You are on page 1of 10

An Energy Efficient Cooperative Clustering

Protocol for Heterogeneous Wireless Sensor


Networks
Md. Abdullah-al Mamun,

Naoshi Nakaya, Yuji koi, Yukari Hagihara
Abstract Topology control in a sensor network balances load on sensor nodes, and increases network scalability and lifetime.
Clustering sensor nodes is an effective topology control approach. In this paper, we developed a low complexity cooperative
diversity clustering protocol for heterogeneous wireless sensor networks. Our proposed approach does not make any
assumptions about the presence of infrastructure or about node capabilities, other than the availability of multiple power levels
in sensor nodes termed as node heterogeneity. This heterogeneity may result from initial setting or uneven power draining
during network operation. In contrast to these approaches, we propose a heterogeneous-aware cooperative clustered (HACC)
scheme for wireless sensor networks based on weighted election probabilities of each node to become a cluster head according
to its residual energy. We then propose a simple modification in the clustering algorithm to exploit virtual MIMO based
cooperative transmission. In place of selecting a single cluster head at network layer, we offered multiple cluster heads in each
cluster to obtain a full diversity gain over long distance communication without sacrificing transmission data rate. Thus we
established virtual MIMO based cooperative transmission in ad-hoc sensor networks. We treated the cooperative cluster head
selection method as a Facility Location Problem (FLP) to gain better energy utilization. Analysis and simulation results show that
cooperative HACC can save a huge amount of energy over the existing clustering protocols, and increase the network lifetime
in terms of both 50% node to die and first node to die. Moreover, this proposal can achieve higher order diversity gain with
improved spectral efficiency when the distance to Base Station is over a given threshold.
Index TermsWireless Sensor Network (WSN), topology control, Base Station (BS), Cluster Head (CH), Cooperative Cluster
Head (CCH), Facility Location Problem (FLP).



1 INTRODUCTION
ENSOR networks are a particular type of ad hoc net
work[1],[3]inwhichalargenumberofsensornodes
aredenselydeployedinsideacertainphenomenonor
environment.ThesensornodessendsenseddatatoBSin
order to monitor the changes in the environment from
timetotime.Networklifetimecanbedefinedasthetime
elapsed until the first node (or the last node) in the net
workdepletesitsenergy(dies).Forexample,inamilitary
field where sensors are monitoring chemical activity, the
lifetimeofasensoriscriticalformaximumfieldcoverage.
Prolonging the network lifetime depends on efficient
management of sensing node energy resource. Energy
consumption is therefore one of the most crucial design
issues in WSN [3], [21]. Energy consumption in a sensor
node can be attributed to either useful or wasteful
sources. A number of protocols have been proposed to
reduce useful energy consumption. The analytical results
for LEACH-type clustered schemes are obtained assum-
ing that the nodes of the sensor network are equipped
with the same amount of energythis is the case of homo-
geneous sensor networks [4]. In this paper, we assume that
a percentage of the node population is equipped with
more energy than the rest of the nodes in the same net-
work this is the case of heterogeneous sensor networks.
As the lifetime of sensor networks is limited there is a
requirement to re-energize the sensor network by time-to-
time adding more sensors during operating period of the
network. These sensor nodes will be equipped with more
energy than the nodes that are already in use, which
creates heterogeneity in terms of node energy [8].
Moreover, virtual MIMO (Multi-Input Multi-Output)
systems are more energy efficient than SISO (Single-Input
Single-Output) systems considering both transmission
energy and circuit energy consumption in Rayleigh fad-
ing channels where there is no line of sight between
transmitter and receiver [11]. Due to the physical size and
energy limitation of sensor nodes, directly deployment of
MIMO in one sensor node is infeasible in practice. How-
ever, through sensor nodes cooperation, virtual MIMO
technique can be implemented in sensor network. For
example, a virtual MIMO based on space-time block code
(STBC) is proposed in [13], where the training overhead
needed for the MIMO transmission is considered. A dif-
ferent scheme, multi-hop with MIMO transmission, is
proposed in [22], where multi-hop MIMO transmission is
used to save communication energy consumption be-
tween cluster heads (CHs) and the base station (BS).


- M.A. Mamun is with the Graduate School of Engineering, Iwate Universi-
ty, Japan, 020-8551
- N.N. is with the Graduate School of Engineering, Iwate University, Japan,
020-8551
- Y.K. is with the Graduate School of Engineering, Iwate University, Japan,
020-8551
- Y.H. is with the Graduate School of Engineering, Iwate University, Japan,
020-8551

