Indonesian
J
our
nal
of
Electrical
Engineering
and
Computer
Science
V
ol.
22,
No.
1,
April
2021,
pp.
407
418
ISSN:
2502-4752,
DOI:
10.11591/ijeecs.v22i1.pp407-418
r
407
Multi-Constraints
based
RPL
objecti
v
e
function
with
adapti
v
e
stability
f
or
high
traffic
IoT
applications
Abdelhadi
Eloudrhiri
Hassani,
Aicha
Sahel,
Abdelmajid
Badri,
El
Mourabit
Ilham
EEA&TI
laboratory
,
F
aculty
of
Sciences
and
T
echniques,
Hassan
II
Uni
v
ersity
,
Casablanca,
Morocco
Article
Inf
o
Article
history:
Recei
v
ed
Oct
9,
2020
Re
vised
Jan
23,
2021
Accepted
Mar
3,
2021
K
eyw
ords:
Combined
metrics
Contiki
OS
RPL
W
orkload
balancing
WSN
ABSTRA
CT
The
internet
of
things
technology
is
classified
as
a
Lo
w
po
wer
and
lossy
netw
ork.
These
kinds
of
netw
orks
require
a
trustw
orth
y
routing
prot
ocol
considered
as
the
back-
bone
for
ma
nagement
and
high
quality
of
service
achie
v
ements.
IPv6
Routing
Proto-
col
for
Lo
w
po
wer
and
lossy
netw
ork
(RPL)
w
as
able
to
g
ain
popularity
compared
to
other
routing
protocols
dedicate
d
to
IoT
for
its
great
fle
xibility
through
the
objec-
ti
v
e
function.
Def
ault
objecti
v
e
functions
implemented
in
the
RPL
core
are
based
on
a
single
metric.
Consequently
,
the
routing
protocol
can’
t
cope
wi
th
dif
ferent
con-
straints
and
sho
w
congestion
issues
in
high
traf
fics.
F
or
that,
we
proposed
in
our
pa-
per
Multi-Constraints-based
Objecti
v
e
Function
with
Adapti
v
e
Stability
(MCAS-OF),
which
uses
no
v
el
strate
gies
for
Radi
o
strength
indicator
,
node
ener
gy
consumption,
hop
count
and
a
designed
w
ork-metric
combination,
ne
w
rank
processing,
and
par
-
ent
selection
procedure.
The
netw
ork
stability
w
as
also
tak
e
n
into
account,
since
the
multi
constraints
can
lead
to
frequent
par
ent
changes,
using
an
adapti
v
e
threshold.
The
proposal,
e
v
aluated
under
the
COOJ
A
emulator
ag
ainst
Standard-RPL
and
EC-OF
,
sho
wed
a
pack
et
deli
v
ery
ratio
impro
v
ement
by
24%
in
high
traf
fics,
a
decrease
in
the
po
wer
consumption
close
to
44%,
achie
v
ed
less
latenc
y
and
DIO
control
messages,
it
also
gi
v
es
a
good
w
orkload
balancing
by
reducing
the
standard
de
viation
of
node’
s
po
wer
consumption.
This
is
an
open
access
article
under
the
CC
BY
-SA
license
.
Corresponding
A
uthor:
Abdelhadi
Eloudrhiri
Hassani,
EEA&TI
laboratory
,
F
aculty
of
Sciences
and
T
echniques,
Hassan
II
Uni
v
ersity
,
Morocco,
Email:
eloudrhiri.abdelhadi@gmail.com
1.
INTR
ODUCTION
IoT
,
considered
as
a
great
connecti
vity
potentiel
for
humanity
,
are
tin
y
sensor
de
vices
[1,
2]
based
netw
orks
with
constraint
and
limited
resources
[3].
Consequently
,
there
is
a
need
for
appropriate
resources
management
and
utilization
to
cope
with
dif
ferent
IoT
appli
cations
and
fields
as
smart
grids
[4,
5],
smart
cities
[6],
industries
[7]
and
healthcare
[8].
Thereby
,
in
order
to
ensure
perfomances
ef
ficienc
y
of
those
netw
orks,
the
researchers
ha
v
e
been
interested
mostly
in
routing
protocols
dedicated
to
IoT
applications.
In
this
conte
xt,
the
most
commonly
is
RPL
designed
by
Internet
Engineering
T
ask
F
orce
(IETF)
[9].
RPL
is
based
on
IPv6
and
uses
the
IEEE802.15.4
at
the
Ph
ysical
(PHY)
and
Medium
access
control
(MA
C)
layers
[10].
P
ack
ets
based
on
IPv6
cannot
fit
in
the
IEEE
802.15.4
protocol.
F
or
this
reason,
6Lo
wP
AN,
IPv6
Lo
w
Po
wer
Personal
Area
Netw
ork
layer
[11],
acts
as
an
adaptation
between
MA
C
and
netw
ork
layers.
W
ith
RPL,
the
paths
are
constructed
once
the
netw
ork
is
initialized.
