TELKOM
NIKA
, Vol. 11, No. 12, Decem
ber 20
13, pp.
7780
~77
8
6
e-ISSN: 2087
-278X
7780
Re
cei
v
ed
Jun
e
28, 2013; Revi
sed Aug
u
st
26, 2013; Accepted Sept
em
ber 7, 201
3
A Novel Adaptive Congestion Avoidance Protocol for
Wireless Sensor Networks
Mei-Wen Hu
ang
1
, Hsu
-
Ju
ng Liu*
2
, Yu-Chang Chen
3
, Wen-Sh
y
o
ng Hsieh
4
1
Dept. of CSE,
NSYSU & Dept
.MIS,
T
a
jen Univ.,
T
a
iw
an
2
Dept. DMD,
Taje
n Univ
ersit
y
,
T
a
i
w
a
n
3,4
Dept. CSIE,
Shu-T
e
Universit
y
, T
a
iw
an
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: m
w
h
u
a
ng@
mail.taj
en.ed
u.tw
l, h
jli
u@ma
il.taje
n.ed
u.t
w
*
A
b
st
r
a
ct
Most rese
arch
es o
n
w
i
rel
e
ss
sensor
netw
o
rk
s (W
SNs) are
focuse
d o
n
how
to sav
e
th
e e
n
e
rgy
of
the se
nsor
nod
es. How
e
ver, i
n
so
me
a
ppl
ica
t
ions of W
S
Ns,
such
as i
n
the
mo
nitori
ng
of a
n
eart
hqu
ake
o
r
a forest fire, it
is
m
u
ch
m
o
re im
po
rtant to
tra
n
smit e
m
er
ge
n
cy data
pack
e
ts to the s
i
nk
n
ode
as s
o
o
n
a
s
possi
ble t
han
to save
pow
er
. T
h
is pa
per
p
r
opos
es a
nov
el Ad
aptiv
e C
ong
estio
n
Avoi
danc
e Protoc
o
l
(ACAP) mo
del
to provide a feasi
b
le
W
S
Ns architectur
e
that w
ill save
the energy of the
sensor n
odes
i
n
the n
o
r
m
al
situ
ation,
but w
i
l
l
t
r
ans
mit
e
m
erg
ency
data
p
a
c
k
ets in
a
n
efficient
man
ner to
the s
i
nk
n
ode
.
T
he si
mul
a
tio
n
analys
is sho
w
s that the ACAP prov
i
des
super
ior p
e
rfor
ma
nce d
u
ri
ng
both n
o
rmal a
n
d
emerg
ency co
nditi
ons.
Ke
y
w
ords
:
w
i
reless se
nsor n
e
tw
orks (W
SN
s), pow
er savin
g
, emerge
ncy traffic
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 Se
n
s
or Netwo
r
ks (WS
Ns) h
a
ve bee
n impl
e
m
ented i
n
m
any appli
c
ati
ons
and
are th
e subj
ect of mu
ch
resea
r
ch recently. T
he m
a
jor
re
stri
ctio
ns of th
e WSNs
are
limited
energy su
ppl
y, limited computing p
o
wer, limit
ed bu
ffer size, and
limited band
width [1]. Mo
st
resea
r
ch on
wirel
e
ss
sen
s
ing networks
(WS
Ns) is fo
cu
sed o
n
ho
w to save th
e ene
rgy of the
sen
s
o
r
node
s (SNs). Howe
ver, in
so
me
appli
c
ation
s
o
f
WSNs, e
s
p
e
cially
duri
n
g
an
eme
r
ge
n
c
y
situation, deli
v
ering data p
a
ckets
to
the
sin
k
n
ode
as
soo
n
a
s
p
o
ssible i
s
mu
ch
more
impo
rta
n
t
than savin
g
p
o
we
r.
Comm
uni
cati
on in the
WSNs
occu
rs i
n
different
ways de
pen
din
g
on the
und
erlying
appli
c
ation
or mission
of th
e net
work. A
s
sho
w
n
in Fi
gure
1,
we
d
e
fined
node
s
A, B and
C
a
s
uplin
k no
de
s
of node
D,
si
nce th
ey are farthe
r
a
w
ay from the
sin
k
node th
an n
o
de D, a
nd n
o
de
D i
s
d
enote
d
as th
e d
o
wnlink
nod
e of
n
ode
s A,
B, an
d C. T
he
dat
a pa
ckets pro
duced
by a S
N
can
be
cl
assified into th
ree d
e
livery t
y
pes:
clo
c
k-d
r
iven, event
-driven
and
q
uery-driven
[2].
Clo
c
k-d
r
iven data
pa
ckets are
gat
hered
by the SN, and are
peri
o
d
i
cally se
nt to the sin
k
nod
e.
