Power Efficient Communication Protocols for Data

Gathering on Mobile Sensor Networks

Chuan-Ming Liu and Chuan-Hsiu Lee
Department of Computer Science and Information Engineering
National Taipei University of Technology
Taipei, Taiwan
cmliu@csie.ntut.edu.tw

Abstract— Wireless sensor networks have attracted much re-

search attention in recent years. One critical issue in wireless
sensor networks is how to gather sensed information in an energy
efficient way since the energy is limited in a sensor node. Many
protocols have been proposed for data-gathering or communi-
cation between wireless sensor nodes. However, most of these
protocols work on static wireless sensor networks. In this paper,
we provide clustering-based and time-driven protocols which
minimize energy dissipation for data-gathering in wireless mobile
sensor networks where the sensor nodes are capable of mobility.
In many applications, the sensor nodes can move either by outside
force or its mobility component. For example, the sensor nodes
attached to moving objects for tracking. Our protocols consist of
three major phases based on LEACH protocol: (1) cluster-head
election, (2) organizing clusters, and (3) message transmission.
First, we consider the node mobility when organizing clusters.
Our protocols will have a sensor node select a proper cluster-
head to join in order to save energy. Then, we consider how to
elect the cluster-heads and provide an algorithm for cluster-head
election. Last, we implement all the protocols and perform the
experiments for evaluating the protocols. The simulation results
show that our protocols are more energy-efficient and make the
system lifetime 40-55% longer than LEACH.
keywords: Wireless Sensor Networks, Mobility, Data Gather-
ing, Energy-Efficiency.
I. I NTRODUCTION
A wireless sensor network consists of a large number of tiny,
low-power, cheap sensors having sensing, data processing, and
wireless communication components [1]. In this work, we
consider the wireless mobile sensor networks where sensor
nodes are capable of moving.
The sensor nodes in a wireless sensor network are deployed
randomly inside the region of interest or close to it. A remote
base station, BS, is engaged to give commands to all the
sensor nodes and gather information from the sensor nodes. In
addition to sensing, the wireless sensor nodes can process the
acquired information and transmit messages to the BS as well
as communicate to each others. A wireless sensor network
scenario is depicted in Figure 1. Because the sensor nodes
are randomly scattered in a sensor field, the wireless sensor
network protocols or algorithms should have the capability of
self-organizing.
Wireless sensor nodes have many limitations, including
modest processing power (8 MHz), little storage, short com-
Base Station
(Sink)
Internet
User
Sensor Field Sensor Nodes
Fig. 1. The architecture of a wireless sensor network in which the sensor
nodes are deployed randomly into the interest area (sensor field) and the BS
(sink) connects to the internet.
munication range (consuming a lot of power), and limited
power source. Among these limitations, the constraint on the
power consumption is an important topic when designing
wireless sensor network protocols. Data gathering (collecting
the sensed information from the sensor nodes or routing
the sensed information) raises an important topic in wireless
sensor networks due to the power limitation of a sensor node.
To have an efficient data gathering protocol in terms of energy
consumption is an on-going research work and is the core of
this paper.
In this paper, we provide a data-gathering protocol in
wireless mobile sensor networks where the sensor nodes are
capable of moving. Many protocols have been proposed for
data gathering or communication between wireless sensor
nodes [2], [3], [4], [5], [6], [7], [8], [9]. To our knowledge,
most of these protocols work on static wireless sensor net-
works. Nevertheless, in many applications, the sensor nodes
can move either by outside force or its mobility component.
II. R ELATED W ORK
Existing protocols for data-gathering can be classified by
the way to organize the sensor nodes into: hierarchy and
non-hierarchy. The non-hierarchical protocols include Directed
Diffusion [4], Flooding [2], [10], Gossiping [11], and SPIN
[2]. On the other hands, the hierarchical protocols include
LEACH [3]and PEGASIS [7]. Furthermore, the communica-
tion patterns for wireless sensor networks take one of two
general forms: (1)Time-driven (periodical) transmissions [3],