S
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 28

In situations where the BS is far from the sensor area,
since the energy dissipation for long haul transmission
dominant the energy consumption, multi-hop MIMO
transmission is unnecessary. All these motivate us to pro-
pose a new WSN communication protocol for prolonging
the lifetime of a network model organized by heterogene-
ous sensor nodes with multiple power level where the
distance to Base Station is over a given threshold.
Contribution: In our algorithm, heterogeneous sensor
nodes (nodes with multiple power levels) organize them-
selves into clusters to form hierarchy for power-aware
communication. One node is selected as cluster head
(CH) in each cluster. The election probabilities of CHs are
weighted by the initial energy of a node relative to that of
other nodes in the network. This provides well distri-
buted energy consumption. We then extend our algo-
rithm to implement space time block code (STBC) based
virtual MIMO cooperative transmission. A simple mod-
ification of distributed clustering algorithm [4], [10] in
network layer is proposed to obtain special diversity in
physical layer. Instead of using only one CH, we propose
L number of CHs within a single cluster. Due to the
broadcast nature of wireless transmission, it is possible
for all CHs within a cluster to receive the same transmis-
sion. After receiving data from the cluster members, L
CHs cooperatively transmit them towards the BS.
After clustering of sensor nodes, all non-CHs (cluster
member in each cluster) perform cooperative cluster
heads (CCHs) selection algorithm to implement vMIMO
technique to achieve better energy savings. CCH selection
algorithm is framed as a Facility Location Problem (FLP),
in which (L-1) numbers of CCHs are selected out of (N/k-
1) numbers of cluster members in each cluster, and then
solved the FLP problem to gain better energy utilization
resulting longer network lifetime. Krivitski et al. [6] men-
tioned that considering only distance is not enough.
Heinzelman et. al. [4] and Santi [1] said that the transmis-
sion energy should include both the sender and receiver
sides. However, in this research we considered the facility
cost, transmission energy of each cluster members that
depends on CHs location, and then trade off these costs
for selecting CCHs.
Paper Organization: The remainder of this paper is
organized as follows. Section 2 briefly surveys related
works. Section 3 presents the network model and states
the problem that we address in this work. Section 4 de-
scribes the detail of proposed protocol and argues that it
satisfies its objectives. In section 5, we analyze the energy
model and energy efficiency of our protocol. Section 6
shows HACC effectiveness via simulations, and compares
it to other existing clustering techniques, e.g., LEACH [4]
and EEHC [8]. Finally, section 7 concludes with directions
of future work.
2 RELATED WORK
Before introducing the proposed algorithm, in this section
we review some closely related prior studies that dealt
with the heterogeneity in energy of sensor nodes and
space time block code (STBC) based vMIMO communica-
tion technique in wireless sensor network. Low-energy
adaptive clustering hierarchy (LEACH) [4], [10] is one of
the most popular hierarchical cluster-based routing pro-
tocols in wireless sensor networks that utilizes rando-
mized rotation of local cluster base station (CH) to evenly
distribute the energy load among the sensors in the net-
works. These sensors organize themselves into clusters
using a probabilistic approach to randomly elect them-
selves as heads in an epoch. The algorithm is run periodi-
cally, and the probability of becoming a cluster head for
each period is chosen to ensure that every node becomes
a cluster head at least once within 1/P rounds, where P is
the desired percentage of cluster heads. This ensures that
none of the sensors is overloaded because of the added
responsibility of being a cluster head. Conventional pro-
tocols such as Minimum Transmission Energy (MTE) and
Direct Transmission (DT) do not assure a balanced and
uniformly use of the sensors respective energies as the
network evolves.
The first work that questioned the behavior of cluster-
ing protocols in the presence of heterogeneity in clus-
tered wireless sensor networks was [10]. In this work
Heinzelman analyzed a method to elect cluster heads ac-
cording to the energy left in each node. The drawback of
this method is that this decision was made per round and
assumed that the total energy left in the network was
known. The assumption of global knowledge of the ener-
gy left in the whole network makes this method difficult
to implement. Even a centralized approach of this method
would be very complicated and very slow, as the feed-
back should be reliably delivered to each sensor in every
round. In [2], Mhatre and Rosenberg presented a cost-
based comparative study of homogeneous and heteroge-
neous clustered wireless sensor networks. They proposed
a method to estimate the optimal distribution among dif-
ferent types of sensors, but again this result is hard to use
if the heterogeneity is due to the operation of the net-
work.
It is well known that transmit diversity technique is
energy efficient transmission protocol. To provide trans-
mit diversity when users cannot support multiple anten-
nas, a new method of transmit diversity for mobile users,
termed as cooperative diversity has been proposed [9].
Various cooperative transmission protocols, their imple-
mentation issues, performance and outage analysis have
been studied in literature [9], [12]. Constructing a virtual
MIMO environment in a distributed wireless sensor net-
work is a great challenge for the research community. The
major challenges are synchronization among the coopera-
tive nodes, delay efficiency, processing overhead, ex-
changing information among the cooperating nodes etc.
Various cooperative MIMO protocols for clustering based
WSN have also been proposed in literature [13], [14], [15],
[16], and [17]. X. Li et al. [13] proposed a STBC-encoded co-
operative transmission scheme for WSNs without perfect syn-
chronization. However, these schemes only involve one hop
MIMO transmission and do not consider specific application of
MIMO in WSN.
In [8], the authors have considered cluster-based archi-
tecture for heterogeneous sensor networks. Their pro-
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 29

posed model consists of two level of hierarchy where the
Base Station is near or inside the network area. The simu-
lation experiments in [8] shows that the algorithm works
fine in a network model with low percentage (10%) of
heterogeneity where the heterogeneous nodes energy is
one or two times higher than normal nodes. But, the algo-
rithm in [8] yields a large unstable region in a network
model with high percentage of heterogeneity where the
heterogeneous nodes energy is higher than two or three
times of low energy nodes. In this paper, we have used a
network model that relaxed the constraints of heterogene-
ity analytically, and then extend our clustering algorithm
to generate vMIMO based cooperative transmission.
Here, we introduce a cross layer approach to obtain high-
er diversity without sacrificing any spectral efficiency. In
the proposed scheme, instead of using cluster member as
cooperative nodes, multiple cluster heads cooperate to
form virtual antenna array so that STBC based MIMO
technique can be implemented to achieve energy efficien-
cy.
The FLP describes how to offer the best service facili-
ties for a large number of cities where each facility has a
cost, and each city has a service cost. The goal is to choose
a subset of all facilities connected with each city that mi-
nimizes the total cost. We introduce the FLP in coopera-
tive cluster head (CCH) selection phase, where CHs are
regarded as cities, and CCHs as facilities. Previous re-
searches on FLP include solution by using Hill Climbing
Heuristic or linear programming [6], [7], [18], [19], [20].
Krivitski et al. [6], [7] considered FLP in sensor networks
by using Hill Climbing method to attack the problem
with the transmission distance and data importance as
main factors of costs. Their algorithm is to select k CHs
from m unmovable CHs. As Hill Climbing method has
always been trapped in local optimal solution, a heuristic
using majority voting mechanism among neighbor sensor
nodes was introduced to give it a better chance to escape
from local optimal solutions
3 NETWORK MODEL
We consider sensor networks with three types of hete-
rogeneous sensor nodes, which collect information and
send them to a BS as shown in figure 1. We assume that
the nodes are randomly distributed over the sensor field
and they form clustering hierarchy for the convenience of
communication. Each cluster contains a cluster head
(CH), (L-1) cooperative cluster heads (CCHs) and several
sensor nodes (SNs). We define the intra cluster transmis-
sion (SNs to CH/CCHs) as local transmission and
CH/CCHs to BS transmission as long-haul transmission.
For channel propagation model, we consider both the free
space model and multi path-fading model [10] depending
on the distance between the transmitter and the receiver.
Let us consider the case where a percentage of the
population of sensor nodes is equipped with more energy
resources than the normal sensor nodes in the network.
Let
adv
m be the fraction of the total number of nodes N,
which are equipped with times more energy than the
node with low energy termed as normal nodes. We refer
to these powerful nodes as advanced node, and
int
m

