A CLUSTER-HEAD SELECTION ALGORITHM FOR WIRELESS SENSOR NETWORKS

foamyflumpMobile - Wireless

Nov 21, 2013 (3 years and 9 months ago)

57 views

A.
Chalak
, S. Misra and M. S. Obaidat

Corresponding Address:

Mohammad S. Obaidat, Fellow of IEEE and SCS

President, Society for Modeling & Simulation International
(SCS) Editor
-
in
-
Chief, International Journal of
Communication Systems, Wiley

Professor of Computer Science and Software Engineering,

Monmouth University, W. Long Branch, NJ 07764, USA

E
-
mail:
Obaidat@monmouth.edu

Website:
www.monmouth.edu/mobaidat



Professor Mohammad S. Obaidat

1

A CLUSTER
-
HEAD SELECTION ALGORITHM
FOR WIRELESS SENSOR NETWORKS

Objectives


The aim of this paper is to present a new energy
-
efficient
clustering approach for WSNs called Improved Minimum
Separation Distance(IMSD).


In order to prolong the network lifetime, energy resources of
each node in WSN need to be effectively managed.


With respect to energy efficiency, the best known protocols
are hierarchical in nature. Energy consumption can be
reduced efficiently by using clustering based hierarchical
routing protocols.


IMSD extends the existing MSD clustering algorithm. The
simulation results show that IMSD is more energy
-
efficient,
implying that it is more effective in prolonging the network
life time.


Prof. Mohammad S. Obaidat

Introduction


Sensors in WSN are small, inexpensive, low
-
power, intelligent and
disposable. The sensor nodes are self
-
configuring and contain one
or more sensors, integrated with wireless communication devices
and data processing components and a limited energy source.


Due to the large number of nodes and the possibly hazardous
environment in which these nodes are deployed, their batteries
are often assumed to be nonreplaced.


The failure of a single node in the network could possibly cause
network partition and dissect a part of the WSN off from the rest
of the network.


Network lifetime is, therefore, dependent on the lifetime of
individual nodes. This raises the issue of energy
-
efficient design of
the network. This research effort focuses on the design of an
energy
-
efficient cluster head selection algorithm for WSNs
.


Prof. Mohammad S. Obaidat

Introduction


In a WSN, there are possibly very large number of sensing
nodes, and a base station.


The sensing nodes have to route data about their
environment to the base station. A sensing node is
sometimes called a source and a base station is sometimes
called a sink.


The sink node collects and interprets the data from all the
source nodes in the network. The sink node may be
connected to a wired network and may not have an energy
limitation. The source nodes, on the other hand, are
dependent on their limited batteries and become dead
when their batteries are completely
exhasuted
.


Prof. Mohammad S. Obaidat

Introduction

Prof. Mohammad S. Obaidat

Sink
Link to other
Network
Nodes
Wireless Sensor Network Area

Introduction


Routing in Wireless Sensor Networks can be
divided into flat
-
based, location
-
based, and
hierarchical
-
based routing.


Flat
-
based routing: All nodes are typically
assigned equal functionalities.


Location
-
based: Positions are exploited to route
data in network.


Hierarchical
-
based: Nodes will play different roles
in network.

Prof. Mohammad S. Obaidat

Introduction


WSN protocols are different from the traditional
wireless protocols due to limited power supply, large
network size, and inaccessible remote deployment
environment.


A lot of energy is saved by using multi
-
hop
communication than direct communication, as in short
range communication due to the fact that consumption
of energy is proportional to the square of the distance.


A routing protocol might be good for continuous data
sensing while it may not perform well where it will have
periodic monitoring.

Prof. Mohammad S. Obaidat

Introduction


Clustering approach is used for wireless communication in
wireless sensor networks. In this case, nodes send their data
to the nearest cluster head which then forwards it to the
base station. In each round, the cluster head node sends out
a beacon, and nodes that hear the beacon join the cluster. If
there are too many nodes in the cluster, the cluster
-
head can
reduce beacon signal strength so fewer nodes will hear it.


On the other hand, if the cluster is too small, the cluster
head can increase its beacon signal strength to increase
cluster membership. Thus, energy efficiency is increased by
using clustering technique.


In this paper, we present the Improved Minimum Separation
Distance (IMSD) cluster head selection algorithm, which is based
on the total average energy available in the network.


