Wireless Sensor Network, 2010, 2, 777-783

doi:10.4236/wsn.2010.210093 Published Online October 2010 (http://www.SciRP.org/journal/wsn/).

An Energy-Balanced Clustering Routing Algorithm for

Wireless Sensor Network
Fengjun Shang, Yang Lei
College of Computer Science and Technology, Chongqing University of Posts and Telecommunications,
Chongqing, China
Email: [email protected]
Received May 19, 2010; revised June 27, 2010; accepted August 4, 2010

Abstract

In this paper, we consider a network of energy constrained sensors deployed over a region. Each sensor node
in such a network is systematically gathering and transmitting sensed data to a base station (via cluster-
heads). This paper focuses on reducing the power consumption of wireless sensor networks. Firstly, we pro-
posed an Energy-balanced Clustering Routing Algorithm called LEACH-L, which is suitable for a large
scope wireless sensor network. Secondly, optimum hop-counts are deduced. Lastly, optimum position of
transmitting node is estimated. Simulation results show that our modified scheme can extend the network
lifetime by up to 80% before first node dies in the network. Through both theoretical analysis and numerical
simulations, it is shown that the proposed algorithm achieves higher performance than the existing clustering
algorithms such as LEACH, LEACH-M.

Keywords: Wireless Sensor Network, LEACH, Energy Efficient, Multiple-Hop

1. Introduction such as the design of low-power signal processing archi-

tectures, low power sensing interfaces, energy efficient
Wireless Sensor Networks (WSNs) can collect reliable wireless media access control and routing protocols, have
and accurate information in distant and hazardous envi- been areas of extensive research in recent years. LEACH
ronments, and can be used in National Defence, Military is one of the first hierarchical routing approaches for sen-
Affairs, Industrial Control, Environmental Monitor, Traf- sors networks, which attempts to improve energy and
fic Management, Medical Care, Smart Home, etc. The routing efficiency of such networks. The idea proposed
sensor whose resources are limited is cheap, and depends in LEACH has been an inspiration for many hierarchical
on battery to supply electricity, so it’s important for Rout- routing protocols, although some protocols have been
ing to efficiently utilize its power. independently developed.
In this paper, we propose an advanced multiple-hop In [2], the author puts forward energy-LEACH and
routing protocol named LEACH-L, whose features can multihop-LEACH protocols called LEACH-M. Energy-
be described as follows: when the cluster-heads are close
LEACH protocol improves the choice method of the
to, they directly communicate with Base station (BS);
cluster head, makes some nodes which have more resi-
when they are far in distance, they telecommunicate by
dual energy as cluster heads in next round. Multihop-
multiple-hop way, and the shortest transmission distance
LEACH protocol improves communication mode from
is limited. The sensors in different areas use different
single hop to multi-hop between cluster head and sink.
frequencies and gaps to communicate with BS. In the last
Simulation results show that energy-LEACH and multi-
part, the authors respectively simulate LEACH [1],
hop-LEACH protocols have better performance than
LEACH-M [2], and LEACH-L on the Matlab. The simu-
LEACH protocols.
lation experiments indicate that: LEACH-L can prolong
In [3], a radio irregularity model is proposed aiming at
the whole network lifetime.
location and topology control. The simulated results show
2. Related Works that radio irregularity has a relatively larger impact on
the routing layer than the MAC layer. It also shows that
Various issues in the design of wireless sensor networks radio irregularity leads to larger localization errors and