is the
proportion of intermediate nodes to the total number of
nodes N with | times more energy than the normal nodes.
The rest of N m m
adv
) 1 (
int
nodes deployed as normal
nodes with initial energy
0
E . So the initial energy of powerful
advanced nodes,
0
) 1 ( E E
adv
+ = o , and for intermediate
nodes,
0 int
) 1 ( E E + = | . We assume /2 o | = .


Fig.1 Network model of cooperative HACC protocol in the presence
of three heterogeneous sensor nodes with one CCH.

Local Transmission: In this case, the distance between
the transmitter and the receiver is relatively small com-
pare to the long-haul transmission. The local communica-
tion channels are considered as additive white Gaussian
noise (AWGN) channel with free space propagation mod-
el. Hence, the baseband equivalent received signal at
node j due to the transmission of node i for symbol n is
given by,
) ( ) ( ) (
) ( ) (
n n s n r
j local ij local j i
q t + = (1)
Where, ) (n
j
q is AWGN sample with N0/2 per dimension
at terminal j,
2
) (

=
ij local ij
d t with
ij
d is the distance
between node i and j, and ) (n s is the signal transmitted
bynodeiwithnormalizedunittransmitpower.
Long-haul transmission: We consider the base station
(BS) is far away from the sensor field. Hence, for long-
haul communication channel the distance between the
transmitter and the receiver is relatively long. We consid-
er the channels between CH/CCHs to BS are subjected to
flat Rayleigh fading with two ray ground propagation
model plus AWGN. The baseband equivalent received
signal at node j due to the transmission of the node i for
the symbol n is given by,
) ( ) ( ) (
) ( ) (
n n s n r
j long ij long j i
q t + = (2)
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 30

Where,
ij
h is the fading coefficient between node i and j
and
4
) (

=
ij long ij
d t . We consider flat Rayleigh fading,
hence
ij
h ismodeledasindependentsamplesofzeromean
complex Gaussian random variable with variance 2 /
2
ij
o
per dimension. Assume that the fading coefficients are
constantoverthetransmissionperiodofawholemessage
block. Because of slow fading channel estimation is
possible at the receivers, and we assume that perfect
channel state information is available at the
correspondingreceiversbutnotinthetransmitters.
4 PROTOCOL DESCRIPTION
In this section we describe our proposed HACC, which
improves the stable region of the clustering hierarchy
process using the characteristic parameters of heterogene-
ity, namely the fraction of advanced nodes (
adv
m ), inter-
mediate nodes (
int
m ) and the additional energy factor
between normal, advanced and intermediate nodes
( | o, ). In order to prolong the network lifetime, HACC
attempts to maintain the constraint of well-balanced
energy consumption. Intuitively, advanced nodes and
intermediate nodes have to become CHs, as well as CCHs
more often than the normal nodes, which is equivalent to
a fairness constraint on energy consumption.

Fig.2 Operation flowchart of an energy efficient cooperative cluster-
ing protocol for wireless sensor networks

The operation flowchart of our scheme is shown in
figure 2. The proposed cooperative HACC protocol oper-
ates in round by round fashion and each operation round
divides into two phases, namely the setup phase and the
steady state phase. In the setup phase, each node decides
whether to become a CH for current round or not. The
decision to become a CH depends on a prior percentage
of cluster heads and the number of times the node has
been a CH, CCH so far. After the clusters are formed,
each CH schedules the sensor nodes within its cluster
through time division multiple access (TDMA) scheme.
After the end of cluster formation, all non-CH nodes per-
form the selection of additional CHs termed as Coopera-
tive Cluster Head to make the long haul transmission
more secure and robust. After the selection of CCHs, each
node informs the selected CH node that it will be a mem-
ber of the cluster. At the same time, it informs whether it
will be a CCH or not.
The overhead of this procedure is just transmitting one
extra bit along with the cluster-joining packet. After-
wards, the sensor network enters the steady state phase,
where it begins sensing and transmitting data to the base
station incorporating virtual MIMO based cooperative
transmission. Each operation round is assigned a fixed
amount of time. After reaching the time limit, the sensor
network will enter another round by re-clustering sensor
nodes and reconfiguring CCHs.

4.1 Setup Phase

4.1.1 Weighted Election Probability of CH and Cluster
formation Algorithm
Since the CHs and CCHs are mainly responsible for data
aggregation and transmission to BS, the way how clusters
form and how CHs and CCHs are selected directly im-
pact the energy dissipation. The LEACH [4, 10] architec-
ture deals with cluster formation and CH selection prob-
lem so that it meets the requirements of low energy dissi-
pation and good load balance. The same demand comes
to network layer of our proposed protocol.
We employed distributed algorithm as in [4, 10], for
clustering sensor nodes and selecting CHs. Here, the sen-
sor nodes elect themselves as cluster heads with respect
to their energy levels autonomously. Each node transmits
data to its closest CH and the CHs perform data aggrega-
tion. Assume an optimal number of clusters k in each
round. Each node can become cluster head with a proba-
bility p
opt
and every node must become cluster head once
every 1/p
opt
rounds. Throughout this paper, we refer to
this number of rounds, 1/p
opt
as epoch of the clustered sen-
sor network. On average, we have N p
opt

clusters and
CHs per round per epoch. Nodes that are elected to be
CHs in the current round can no longer become CHs in
the same epoch. The non-elected nodes belong to the
set G , and in order to maintain a steady number of CHs
per round, the probability of nodes G e to become a clus-
ter head increases after each round in the same epoch.
The decision is made at the beginning of each round by
each node G ne independently choosing a random num-
ber between 0 to 1. If the random number is less than a
threshold for node n, ) (n T , then the sensor node becomes
a cluster head in the current round.
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 31