In an e
m
erg
ency
situatio
n, wh
en the
sen
s
in
g dat
a
is ove
r
the
pre
s
et th
re
sh
old value, th
e
sen
s
in
g nod
e
send
s the ev
ent-d
riven da
ta packets
to
the sin
k
nod
e
as soon a
s
possibl
e. SNs
may
se
nd qu
ery-d
r
iven da
ta
pa
ckets b
a
ck
to
th
e
si
nk. Event-dri
ven data
pa
ckets an
d q
u
e
r
y-
driven
pa
ckets h
a
ve hi
ghe
r prio
ritie
s
to
b
e
de
live
r
ed
th
an the
cl
ock-driven
data.
For th
e q
uery
-
driven
data p
a
ckets the q
u
e
rying m
ode
of the si
nk is
set a
s
a
n
em
erge
ncy
situa
t
ion that mu
st
be re
sp
ond
e
d
to as
so
on
as p
o
ssibl
e
, and thu
s
th
e que
ry-d
rive
n data i
s
a
s
signed
as
eve
n
t-
driven data.
Senso
r
net
works u
s
u
a
lly operate un
de
r light
load a
nd be
com
e
a
c
tive in re
sp
onse to
the monito
re
d event. The
ene
rgy con
s
traint an
d lo
w buffe
r si
ze
are th
e two
most imp
o
rta
n
t
probl
em
s in SNs. It is hard to determin
e
the sen
s
ing
interval (
τ
) re
quire
d for a SN to sen
s
e a
n
d
transmit clo
c
k-d
r
iven d
a
ta
packet
s
to the sin
k
no
de.
This i
s
due to
the fact that if the value of
τ
is too sh
ort, the SN will p
r
odu
ce a
nd send dat
a p
a
ckets to the si
nk no
de mo
re freque
ntly.
Con
s
e
quently
, the e
nergy o
f
the SN
will
be
con
s
u
m
ed
more
quickly
. On th
e oth
e
r han
d, the
SN
can n
o
t se
nd
the real time
status to the
sink
nod
e if the se
nsi
ng in
terval is too l
ong. The oth
e
r
importa
nt pro
b
lem in th
e d
e
sig
n
of the
WSNs rout
in
g protocol is t
he lo
w buffe
r
size. In ge
neral,
SNs
will n
o
t sen
d
la
rge
a
m
ount of d
a
ta pa
ckets
to
the sin
k
node
duri
ng n
o
rm
al situatio
n. F
o
r
example, in the monitoring of forest fires,
the SN
will send periodica
lly (clock-driven) data
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
A Novel Ad
ap
tive Co
nge
sti
on Avoid
a
n
c
e
Protocol fo
r Wirel
e
ss Sen
s
or… (M
ei-Wen Hu
ang
)
7781
packet
s
to th
e sin
k
n
ode,
contai
ning i
n
formatio
n re
g
a
rdin
g lo
catio
n
, humidity a
nd tempe
r
atu
r
e.
In the event-driven mo
del,
when the
se
nse
d
tempe
r
a
t
ure is hig
h
e
r
than the thre
shol
d value set
in the SN, th
e trigg
e
re
d S
N
will
delive
r
a large a
m
ou
nt of eme
r
ge
ncy pa
ckets to the
sin
k
no
de,
whi
c
h m
a
y cause
con
g
e
s
tion. In thi
s
situation,
the
e
m
erg
e
n
c
y pa
ckets th
at a
r
e far away from
the si
nk nod
e
may not
be
delivere
d
su
cce
ssfully,
sin
c
e th
ey have
to com
pete
with clo
c
k-d
r
iv
en
data packet
s
gene
rated in
the downli
n
k
path. Con
s
eq
uently, in real time, the sink node m
a
y be
unabl
e to respond to the e
n
tire WS
Ns
si
tuation.
The remain
d
e
r of thi
s
pap
er is organi
zed a
s
follo
ws. Section 2
discusse
s the
relevant
literature and
se
ction 3 d
e
t
ails the alg
o
r
ithm
of the p
r
opo
se
d ACA
P
for the WS
Ns. Se
ction
4
explain
s
the perform
an
ce simulatio
n
of the
ACAP, and finally in section
5 we dra
w
our
con
c
lu
sio
n
s.
2. Relate
d Litera
ture
Many rese
arche
s
have p
r
opo
sed to
sol
v
e
the ab
ove
de
scribe
d p
r
oblem
s. Th
e
buffer-
based co
nge
stion avoida
n
c
e algo
rithm
[3] is an e
fficient sch
eme to prevent the
occu
rren
ce
of
con
g
e
s
tion in
the WS
Ns.