The
nodes
aims
to
set
up
destination-
oriented
directed
ac
yclic
graph
(DOD
A
G),
a
tree
routing
topology
,
using
four
principal
ICMPv6
messages
:
J
ournal
homepage:
http://ijeecs.iaescor
e
.com
Evaluation Warning : The document was created with Spire.PDF for Python.
408
r
ISSN:
2502-4752
DOD
A
G
Information
Object
(DIO)
holds
informations
that
enable
nodes
to
kno
w
the
instance,
configuration
and
preferred
parent,
it
managed
by
the
trickle
algorithm
and
necessary
for
Multiple
Points
to
Point
(MP2P)
and
Point-to-Point
(P2P)
communications
[12].
DOD
A
G
Information
Solicitation
(DIS)
message
as
a
DIO
request
from
neighbours.
DOD
A
G
Adv
ertisement
Object
pack
ets
(D
A
O)
used
to
collect
do
wnw
ard
topology
informations
and
D
A
O-A
CK
as
a
response
to
a
D
A
O
message.
The
aim
of
DOD
A
G
topology
is
to
steer
data
pack
ets
to
one
or
multiple
sink.
Routing
paths
are
created
using
an
objecti
v
e
function
(OF).
In
the
core
of
RPL,
tw
o
OF’
s
are
proposed
namely
the
minimum
rank
with
h
ysteresis
objecti
v
e
function
(MRHOF)
[13]
based
on
e
xpected
transmission
counts
(ETX)
as
a
routing
metric
and
OF0
[14],
whi
ch
is
based
on
hop
count.
The
paths
in
MRHOF
are
based
on
the
link
quality
metric
calculated
by
broadcasting
probe
pack
ets
between
the
sender
and
recei
v
er
nodes
at
time
interv
als.
On
the
other
hand,
the
routing
paths
in
OF0
are
based
on
node
metric
that
aims
to
each
node
in
the
netw
ork
to
kno
w
its
position
related
to
sink.
Ho
we
v
er
,
the
tw
o
OF
tend
to
minimize
the
cost
of
their
metrics
which
causes
non-optimized
routes
due
to
taking
into
account
a
single
constraint
.
Furthermore,
the
y
are
not
ef
fecti
v
e
for
high
traf
fic
applications
and
load
balancing
between
netw
ork
nodes.
Considering
these
dra
wbacks,
we
ha
v
e
proposed
Multi-Constraints
based
objecti
v
e
function
with
Adapti
v
e
Stability
(MCAS-OF),
which
uses
no
v
el
strate
gies
for
link
and
node
metrics
combination,
ne
w
rank
processing
and
parent
selection
procedure.
The
netw
ork
stability
issue
is
also
tak
en
into
account
since
the
multi
constraints
can
lead
to
frequent
parent
changes.
Most
of
the
researchers
induce
a
constant
threshold,
consequently
,
it
decreases
t
he
chances
for
nearest
nodes
from
sink
to
change
their
parents
e
v
en
if
the
y
are
congested
or
ha
v
e
a
lot
of
w
orkload.
F
or
that
purpose,
our
objecti
v
e
function
uses
an
adapti
v
e
threshold
based
on
ranks.
The
main
contrib
utions
of
this
paper
are
summarized
as
follo
ws:
a.
Impro
v
e
the
RP
L
QoS
by
a
ne
wly
designed
objecti
v
e
function
MCAS.
Our
proposal
combines
multi-
metrics
chosen
to
cope
with
dif
ferent
constraints
while
respecting
the
objecti
v
e
function
con
v
er
gence.
b
.
Consider
the
netw
ork
stability
in
routing
by
a
proposed
adapti
v
e
threshold
based
on
rank’
s
node.
c.
A
simulation
under
COOJ
A
of
MCAS-OF
compared
to
Standard-RPL
and
EC-OF
in
term
of
PDR,
Ener
gy
consumption,
Latenc
y
,
DIO
control
messages
and
standard
de
viation.
The
rest
of
this
paper
is
or
g
anized
as
follo
w
.
In
Section
2.
we
present
the
related
w
orks
with
RPL
routing
protocol.
In
Section
3.
we
present
the
proposed
objecti
v
e
function
MCAS.
In
Section
4.
we
report
the
performance
e
v
aluations
results
and
discussion,
finally
a
conclusion
is
gi
v
en
in
Section
5.
2.
RELA
TED
W
ORKS
In
this
section,
we
present
some
related
researches
to
our
w
ork
that
impro
v
e
the
objecti
v
e
functions
performances
for
RPL
protocol
in
dif
ferent
IoT
applications.
In
[15],
authors
proposed
an
e
xtented
m
etric
based
on
the
e
xpected
transmission
count
called
Sigma-ETX
which
is
the
standard
de
viati
on
v
alue
of
ETX
by
number
of
hops
to
sink.
This
technique
a
v
oid
the
long
hops
issue
that
cause
bottlenecks
in
high
netw
ork
densities.
Also
EL
T
metric,
i.e.