As we explained in figure 1, all heterogeneous nodes
are distributed over the sensor field and the initial energy
of normal nodes, powerful advanced nodes and interme-
diate nodes are assumed as follows:

+ =
+ =
=
0 int
0
0
) 1 (
) 1 (
E E
E E
E E
adv
nrm
|
o
(3)
Here we also assume /2 o | = . However, the total ini-
tial energy of the system is increased by the introduction
of both powerful advanced and intermediate nodes:
) 1 (
) 1 (
) 1 ( ) 1 (
int 0
0 int
0 0 int
| o
|
o
+ + =
+ +
+ + =
m m E N
E m N
E m N E N m m E
adv
adv adv total
(4)
The overall energy of the network is increased by a
fraction of ) 1 (
int
| o + + m m
adv
. In order to optimize
the stable region of the system, the new epoch must be-
come equal to
) 1 ( 1
int
| o + + m m p
adv opt
, because the sys-
tem has ) (
int
| o + m m
adv
times more energy than the
homogeneous settings. Note that the new heterogeneous
setting (with powerful and normal nodes) has no effect on
the spatial density of the network so the priori setting
of
opt
p does not change. The number of CHs per round
per epoch will also be constant at N p
opt
. If the same
threshold is set for both normal and powerful (advanced
and intermediate) nodes then there is no guarantee that
the number of cluster heads per round per epoch will be
same as N p
opt
. So the constraint of N p
opt
cluster heads
per round is violated.
Our approach is to assign a weight to the optimal
probability p
opt
. This weight must be equal to the initial
energy of each node divided by the initial energy of the
normal node. Let us define as p
nrm
the weighted election
probability for normal nodes, p
adv
the weighted election
probability for the advanced nodes, and p
int
the weighted
election probability for the intermediate nodes. Virtually
there are ) 1 (
int
| o + + m m N
adv
nodes with energy
equal to the initial energy of a normal node.
In order to maintain the minimum energy consump-
tion in each round within an epoch, the average number
of CHs per round per epoch must be constant and equal
to N p
opt
. In heterogeneous scenario the average number
of CHs per round per epoch is equal to
) 1 (
int nrm adv
p m m N + + | o (because each virtual node
has the initial energy of a normal node.) The weighed
probabilities for normal, advanced and intermediate
nodes are set to as, respectively:
| o + +
=
int
1 m m
p
p
adv
opt
nrm
(5)
) 1 (
1
int
o
| o
+
+ +
=
m m
p
p
adv
opt
adv
(6)
) 1 (
1
int
int
|
| o
+
+ +
=
m m
p
p
adv
opt
(7)
ToguaranteethatthesensornodesmustbecomeCHs
as we have assumed above, we must define a new thre
shold for the election processes. The threshold T(nnrm),
T(nint), T(nadv) for normal, intermediate and advanced
noderespectivelybecomes:

' e

=
Otherwise 0
G if
)
1
( 1
) (
nrm
nrm
nrm
nrm
nrm
n
p
mod r p
p
n T
(8)
Where r is the current round, G' is the set of normal
nodes that has not became cluster heads within the last
nrm
p
1
rounds of the epoch, and T(n
nrm
) is the threshold ap-
plied to a population of ) 1 (
int
m m N
adv
, normal nodes.
The same analogy follows for the intermediate and ad-
vanced nodes,

' ' e

=
Otherwise 0
G if
)
1
( 1
) (
int
int
int
int
int
n
p
mod r p
p
n T
(9)

' ' ' e



=
Otherwise 0
G if
)
1
( 1
) (
adv
adv
adv
adv
adv
n
p
mod r p
p
n T
(10)
Fromequations(5),(6),and(7),theaveragetotalnumber
ofclusterheadsperroundwillbe:
opt adv adv int int nrm adv
p N p m N p m N p m m N = + + ) 1 (
int

(11)

4.1.2 CCH Configuration Algorithm
After the end of CH selection, our algorithm performs the
CCH configuration. CCH configuration problem is consi-
dered as Facility Location Problem (FLP), the goal is to
find a good solution to incorporate virtual MIMO based
cooperative transmission for the long-haul communica-
tion, and achieve better load balance among the hetero-
geneous sensor nodes. We associate each non-CH node
with two quantities: transmit energy and remaining ener-
gy. Using the cost function we designed, we are able to
select(L1)CCHsoutof(N/k1)numbersofclustermem
bers(SNs)ineachclusterbasedonpredefinedvalueofL.
The facility cost of each nonCH is defined as ejBS/ej,
where ejBSindicates transmits energy from SNjto BS, and
ejindicatestheremainingenergy of
j
SN .ejBS/ej gives us the
informationabouttheimpactleveloftransmissiononthe
remainingenergy.Ifthevalueishigh,i.e.,thecostishigh,
itmeansthatthetransmissionwillhaveagreatimpacton
remaining energy, and we show less favor on selecting
suchsensornode(clustermember)asCCHs.Astheresi
dualenergyofnormalnodesissmallercomparetothatof
powerful nodes, the facility cost is likely to increase be
causeeverytransmissionbecomesmoreandmoreexpen
sive relative to the remaining energy of normal SNs
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 32