0-7803-8521-7/04/$20.00 © 2004 IEEE 4635

0-7803-8521-7/04/$20.00 (C) 2004 IEEE


Base Station
Cluster-head
Cluster
Fig. 2. A snapshot of a hierarchical protocol for wireless sensor networks
where the non-cluster-head sensor nodes (unfilled circles) send the sensed
information to the cluster-heads (filled circles) and the cluster-heads fuse the
information received and transmit the fused message to the BS.
[7] and (2) Event-driven transmissions [4], [6], [2], [11], [8],
[12], [10].
LEACH (Low-Energy Adaptive Clustering Hierarchy) is a
clustering-based and time-driven data gathering protocol. The
operation of LEACH is broken up into R rounds. Sensor nodes
take turns to be cluster-heads exactly once among R rounds.
After R rounds, all the sensor nodes are again possible to be
cluster-heads once in the next R rounds. In each round, after
the cluster-heads have been decided, each cluster-head will
send a ”cluster-head-advertisement” message by broadcasting.
Sensor nodes listen to the advertisements and join the ”closest”
cluster-head. When the clusters are organized, cluster-heads
aggregate data and then transmit information to the BS.
Figure 2 captures the operation of LEACH in a round and
presents a hierarchical mechanism. Recent research on the data
gathering, such as PEGASIS [7], follows LEACH protocol and
uses LEACH as a benchmark. Our work follows this trend and
is to develop a clustering-based data-gathering protocol for
wireless mobile sensor networks, which consists of three major
phases in one round: (1) cluster-head election, (2) organizing
clusters, and (3) message transmission phases.
III. P ROBLEMS
We first provide a mechanism to organize clusters for sensor
nodes with mobility on the organizing clusters phase. Then,
we discuss how the LEACH protocol elect cluster-heads and
provide an algorithm to elect cluster-heads in the cluster-
head election phase. Based on the mechanism to organize
the clusters, using the two different algorithms for cluster-
head election makes different protocols for data gathering on
mobile sensor networks. We measure the energy efficiency by
the system lifetime in terms of number of rounds which a
system experiences. We consider the wireless mobile sensor
networks where:
• The BS is fixed and located far away from the sensor
nodes.
• All the sensor nodes are homogeneous and power limited.
• Each sensor node is equipped with a GPS (Global Posi-
tioning System) device [13].
• Each sensor node is capable of moving: each sensor
node has a constant speed and fixed direction. Note that,
0-7803-8521-7/04/$20.00 © 2004 IEEE
0-7803-8521-7/04/$20.00 (C) 2004 IEEE
different sensor nodes may have different directions.
• All sensor modes are time-synchronized [14].
We use the radio model discussed in [3] where the radio
dissipates Eelec = 50nJ/bit to run the transmitter or receiver
circuitry and
amp = 100pJ/bit/m2 for the transmitter
amplifier. The radios have power control and can expend the
minimum required energy to reach the intended recipients. We
also assume an r2 energy loss due to channel transmission.
The transmission cost and receiving cost for a k-bit message
and a distance d using this radio model is
2
• Transmission: ET x (k, d) = Eelec × k +
amp × k × d ,
and
• Receiving: ERx (k, d) = Eelec × k.
Receiving data is also a high cost operation; therefore, the
number of receptions and transmissions should be minimized
to reduce the energy cost of a application.
We observe that the node mobility plays an important
role when organizing the clusters. If a sensor node neglects
its node mobility when deciding which cluster-head to join,
it may select a cluster-head with a longer distance to it;
therefore, consumes more power. Recall that each sensor node
is equipped with a GPS device. The problem we explore in this
paper is, for each sensor node, how to use the information from
the attached GPS device to decide which cluster it belongs to
properly. We provide a mechanism to have each node make it
own decision properly such that it has less power consumption
and can live longer in the system.
Besides, we observe that if the number of cluster-heads is
small in a round, these cluster-heads needs to consume more
energy to communicate with the BS for large amount of data. If
the mechanism for cluster-head election generates the cluster-
heads unevenly in each round, some sensor nodes will use
up all of its energy quickly. The mechanism for cluster-head
election in LEACH has such a drawback. We work towards to
have the algorithm for cluster-head election which make the
number of cluster-heads in each round evenly.
IV. C LUSTERING WITH M OBILITY (CM)
This section describes how to organize the clusters when
the sensor nodes can move. Having the location information
from the attached GPS device, each mobile sensor node can
calculate its speed and direction. We consider the speed of
each sensor node is a constant. Our mechanism for clustering
sensor nodes having mobility improves the energy efficiency;
therefore, makes the system live longer.
After the cluster-heads in the current round have been
decided, the cluster-heads then broadcast an advertisement
message to all the sensor nodes to recruit the sensor nodes.
This advertisement message also includes the cluster-head’s
position, direction, and speed. Every non-cluster-head node
must listen to such a broadcast after the cluster-head election
has been done. After receiving the advertisement message,
each sensor node needs to select a proper cluster-head to join.
In our mechanism, each non-cluster-head node considers the
distances from itself to all the cluster-heads. The algorithm
4636

