od
uced, a pai
r of
”pa
r
ent” solut
i
ons i
s
sele
ct
ed for
bre
edi
ng from
the p
ool
sele
cted p
r
ev
iously. By produ
cing a ”child” solu
tion
usin
g the ab
ove method
s of crossove
r an
d
mutation, a new solution
is create
d
whi
c
h typica
ll
y share
s
ma
ny of the chara
c
teri
stics of its
”pa
r
ent
s”. Ne
w parents a
r
e sele
cted fo
r each ch
il
d, and the pro
c
e
ss
contin
u
e
s until a ne
w
popul
ation of solutio
n
s of a
ppro
p
ri
ate si
ze is gen
erate
d
.
4.3. Qos Violation Dete
cti
on
The route bre
a
k ca
n not be detected e
a
sily
. The co
mmon app
ro
ach u
s
ed in most of
existing ad h
o
c ro
uting p
r
otocol
s is by
waiting for a
neighb
or tim
eout, i.e., the
hello messa
ge
from a n
e
igh
bor
doe
s not
arrive to th
e
node
on tim
e
. Whe
n
nei
g
hbor tim
eout
is di
scovere
d
,
a
route e
r
ror m
e
ssag
e is
se
nt to the sou
r
ce notif
ying a
bout the brea
k. Ho
weve
r this ki
nd of ro
ute
brea
k
detecti
on meth
od n
o
rmally ta
ke
s seve
ral
se
co
nds,
whi
c
h i
s
not de
sira
ble
to time se
nsit
ive
QoS flows. In our app
roa
c
h, we utilize
the bandwid
th rese
rvatio
n timeout at the destin
a
tio
n
to
sign
al possibl
e route brea
ks. If the destination fails
to
receive d
a
ta
packets of a
n
active flow 21
before it
s reservation timeout, r
oute recovery
will be
triggered at
the destination. Usi
ng
this
method, we
can dete
c
t b
o
th types of QoS vi
olations at the same time an
d handle the
m
identically. The neig
hbo
r l
o
st dete
c
tion
will also
be
use
d
in case
the destin
a
tion that initiated
instant
re
cov
e
ry can n
o
t reach t
he
so
u
r
ce
be
ca
use
of netwo
rk pa
rtition or pa
cket loss. Whe
n
a
node dete
c
ts
that the down
link nod
e of a rese
rved r
o
u
t
e is lost, it will send a ro
ute erro
r pa
cke
t
,
with its curre
n
t route
seq
uen
ce nu
mb
er, to the co
rre
sp
ondi
ng
uplin
k nod
e. The route e
r
ror
packet is then forwa
r
ded
upward
s
to the sou
r
ce to
indicate the occurren
ce of the route bre
a
k.
As a consequence, the reserved
bandwi
dth of the flow will be re
leased at the forwardi
ng nodes.
4.3. QoS Violation Recov
e
r
y
To provide in
stant route a
dapt
ivity, we
use de
stinati
on initiated route recovery
. After
the QoS violations a
r
e de
tected, the d
e
stinatio
n will
incre
a
se its route sequ
en
ce num
be
r and
broa
dcast an unsolicite
d
ro
ute reply packet, also
call
e
d
route updat
e packet, back to the source.
The route u
pdate p
a
cket
is tre
a
ted i
n
the sa
me
manne
r a
s
a route
re
qu
est pa
cket
with
admissio
n
co
ntrol an
d loo
p
preve
n
tion
mech
ani
sm,
but in the reverse di
re
ction
from de
stina
t
ion
to source. Upon re
ceivin
g the first in time
route update p
a
cket with appropriate
seq
u
ence
numbe
r, the
sou
r
ce switches the flo
w
in que
st
ion
to the reverse route o
n
which the
upd
ate
arriv
e
s.
On the othe
r
hand, a late
route upd
ate
packet
o
r
a route erro
r pa
cket, with the
valid
route sequ
en
ce, sign
als th
e occurren
ce
of QoS
violation and the
failure of route recovery. In
su
ch case, the appli
c
ation
can eith
er d
e
c
ide to
contin
ue tran
smittin
g
the flow wit
h
the ab
sen
c
e
of QoS guara
n
tee or
suspe
nd the flow a
nd try later.
5. Simulation Experimen
t
of EGWRA
5.1. The Desi
gn of Simulation Experim
e
nt
We u
s
e the
simulation software mad
e
b
y
a
Ns 2.30
pairs rel
a
y co
operation st
rategy
to con
s
truct t
he expe
rime
nt. The topol
ogy of ex
pe
ri
ment WNN can be
de
scri
bed li
ke figu
re 4,
whi
c
h divided
into two part
s
, mesh network a
nd un
de
rlay netwo
rk.