(nodeswithlowenergy).Thefacilitycostfornormalnode
canbecalculatedasfollows,
jo
jBS
nrm
e
e
FC = (12)
Where ejBS indicates transmits energy from SNj to BS
and
jo
e indicates the residual energy of normal sensor
nodes. Since, advanced and intermediate nodes consist
and times more energy than the normal nodes, the facil-
ity cost (FC) for powerful non-CH nodes is lower than the
normal non-CH nodes.
) 1 ( o +
=
jo
jBS
adv
e
e
FC
(13)
) 1 ( | +
=
jo
jBS
int
e
e
FC
(14)
Where
adv
FC and
int
FC define the facility cost for ad-
vanced non-CH nodes and intermediate non-CH nodes
respectively. Because of
nrm int adv
FC FC FC < < , powerful
advanced and intermediate nodes perform the duty of
CCHs more often than the normal nodes. By considering
the facility cost, a proper load balance can be estimated
between normal nodes and powerful nodes to reduce the
total energy dissipation. Through the consideration of
transmission impact as defined by facility cost, we can
reach a balanced energy distribution compared to that of
considering mainly on average remaining energy. This is
because when the average remaining energy is the only
considered factor, some CCH may be selected to be active
due to higher remaining energy level. However, if the
transmit distance of the chosen CCH is long, i.e., the
transmit energy dissipation is high, the choice may not be
the best one. Our system will favor a low facility cost;
even the remaining energy level might not be the highest.
Therefore, (L-1) CCHs are elected out of (N/k1) num
bersofclustermembers(SNs)with lower value of facility
cost. If a pair of CH/CCH is in the same place or very
close to each other (less than the half wavelength) then
the system will not achieve full diversity gain. CH can
avoid this situation by setting a predefined threshold val-
ue of facility cost. Depending on the number of nodes in
the cluster, the CH creates a TDMA schedule and sends
this schedule to the CCHs and all other sensor nodes. At
the same time, the CH informs the cooperative CH selec-
tion. As soon as the CCHs received the schedule from the
CH, they will broadcast an acknowledgement signal to-
wards all sensor nodes. The SNs can adjust their trans-
mission power according to the strength of the broadcast
signals from CH & CCHs. Therefore, CH & CCHs can
receive the local transmission reliably. The overhead of
this procedure is about transmitting few extra bits along
with the TDMA schedule.

4.2 Steady State Phase
The sensor nodes start transmitting data and due to the
broadcast nature of wireless transmission, CHs and all
the CCHs will receive these transmissions. Similar to
LEACH protocol we consider all the sensor nodes trans-
mit information in their allocated time slot. We also con-
sider that the neighboring clusters are using different or-
thogonal channels to avoid the inter-cluster interference.
After receiving the information from all sensor nodes
CHs and CCHs perform data aggregation and some other
signal processing if necessary, prior to transmit them to
the BS using the proper signaling structure of cooperative
transmission. After a predefined period, the sensor net-
work will go into another new round for re-clustering
and re-configuration. If a sensor node is dead somehow,
the node will be excluded from network in the next
round.

4.3 Cooperative Transmission
Inconventionalcooperativetransmissionprotocols[5],[9],
[12], [17] relay terminals have to process their partners
received signals. However, current limitations in radio
implementation preclude the terminals from full duplex
operation, i.e., transmitting and receiving at the same
timeandfrequencyband[9];thus,conventionalcoopera
tive protocols ensure halfduplex operation. But, our
crosslayer design approach overcomes this major limita
tion of cooperation technique. In this proposal, CH and
CCH are gathering data independently from the sensor
nodes at the same time. Therefore, CH and CCH do not
requireexchanginginformationamongthemforcoopera
tion. Existing virtual MIMO based cooperative protocols,
for examples [11], [13], [14], [15], and [16] require ex
changinginformationamongthecooperatingnodes.This
fact allows us to use a full rate transmission similar to
direct transmission using LEACH. We propose a virtual
MIMO communication architecture with distributed
space time block code (DSTBC) where CH and CCH
transmitatthesametimeandsamefrequency[17].Hence,
our proposal can offer same diversity order as existing
protocols [11], [13], [14], [15], and [16] with higher spec
tralefficiency.Incaseofperfectlocalcommunication,the
receivedsignalofDSTBChastheformofMbranchmax
imal ratio combining (MRC). The bit error rate (BER)
probability of DSTBC signaling structure for BPSK mod
ulation in terms of longhaul signal to noise ratio (SNR)
derivedin[23]andcanbegivenas,
| |
k
L
k
L
k
k S
Pb
f
f
) 1 (

) 1 (
)] 1 ( [
2
1
1
0
2
1
+
|
|
.
|

\
| +
=

=
(15)

Where, long long 1 + = with average long-
haul received SNR long . Numerical BER expression of
equation (15) is also verified with simulations for cooper-
ative HACC with L=2. Simulation and numerical results
show that proposed cooperative HACC protocol outper-
forms the LEACH [4] and EEHC [8]. For simplicity, we
explain the BER performance for BPSK modulation only.
Similar results for other modulations are available in the
literature [23] and can be easily adapted with our propos-
al.
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 33

5 ENERGY MODEL
Wireless sensor networks are highly energy limited;
therefore, we need to consider all the sources of energy
consumption while comparing the protocols. The total
energy consumption of proposed HACC protocol can be
dividedintothreemajorparts:theenergyconsumptionin
cluster setup, energy consumption in local transmission
and energy consumption in longhaul transmission. For
simplicity of analysis, we include cluster setup energy in
local communication energy. Each of the energy compo
nents can be divided into two parts: transmission energy
and the transmitter & receiver circuit energy dissipation.
The transmission energy is the energy required at the
transmit amplifier to achieve an acceptable SNR and the
circuit energy dissipation is the energy dissipated by the
transmitterandreceiverelectronics.Thetotalenergyequ
ationisgivenby,
long c long t local c local t
long local
E E E E
E E E

+ + + =
+ =

(16)
where, the subscripts c represents circuit energy dissipa
tion, t represents transmission energy, long represents
longhaul communication and localrepresents local com
munication.
Localcommunication:A free space propagation model
with AWGN is considered for local communication. The
receivedsignalsattheCHsandCCHsaredefinedbyeq
uation (1). According to the Friis free space equation [23]
the received power at the receiver antenna can be given
as,
f
r t t
r
L d
G G P
d P
2
2
) 4 (
) (
t

= (17)
Where,Ptisthetransmitpower,Gt&Graretransmitting&
receiving antenna gain, is the carrier wave length, d is
the distance between transmit and receive antenna and
f
L is the system loss factor not related to propagation.
Forthelocalcommunication,weconsiderthatthesensor
nodes can adjust their transmit power by estimating the
distance. If the bit rate of the system is Rb, the transmit
powercanberepresentedas,