As shown in
Figure 1,
the downs
tream traffic
from
SNs
toward t
h
e
sin
k
is many
-to-one m
u
lti-hop converg
ent. Due
to the co
nverg
e
n
t nature of the do
wn
stre
am
traffic, co
nge
stion i
s
mo
re proba
ble t
o
occu
r in t
he do
wn
stre
am directio
n. The
key fo
r
con
g
e
s
tion a
v
oidan
ce in
[3] is to
ma
ke
su
re th
at th
e uplin
k
se
nsor n
ode
s
(USN
i
)
s
e
n
d
data
packet
s
to their do
wnlin
k
sen
s
o
r
nod
e
D only wh
en
node
D ha
s the buffer
spa
c
e to hold th
e
packet. Th
erefore, in
ca
se
of cong
e
s
tion, the
u
p
link nod
es
(US
N
i
) sh
oul
d
re
du
ce da
ta
forwa
r
di
ng to their do
wnlin
k node (n
ode
D).
Figure 1
.
Buffer-ba
s
ed Con
gestion Avoid
ance Detectio
n in [3
]
The
hybrid cl
uster (HC) WSNs
mod
e
l
[5
], based on the clu
s
ter-ba
sed
WSNs scheme,
provide
s
diff
erent p
r
io
ritie
s
of data
pa
ckets
a
nd d
e
livers th
e h
i
gher
pri
o
rity packet
s
in
a
con
g
e
s
ted
co
ndition. However, the
HC WSNs a
dop
ted fixed-tim
e
interval fo
r rep
o
rting
da
ta
packet
s
can
not prevent b
u
ffer cong
esti
on well in
th
e SN
whe
n
a
burst of em
erge
ncy eve
n
t
-
driven traffic
occurs. In the ev
ent-to-sin
k
reliable tra
n
sport (ESRT
)
[6], a sensor
node pla
c
e
s
a
con
g
e
s
tion-n
o
tification (CN) bit in the
packet he
ad
er when its
buffer is n
e
a
r
ly full. The sin
k
node
re
cal
c
ul
ates a
n
y ne
w rep
o
rting
dat
a pe
riodi
cally
and
will reco
gnize the S
N
by its CN bit.
Con
g
e
s
tion
and avoid
a
n
c
e in SNs (CO
D
A)
[
7
] provide
s
an ope
n-l
oop ho
p-by-hop
backp
re
ssure
mechani
sm,
and a
clo
s
e
-
loop
multis
o
u
rce regul
ation me
cha
n
ism to solve t
he
c
o
ng
es
tio
n
.
Xu
a
n
d
C
h
r
i
s
t
os
[8
] pr
op
os
ed
a
d
y
nami
c
slee
p-time
co
ntrol in
event
-driven
WS
Ns,
but did not provide different priority traffic se
rvices. T
he Adaptive Samp
ling Protocol (ASP) [9]
prop
ose a
scheme
whi
c
h
can
dyna
mically eliminat
e
the redu
ndan
cy an
d e
s
tim
a
te the
defici
ent
data ba
sed
o
n
learned
rel
a
tions in
a way to
ensu
r
e
low an
d bal
anced en
erg
y
consumptio
n.
Alireza, etc.
[10] propo
se
an
aware a
nd p
uni
shme
nt ba
sed
co
operative ad
aptive sampli
n
g
techni
que
to
sati
sfy both
network life
-
time a
nd
da
ta quality re
quire
ment
s.
The S
R
PL [
11]
pre
s
ent
a
sy
nch
r
on
ou
s
rich p
r
eam
ble li
stenin
g
protocol
for WS
Ns d
eploye
d
in
the
agri
c
ultu
re
can
opy to
re
duce the
po
wer con
s
um
ption. Hii
Pei
-
Ch
eng,
etc.
[12] pro
p
o
s
e
d
an i
n
tegrated
mobile
h
ealt
h
ca
re
m
onito
ring system combi
n
ing WSN
and CDMA
(Co
de Division
Multip
le
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 12, Dece
mb
er 201
3: 778
0 – 7786
7782
Access) technology which achiev
e the key drivers
of mobility,
fl
exibility, conveniency, and
indep
ende
ncy.
In this paper, we p
r
o
p
o
s
e a
Hybrid
Flex
ible Con
gestio
n
Avoid
ance Proto
c
o
l
, called
the ACAP, which p
r
ovide
s
a hybrid p
r
i
o
rity fo
r pa
cket delivery a
nd uses
a flexible se
nsi
n
g
interval acco
rding to the bu
ffer length of the SNs
in order to prol
on
g the lifetime
of the SNs an
d
to transmit ev
ent-d
riven dat
a packet
s
to the sin
k
no
de
as soon a
s
p
o
ssible.
Figure 2.