Expected
Lifetime,
w
as
designed
by
authors
in
[16]
that
consist
of
the
node
remaining
lifetime
before
to
be
out
of
service.
It
is
injected
in
RPL
based
on
multipath,
which
leads
to
enhance
the
netw
ork
relia-
bility
despite
an
additional
delay
induced.
Node’
s
remaining
ener
gy
metric
w
as
designed
by
authors
in
[17]
on
which
the
objecti
v
e
function
is
based
to
select
preferred
parents.
This
proposal
has
pro
v
ed
its
ability
to
impro
v
e
the
o
v
erall
netw
ork
lifetime
b
ut
doesn’
t
cope
with
the
reliability
since
it
can
choose
lossy
links.
Minimizing
the
delay
of
recei
ving
pack
ets
by
the
sink
w
as
considered
by
authors
in
[18]
using
the
A
V
GDELA
Y
as
a
rout-
ing
metric.
This
technique,
ran
with
a
lo
w
duty
MA
C
c
ycle,
sho
wed
a
decrease
in
terms
of
end-to-end
delay
b
ut
don’
t
address
the
netw
ork
reliability
at
all.
The
fuzzy
logic
approach
w
as
considered
by
authors
in
[19].
Indeed,
the
y
proposed
COOF
,
i.e.
a
no
v
el
objecti
v
e
function
conte
xt-oriented
objecti
v
e
function,
that
considers
tw
o
designed
metrics
respecti
v
ely
remaining
ener
gy
(RE)
and
queue
fluctuation
inde
x
(QFI)
for
smart
cities
application
based
IoT
requirements.
Using
the
same
approach
in
[20],
the
authors
proposed
fuzzy
objecti
v
e
function
called
EC-OF
.
It
combines
ETX
and
node
ener
gy
consumption
while
hop
count
as
node
redirection
metric.
Results
sho
wed
that
the
proposal
impro
v
es
the
RPL
performances
in
terms
of
pack
et
deli
v
ery
ratio,
la-
tenc
y
,
con
v
er
gence,
po
wer
consumption
including
the
netw
ork
lifetime
ag
ainst
MRHOF
.
Simil
arly
,
the
authors
in
[21]
proposed
DQCA-OF
that
combines
the
same
three
metrics
considering
dif
ferent
performances.
The
method
sho
wed
a
significant
decrease
in
delay
and
a
high
PDR
in
lo
w
density
.
The
authors
in
[22]
proposed
Fuzzy
Logic
Based
Ener
gy
A
w
are
(FLEA)
for
RPL
objecti
v
e
function
based
on
residual
node
ener
gy
,
e
xpected
Indonesian
J
Elec
Eng
&
Comp
Sci,
V
ol.
22,
No.
1,
April
2021
:
407
–
418
Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian
J
Elec
Eng
&
Comp
Sci
ISSN:
2502-4752
r
409
transmission
count
ETX
and
traf
fic
load.
it
is
used
for
calculating
the
rank
increase
parameter
for
node’
s
rank.
The
proposal
sho
ws
a
increase
of
PDR
around
2%
to
5%
and
lifetime
around
10%.
The
conte
xt-a
w
areness
is
also
considered
during
no
v
el
objecti
v
e
functions
proposal.
In
[23],
authors
designed
Scalable
Conte
xt-A
w
are
for
Agricultural
en
vironmental
monitoring.
SCA
OF
is
based
on
ener
gy
,
relia
b
i
lity
,
resource,
and
rob
ustness
routing
metrics.
The
proposal
Performances,
v
erified
using
real
and
simulation
tests,
can
pro
vide
the
desired
adv
antages
in
terms
of
reliability
,
high
ef
ficienc
y
,
and
netw
ork
lifeti
me
e
xtension.
Congestion
problem
due
to
the
high
number
of
forw
arded
pack
ets
w
as
addressed
in
[24]
by
proposing
F
orw
arding
T
raf
fic
Consciousness
objecti
v
e
function,
which
combines
hop
count,
rssi
and
a
ne
wly
designed
FTM
metric.
The
proposed
method
sho
wed
a
pack
et
deli
v
ery
ratio
increase
respecti
v
ely
with
2%
and
11%
in
lo
w
and
high
traf
fics,
considerably
reduces
the
po
wer
consumption
with
approximately
47%
as
well
as
it
achie
v
es
a
good
balance
of
traf
fic
man-
aged
by
the
relay
nodes.
The
additi
v
e
and
le
xical
approachs
were
also
the
base
of
se
v
eral
objecti
v
e
function
optimizations.
Indeed,
authors
in
[25]
proposed
a
ne
w
objecti
v
e
function
based
on
additi
v
e
combination
of
node
and
link
metrics
respecti
v
ely
ETX
and
Ener
gy
consumption
along
the
routing
paths.
The
y
proposed
tw
o
w
ay
of
combination,
i.e.
with
weights
and
non
weighted
methods.
Results
sho
w
that
the
WCMOF
and
NWC-
MOF
can
increase
the
rel
iability
,
maximize
the
netw
ork
life
time
and
reduce
the
parent
changes.