2
d R P
b local mp t
=

c (18)

Where, mplocalis the energy required at the transmit am


plifier to achieve an acceptable SNR at the receiver. Now
fromEqns.(17)and(18)wecanget,

2
2
_
) 4 (

t
c
r t b
th r
local mp
G G R
P
=

(19)
Where, Pr_th is the threshold value of the received power
for error free reception at the receiver. We consider the
same model proposed inLEACH protocol. In [4],the au
thors assumed that for error free reception destination
requires a SNR of 30 dB with a noise figure of 82 dBm.
Hence, Pr_th_local > 30 82 = 52 dBm6.3 nW . Considering
thesametrendof[4],i.e.,Gt=Gr=1,=0.328,Lf=1and
Rb=1Mbps,wecancalculateas,mplocal=10pJ/bit/m
2
.
Longhaulcommunication:Forlonghaulcommunication,
we consider a two ray propagation model and the re
ceivedsignalsattheCHsandCCHsaredefinedbyequa
tion (2). According to the tworay ground propagation
equation [23], the received power at the receiver antenna
canbegivenas,

4
2 2
) (
d
h h G G P
d P
r t r t t
r
= (20)
Where, ht and hr are the height of the transmitting and
receivingantennaaboveground.Ifthebitrateofthesys
temisRb,thetransmitpowercanbewrittenas,
4
d R P
b long mp t
=

c (21)
Where,mplongistheenergyrequiredatthetransmitam
plifiertoachieveanacceptableSNRatthesinknode.
NowfromEquations(20)and(21)wecanget,
2 2
_
r t r t b
th r
long mp
h h G G R
P
=

c (22)
Where,Pr_thisthethresholdvalueofthereceivedpower.
Incaseofthelonghaultransmission,weconsiderthatthe
channelsaresubjectedtotheRayleighfading.Hence,the
adjustmentofthetransmitpowertoachieveanerrorfree
communication requires channel state information (CSI)
at the transmitting nodes (CHs and CCHs) which is very
difficult to obtain in practical case. In this paper, we as
sume that CSI is available only at the corresponding re
ceivers. Therefore, we consider that the CHs and CCHs
transmit data with an average power to achieve an arbi
trarysmallBERatthesinknode.From oursimulation,it
isclearthattheperformanceoftheproposedHACCpro
tocolincreaseslogarithmicallyasthetargetBERdecreases.
We chose a moderate value of 10
4
as target BER for our
energy analysis. From our analysis we found that a BER
of10
4
canbeachievedbytheLEACHandEEHCat35dB
and HACC protocol at 20 dB SNR for L = 2. The corres
pondingreceivedpowerfortheLEACHprotocolscanbe
givenas,
EEHC LEACH
long th r
P
/
_ _
>35+(82)=47dBm20nW.Now
considering the similar parameters of local communica
tionalongwithht=hr=1m,wecanget,
4 /
_
/ / 02 . 0 m bit pJ
EEHC LEACH
long mp
~ c
Usingthesimilarapproach,wecancalculatetheener
gy required at the transmit amplifier for the proposed
HACC protocol corresponding to the SNR required to
achieveaBERof10
4
asfollows,
2 L for / / 00063 . 0
4
_
= ~ m bit pJ
HACC
long mp
c
The circuit energy dissipation for local and longhaul
communication, Eclocal and Eclong are modeled similarly as
[4] and given in Table 1. We also consider data aggrega
tionmethodatbothCHsandCCHsthatcancompressthe
datawitharatioofr:1[3].
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 34

TABLE 1
Simulation Parameters
6 PERFORMANCE EVALUATION
We perform a clustered wireless sensor network simula-
tion to evaluate the performance of HACC protocol with
both of EEHC [8] and LEACH [4]. MATLAB is used to
implement the simulation. Assume all sensor nodes have
data of block size 4000 bits. We consider 100 node net-
works as shown in figure 1. Here the nodes are randomly
distributed over the area M M , and BS node is located Y
meters from the nearest sensor node (x=M/2, y=Y). We
consider each normal node initially starts with 0.5 Joule of
energy. Each advanced and intermediate node consist
(1+) and (1+) times energy than the initial energy of
normal sensor node respectively. In our simulation set up,
we considered multiple power level of sensor nodes
termed as heterogeneity of nodes energy. We assumed
15% of the nodes be advanced nodes, and 15% of the
nodes be intermediate nodes with additional energy fac-
tor 5 . 1 and 3 = = | o respectively. We set our parameter
as m
adv
=0.15, =3, and m
int
=0.15, =1.5.
Here, we compare the energy required for local com
munication of cooperative HACC protocol with LEACH
and EEHC. Local communication includes the cluster
setup procedure and proposed HACC protocol introduc
es some overhead for selecting the CCHs and synchroni
zation among CHs and CCHs. To compensate this over
head we consider cooperative HACC protocol transmits
extra 200 bits along with the data packet of 4000 bits in
localcommunication.Figure3showstheamountofener
gy dissipated in the local communication for HACC,
EEHC [8] and LEACH [4] with the same heterogeneous
settings. Simulation result indicates that cooperative
HACC protocol consumes higher energy than that of the
EEHC and LEACH in the local communication. The
sources of this extra energy consumption are the in
creased local communication distance and the extra re
ceiving operations of CCHs. This extra power consump
tionofthecooperativeHACCprotocolinthelocalcom


Fig.3 Comparison of local communication energy consumption for
HACC, EEHC [8] and LAECH [4] in the presence of heterogeneity
with m
adv
=15%, =3, and m
int
=15%, =1.5.

Fig.4 Comparison of long-haul communication energy consumption
for HACC, EEHC and LAECH in the presence of heterogeneity with
m
adv
=15%, =3, and m
int
=15%, =1.5.