Prof. Mohammad S. Obaidat

Terminology



LEACH

Low

Energy

Adaptive

Clustering

Hierarchy


MSD

Minimum

Separation

Distance


MST


Minimum

Spanning

Tree


BS


Base

Station/Sink


CH


Cluster

Head



dec_MSD

Decrease

MSD

by

10
%


LCF


Link

Cost

Factor


SEP


Stable

Election

Protocol


gap


decrease

MSD

by

10
%

Prof. Mohammad S. Obaidat

RADIO COMMUNICATION MODEL


In a WSN, there are two types of energy costs; fixed cost
in the electronics when transmitting or receiving a
message, and variable energy costs which are
proportional to the distance of transmission.


In our work, we consider a simple model where the
radio dissipates
𝐸
𝑒

𝑒
𝑐

= 50
nJ
/bit to run the transmitter
or receiver
circuitry and

𝑎

𝑝
= 100
pJ
/bit/

2 for the
amplifying signal.


These parameters are slightly better than the current
state
-
of
-
the
-

art in radio design. We also assume d
2

energy loss due to channel transmission.

Prof. Mohammad S. Obaidat

RADIO COMMUNICATION MODEL


Thus, to transmit a k bit message with distance d and
considering free space model, the radio model as shown
in Figure 1 expends:









Prof. Mohammad S. Obaidat

and to receive this message, the radio expends:

RADIO COMMUNICATION MODEL


For these parameter values, receiving a message is not a
low cost operation; the protocols should thus try to
minimize not only the transmit distances, but also the
number of transmit and receive operations for each
message.


An inefficient use of the available energy leads to poor
performance and short life cycle of the network.


Energy in these sensors is a scarce resource and must be
managed in an efficient manner. The most significant
factor in the development of WSN is typically power. It is
necessary that the transmission distances of nodes is
reduced in order to conserve energy.


Prof. Mohammad S. Obaidat

RADIO COMMUNICATION MODEL

Prof. Mohammad S. Obaidat

Background


Low

Energy

Adaptive

Clustering

Hierarchy

(LEACH)

uses

a

self
-
configuration

approach

and

reduces

much

energy

consumption
.



It

does

not

consider

the

distribution

of

nodes
.



Cluster

head

nodes

are

distributed

unequally
.



Data

collected

by

cluster
-
head

directly

forwarded

to

base

station
.



As

cluster
-
heads

are

far

away

from

the

base

station
.

Direct

transmission

is

bad

for

energy

conservation
.



Clusters constructed by LEACH.


Black nodes represent cluster
-
heads

Prof. Mohammad S. Obaidat

Background (contd.)


Cluster

constructing

method

has

to

be

changed
.



The

communication

between

cluster
-
heads

and

base

station

also

needs

to

be

modified
.



In

this

work,

we

present

a

modification

of

LEACH’s

cluster
-
head

selection

algorithm

to

further

reduce

the

total

energy

dissipation

of

sensors


Our

cluster

constructing

method

avoids

the

uneven

member

distribution

for

clusters
.




Clusters constructed by LEACH MSD and MST.


Prof. Mohammad S. Obaidat

PROPOSED ALGORITHM: IMSD


Low Energy Adaptive Clustering Hierarchy (LEACH)

is a
hierarchical routing protocol. It is an energy
-
efficient protocol, as
it group nodes into clusters.


It uses Cluster Head (CH) to fuse data before transmitting to the
Base Station (BS).


But LEACH has some limitations:



First the geographical distribution of the CHs severely influences
the overall energy consumption of the network. Spreading the
CHs more evenly results in prolonging the lifetime of the
network.


Second, in WSNs asymmetric communication is possible. That is,
the base station can reach all the sensor nodes directly. While
some sensor nodes cannot reach the base station directly,
therefore these sensor nodes need help from other nodes to
forward their data. Hence routing schemes are necessary.


Prof. Mohammad S. Obaidat

PROPOSED ALGORITHM: IMSD


Third, as the network size increases, the transmission
distance within the cluster increases. This results in
an increase in energy consumption.


In this paper, we present an energy
-
efficient cluster
head selection algorithm to overcome the above
three limitations.


Prof. Mohammad S. Obaidat

PROPOSED ALGORITHM: IMSD


The geographical distribution of the CHs severely
influences the overall energy consumption of the
network.


For prolonging the lifetime of the network, CHs should
be spread evenly. Use of improved minimum separation
distance between CHs improves the network lifetime.


The smallest distance allowed between CHs is minimum
separation distance. The minimum separation distance
can be smaller than the separation distance between
CHs, but should not be larger.