Smart
grid
application
requirements
were
considered
by
authors
in
[26].
The
authors
designed
OFQS
objecti
v
e
function
based
on
po
wer
state
metric,
e
xpected
transmission
count
and
delay
.
The
stability
w
as
consi
dered
by
t
he
h
ys-
teresis
concept
as
in
MRHOF
.
Results
sho
wed
impro
v
ement
in
terms
of
end-to-end
delay
,
netw
ork
lifetime
and
pack
et
deli
v
ery
ratio.
The
authors
proposed
le
xicographic
and
additi
v
e
a
p
pr
o
a
ches
in
[27]
to
combi
ne
e
xpected
transmission
count,
hop
count
and
a
v
ailable
ener
gy
metrics
in
EHA
OF
.
The
results
sho
wed
better
performance
in
terms
of
ener
gy
consumption,
netw
ork
latenc
y
and
pack
et
deli
v
ery
ratio
ag
ainst
MRHOF-ETX
and
OF0.
In
[28],
authors
ha
v
e
been
interested
in
the
congestion
of
the
path
caused
by
the
b
uf
fer
nodes
occupanc
y
.
F
or
that
purpose,
Congestion-A
w
are
Objecti
v
e
Function
CA-OF
were
be
proposed
which
consider
ETX
metric
at
a
lo
w
data
rate
while
the
b
uf
fer
occupanc
y
is
considered
at
a
high
data
rate.
Figure
1
[29]
summarizes
the
structure
of
dif
ferent
researches
related
to
RPL
protocol.
Figure
1.
Dif
ferent
researches
related
to
RPL
protocol
3.
RESEARCH
METHOD
The
QoS
of
fered
by
a
routing
protocol
for
IoT
based
on
wireless
sensor
netw
orks
is
one
of
the
sub-
stantial
design
concerns.
Besides,
its
multiple
requirements
meeting
allo
ws
it
to
be
widely
used
for
dif
ferent
applications.
As
mentioned
before,
the
RPL
objecti
v
e
function
is
mainly
responsible
for
satisfying
the
appli-
cation
prerequisites
according
to
the
constraint
metrics
used
i
n
the
preferred
parents
selection.
These
metrics
are
commonly
classified
into
tw
o
cate
gories.
The
first
cate
gory
considers
the
quality
of
links
between
nodes
as
rssi,
delay
and
e
xpected
transmission
count
ETX,
while
the
second
cate
gory
re
groups
node
parameters
such
as
po
wer
consumption,
residual
ener
gy
,
number
of
hops,
queue
utilization.
In
a
standard
manner
,
RPL
emplo
ys
an
objecti
v
e
function
with
one
metric
to
minimize.
This
lead
to
cope
wi
th
limited
constraints,
which
certainly
in-
flict
poor
performances
in
term
of
v
arious
requirements.
F
or
instance,
the
objecti
v
e
funct
ion
based
on
e
xpected
transmission
count
ETX
can
sho
w
lo
w
pack
et
losses,
b
ut
ignores
the
node’
s
ener
gy
consumption
management.
Also,
based
on
the
hop
count
metric,
the
RPL
routing
protocol
can
minimize
the
hops
to
the
destination
then
MCAS
RPL
objective
function
for
high
tr
af
fic
IoT
applications
(Abdelhadi
Eloudrhiri
Hassani)
Evaluation Warning : The document was created with Spire.PDF for Python.
410
r
ISSN:
2502-4752
achie
v
es
a
high
le
v
el
of
ener
gy
preserv
ation
b
ut
can’
t
cope
with
the
ne
tw
ork
reliability
.
Furthermore,
the
ob-
jecti
v
e
functions
defined
in
the
RPL
core
are
not
optimized
for
high
traf
fic
applications
and
considering
load
balancing
between
netw
ork
nodes.
T
o
o
v
ercome
these
challenges,
we
ha
v
e
proposed
Multi-Constraints
based
objecti
v
e
function
with
Adapti
v
e
Stability
(MCAS-OF),
which
uses
no
v
el
strate
gies
for
link
and
node
metrics
combination,
ne
w
rank
processing
and
parent
selection
procedure.
The
netw
ork
stability
issue
is
also
tak
en
into
account
since
the
multi
constraints
can
lead
to
frequent
parent
changes.
Most
of
the
researchers
induce
a
constant
threshold,
consequently
,
it
decreases
the
chances
for
nearest
nodes
from
sink
to
change
their
parents
e
v
en
if
the
y
are
congested
or
ha
v
e
a
lot
of
w
orkload.
F
or
that
purpose,
our
objecti
v
e
function
uses
an
adapti
v
e
threshold
based
on
ranks.
3.1.
Multi-Constraint
Metrics
with
W
orkload
Consideration
Reliability
,
ener
gy
consumption,
w
orkload
management,
shortest
path
are
the
most
influential
parame-
ters
on
the
QoS
of
IoT
applications.