(a) Clock-driven model and (b) Hybrid model
3. Adap
tiv
e
Cong
estion
Av
oidance Protoc
ol (ACAP) fo
r WSNs
In WSNs
, a SN may transmit three types
of
d
a
ta to t
he
sink no
de;
clo
c
k-d
r
iven,
event-
driven an
d q
uery-driven d
a
ta packet
s
. Gene
rally
, the sen
s
o
r
no
des g
a
ther t
he clo
c
k-d
r
iven
data packet
s
and se
nd the
m
periodi
call
y to
the sink
node, as
sho
w
n in Figu
re
2(a
)
. In order to
prolo
ng the lif
etime of the SNs in th
e ACAP, as
sho
w
n in the Fi
g
u
re 2
(
b
)
, the event-d
riven
data
packet
s
may
be p
r
odu
ce
d
or n
o
t dep
en
ding o
n
the
sensi
ng result in ea
ch time i
n
terval
τ
. As
we
kno
w
, the
e
nergy to
tra
n
s
mit a
data
packet i
s
mu
ch
more tha
n
the
se
nsi
n
g en
ergy. F
o
r
example, the
value of
α
is 5 in Figure 2
(
b),
which m
ean
s
that the
r
e are on
e cl
ock-d
r
iven d
a
t
a
packet and
α
times event-driven
data
p
a
ckets bei
ng
se
nsed i
n
a
ro
und. A
clo
c
k-d
r
iven
dat
a
packet
will
b
e
p
r
odu
ce
d o
n
ce
in
ea
ch
round,
and
th
e event
-drive
n data
pa
cke
t
s will
only
b
e
sent when th
e pre
s
et thre
shol
d in the
SN is
trig
gered. It is obvious that the
SNs
can
sav
e
energy in a non-eve
n
t con
d
ition.
3.1. Cons
tru
c
tion Phas
e
We
con
s
id
er
the ca
se
of multipath ro
utin
g to
the si
nk i
n
the ACAP. A sen
s
o
r
, SNx, has
to build the li
sts of its d
o
wnlink n
e
igh
b
o
r
nod
es
a
nd
uplin
k its nei
ghbo
r no
de
s. Dx is the list
of
the do
wnli
nk
neigh
bor no
d
e
s. S
N
x can
use
the
s
e
no
des to fo
rward pa
ckets to
the si
nk no
de
,
and Ux is the
list of the uplink nei
ghb
or
node
s. SN
x may be use
d
by the node
s in Ux to be the
packet fo
rwarding n
ode. T
here
are fou
r
fields in
clu
d
e
d
in a b
e
a
c
on
messa
ge, in
cludi
ng b
e
a
c
on
numbe
r
(bn
)
, bea
con
stat
us
(b
s),
sen
s
or n
ode i
d
(Sn), and
ho
p
co
unter (h
c). The b
e
a
c
o
n
numbe
r i
s
i
n
i
t
iated by the
sin
k
nod
e
which
to m
a
ke
out e
a
ch
of the b
e
a
c
on
messag
e. Th
e
bea
con
statu
s
h
a
s t
w
o
di
fferent ph
ase
s
: “B”
indi
cat
e
s th
e be
gin
n
ing of
a n
e
w
b
eacon
an
d
inform
s the
receivin
g n
o
d
e
s to
o
perate
the
build
in
g
pro
c
e
s
s of th
e list
s
of the
neigh
bors.
“E
”
indicates the
ending
of a
bea
con a
nd i
n
form
s the receivin
g nod
es to up
date
the lists of t
h
e
neigh
bors. T
he ho
p count
er i
s
u
s
ed to
indicat
ed that
the bea
co
n
messag
e is forwarding fro
m
the downli
n
k nodes o
r
backwarding
from the uplin
k node
s. A
forwa
r
di
ng bea
con
is
reb
r
oa
dcaste
d with th
e
sensor
nod
e i
d
of
the
re
broad
ca
sting n
ode a
nd th
e
hop
co
unte
r
is
increa
sed by
one. The backwardi
ng b
eacon will
b
e
terminated.