Prof. Mohammad S. Obaidat

PROPOSED ALGORITHM: IMSD

Prof. Mohammad S. Obaidat

PROPOSED ALGORITHM: IMSD

Prof. Mohammad S. Obaidat


An extension to LEACH is used in the proposed routing
technique. The cluster formation carried out by the base
station uses a centralized cluster formation algorithm.


The same steady
-
state protocol as LEACH is used in the
proposed protocol. The current location & energy level of each
node is informed to the base station during the set
-
up phase.


To determine the CH and the cluster for that round, centralized
cluster formation algorithm is run at base station.


All the nodes in the network receive the information that is
broadcast by base station after the creation of clusters. Before
a node goes to sleep mode until it is time to transmit data to
the CH, i.e., until the arrival of next slot, determination of its
local TDMA is essential for each of the nodes, except the CH.


PROPOSED ALGORITHM: IMSD

Prof. Mohammad S. Obaidat


From a list of eligible nodes, a CH is randomly chosen in the CH
selection part.


In the network, the average energy of remaining nodes is
calculated to determine which nodes are eligible. The nodes with
energy level above the average are eligible, in order to spread
the load evenly.


A new node has to be chosen to guarantee the minimum
separation distance if a node that has been randomly chosen is
too close, i.e., within the range of the minimum separation
distance from all other chosen CHs.


If the desired number of CHs is not formed, then the minimum
separation distance is decremented by 10% and the same
algorithm is implemented until the desired number of CHs is
attained.

PROPOSED ALGORITHM: IMSD

Prof. Mohammad S. Obaidat


Clusters are created the same way as in the
Low
Energy Adaptive Clustering Hierarchy (LEACH)

protocol, generally with at least the minimum
separation distance when all CHs have been chosen
and separated.


Asymmetric communication and routing in sensor
network (AROS) architecture is used for simulation
purpose [6] [7] [8].


The data is transmitted by CH to base station when
the CH gathers data from its cluster nodes. This is
based on AROS.


Performance Evaluation

Prof. Mohammad S. Obaidat


We have implemented cluster head selection
algorithms for WSN with different protocols LEACH,
MSD and IMSD. The implementations have been
carried out in MATLAB.


The following
assumptions

were made:

1.
Sensors and the base station are all stationary after
deployment; t
he BS is fixed at a distance from the
sensor nodes.

2.
All nodes are homogeneous
and energy constrained
with uniform energy.

1.
Each sensor node initially has the ability to transmit
data
directly to base station.

Performance Evaluation

Prof. Mohammad S. Obaidat

4. Each node is assigned a unique identifier (ID).

5. Nodes can adjust their transmission power according to
the
distance from the receiver.

6. A node belongs to only one cluster, but may change its
cluster
during each round.

7. The nodes have initial uniform energy, of 0.5 Joules.

8. Each sensor node is location aware and the base station
has
strong computation power.

9. All sensors have enough energy to communicate with the
sink
directly at the initial stages.

10. Base station has location and energy info of all nodes.

11. The BS is fixed at a distance from the sensor nodes.

Also, the
Base station has "unlimited" power supply and high calculation
capacity.






Performance Evaluation

Prof. Mohammad S. Obaidat


The node deployment used in the simulation
study is shown in Figure 2.


The nodes are randomly distributed across the
sensor field. The base station is situated inside
the field.


Table I indicates the other parameters as set
during simulation.

Performance Evaluation

Prof. Mohammad S. Obaidat

Proposed Approach



Minimum Separation Distance (MSD)



Minimum Spanning Tree within Cluster


Prof. Mohammad S. Obaidat

Minimum Separation Distance (Algorithm)

MSD


= Minimum Separation Distance

dc = Number of desired cluster heads,

energy(n) = Remaining energy for node n

avg

=


eligible = { n | energy(n) ≥
avg

}

CH = { }

while ( |CH| < dc )






add(
n,CH
)


remove(
n,eligible
)


else



add(
n,CH
)



remove(
n,eligible
)



endif

endwhile




nodes
alive
of
number
n
energy

)
(
MSD
n
m
dist
CH
m
eligible
n
n
If





))
,
(
,
(
:
Prof. Mohammad S. Obaidat

Minimum Separation Distance (Modified Algorithm)

while (( |CH| < dc ) && (gap>10))


while (( |CH| < dc ) && (n<
count|eligible
|)) do









add(n, CH)