The MA
C lay
e
r of
each m
obile node
or
AP model uses the default IEEE 802.11 DCF
module
provided by
OPNET with mo
di
fication
s
for
multihop
con
nectio
n
supp
ort. Each
mo
bile
node or AP has the same
transmissio
n rang
e of
200 meters and
raw data rate
of 2Mb/s. Th
e
WM
R mod
u
le
is in
sert
ed o
n
top of the M
A
C layer m
o
d
u
le. In the WMR mo
dule,
we impl
eme
n
t
ed
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Novel
QoS Routin
g Algorithm
in Wirele
ss M
e
sh Net
w
orks (Li
u
Hu
i)
1661
a send
er buff
e
r of 64 packets to buffer data packet
s
waiting for a route, e.g., packets for wh
ich
route di
scove
r
y has sta
r
te
d, but no repl
y has ar
rived
yet. Upon discovery failu
re
, the source
will
back-off for
a dou
ble tim
e
interval a
n
d
try agai
n
until the retry limit of 3 is rea
c
hed. T
h
e
maximum pa
cket size u
s
e
d
in the tempora
r
y
band
wi
dth reservati
on is set to 1024 Bytes. The
hello interval
of 1 second i
s
used with n
e
ighb
or timeo
u
t set to 3 se
con
d
s.
All the node
s are
stationa
ry during th
e
simu
latio
n
pe
riod of 3
00
seco
nd
s. The
r
e are
10 external flows with the sou
r
ce node
s sp
re
ad ran
d
o
mly among the 40 node
s. Since we are
intere
sted in the perfo
rma
n
c
e of WM
R in
a mesh
ne
two
r
k
,
th
e
c
o
nne
c
t
io
ns
to
th
e w
i
r
e
d
ne
tw
ork
are n
o
t inclu
ded in the
si
mulation. Th
e traffic sin
k
module in t
he AP model
will re
ceive
the
external data
packet
s
as if they are forwa
r
ded to the wi
red net
wo
rk.
To simul
a
te stream me
dia
appli
c
ation
s
we
u
s
e CBR (co
n
sta
n
t
bit-rate)
flo
w
s wi
th
10
packet
s
per seco
nd and fixed data pa
cket size of
102
4 bytes. The sou
r
ces
will keep gen
erating
data pa
ckets
at the fixed ra
te throug
hout
the si
mul
a
tio
n
. All the flows have th
e sa
me end
-to-en
d
delay bou
nd.
Figure 4.The
topology of e
x
perime
n
t network
Simulation is using
ro
utin
g EGWRA. Collabo
ration
d
e
fined level i
ndex nod
e fo
r the
netwo
rk in a
n
y
of NARMea
ning the a
c
tu
al relay nod
e
s
and the n
u
m
ber of pa
ckets followi
ng
th
e
packet in
whi
c
h the
numb
e
r of
requi
re
dVolume
rati
o. NAR
∈
[0,
1
], the great
er the valu
e
of a
node'
s
NAR.
HelpT
he
high
er the
deg
re
e
of coll
abo
rati
on involved i
n
rel
a
y.Simulation sce
n
a
r
i
o
is
set as follows. number of node
s N for a
50, each sect
ion Point spe
ed in the 10
~ 20m/ s were
evenly distrib
u
ted within the mobile ra
nge
Wai as a
500m × 50
0m, each no
de in a session
immediately after the end ofInitiate another sess
ion
carved, and th
e duration of each se
ssi
on
5-
10swere
unif
o
rmly di
stributed withi
n
. These
setti
ngs will ensure
th
at each node in t
h
e
netwo
rkInte
ra
ction b
e
twe
e
n
the hig
her t
he proba
bilit
y of the relay.
Each
se
ssi
on
is a fixed-sp
eed
transmissio
n
R
ate of 1 Mb / s data stream, pa
ck
e
t
size is 5
1
2by
t
e
.
E
a
ch
simulat
i
on t
r
ue
contin
uou
s 9
00s, 50 time
s in the implement
ation of the statisti
cal
averag
e dem
and.
5.2. Result
Analy
s
is of Simulation
Accrodi
ng to
the pa
ramete
r we sta
r
t the
experim
ent and
the sim
u
l
a
tion
re
sult can
be
descri
bed li
ke figure
4. At fixed interva
l
s of
time, t = 1-75, move
ment occu
rs
by updating t
h
e
spe
ed an
d d
i
rectio
n of each M
N
.We have ch
oo
se
the tuning para
m
eter u
s
ed to vary
the
rand
omn
e
ss.
the speed a
nd dire
ction
are cho
s
en from a ran
d
o
m
Gaussia
n
distrib
u
tion with
mean eq
ual to zero and
standa
rd deviat
i
on equal to
o
ne. For a ra
n
dom ch
osen i
n
stant t in total
simulatio
n
time, our pro
p
o
se
d approa
ch extra
c
t 21 MNs as a st
able co
re in term
s of mobility
whi
c
h is ma
rked with a re
a
d
dot.
Evaluation Warning : The document was created with Spire.PDF for Python.