The sen
s
o
r
node id in the
forwa
r
di
ng b
e
a
co
n is put in
to the list
of Dx, and
th
e sensor nod
e
id
in
the ba
ckwardin
g
b
e
a
c
o
n
will be put into the list of Ux. The beaco
n
messag
e is periodi
cally b
r
oad
ca
sted b
y
the sink no
de
to help the se
nso
r
nod
e to maintain the l
i
st of neighb
o
r
s, Ux an
d Dx
.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
A Novel Ad
ap
tive Co
nge
sti
on Avoid
a
n
c
e
Protocol fo
r Wirel
e
ss Sen
s
or… (M
ei-Wen Hu
ang
)
7783
3.2. Conge
stion Av
oidan
ce Phase
In orde
r to prevent a SN from being
con
ges
te
d by a burst of eme
r
g
ency data p
a
ckets,
the co
nge
stio
n dete
c
tion
a
nd avoid
a
n
c
e
schem
e
a
r
e
involved in
each SN. T
h
e value
of the
buffer len
g
th
is u
s
ed to d
e
tect if a SN m
a
y become
conge
sted d
u
e
to an exce
ssive amount of
incomi
ng dat
a
pa
ckets. While
the dete
c
ted buffer
l
e
ngth b
k
rea
c
h to the val
u
e of Bmax a
s
shown i
n
the Figure 3, the SN
will
send out
an i
m
plicit ACK
[4], including a congestion
notification (CN), to the uplink SNs in
orde
r
to stop
the downlin
k SNs tra
n
smitting data pa
cket
contin
uou
sly in the next time interval. Th
e bk e
s
timati
on is b
k
= b
k-1
+
ᇞ
b
k
, where
ᇞ
b
k
=(
b
k-1
- b
k-2
).
This mea
n
s
that the p
r
e
d
i
cted val
ue
o
f
bk i
s
the
sum of th
e p
r
ese
n
t buffe
r
size a
nd the
incr
ea
se in b
u
f
f
e
r si
ze in
the previo
us ti
me interval.
3.3. Sensing-interv
al Adjustmen
t
-pha
se
In the co
nge
stion avoi
dan
ce p
h
a
s
e, th
e se
ndi
n
g
of
the do
wn
stre
am data p
a
ckets
will
be pa
used
whe
n
the
bu
ffer length
is pre
d
icte
d to
be
rea
c
hi
ng
the value
o
f
Bmax. This
avoidan
ce
scheme p
r
ote
c
ts a se
nsor n
ode from ov
e
r
flowin
g, but it will also inf
o
rm the upli
n
k
node
s to
blo
c
k tra
n
smi
s
si
on m
e
ssag
e. Wh
en th
e
b
u
ffer le
ngth i
s
b
e
twe
en th
e value
of B
m
in
and Bmax, a
s
sho
w
n in Fi
gure
3, the sensi
ng inte
rval (
τ
) i
s
ad
ap
tive between
τ
min and
τ
ma
x
according to
the mea
s
ure
d
buffer len
g
t
h b. Longe
r sen
s
ing i
n
te
rval provid
e
a lowe
r po
wer
con
s
um
ption.
In the sen
s
in
g interval
adj
ustment
pha
se, the sen
s
in
g interval
(
τ
) will
be adjust
e
d
according
to
the buffer le
ngth b
e
fore
the b
u
ffer ov
erflows. T
he
value of
τ
wil
l
be in
crea
se
d
whe
n
the b
u
ffer len
g
th is
p
r
edi
cted to in
cre
a
se an
d will be de
crea
sed when th
e
buffer len
g
th i
s
decrea
s
in
g, as sh
own in Fi
gure 3. Figu
re 4 sh
o
w
s th
e operation flow chart of the con
g
e
s
tio
n
avoidan
ce
p
hase an
d fle
x
ible se
nsi
n
g
peri
od
adju
s
tment pha
se
in a
se
nsor node. S
e
n
s
i
ng
interval (
τ
) wi
ll be cha
nge
d
in each
rou
n
d
as p
e
r the
followin
g
rule
according to
the predi
cte
d
length of the buffer (b
).
Ca
se 1.
mi
n
b
B
: The SN will notify its upstream SN
s that the sensing interval (
τ
) is
τ
min
.
The S
N
will
transmit
both
o
f
the cl
ock-dri
v
en dat
a
pa
ckets an
d eve
n
t-drive
n
data
pa
cket towa
rd
the downlin
k node
s.
Ca
se 2.
max
min
B
b
B
: The SN
will bl
ock to transmit
cl
ock-driven da
ta packets
and
the event-driven me
ssa
ge
sen
s
ing
interval
(
τ
) is p
r
olon
g to
τ
min
+
ᇞ
τ
, w
h
er
e
ᇞ
τ
=
)
(
*
min
max
min
max
min
B
B
B
b
Ca
se 3.
b
B
max
: Wh
en the b
u
ffer length
rea
c
h
e
s, o
r
is more than B
max
, the SN
will
send a congestion notification
(CN) to its upstream SNs, and
the upstream SNs will stop
prod
uci
ng dat
a packet
s
.
F
i
g
u
r
e
3
.
Sensing Interval (
τ
) Adjustment in
the Sensor Node
F
i
g
u
r
e
4
.