Copyright © 2010 SciRes. WSN

778 F. J. SHANG ET AL.

makes it harder to maintain communication connectivity The p is a cluster-head section probability, used dur-
in topology control. ing cluster creation.
In [4], author addresses the problem of clustering in The round is the time, which all cluster members com-
WSNs, subject to upper bounds on the maximum latency, municate with cluster heads and cluster heads communi-
the energy consumed by intermediate nodes, and clusters cate with BS.
size. Those constraints are necessary for the reliability of
the system and for extending its lifetime. 3.1. Problem Formulation
In [5], author proposes a novel energy efficient clus-
tering scheme for single-hop wireless sensor networks. A This paper proposes a modification of LEACH cluster-
novel cost function is introduced to balance the load head selection algorithm to reduce energy consumption.
among the cluster heads and prolongs the network life- For a microsensor network, we make the following as-
time significantly against the other clustering protocols sumptions:
such as LEACH.  The base station (BS) is located among the sensors
In [6], an Energy-Efficient Unequal Clustering is pro- and is stationary.
posed for multihop sensor network. Simulation results  All nodes in the network are homogenous and en-
show that the unequal clustering mechanism balances the ergy-constrained.
energy consumption well among all sensor nodes and  All nodes are able to reach BS in one hop.
achieves an obvious improvement on the network life-  Nodes have location information.
time.  Propagation channel is symmetric.
In [7], a novel multicast protocol, uCast is proposed Cluster-heads perform data compression.
for energy efficient content distribution in sensor net- This paper presents an improvement to the LEACH’s
works. The uCast support a large number of multicast
cluster-head selection algorithm by introducing distance-
sessions, especially when the number of destinations in a
based factor, called LACH-L. In LEACH-L, the clusters
session is small.
are re-established in each round. Each of these rounds
In [8], a novel energy-efficient protocol is proposed,
consists of a set-up and a steady-state phase. New cluster
called Route Maintenance based-on Energy Threshold
heads are elected in each round and as a result the load is
and extends the lifetime of nodes and enhance the effi-
well distributed and balanced among the nodes in the
ciency of the entire network.
network. Moreover, each node transmits to the closest
LEACH-C uses a centralized algorithm to form clus-
cluster-head to split the communication cost to the sink.
ters [9]. A non-autonomous cluster-head selection is again
Only the cluster head has to report to the sink and may
the main disadvantage of this algorithm. Moreover, it re-
consume a large amount of energy, but this happens pe-
quires the location information of all nodes in the net-
work. However, location information in mobile wireless riodically for each node. During the set-up phase clus-
networks is only available through GPS or a location- ter-heads are determined and the clusters are organized.
sensing technique, such as triangulation which requires During the steady-state phase data transfers to the base
additional communication among the nodes. station occur.
In [10], DCHS algorithm is proposed. In this algo- We use a simplified model shown in Figure 1 for the
rithm, in order to extend the lifetime, a parameter En_ radio hardware energy dissipation. Both the free space
current/En_max is introduced. Nevertheless, it has a cru- ( d 2 power loss) and the multi-path fading ( d 4 power loss)
cial disadvantage: After a certain number of rounds the channel models are used in the model, depending on the
network is stuck, although there are still nodes available distance between the transmitter and receiver. Transmis-
with enough energy to transmit data to the base station, sion ( ETx ) and receiving costs ( ERx ) are calculated as
so DCHS algorithm need be improved. follows [9]:
 lE  l fs d 2 , d  d o
3. LEACH-L Algorithm ETx (l , d )   elect (1)
lEelect  l mp d , d  d o
4

In order to explain our idea, we fist define some parame- where d is the distance between the transmitter and the
ters as follows. receiver.
The Eelect is consumed energy per bit. Where is:
The  fs and  mp are consumed energy per bit and
 fs
per area respectively. d0  (2)
The do is a distance, which decide the parameter value  mp
both  fs and  mp . The following equation shows the amount of energy
The EDA is the energy required for data aggregation. required to receive each message.

Copyright © 2010 SciRes. WSN

 F. J. SHANG ET AL. 779

that there is a linear network model, as shown in Figure 2.