Accordingly
,
it
is
important
to
determine
ho
w
these
four
paramete
rs
should
be
combined
to
mak
e
the
best
rout
ing
decisions.
In
this
conte
xt,
we
ha
v
e
proposed
in
our
paper
an
additi
v
e
combination
of
four
metrics
namely
rssi,
node
ener
gy
consumption,
hop
count
and
a
designed
w
ork-metric.
Ho
we
v
er
,
these
metrics
should
be
chosen
in
such
a
w
ay
that
all
must
be
minimized
to
a
v
oid
loops
in
routing
and
respect
the
objecti
v
e
function
con
v
er
gence.
Number
of
hops
is
the
only
one
among
the
four
metric
which
is
announced
by
neighboring
nodes.
F
or
that
purpose,
we
ha
v
e
modified
the
s
tructure
of
the
DIO
control
message
precisely
in
the
metric
container
part.
Indeed,
at
reception
of
DIO
message
by
a
node
i,
it
process
the
hop
count
metric
follo
wing
Equation
1.
H
C
(
i
)
=
M
inhc
I
ncr
ement
if
n
=
sink
hc
(
n
)
+
M
inhc
I
n
c
r
ement
if
n
6
=
sink
(1)
Where
n
is
the
neighbor
sus
ceptible
to
be
the
preferred
parent
of
node
i,
HC(n)
reprensent
the
number
of
hops
to
the
sink
adv
ertised
in
the
neighbor
DIO
whereas
MinHC
Increment
is
a
scalar
v
alue
equal
to
256.
Ener
gy
consumption
is
calculated
by
a
node
i
using
the
po
wer
trace
tool
follo
wing
Equation
2,
it
is
the
summation
of
transmission,
reception,
cpu,
and
lpm
po
wer
consumptions
in
each
state.
E
C
(
i
)
=
T
cpu
5
:
4
+
T
tx
58
:
5
+
T
r
x
64
:
5
+
T
LP
M
0
:
1635
R
T
I
M
E
R
AR
C
H
S
E
C
O
N
D
V
ol
tag
e
(2)
Where
T
cpu,
T
tx,
T
rx,
T
LPM
are
respecti
v
ely
the
ticks
numbers
when
the
node
is
processing
at
the
cpu
le
v
el,
transmitting,
listening
or
going
to
lo
w
po
wer
mode
follo
wig
the
MA
C
protocol,
the
numerical
parameters
are
nominal
v
alues
pro
vided
in
the
Sk
ymote
datasheet,
R
TIMER
ARCH
SECOND
represents
the
number
of
ticks
per
second
equal
to
32768
and
V
oltage
is
the
initial
battery
v
alue
equal
to
3V
.
W
ork-metric
is
processed
follo
wing
Equation
3,
such
as
it
the
summation
of
the
number
of
sent
data
pack
ets
and
the
number
of
D
A
O
control
messages.
Indeed,
the
D
A
O
is
unicast
from
children
to
the
parent
node.
F
or
that,
a
node
recei
ving
more
D
A
O
implies
that
it
manage
more
nodes.
Then,
the
w
ork-metric
gi
v
es
an
idea
of
the
w
orkload
of
each
node,
and
should
be
minimizable.
w
or
k
metr
ic
(
i
)
=
D
AT
A
sent
pack
ets
+
D
AO
r
eceiv
ed
(3)
Counts
the
number
of
sent
pack
ets
at
the
IP
layer
.
It
includes
the
number
of
pack
ets
containi
ng
the
data
sent
by
the
node
i
and
the
D
A
O
control
messages.
The
sent
data
are
equal
b
e
tween
all
nodes
b
ut
the
D
A
O
are
not.
Indeed,
when
a
node
has
a
lot
of
childrens,
it
recei
v
es
more
D
A
O
control
messages
which
gi
v
es
an
idea
of
the
number
of
childrens
then
the
w
orkload
imposed
to
each
node.
F
or
this
reason,
we
designed
this
metric
that
aims
to
balance
the
load
between
all
the
nodes.
Radio
signal
str
ength
indicator(rssi)
represent
a
link
metric
measured
by
a
node
i
through
the
CC2420
radio
follo
wing
Equation
4.
r
ssi
(
i
)
=
r
ssi
r
eg
(
i
)
empir
ical
ad
j
ustment
(4)
Where
rssi
re
g(i)
is
a
8
bits
re
gister
that
gi
v
es
the
signal
strength
between
node
i
and
its
neighbor
.
The
antenna
v
ariation
of
fset
is
represented
with
empirical
adjustment
equal
to
45
pro
vided
from
the
CC2420
radio
datasheet.
Since
the
rssi
m
etric
gi
v
es
v
alues
between
0
and
-110
dBm
respecti
v
ely
for
good
and
bad
signal
strength,
this
metric
must
be
maximized.
T
o
k
eep
the
objecti
v
e
function
con
v
er
gence,
we
w
ork
ed
with
its
absolute
v
alue
to
follo
w
the
minimizing
rule
as
other
combined
metrics.
Indonesian
J
Elec
Eng
&
Comp
Sci,
V
ol.