Operation Flow Chart in a Sens
or
Node
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 12, Dece
mb
er 201
3: 778
0 – 7786
7784
Table 1. Simulation Para
meters
Value
Net
w
orks size
100m*100m
Num
b
er o
f
sen
s
i
ng
node
s(N
sen
sin
g
)
100
Initial
e
n
erg
y
of the
sensor no
de
2J
Buf
f
er size (B
) i
n
SN
22
packets
Processin
g
del
a
y
50
μ
s
Loca
t
io
n of Si
n
k
(50,50)
Data p
acket
size
512
b
y
t
e
s
Trans
m
issio
n
r
a
te (R
SN
)
512
kbps
Processin
g
energ
y
(
E
am
p
+E
el
ec
)
50 nJ/bit
Energ
y
o
f
se
nsi
ng
(E
sen
sin
g
)
0.05 nJ/bit
En
e
r
g
y
of
da
t
a
aggre
g
ati
on (E
da
)
5 nJ/bit/signal
B
min
10
packets
B
max
20
packets
Minim
un se
nsin
g int
e
r
v
al
(
τ
)
0.1 sec
Maximu
m se
nsi
ng in
ter
v
al
(
τ
)
0.5 sec
α
5
4. Performan
ce Analy
s
is
4.1. Simulation En
v
i
ronment
For evalu
a
tin
g
the perfo
rmance of the
pr
opo
se
d ACAP, the NS2 simulato
r [14] wa
s
applie
d to the environm
ent with t
he para
m
eters listed
as follo
ws.
4.2. Simulation Analy
s
is
The p
a
cket d
e
livery ratio,
accumul
a
ted
numbe
r of d
r
oppe
d pa
cket
s, and
the n
u
mber
of
alive no
de
s
are th
e
key f
a
ctors in eva
l
uating
th
e p
e
rform
a
n
c
e
o
f
the WS
Ns.
A lowe
r
pa
cket
delivery ratio
or a higher accumulatio
n
of
droppe
d
packets me
an that more
packets were
blocke
d or d
r
opp
ed in th
e forwardi
ng
path. In ord
e
r to make
sure that the
sink n
ode
can
maintain the
actual
state of all of the sen
s
o
r
netwo
rks, the se
nsor nod
es m
u
st sen
d
more
packet
s
to
th
e si
nk n
ode.
In the
follo
wing
simu
latio
n
s,
we
will
measure the
pa
cket d
e
livery
ratio, the
a
c
cumulated
d
r
o
pped
pa
ckets and
the
num
ber of alive
n
ode
s in
the
p
r
opo
se
d A
C
AP
scheme. Th
e simulatio
n
result
s will
be comp
ared with the
perform
an
ce of the CODA
backp
re
ssure
sch
eme [7] a
nd the no-
co
n
gestio
n
co
ntrol (NCC).
Figur
e 5 sh
ow
s the pa
c
k
et deliv
ery
ratio in ACA
P
, NCC, an
d CO
DA un
der the
environ
ment
listed in
Tabl
e 1 for the d
i
fferent
rate
s of event pa
ckets. Th
e si
mulation
re
sult
sho
w
s that
th
e delive
r
y ratio of th
e eve
n
t
-driven
pa
ckets in
the
ACAP is
better then th
at of th
e
CO
DA an
d
the NCC be
cau
s
e
it sa
crifices
th
e
c
l
oc
k-
dr
ive
n
p
a
c
k
e
ts
un
de
r
c
o
ng
es
tion
con
d
ition
s
. In addition, the
packet deliv
ery ratio in th
e NCC d
e
cre
a
se
s d
r
amati
c
ally wh
en th
e
event-data
ra
te is over 30
%.
In an em
erge
ncy situ
ation,
the event-dri
v
en
data pa
ckets will be
b
l
ocked or dro
pped
from the
forwarding
path
if no
con
g
e
s
tion
co
ntrol
schem
e i
s
i
n
volved. The
ba
ckpre
s
su
re
scheme in CODA or the
congestion avoidance
in ACAP will prevent the intermediate nodes
from a
buffe
r
overflow,
but
the eme
r
g
e
n
c
y event-driv
en d
a
ta p
a
ckets to th
e n
o
des will
al
so
be
discarded. T
he se
nsi
ng i
n
terval adj
usting sc
hem
e in the ACAP can
solve these proble
m
s.
Figure 6
sho
w
s the
accu
m
u
lated
dro
ppe
d pa
ckets
in t
he ACAP, th
e
NCC,
and
th
e CODA
whe
n
the event-d
ata rate i
s
50%
. The sim
u
lation re
sult
sho
w
s th
at the ACAP dro
p
s
substa
ntially le
ss
event-d
riven data
pa
cket
s
and cl
ock-dri
v
en data pa
ckets t
han the
NCC and
CODA sch
e
me
s.