Fig.5 Average total energy dissipation per round for HACC, EEHC
and LAECH in the presence of heterogeneity with m
adv
=15%, =3,
and m
int
=15%, =1.5.
0 5 10 15 20 25 30 35 40 45 50
0.02
0.03
0.04
0.05
0.06
0.07
0.08
Number of Rounds
T
o
t
a
l

E
n
e
r
g
y

D
i
s
s
i
p
a
t
i
o
n

i
n

E
a
c
h

O
p
e
r
a
t
i
n
g

R
o
u
n
d

(
J
)


HACC (L=2)
EEHC (L=1)
LEACH (L=1)
0 100 200 300 400 500 600 700 800 900
0
5
10
15
20
25
30
Time in Rounds
A
v
e
r
a
g
e

L
o
c
a
l

E
n
e
r
g
y

D
i
s
s
i
p
a
t
i
o
n

i
n

J
o
u
l
e
s

(
j
)
Energy Dissipation for Local Communication


HACC (L=2)
EEHC (L=1)
LEACH (L=1)
0 100 200 300 400 500 600 700 800 900
0
2
4
6
8
10
12
14
16
18
20
Time in Rounds
A
v
e
r
a
g
e

L
o
n
g
-
h
a
u
l

E
n
e
r
g
y

D
i
s
s
i
p
a
t
i
o
n

i
n

J
o
u
l
e
s

(
J
)
Energy Dissipation for Long-haul Communication


HACC (L=2)
EEHC (L=1)
LEACH (L=1)
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 35

municationisnegligiblecomparedwiththediversitygain
achieved in the longhaul communication with reduced
energy consumption shown in figure 4. For cooperative
HACC protocol the target BER 10
4
can be achieved at 20
dB SNR which is much more lower than LEACH and
EEHC. This ensures that the energy dissipation for long
haul data transmission for cooperative HACC can be re
duced10timesthanthatofLEACHandEEHC,shownin
figure4.Figure5representstheaveragetotalenergydis
sipation in each operating round. It depicts that total
energy consumption for LEACH and EEHC is two times
higher than that of cooperative HACC in almost every
round.Itresultslongernetworklifetime.
Fig.6 Comparison of network lifetime between HACC, EEHC and
LAECH in the presence of heterogeneity with madv=15%, =3, and
mint=15%,=1.5.

Figure6showsthecomparisonofnetworklifetime,in
termsofboth50%nodestodieandfirstnodetodie,ofa
heterogeneoussensornetworksthatuseHACC,EEHC[8]
and LEACH [4] in the presence of same heterogeneous
settings (madv=15%, =3, and mint=15%, =1.5). We define
thenetworklifetimeintotwoterms,stabilityperiod(time
intervalbeforethedeathofthefirstnode)andinsatiabili
ty period (time interval between the death of the first
node until the death of the last alive node). Figure 6 re
veals that not only the system lifetime is extended; the
system also achieves a better load balance. The first sen
sor node died at about 2570
th
rounds for the proposed
cooperative HACC protocol, which is longer than 661
th
rounds of EEHC and 469
th
rounds of LEACH. It is ob
vious that proposed HACC can significantly increased
thestabilityperiodmorethan2.883timescomparewith
EEHC and more than 4.47 4 times than LEACH in
termsofboth50%nodestodieandfirstnodetodie.This
improvement in network lifetime is achieved due to the
diversity gain of the longhaul transmission. Moreover,
Due to the well balanced energy consumption between
heterogeneous sensor nodes, the instability period (time
intervalbeforethedeathofthelastsensornode)ofcoop
erative HACC is shorter than that of EEHC and LEACH.
It indicates, advanced and intermediate nodes follow the
death process of normal sensor nodes and ensured well
balancedenergyconsumption.
Fig.7 Throughput comparison between HACC, EEHC [8] and
LEACH[4]inthepresenceofheterogeneitywithmadv=15%,=4,and
mint=10%,=2.

Figure 7 shows the throughput, received amount of
data packets in Kbits, from sensor nodes to Base Station.
We observe that the throughput of HACC is significantly
larger than EEHC [8] and LEACH [4] in the stable region
and for most of the unstable region. Our simulation re-
sults conforms that cooperative HACC protocol can re-
duce huge energy consumption with the same bit error
rate, spectral efficiency and delay requirements compared
with other existing protocol both of homogeneous and
heterogeneous. Saving the energy equivalently prolongs
the network lifetime which is a prime performance crite-
rion of WSN.
7 CONCLUSION
In this paper, a cluster-based virtual MIMO transmission
scheme is proposed to prolong the network lifetime of
heterogeneous WSN. In proposed protocol, we
considered multiple energy levels of sensor nodes, and
instead of using cluster member as cooperating nodes,
multiple CHs cooperate to form virtual antenna array so
that STBC based MIMO technique can be implemented to
achieve transmission energy savings. Based on the
communication energy consumption model, the overall
energy consumption model of the proposed scheme is
derived both for local and long-haul communication.
Moreover, weighted election probability of CH selection,
and FLP based CCH selection method guaranteed the
well balanced energy consumption among heterogeneous
sensor nodes. Numerical and simulation results together
show that the proposed HACC can prolong the sensor
network lifetime greatly when the distance to sink is
aboveathreshold,especiallyinsituationswherethesink
isfarfromthesensorarea.
In this paper, we investigate the network topology
considering single cooperating node (CCH); in future we
intend to explore our analysis with different numbers of
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
0
10
20
30
40
50
60
70
80
90
100
110
Time in Rounds
N
u
m
b
e
r

o
f

N
o
d
e
s

A
l
i
v
e


HACC (L=2)
EEHC (L=1)
LEACH (L=1)
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
0
50
100
150
200
250
300
350
400
450
Time in Rounds
D
a
t
a

i
n

K
b
i
t
s



HACC (L=2)
EEHC (L=1)
LEACH (L=1)
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 36