22,
No.
1,
April
2021
:
407
–
418
Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian
J
Elec
Eng
&
Comp
Sci
ISSN:
2502-4752
r
411
3.2.
Designed
MCAS
Objecti
v
e
Function
During
the
process
of
parent
sel
ection,
nodes
are
based
on
ranks
calculated
by
the
objecti
v
e
f
u
nc
tion.
Neighbors
with
lo
wer
rank
are
elected
by
the
concerned
node
to
be
its
preferred
parent.
Unlik
e
standard
objecti
v
e
functions
based
on
a
single
metric,
MCAS-OF
uses
a
set
of
metrics
combined
for
rank
processing.
Indeed,
at
a
DIO
reception
from
a
neighbor
,
the
node
i
calculates
the
number
of
hops,
w
ork-metric,
measure
its
ener
gy
consumption
and
the
RSSI
at
the
MA
C
la
yer
.
The
combination
of
these
metrics
is
done
linearly
with
dif
ferent
weights
follo
wing
Equation
5.
C
ombined
(
r
ssi;
ener
g
y
;
w
or
k
;
hc
)
=
r
ssi
(
i
)
+
ener
g
y
(
i
)
+
w
or
k
metr
ic
(
i
)
+
hc
(
i
)
(5)
The
weights
are
distinguished
bet
ween
constant,
complementary
,
and
influencing.
In
this
paper
,
the
hop
count
weight
is
related
to
a
unit
constant
v
alue,
rssi,
and
ener
gy
consumption
ha
v
e
a
complementary
weights
equal
to
1
(
,
)
while
the
w
ork
metric
ha
v
e
an
influencing
weight
empirically
defined
(
).
Subse-
quently
,
the
node
i
e
xtract
the
rank
of
neighbor
adv
ertised
in
the
DIO
message
and
process
its
rank
according
to
Equation
6.
R
ank
(
i;
n
)
=
R
ank
(
n
)
+
C
ombined
(
r
ssi;
ener
g
y
;
w
or
k
;
hc
)
(6)
Where
Rank(n)
is
the
rank
of
the
neighbor
candidate
to
be
the
preferred
parent
through
node
i
could
forw
ard
pack
ets
to
sink.
At
this
stage,
once
the
node
has
calculated
its
rank
based
on
the
DIO
sender
,
if
it
is
the
first
round
for
netw
ork
establishment,
the
node
chooses
it
as
the
preferred
parent.
Otherwise,
a
rank
comparison
between
old
and
candidate
parent
is
done,
if
it
is
higher
,
the
candidate
parent
is
discarded.
Else,
considering
the
adapti
v
e
stability
,
if
it
is
lo
wer
than
the
preferred
parent
rank
with
the
adapti
v
e
threshold
v
alue,
then
the
candidate
parent
is
retained.
The
threshold
is
a
mandatory
parameter
for
reducing
the
preferred
parent
changes.
F
or
that,
the
adapti
v
e
threshold
defined
in
Equation
7
is
used
for
MCAS
objecti
v
e
function
in
order
to
gi
v
e
routing
path
stability
on
one
hand,
and
on
the
other
hand,
increase
the
chances
for
nodes
near
to
sink
to
change
their
preferred
parent.
Adaptiv
e
T
hr
eshol
d
(
i;
n
)
=
r
ank
(
i
)
+
r
ank
(
n
)
2
+
M
inhc
I
n
c
r
ement
(7)
Finally
,
if
the
condition
for
the
parent
change
is
fulfilled,
the
node
updates
its
rank
and
the
hop
count
metric
in
the
DIO
message,
then
broadcast
a
ne
w
one
to
adv
ertise
neighbors
that
ther
e
is
a
change
in
the
routing
path.
Algorithm
1
sho
ws
the
proposed
MCAS
objecti
v
e
function
algorithm
implemented
in
the
RPL
core,
while
Figure
2
summarizes
the
ne
w
process
of
routing
decision.
Figure
2.
Process
of
preferred
parent
selection
with
MCAS-OF
MCAS
RPL
objective
function
for
high
tr
af
fic
IoT
applications
(Abdelhadi
Eloudrhiri
Hassani)
Evaluation Warning : The document was created with Spire.PDF for Python.
412
r
ISSN:
2502-4752
Algorithm
1:
Proposed
MCAS
objecti
v
e
function
algorithm
When
node
i
recei
v
e
a
DIO
message
from
neighbor;
if
(neighbor
!