The cluster-head which is nr far from the BS, if direct
communication protocol is adopted, so the total energy
consumed by network is:
Edirect  ETx (k , nr )  ETx  elect (k )  ETx  amp (k , nr )
If multiple-hop transmission is adopted, the total en-
ergy consumed by network is:
Emultihop  nETx  elect (k )  (n  1) ETx  elect (k )
+ nETx  amp (k , nr )
 (2n  1)kETx  elect  nETx  amp (k , r )
Figure 1. Radio energy dissipation model. This paper adopts the parameters of Table 1 in the fif-
th chapter. There are following conclusions:
ERx ( L)  Eelect L (3) When r  d 0 , so
with L as the length of the message in bits, d as the dis- Edirect  Emultihop 
tance between transmitter and receiver node. A sensor n 4 r 4 k  mp  2(n  1)kEelect  nr 4 k  mp
node also consumes EDA amount of energy for data
aggregation. We assumed that the sensed information is When nr  d 0 ，so
highly correlated, thus the cluster-head can always ag- Edirect  Emultihop 
gregate the data gathered from its members into a single n 2 r 2 k  fs  2(n  1)kEelect  nr 2 k fs  0
length-fixed packet.
When r  d 0 , nr  d 0 , so
3.2. Problem in Single-hop Routing Edirect  Emultihop 
n 4 r 4 k  mp  2(n  1)kEelect  nr 2 k  fs  0
When the scope of WSN gets larger, the diversity of en-
ergy consumption among cluster-heads of LEACH as d 0 d 2  40000n 2 (n  1) 14
When r  ( ) , so
well gets larger. n 4n 2 d 2
Let’s suppose that there are cluster-heads S (i ) which n 4 r 4 k  mp  2(n  1)kEelect  nr 2 k  fs  0
is r far from the BS and S ( j ) which is 3r far from the
d 0 d 2  40000n 2 (n  1) 14
BS. When r  ( ) , so
When r  d 0 , the energy consumption of S (i ) is listed n 4n 2 d 2
as n 4 r 4 k  mp  2(n  1)kEelect  nr 2 k  fs  0
ETx (k , r )  kEelect  k  mp r 4 Therefore, when the nodes are close to the BS, directly
The energy consumption of S(j) is written as communicate routing can exhibition a good result; when
ETx (k , r )  kEelect  81k  mp r 4 the transmission distance is long, under limiting each
In the formula, when the circuit consumption is the transmitting distance the multiple-hop strategy can re-
same, the energy consumed by transmit amplifier of S(i) duce the energy consumption of WSN.
is 8l times as much as that of S(j) so WSNs in large
scope should adopt multiple-hop routing protocols. 3.4. Communication Conflicts

3.3. Introducing Multiple-Hop Routing The communication between the cluster-heads and the BS
should adopt Collision Avoidance when the network be-
The data collected by myriad sensors flow to a small comes large and the number of sensors increase, or it will
number of BS, which leads to the energy of sensors give rise to the occurrence of conflicts when the clus-
which near to basic be used up quickly and leads to the ter-heads are sending data. Therefore, sensor networks in
break of networks, so the question existing in many-to- large scope should adopt certain strategies to reduce the
one networks is viewed as the hot spot [11]. Meanwhile,
only considering the shortest route will result in the geo-
graphic superiority of individual sensor embodied in the
multiple routes, and this is proved in LEACH-M. Be-
sides, multiple-hops routing cuts down the communica-
tion consumption, but increases the circuit energy con-
sumption. In order to go on the research, let’s suppose Figure 2. Linear network model.

Copyright © 2010 SciRes. WSN

780 F. J. SHANG ET AL.

occurrence of conflicts. Figure 2, we have

r2  r  r1 (6)
3.5. Estimating Hop-Counts When node 2 transmits data to node 1, the energy con-
sumption is written as
CHs can send their data via just one (high-energy) trans- Enode1  k  fs r12  kEelect
mit of data to the base station or via a multi-hop scheme
where each data message must go through n (low energy) Enode 2  k  fs (r12  r22 )  2kEelect
transmits and n receives. Depending on the relative costs The total energy consumption is as follows
of the transmit amplifier and the radio electronics, the (7)
Etot  2k  fs (r12  r22 )  3kEelect
total energy expended in the system might actually be
greater using multi-hop routing than direct transmission We replace (6) into (7),
to the base station. Etot  2k  fs ((r  r2 ) 2  r22 )  3kEelect
To illustrate this point, consider the linear network And the optimum node position is computed as below:
shown in Figure 1, where the distance between the no- r2  r / 2
des is r. If we consider the energy expended transmitting
a single l-bit message from a node located a distance nr In a similar way, when r31  r / 3 and r32  2r / 3 ,
from the base station using the direct communication the energy consumption is minimum, etc.
approach via one hop and Equations (1) and (3), we have: According above computation, optimum transmitting
distance is
n 1 2
ETx  l fs ( r )  lEelect (4) d
 d opt   (  opt  1)
 opt
Where  is the number of hops. Thus, the total en-
Given coordinate of S(i) be (x, y), optimum node co-
ergy is:
ordinate is listed as
Etot   ETX   ERX
(n  1) 2 r 2  x
 2l  Eelect  l fs (5)  xopt    (  opt  1)
  opt
（8）