h3 h2
dh1 dh2
3
2
dh1
2
dh2 3 dh2
dh1 v superround
dh11
4
h4 dh2
4
h1
Fig. 3. A portion of a wireless mobile sensor network where, in some round,
sensor node v is a non-cluster-head and h1 , h2 , h3 , and h4 are cluster-heads;
arrows represent sensor nodes’ directions and speeds; solid (dashed) lines
represent the distances from v to all the cluster-heads at time t1 (t2 ).
works as follows. Suppose, at time t1 , nodes h1 , h2 , · · · , hn
have been elected as the cluster-heads and the next round will
begin at time t2 . For a non-cluster-head node v, since the
advertisement messages include the location information of
the cluster-head nodes’, sensor node v can calculate all the
distances from sensor node v to all the cluster-heads at time
t1 and t2 , respectively. Let d1hi and d2hi be the distances from
sensor node v to the cluster-head node hi at time t1 and t2 ,
respectively, for i = 1, 2, · · · , n. Suppose further that δdhi =
d2hi - d1hi and δt = t2 − t1 . We then consider the value
δdhi
DHi = (T − t1 ) + d1hi , (1)
δt
where T is in [t1 , t2 ] and i = 1, 2, · · · , n. The value of DHi
represents the distance from node v to the cluster-head hi at
time T . Sensor node v will use the DH value at some time
between t1 and t2 to decide which cluster to join. Figure IV
gives a picture of the method we described.
From Equation 1, we observe that the value of T impacts
the cluster organization. We refer T as the clustering factor.
Note that, when T = t1 , the way to organize clusters is the
same as the way in LEACH. Recall that, in each round, we
mainly have three phases. The message transmission phase is
the last phase and will consume most of the power in a round.
We therefore consider to set T larger than or equal to t1 + 34 δt
but not greater than t2 . After T has been set, sensor node v
will select the cluster-head, hj , to join where
n
DHj = min{DHi }.
i=1
V. A LGORITHMS FOR C LUSTER -H EAD E LECTION
This section presents an algorithm for cluster-head election.
LEACH provides a rule to elect cluster-heads and can be
fully applied to a distributed environment like sensor networks.
However, we observe that the difference in the number of the
cluster-heads between the rounds may be large. It also happens
that there is no cluster-heads at all in some rounds. If there
is no cluster-head in a round, every sensor node sends the
information directly to the BS; therefore, consumes a large
0-7803-8521-7/04/$20.00 © 2004 IEEE
0-7803-8521-7/04/$20.00 (C) 2004 IEEE
:invalid round
:valid round
:superround
r:round counter
r i-1 i i+1 i+2 i+3 i+4 i+5
j j+1 j+2 j+3
(a) relation between super round and invalid round as well as
valid round; algorithm using round counter r to elect cluster-
heads; system using the number of super (valid) rounds as the
lifetime.
invalid rounds
valid round
cluster-head
organizing
clusters
m essage transm ission
election phase phase
phase
superround
(b) A super round consists of the three major phases; in par-
ticular, the cluster-head election phase may contain a number
of consecutive invalid round before the valid round.
Fig. 4. The concepts of invalid round, valid round, and super round.
amount energy. We provide an algorithm which avoids the
case that there is no cluster-head in a round. The experiment
work in the next section shows that the algorithm perform
better than the mechanism used in LEACH in terms of energy
efficiency.
Cluster-Head Election with Invalid Rounds
The basic idea of the algorithm is to give up a round when
there is no cluster-head elected in that round. If a round i has
no cluster-head elected, we refer round i as an invalid round;
otherwise, round i is a valid round. Suppose that the system is
about to start round i and moves to the cluster-head election
phase of round i. There are two cases to consider:
• If there is no cluster-head elected in the cluster-head
election phase, round i is invalid and the system moves
to the next round, round i + 1, without going through
organizing cluster and message transmission phases.
• If some sensor nodes are elected as the cluster-heads in
the first phase, the algorithm moves to the organizing
clusters and message transmission phases.
The process for electing cluster-heads is similar to the one
in LEACH. However, no matter what the round is (valid or
invalid), the value of r still increases by 1 for each round.
Figure 4(a) gives an overview about the concepts of different
kinds of rounds.
In an invalid round, a sensor node only consumes energy
for local computation and listening to the broadcast in a short
4637