=
NULL)
then
rssi
=
abs(rssi
measure());
ener
gy
=
po
wer
trace();
hc
=
neighbor
.dio.hc
+
Minhc
Increment;
w
ork
metric
=
number
sent
data
+
Number
D
A
O
recei
v
ed;
Combined(rssi,ener
gy
,w
ork,hc)
=
*rssi
+
*ener
gy
+
*w
ork
metric
+
hc;
Possible
rank
=
neighbor
.dio.rank
+
Combined(rssi,ener
gy
,w
ork,hc);
Adapti
v
e
Threshold
=
P
ossibl
e
r
ank
+
P
r
ef
er
r
ed
par
ent:r
ank
2
+
Minhc
Increment;
end
if
((Pr
eferr
ed
par
ent
==
NULL)
jj
(neighbor
.dio.r
ank
+
Combined(r
ssi,ener
gy
,work,hc))
<
(Pr
eferr
ed
par
ent.r
ank
+
Adaptive
Thr
eshold))
then
Preferred
parent
=
neighbor;
rank
=
Possible
rank;
hc
=
hc;
else
/*
K
eep
old
pr
eferr
ed
par
ent
e
xit;
end
/*
Update
the
ne
w
DIO
messa
g
e
dio.mc.hc
=
hc;
dio.rank
=
rank;
/*
Br
oadcast
DIO
messa
g
e
4.
RESUL
T
AND
DISCUSSION
4.1.
Simulation
Settings
The
performance
e
v
aluation
of
the
propose
d
objecti
v
e
f
u
nc
tion
is
based
on
s
imulation
carried
on
the
Cooja
simulator
,
using
embedded
platforms
with
operating
system
Contiki
2.7
[30].
Cooja
is
a
simula-
tor/emulator
widely
used
by
the
IoT
researchers
to
test
their
proposed
w
orks.
The
netw
ork
consists
of
25
or
50
client
nodes
with
a
unique
sink.
Each
node
generates
a
1
to
20
pack
ets
per
minute
to
test
our
w
ork
in
multiple
scenarios
especially
in
high
traf
fics.
Cooja
pro
vides
as
one
of
the
radio
models,
the
UDGM
i.e.
unit
disk
graph
medium
(UDGM),
that
adds
wireless
medium
losses,
which
has
been
used
to
cope
with
a
realistic
simulation
en
vironment.
The
looseness
in
the
medium
is
relati
v
e
to
the
distance
between
de
vices
in
the
netw
ork.
In
our
scenario,
we
set
70
meters
as
the
transmission
range
with
100
meters
for
the
interference
range.
The
nodes
are
randomly
distrib
uted
in
an
area
of
200
×
200
m²,
while
the
sink
is
placed
a
w
ay
from
the
monitored
netw
ork.
The
platforms
that
are
used
in
the
simulation
are
the
Sk
y
mote
type,
which
has
MSP430
as
a
microcontroller
with
2.4
GHz
wireless
transcei
v
er
Chipcon
CC2420.
The
motes
run
Contiki
2.7
OS
and
conform
with
com-
munication
protocol
IEEE
802.15.4.
The
simulations
are
performed
o
v
er
600s
for
e
v
ery
en
vironmental
setup.
Simulation
settings
are
summarized
in
T
able
1.
T
able
1.
Simulation
Settings
Netw
ork
simulator
Cooja
Embedded
operating
system
Contiki
2.7
Radio
en
vironment
Unit
disk
graph
medium
-
DL
Emulated
nodes
Sk
y
motes
Netw
ork
area
200
x
200
m²
Deplo
yment
of
nodes
Random
Number
of
sinks
1
Number
of
senders
25,50
Generated
pack
ets
1,5,15,20
ppm
T
ransmission
/
interference
ranges
70/100
m
Simulation
time
600s
Indonesian
J
Elec
Eng
&
Comp
Sci,
V
ol.
22,
No.
1,
April
2021
:
407
–
418
Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian
J
Elec
Eng
&
Comp
Sci
ISSN:
2502-4752
r
413
4.2.
P
erf
ormances
Ev
aluation
4.2.1.
P
ack
et
Deli
v
ery
Ratio
Figure
3a
and
3b
sho
w
that
in
lo
w
traf
fic,
whate
v
er
the
density
,
the
three
objecti
v
e
functions
ha
v
e
almost
identical
result
or
a
slight
dif
ference.
But
when
the
traf
fic
increase,
MCAS-OF
pro
vides
an
a
v
erage
PDR
increase
with
5.5
%
and
16
%
in
a
density
of
25
nodes,
14
%
and
24
%
in
a
densi
ty
of
50
nodes,
ag
ainst
Standard-RPL
and
EC-OF
respecti
v
ely
.
These
results
can
be
e
xplained
by
the
f
act
that,
in
high
traf
fic,
the
congestion
in
paths
frequently
occurs
with
standard
objecti
v
e
functions
that
minimize
a
single
metric.
While
our
proposal,
in
addition
to
the
four
metrics
combination,
the
w
ork-metric
can
influence
routing
path
selection
by
the
dispersion
of
node’
s
w
orkload
then
decrease
the
probability
of
congestion.
The
other
reason
is
the
o
v
erall
netw
ork
stability
pro
vided
by
the
adapti
v
e
stability
threshold
which
limits
the
frequent
parent
changes
then
unnecessary
pack
ets
loss.
(a)
(b)
Figure
3.