Figure 7 sho
w
s th
e num
b
e
r of no
de
s whi
c
h k
eep
a
live in the si
mulating p
r
o
c
ess with
50 pe
rcent o
f
the SNs
ra
ndomly repo
rting ev
ent-d
ri
ven data pa
ckets. T
he A
C
AP ca
n ke
ep
more no
de
s alive beca
u
se it is able to redu
ce the
generating of ev
ent-driv
en data pa
cket
s
whe
n
the do
wnlin
k no
des
are in cong
estion.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
A Novel Ad
ap
tive Co
nge
sti
on Avoid
a
n
c
e
Protocol fo
r Wirel
e
ss Sen
s
or… (M
ei-Wen Hu
ang
)
7785
F
i
g
u
r
e
5. The Packet Delivery Ratios for
Different Event-data Rates
Figure 6
.
Accumulated Number of Packets
Lost in a 50% Event-data
Rate
To p
r
olo
ng th
e life
of the
WSNs un
der no
rmal
(n
on-event)
co
ndition
s i
s
one
of th
e ma
i
n
goal
s of th
e
ACAP mod
e
l. It is
ea
sy to
prolo
ng
th
e lif
e of
WSNs
b
y
extending
the valu
e of
α
in
the ACAP model. The trad
eoff is that th
e system
can
not resp
ond
to the sensi
n
g information
in
real time if the value of
α
is large.
Figure 8 sh
ows the si
mulation result
that co
mpares the
accumul
a
te
d ene
rgy
con
s
um
ption of
all
sen
s
o
r
node
s unde
r norm
a
l condit
i
ons. With
th
e
value of
α
s
e
t as
3, the
ACAP ca
n re
duce the e
n
e
r
gy consumpt
ion an
d ex
te
nd the life ti
me of the S
N
s to
a g
r
eat
er
extent than the CO
DA ba
ck
pre
s
su
re a
nd NCC
sche
mes.
F
i
g
u
r
e
7. Comparison of the Number of Nodes
kept Alive
in
the WSNs
Figure 8
.
Comparison of the Accumulated
Energy Consumption under Normal
Conditions
5. Conclusio
n
Saving en
erg
y
and d
e
livering em
erg
e
n
c
y data
pa
ckets to the
si
n
k
n
ode
as so
on a
s
possibl
e a
r
e
the two
imp
o
r
tant i
s
sue
s
i
n
mo
st
WSN appli
c
atio
ns.
The
ACAP
prop
ose
hybrid
and flexible
algorith
m
incl
uding
differe
nt prio
riti
es d
a
ta pa
ckets i
n
ord
e
r to
re
port em
erg
e
n
c
y
messag
e a
s
soo
n
as p
o
ssible wh
en co
nge
stion o
c
curs. Th
e sim
u
lation re
sult
s sh
ow that t
h
e
contri
bution
s
of the propo
sed ACAP WSN model
can
be de
scribe
d as follows: firstly, in order to
prolo
ng
th
e
lif
etime
of
th
e SNs, we
u
s
e a
hyb
r
id rep
o
r
ting data pa
ckets
to
th
e si
nk; se
con
d
ly, in
orde
r to red
u
c
e the data p
a
ckets lo
se, we u
s
e t
he flexible sen
s
in
g perio
d strat
egy. Finally, the
ACAP gua
ra
ntees that th
e eme
r
g
e
n
c
y pa
ckets
will
be tra
n
sfe
r
re
d to the
si
nk
node
a
s
soo
n
as
possibl
e by
way of the t
w
o p
r
io
rities
packet M
a
r
k
er.
The
sim
u
l
a
t
i
on r
e
sult
s
sho
w
e
d
t
hat
t
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 12, Dece
mb
er 201
3: 778
0 – 7786
7786
prop
osed A
C
AP has a
su
perio
r p
e
rfo
r
mance for
en
ergy efficie
n
cy and real ti
me monito
rin
g
in
WSN
s
.
Referen
ces
[1]
Jenn
ifier Yick,
Bis
w
an
ath M
u
kher
jee, D
i
pa
k Ghosal.
W
i
reless se
nsor
netw
o
rk surve
y
. Computer
Net
w
orks. 20
0
8
; 52: 229
2-23
30.
[2]
Cho
ngg
aa
ng
W
ang, Kazem
Sohra
b
y
, V
i
ctor La
w
r
ence
,
Bo Li, Yue
m
ing H
u
.
Pri
o
rity-bas
ed
Con
gestio
n
Co
ntrol i
n
W
i
re
les
s
Sens
or N
e
tw
orks
. Proce
edi
ngs
of the IEE
E
Internati
o
n
a
l
Confer
enc
e
on Sens
or Net
w
o
r
ks, Ubi
quit
ous, and T
r
ustw
o
r
th
y
Com
put
ing. 20
06; 22-
3
1
.