And the optimum number of hops is computed as be- y  y
 (  opt  1)
low:  opt  opt

dEtot d2
 2lEelect  l fs 2  0
d  opt 4. Simulation Results
 fs d
  opt  d  This paper makes Matlab as experiment platform, and
2 Eelect 100 goes on imitation in two scenes respectively. The para-
This shows that, when transmission energy is on the meters of imitation experiments see the following Table 1,
same order as receive energy, which occurs when trans- in which the first ten parameters are same to the docu-
mission distance is short, direct transmission is more en- ment.
ergy-efficient than multi-hop routing. Thus we use direct Figure 4 and Figure 5 are the network’s life cycle in
transmission communication among CHs and the base two scenes. In the scene 1, the Fist Node Death time (FND)
station. of LEACH-L is 2.5 times more than that of LEACH, and
the Half Nodes Death (HND) time of LEACH-L is about
3.6. Selecting Optimum Transmitting Node 1.1 times more than that of LEACH. In the scene 2, LEA-
CH-L’s FND and HND are 12 times and 1.8 times as mu-
To illustrate this point, consider the linear network shown ch as that of LEACH. Besides, the last node dead time of
in Figure 3. LEACH and LEACH-M are longer than that of LEACH-
Where r1  d 0 is distance between BS and node 1, L. The reason is that LEACH-L makes power equally
d 0  r2  2d 0 is distance between node 2 and node 1, r distribute among all sensors, therefore, in the pre-period,
is distance between BS and node 2. We select node 1 to the network’s activity nodes and cover areas of LEACH-L
transmit data, when node 2 transmits data to BS. From are greatly larger than that of LEACH-M and LEACH.
Figure 6 and Figure 7 are the sum of data packets
which is the cluster-heads sending to the BS in two dif-
r1 r2 node 2
BS node 1 ferent scenes. In the scene 1, before HND, the sum of data
packets transmitted by LEACH-L is 1.3 times as much as
Figure 3. Relay node network model. that of LEACH and 1.4 times as much as that of LEACH–

Copyright © 2010 SciRes. WSN

 F. J. SHANG ET AL. 781

Table 1. Simulation parameters. M. In the scene 2, before HND, the data transmitted by
scene 1 scene 2
LEACH-L is 2.4 times that of LEACH and 2.1 times that
of LEACH-M.
The scope 300 × 300 m 500 × 500 m
Figure 8 is relation between energy consumption and
The number of sensors 900 2500 round among LEACH, LEACH-M, LEACH-L in 300 ×
The location of BS (150,150) (250,250) 300 scene. From Figure 8, it is known that LEACH-L
The initial energy 0.5 J 0.5 J can save energy consumption of network when LEACH-
E 1J 1J L is compared with LEACH-M and LEACH. After a
while, it saves energy in LEACH and LEACH-M better
The length of packets 4000 4000
-8
than LEACH-L. In about the 400th round, total energy
ETX 5 × 10 5 × 10-8 consumption of network in LEACH-L is more than LE-
ERX 5 × 10 -8
5 × 10-8 ACH and LEACH-M, because some nodes in LEACH
εfs 10-11 10-11 and LEACH-M are dead with time going.
εmp 1.3 × 10 -15
1.3 × 10-15 Figure 9 is relation between energy consumption and
round among LEACH, LEACH-M, LEACH-L in 500 ×
EDA 5 × 10-9 5 × 10-9
500 scene. From Figure 9, it is known that LEACH-L
P 0.1 0.1 can save 50% energy consumption of network in the 100th
M 0.1 0.1 round when LEACH-L is compared with LEACH-M and
D 70 m 70 m LEACH. In about the 400th round, total energy consump-
Restriction_distance 30 m 70 m tion of network in LEACH-L is more than LEACH and
Max_distance 87 m 87 m

Figure 6. Received packets over rounds in 300 × 300 m.

Figure 4. Dead nodes over rounds in 300 × 300 m.

Figure 5. Dead node over rounds in 500 × 500 m. Figure 7. Received packets over rounds in 500 × 500 m.

Copyright © 2010 SciRes. WSN

782 F. J. SHANG ET AL.

Figure 8. 300 × 300 energy consumption among LEACH,

LEACH-M and LEACH-L. Figure 11. 500 × 500 energy consumption of LEACH-L in
different areas.

LEACH-M, because some nodes in LEACH and LEA-

CH-M are dead with time going.
Figure 10 is relation between energy consumption and
rounds in different area in LEACH-L. It is obvious that
in two areas energy consumption is balancing in 300 ×
300 scene.
Figure 11 is relation between energy consumption and
rounds in different area in LEACH-L. It is obvious that
in two areas energy consumption is balancing in 500 ×
500 scene.