endif



endif



n=n+1


endwhile



gap = gap* 0.9



n=2

end_while

then
MSD
n
m
dist
CH
m
eligible
n
n
if





))
,
(
,
(
:
then
CH
in
not
eligible
n
n
if
)
:
(


Prof. Mohammad S. Obaidat

Minimum Separation Distance
(Modified Algorithm)

avg_energy

=

6
.
7


MSD

=

5
m





CH

=

{

}

dc

=

3

N

=

10

eligible={C,

F,

J,

D,

I}

A

G

E

B

I

C

D

F

J

H

10

4

6

8

4

6

8

9

7

5

Prof. Mohammad S. Obaidat

Minimum Spanning Tree






Replacement of direct
communication to MSTs in intra
-
cluster is the main idea.



For large network area using direct
transmissions is not energy efficient.



Average transmission distance of
each node can be reduced by using
MSTs instead of direct transmissions



Thus the energy dissipation of
transmitting data is reduced.


Direct communication in LEACH

Prof. Mohammad S. Obaidat

Minimum Spanning Tree (Contd.)






All nodes including the CH are connected by a
MST


CH as the leader collects the data from the
whole tree.


CHs uses Relay nodes to forward the data to the
BS.


As network area is larger, the reduced
transmission distance is greater.


In the direct transmission, information of
routing path is simple, and each node only
needs to know its CH


But in trees, each node must know the next
node that it would send data to.

MST communication in LEACH

Prof. Mohammad S. Obaidat





Forming Minimum Spanning Tree


Every node in network
broadcasts HELLO message with
a certain frequency.


Nodes receiving HELLO message
respond back with RESPONSE
message with network
-
id,
energy
-
remaining, neighbor
-
id’s,
distance between neighbors.


With this information every node
creates a spanning tree for its
cluster. And informs neighbors
the next node that it would send
data to.



Minimum Spanning Tree (Contd.)

Prof. Mohammad S. Obaidat





d < MSD

d = distance between two node

MST

d ≥ MSD

Proposed Scheme

Prof. Mohammad S. Obaidat

Radio Energy Model





Transmit
Electronics

Transmit
Amplifier

k
E
elec

*

2

*


*

d
k
amp

)
(
d
E
Tx
k bit packet

d

Receive
Electronics

Rx
E
k
E
elec

*

k bit packet

Transmitter

Receiver

Radio Energy Model

Prof. Mohammad S. Obaidat

Radio Energy Model






In our work, we assume a simple model where the radio
dissipates





While estimating the transmit energy, we assumes d
2

energy
loss in channel transmission.



Operation


Energy

Dissipated

Transmitter

Electronics


Receiver

Electronics




50

nJ/bit

Transmit

Amplifier


100

pJ/bit/m
2

)
(
k
E
elec
Tx

)
(
k
E
elec
Rx

)
(
elec
elec
Rx
elec
Tx
E
E
E




amp

Radio Characteristics

Prof. Mohammad S. Obaidat

Radio Energy Model





)
,
(
)
(
)
,
(
d
k
E
k
E
d
k
E
amp
Tx
elec
Tx
Tx




2
*
*
*
)
,
(
d
k
k
E
d
k
E
amp
elec
Tx



)
(
)
,
(
k
E
d
k
E
elec
Rx
Rx


k
E
d
k
E
elec
Rx
*
)
,
(

Thus, to transmit a k
-
bit message a distance d using our radio
model, the radio expends:

(
1
)

and to receive this message, the radio expends:

(
2
)

Prof. Mohammad S. Obaidat

Radio Energy Model






For

these

parameter

values,

receiving

a

message

is

not

a

low

cost

operation
;

the

protocols

should

thus

try

to

minimize

not

only

the

transmit

distances,

but

also

the

number

of

transmit

and

receive

operations

for

each

message
.

Prof. Mohammad S. Obaidat

Performance Evaluation

Prof. Mohammad S. Obaidat

Performance Evaluation

Prof. Mohammad S. Obaidat


Simulation area: Defines the dimensions of the
sensor
field. We considered a field measuring 100 x
100 m
2
.


Number of nodes: The model of sensor network that
was considered consisted of 20/50/100/200/350/500
nodes randomly deployed into the field for 6 different
experiments.


Radio propagation model: Used to predict the
distance
from the signal power of each node. Free
space model and multi path fading model are used in
this simulation.

Performance Evaluation

Prof. Mohammad S. Obaidat


Energy model
-

The energy model, which is a node
attribute, represents the amount of energy in a node.
Energy consumption takes place for each and every
packet transmitted (
tx
-
power) or received (
rx
-
power).