P
ack
et
Deli
v
ery
ratio
vs
number
of
pack
ets
per
minute,
(a)
Density
of
25
nodes,
(b)
Density
of
50
nodes
MCAS
RPL
objective
function
for
high
tr
af
fic
IoT
applications
(Abdelhadi
Eloudrhiri
Hassani)
Evaluation Warning : The document was created with Spire.PDF for Python.
414
r
ISSN:
2502-4752
4.2.2.
A
v
erage
P
o
wer
Consumption
In
term
of
node’
s
a
v
erage
po
wer
consumption,
Figure
4a
and
4b
sho
w
that
our
proposal
consumes
less
than
the
other
objecti
v
e
functions.
Indeed,
MCAS-OF
decreases
the
po
wer
consumption
with
21
%
and
32
%
in
a
density
of
25
nodes,
40
%
and
44
%
in
a
density
of
50
nodes,
compared
to
Standard-RPL
and
EC-OF
respecti
v
ely
.
The
reason
for
less
consumption
is
the
hop
count
and
ener
gy
metrics
introduced
in
rank
processing,
so
nodes
can
a
v
oid
unnecessary
retransmissions
of
pack
ets
due
to
long
hops
while
considering
also
the
ener
gy
consumption.
The
other
reason
is
the
rssi
metric
which
allo
ws
the
parent
selection
where
the
signal
strength
between
them
is
good,
also
it
limits
ener
gy
losses
in
the
pack
et
transmission
o
v
er
poor
link
quality
.
(a)
(b)
Figure
4.
A
v
erage
po
wer
consumption
vs
number
of
pack
ets
per
minute,
(a)
Density
of
25
nodes,
(b)
Density
of
50
nodes
4.2.3.
Standard
De
viation
of
Nodes
P
o
wer
Consumption
The
Standard
de
viation,
as
a
statistical
parameter
,
is
included
in
the
performance
e
v
aluation
to
in-
v
estig
ate
on
the
po
wer
consumption
distrib
ution
between
nodes.
It
is
defined
as
the
de
viation
of
the
ener
gy
consumption
of
each
in
the
netw
ork
related
to
their
a
v
erage.
A
glance
on
Figure
5a
and
5b
sho
ws
that
our
proposal
significantly
decreases
the
dif
ference
between
the
po
wer
consumption
of
nodes
and
their
a
v
erage.
Indonesian
J
Elec
Eng
&
Comp
Sci,
V
ol.
22,
No.
1,
April
2021
:
407
–
418
Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian
J
Elec
Eng
&
Comp
Sci
ISSN:
2502-4752
r
415
These
results
can
be
e
xplained
thanks
to
the
w
ork
metric
introduced
in
the
combined
metrics,
which
fulfills
its
load-balancing
role
between
all
the
nodes,
e
v
en
for
those
close
to
the
sink.
(a)
(b)
Figure
5.
Standard
de
viation
of
nodes
po
wer
consumption
vs
number
of
pack
ets
per
minute,
(a)
Density
of
25
nodes,
(b)
Density
of
50
nodes
4.2.4.
DIO
Contr
ol
P
ack
ets
DIO
control
pack
ets
are
used
by
RPL
protocol
to
create
and
maintains
the
netw
ork
topology
.
As
mentionned
before,
e
v
ery
node
broadcasts
DIO
periodically
using
the
trickle
algorithm
to
update
nodes
status.
More
the
netw
ork
is
stable,
the
DIO
sent
frequenc
y
will
be
reduced.
The
results
in
Figure
6a
and
6b
sho
w
that
proposed
MCAS-OF
generates
less
DIO
compared
to
the
other
objecti
v
e
functions.
This
is
because
the
proposal
uses
the
adapti
v
e
stability
threshold
that
gi
v
e
more
stability
in
routing
paths.
MCAS
RPL
objective
function
for
high
tr
af
fic
IoT
applications
(Abdelhadi
Eloudrhiri
Hassani)
Evaluation Warning : The document was created with Spire.PDF for Python.
416
r
ISSN:
2502-4752
(a)
(b)
Figure
6.
DIO
control
pack
ets
vs
number
of
pack
ets
per
minute,
(a)
Density
of
25
nodes,
(b)
Density
of
50
nodes
4.2.5.
Latency
Latenc
y
is
defined
as
the
total
delay
that
mak
e
sent
pack
ets
by
a
node
in
the
netw
ork
to
be
succes
fully
recei
v
ed
by
the
sink.
Compared
to
other
objecti
v
e
functions
tak
en
in
our
performances
e
v
aluation,
MCAS-OF
gi
v
es
less
delay
duration
as
sho
wn
in
Figure
7a
and
7b
especially
when
the
number
of
pack
ets
sent
per
minute
increase.
The
result
s
can
be
e
xplained
by
that
our
method
can
relie
v
e
congestion
that
may
occur
in
high
traf
fics,
allo
wing
for
f
aster
package
deli
v
ery
.
(a)
Indonesian
J
Elec
Eng
&
Comp
Sci,
V
ol.
22,
No.
1,
April
2021
:
407
–
418
Evaluation Warning : The document was created with Spire.PDF for Python.