[3]
Shig
ang
Ch
en
, Na Ya
ng. C
ong
estio
n
Avo
i
da
nce B
a
sed
on
Lig
h
t
w
e
i
g
h
t Buffer Man
agem
ent i
n
Sensor N
e
t
w
or
ks.
IEEE
Trans. On Parallel a
nd Distribution System
s
. 20
06
; 17(9): 934-9
4
6
.
[4]
W
ang C, Li B, Sohrab
y K, D
anes
hman
d M, Hu Y.
Upstream cong
estio
n
control in
w
i
r
e
less sens
or
net
w
o
rks thro
u
gh cross-l
a
yer
optimiz
ation.
I
EEE Journal on Select
ed Area in Communic
a
tions
. 2
007;
25(4): 78
6-7
9
5
.
[5]
Mei-W
en H
uan
g, Hsu-Jun
g
Li
u, W
en-Sh
yo
n
g
Hsie
h. A H
y
brid Protoc
ol f
o
r Cluster-
bas
ed W
i
rel
e
ss
Sensor
Net
w
o
r
ks.
The thir
dteenth
IEEE A
s
ia-P
acific
C
o
mp
uter Syste
m
Arc
h
itectur
e
Co
nferenc
e
.
(ACSAC). 200
8; 243-2
47.
[6]
O Akan, I Ak
yi
diz. Eve
n
t-to-Sink R
e
li
ab
le T
r
ansp
o
rt in
W
i
reless S
ens
or
Net
w
orks. IEE
E
T
r
ans. on
Net
w
orki
ng. 20
05; 100
3-1
016.
[7]
CY W
an, SB
Eisenm
an.
AT
Ca
mp
be
ll. C
O
DA: Con
gest
i
on
Detecti
on
and
Avoi
da
nc
e in
Se
nsor
Netw
orks
. Proc. ACM SenS
y
s
.
2003: 2
66-2
7
9
.
[8]
Xu N
i
n
g
, Chri
stos G Cassandras.
Dyn
a
m
i
c
Sleep T
i
me
Contro
l in Eve
n
t-Driven W
i
re
l
e
ss Sensor
Networks.
Proceed
ing
of the 45th Confer
enc
e on
Dec
i
sio
n
& Control. 20
0
6
; 2722-
27
27.
[9]
Gergel
y, R
o
ll
and. Ad
aptiv
e sle
ep sc
hed
ul
i
ng
prot
ocol i
n
w
i
r
e
less s
ensor
net
w
o
rks.
Infoco
mmun
ica
t
ions Jour
nal
. 2
010; 65: 2
0
-30.
[10]
Alireza M
a
sou
m
, Nirvana Me
ratnia, Z
ahra
T
aghik
haki, Pa
ul JM Havi
ng
a
.
Re
w
a
rd
and
Punis
h
men
t
base
d
Co
oper
ative Ada
p
tive
Sampli
ng i
n
W
i
reless Se
ns
or Net
w
orks.
Intelli
ge
nt Sens
ors, Senso
r
Netw
orks and Informatio
n
Pro
c
essin
g
.
201
0: 145-
150.
[11]
Hu
Si
qua
n, Ya
ng
Ji
n
y
an
g, W
ang
Hai
ou, Sh
e
Ch
und
on
g, W
ang Ju
nfen
g. A Lo
w
P
o
w
e
r
MAC Protoco
l
for Wireless
S
ensor
Net
w
o
r
k
in
Agric
u
lture
Ca
nop
y
Mon
i
torin
g
.
Sens
or Letters.
20
11; 9(3):
1
235-
124
1.
[12] Hii
Pei-
Che
ng, T
oh
Sing-Hui,
Chu
ng Wan-Y
oun
g. Emerge
nc
y
H
a
n
d
li
ng
Guides i
n
a Wireless Se
nsor
Net
w
ork and Cod
e
Div
isio
n Multipl
e
Acces
s
T
e
chnolog
y Based
M
obi
le Healt
h
care
S
ystem.
Senso
r
Letters.
201
1; 3(2): 1-10.
[13]
Ke
w
e
i Sha, W
e
iso
ng Sh
i. Mo
deli
ng th
e lif
eti
m
e of W
i
reless
Sensor
Net
w
o
r
ks.
Sensor L
e
tters
. 2005;
9(1): 428-
43
2.
[14]
K F
a
ll, K Varadhan. ns notes
and d
o
cume
ntation. T
he VIN
T
Project, UC
Berkel
e
y
, LB
L, USC/ISI, and
Xer
o
x PARC, a
v
aila
bl
e from http://
w
w
w
.
isi.e
d
u
/nsam/ns. 20
10.
Evaluation Warning : The document was created with Spire.PDF for Python.