Communication model
-

The communication
between the
nodes is considered one way. A cluster
head advertises itself to the common nodes which in
turn join and transmit data to the CH. Efficient
utilization of the energy resources of the sensor nodes
increases the lifetime of the network. In the ideal
network, all sensor nodes would live exactly the same
period of time.


Assumptions



Sensors and the base station are all stationary after
deployment.



All nodes are homogeneous and have the same
capabilities.



Each sensor node initially has the ability to transmit data
directly to base station.



Each node is assigned a unique identifier (ID).



A node belongs to only one cluster, but may change its
cluster during each round.



Each sensor node is location aware and the base station
has strong computation power.



All sensors have enough energy to communicate with the
sink directly at the initial stages.

Prof. Mohammad S. Obaidat

Simulation tool



In

this

paper,

we

have

analyzed

the

cluster

head

selection

algorithms

for

WSN

with

different

protocols

LEACH,

MSD

and

IMSD
.





We

used

MATLAB

for

the

analysis

which

is

good

for

analyzing

at

the

Physical

layer
.


Matlab

is

a

computing

platform

that

enables

various

simulation

projects
.

It

has

been

particularly

used

for

PHY

layer

study

of

wireless

communications
.


Prof. Mohammad S. Obaidat

Simulation tool (cont…)



What

does

wireless
-
matlab

provide?



Radio

propagation
:

free

space,

two
-
ray,

and

lognormal

shadowing



Mobility
:

random

waypoint

model



PHY
:

broadcast,

dynamic

transmission

rate

and


power



MAC
:

IEEE

802
.
11

(CSMA/CA

and

RTS
-
CTS
-

DATA
-
ACK
)



NET
:

ad

hoc

routing
.

Prof. Mohammad S. Obaidat

Simulation Parameters

ITEM DESCRIPTION

SPECIFICATION


Simulation area

100m x 100m


Channel type

Wireless channels


Radio propagation model

Free space model and Multi path fading
model


Number of nodes

20/50/100/200/350/500


Energy model

Battery


Communication model

Bi
-
direction


Prof. Mohammad S. Obaidat

Results and Analysis


We

have

performed

the

simulation

in

order

to

determine

how

much

we

can

lower

the

energy

consumption

in

the

sensor

network



The

obtained

results

of

IMSD

through

simulation

are

compared

with

the

simulated

results

of

LEACH,

SEP

and

MSD
.


The

metrics

used

are
:

1.

No. of rounds of operations; a o
ne round is specific
time slice (or period) after which the algorithm selects
new cluster heads for next round/period.
So,
we
calculate the total number of rounds the network
sustains with energy dissipation at each round.

2.
Average

energy

consumed,

3
.

No
.

of

nodes

alive,

etc
.



Prof. Mohammad S. Obaidat

Performance Evaluation Results

Prof. Mohammad S. Obaidat


In Figures 3 and 4, we see that a more efficient
utilization of power makes the sensor network stay
alive a longer period of time.


We can also see that as soon as the sensor nodes in
the network start to die, the whole network dies
shortly.


Figure 5, shows the average energy consumed for
every 200 rounds of simulation. It also shows how
IMSD average energy sustains for longest duration.

Performance Evaluation Results

Prof. Mohammad S. Obaidat

Performance Evaluation Results

Prof. Mohammad S. Obaidat

Performance Evaluation Results

Prof. Mohammad S. Obaidat

Concluding Remarks

Prof. Mohammad S. Obaidat


In this paper, we have presented the improved MSD
algorithm; having improved minimum separation
distances between CHs.


We have performed simulations in order to be able to
determine how much we can lower the energy
consumption in the sensor network by separating the
CHs, i.e., by distributing the CHs through the whole
network.


We have shown that using improved minimum
separation distance between CHs improves energy
efficiency, and also the life
-
time of the network.

Concluding Remarks

Prof. Mohammad S. Obaidat


By using IMSD algorithm between CHs we make the
network live longer.


The number of clusters used together with the IMSD
affects the energy consumption.


Future extensions of this work includes more
thorough analysis in more scenarios with varying
numbers of sensor nodes and network sizes, as well as
evaluating alternative algorithms for cluster head
selection.


A comparison between the improved minimum
separation distance algorithm and the other available
algorithms will be
considered in the future.

Thank You

53

Prof. Mohammad S. Obaidat