CCHtofindtheoptimalnumberofcooperatingnodes(L).
ACKNOWLEDGMENT
This research work has been partially supported by the
MEXT: Ministry of Education, Culture, Sports, Science,
and Technology, Monbukagakusho (Japanese govern-
ment) scholarship.
REFERENCES
[1] PaoloSanti,TopologyControlinWirelessAdHocandSensor
Networks,IstitutodiInformaticaeTelematicadelCNR,Italy2005.
[2] V.MhatreandC.Rosenberg,Homogeneousvs.heterogeneous
clustered sensor networks: A comparative study, in Proceed
ings of 2004 IEEE International Conference on Communications
(ICC2004),June2004.
[3] I.F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayir
ci,Asurveyonsensornetworks.IEEEcommunicationMaga
zine,vol.40,no.8,Aug.2002,pp.6873.
[4] W. Heinzelman, A. Chandrakasan,H. Balakrishnan, Energy
efficient communication protocol for wireless microsensor net
works,in:Proceedingsofthe33rdHawaiiInternationalConference
onSystemSciences(HICSS00),January2000.
[5] A. Hedayat, T. E. Hunter, A. Nosratinia, Cooperative Com
municationinWirelessNetworksIEEEcommunicationMaga
zine,vol.42,no.10,Oct.04,pp.6873.
[6] DenisKrivitski,AssafSchuster,andRanWolff,ALocalFacili
ty Location Algorithm for Sensor Networks , DCOSS 2005:
MarinadelRey,CA,USA,pp.368375.
[7] Denis Krivitski, Assaf Schuster, and Ran Wolff, Local Hill
Climbing in Sensor Networks , SIAM International Conference
onDataMining,April2123,2005,pp.3847.
[8] Dilip Kumar, Trilok C. Aseri, R.B. Patel, EEHC: Energy effi
cient heterogeneous clustered scheme for wireless sensor net
works, in: Elesevier, Computer Communication Journal 32
(2009)662667,2009.
[9] J.NLaneman,.D.N.CTse,G.W.Wornell,Cooperativediversi
ty in wireless networks: Efficient protocols and outage beha
viorIEEETrans.onInformationTheory,vol50,Issue12,Dec.
2004Pages:30623080.
[10] W.HeinzelmanApplicationSpecificProtocolArchitecturesfor
Wireless Networks, Ph.D. dissertation, Mass. Inst. Technol,
June2000.
[11] S. Cui, A. J. Goldsmith, and A. Bahai, Energyefficiency of
MIMOandcooperativeMIMOtechniquesinsensornetworks,
IEEEJourn.Select.Areas.Commun,2004,pp.10891098.
[12] T. E. Hunter and A. Nosratinia, Diversity through Coded Co
operation IEEE transactions on wireless communications, vol.
5,no.2,FEB.2006,pp.283289.
[13] Xiaohua Li, M. Chen, and W. Liu, Application of STBC
Encoded Cooperative Transmissions in Wireless Sensor Net
works, IEEE signal proce. letters, vol. 12, no. 2, Feb2005, pp.
134137.
[14] A. del Coso, U. Spagnolini, C Ibars, Cooperative distributed
MIMO channels in wireless sensor networks, IEEE Journ. Se
lect.Areas.Commun.Vol25,no.2,2007.pp402414.
[15] S.K.Jayaweera,VirtualMIMObasedcooperativecommunica
tion for energy constrained wireless sensor network, IEEE
Trans.onwirelesscommunications,vol.5,no.5,May2006,pp.
948989.
[16] W. Cheng; K. Xu; W. Liu; Z. Yang; Z. Feng; An Energy
EfficientCooperativeMIMOTransmissionSchemeforWireless
SensorNetworks.WiCOM2006,Sept.2006,pp.14.
[17] J. N. Laneman, and G. W. Wornell Distributed SpaceTime
Coded Protocols for Exploiting Cooperative Diversity in Wire
lessNetworksIEEETransconInf.Theory,vol.49,no.10,Oct
2003,pp.24152425.
[18] K.Jain,M.Mahdian,andA.Saberi,ANewGreedyApproach
for Facility Location Problems, Proc. of the 34
th
Annual ACM
SymposiumonTheoryofComputing,2002,pp.731740.
[19] Guha and Khuller, Greedy Strikes Back: Improved Facility
LocationAlgorithms,InSODA:ACMSIAM,1998.
[20] Madhukar R. Korupolu, C. Greg Plaxton, and Rajmohan Raja
raman,Analysis of a Local Search Heuristic for Facility Loca
tionProblems,InProc.ACMSIAM,1998,pp.110.
[21] J. M. Kahn, R. H. Katz and K. S. J. Pister, Next Century Chal
lenges:MobileNetworkingforSmartDust,intheProceedings
of 5
th
Annual ACM/IEEE International Conference on Mobile
Computing and Networking (MobiCom 99), Aug. 1999, pp.
271278.
[22] Y.Yuan,M.ChenandT.Kwon,ANovelClusterBasedCoope
rativeMIMOSchemeforMultiHopWirelessSensorNetworks,
EURASIP Journal on Wireless Communications and Network
ing,vol2006,Issue2,April2006,pp.19.
[23] John G. Proakis, Digital communications, fourth edition,
McGrawHillInternationaledition2001.


Md. Abdullah-al Mamun obtained a Master of Engineering (M.E.)
degree in 2009 from the department of Computer and Information
Sciences, Iwate University, Japan. Currently, he is a Ph.D. candidate
in Graduate School of Engineering, Iwate University. His research
interests include Wireless Networks, Wireless Sensor Networks,
Cooperative communications, Multi Channel operation, pattern rec-
ognition and its applications.

Naoshi Nakaya obtained a Ph.D. in Mathematical Science, Depart-
ment of Science and Engineering, Saitama University, Japan. Pre-
sently he is working as an assistant professor in the Iwate University,
Japan. His research interests are computer networks, wireless net-
works security, computer viruses, and Evolutionary algorithms.

Yuji Koi worked as Department Chief, Mitsubishi electric corpora-
tion, Information Technology R&D Audio-Visual Information Tech-
nology. He obtained a Ph.D. degree in Information Engineering from
Tohoku University, Japan. Presently he is working as a professor in
the Iwate University, Japan. His research interests are wireless sen-
sor networks, wireless networks security, Unknown computer virus-
es, and Remote control protocols.

Hagihara Yukari graduated from the Department of Information on
System Fundamentals, the University of Electro-Communications,
Japan. She is working as a technical staff in the Iwate University,
Japan. Her research interests are wireless networks, mobile robot,
and computer vision.

JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 37

You might also like