5. Conclusions

We have proposed a routing protocol named LEACH-L.

Figure 9. 500 × 500 energy consumption among LEACH, The experiments demonstrate that in the larger scope
LEACH-M and LEACH-L. sensor networks, Comparing with LEACH and LEACH-
M which is the multiple-hop protocol only considering
the distance, LEACH-L can balance network load, re-
duce the network’s energy consumption, enhance the net-
work’s data collecting precision and extend the net-
work’s life cycle, certainly, LEACH-L demands each sen-
sor node to record its own location information and the
information of candidate routing cluster-heads, increas-
ing the storage requirements of sensor nodes. But, com-
pared with balancing the nodes’ surplus energy and pro-
longing the life time of network, LEACH-L is valuable.

6. Acknowledgements
This work is supported by the Chongqing Natural Sci-
ence Foundation under Grant No. 2009BB2081. The Pro-
ject Sponsored by the Scientific Research Foundation for
Figure 10. 300 × 300 energy consumption of LEACH-L in the Returned Overseas Chinese Scholars, State Education
different areas. Ministry.

Copyright © 2010 SciRes. WSN

 F. J. SHANG ET AL. 783

7. References 27-36.
[7] Q. Cao, T. He and T. F. AbdelZaher, “uCast: Unified Con-
[1] W. Heinzelman, A. Chandrakasan and H. Balakrishnan, nectionless Multicast for Energy Efficient Content Dis-
“Energy-Efficient Communication Protocol for Wireless tribution in Sensor Networks,” IEEE Transactions on
Microsensor Networks,” Proceedings of 33rd Interna- Parallel and Distributed Systems, Vol. 18, No. 2, 2007,
tional Conference on System Sciences, Hawaii, 2000, p. 10. pp. 240-250.
[2] X. N. Fan and Y. L. Song, “Improvement on LEACH Pro- [8] L. Hu, Y. Li and Q. B. Chen, “A New Energy-Aware
tocol of Wireless Sensor Networks,” Proceedings of 2007 Routing Protocol for Wireless Sensor Networks,” Inter-
International Conference on Sensor Technologies and national Conference on Wireless Communications, Net-
Applications, Valencia, 2007, pp. 260-264. working and Mobile Computing, Shanghai, 2007, pp.
[3] G. Zhou, T. He, S. Krishnamurthy and J. A. Stankovic, 2444-2447.
“Models and Solutions for Radio Irregularity in Wireless [9] W. Heinzelman, A. Chandrakasan and H. Balakrishnan,
Sensor Networks,” ACM Transactions on Sensor Net- “An Application-Specific Protocol Architecture for Wire-
works, Vol. 2, No. 2, 2006, pp. 221-262. less Microsensor Networks,” IEEE Transaction on Wire-
[4] B. Aoun and R. Boutaba, “Clustering in WSN with La- less Networking, Vol. 1, No. 4, 2002, pp. 660-670.
tency and Energy Consumption Constraints,” Journal of [10] M. J. Handy, M. Haase and D. Timmermann, “Low En-
Network and Systems Management, Vol. 14, No. 3, 2006, ergy Adaptive Clustering Hierarchy with Deterministic
pp. 415-439. Cluster-Head Selection,” Proceeding of the 4th IEEE
[5] M. Ye, C. F. Li, G. Chen and J. Wu, “EECS: An Energy Conference on Mobile and Wireless Communications
Efficient Clustering Scheme,” 24th IEEE International Networks, Stockholm, 2002, pp. 368-372.
Performance, Computing, and Communications Confer- [11] G. Smaragdakis, I. Matta and A. Bestavros, “SEP: A Sta-
ence, Phoenix, 2005, pp. 535-540. ble Election Protocol for Clustered Heterogeneous Wire-
[6] G. Chen, C. F. Li, M. Ye and J. Wu, “An Unequal Cluster- less Sensor Networks,” Proceeding of the International
Based Routing Strategy in Wireless Sensor Networks,” Workshop on Sensor and Actor Network Protocols and
Chinese Journal of Computers, Vol. 30, No. 1, 2007, pp. Applications, Boston, No. 4, 2004, pp. 660-670.

Copyright © 2010 SciRes. WSN