time. The energy consumption in an invalid round is very small
compared with the one in a valid round in which the system
consumes much energy on message transmission. Note that no
advertisement message will be received for a sensor node in an
invalid round. Since the invalid round contains only one pass
for cluster-head election and the system may have a number of
consecutive invalid rounds before moves to a valid round, we
may include the consecutive invalid rounds into the cluster-
head election phase of a super round as shown in Figure 4(b).
In other words, the super round consists of one valid round and
possibly some consecutive invalid rounds in the cluster-head
election phase. From the system viewpoint, the BS collects
information and synchronizes all the sensor nodes based on
the super round. Therefore, the time for each super round is
fixed. If the number of invalid rounds in a super round is large,
it will cause a delay on the time for message transmission and
may results in losing sensed information in that super round
since the next (super) round may start before the message
transmission completes.
VI. E XPERIMENTS
In this section, we present the simulation for the algorithms
we provided in this paper and compare them with LEACH.
We use Visual C++ 6.0 on Window XP as the platform and
implement LEACH and our protocols in C++. The sensor
nodes are capable of mobility in our simulation. We discuss the
energy efficiency by considering how long the whole system
can last in terms of number of rounds. Note that we count
the number of super (valid) rounds as the system lifetime
for protocol CM-IR since the energy consumption for invalid
round is too small and can be neglected. We first implement the
protocol CM using the mechanism, discussed in Section IV,
which organizes the clusters by considering the node mobility.
The cluster-head election in protocol CM is the same one
as LEACH. Then, based on the mechanism for organizing
clusters, we implement the protocol CM-IR which uses the
cluster-head election with invalid rounds. The experimental
result shows that all these protocols, CM and CM-IR make
the whole system live longer. We will first discuss the system
lifetime and make a comparison between these protocols.
Then, we further discuss the impact of the clustering factor
T used in organizing clusters phase.
In the experiments, all the sensor nodes use the same
velocity but arbitrary fixed direction to move in a fixed range.
We consider two coverage areas: 100m × 100m and 200m ×
200m respectively. There are 100 mobile sensor nodes in each
system. The speed of each node is 0.5m/sec and the initial
energy in a sensor node is 0.5 or 1.0 J. We have each sensor
node send a 2000-bit data packet to the base station in each
round. The period of one round is 5 seconds and the cluster
factor T varies from 0 to 5 seconds. Initially, the sensor nodes
are assumed to be scattered randomly in the interested area.
For the cluster-head election, the initial probability P of a
sensor node to be a cluster-head is 0.05 when we consider
LEACH, CM, or CM-IR protocols.
0-7803-8521-7/04/$20.00 © 2004 IEEE 4638
0-7803-8521-7/04/$20.00 (C) 2004 IEEE
System Lifetime(200m x 200m;1J)
100
90 : LEACH
: CM
: CM−IR
80
accumulated number of dead nodes
70
60
50
40
30
20
10
0
250 300 350 400 450 500
round
550 600 650 700 750 800
Fig. 5. The system lifetime of the wireless mobile sensor network which
has 100 sensor nodes in the range of 200m× 200m; each sensor node has
initial energy 1.0J.
System Lifetime
We first consider the energy efficiency by measuring the
system lifetime in terms of the total number of (valid) rounds
which a wireless mobile sensor network system experiences.
Figure 5 shows the system lifetimes when each sensor node
has initial energy 1.0 J in areas of 200m× 200m. The dead
node v represents that the energy of v has been used up. We
set the clustering factor T = 4. The reason for selecting such
a value for T will be discussed later.
From the plots, we learn that our protocols make the system
live longer than LEACH does. The result also shows that
protocol CM-IR is better than protocol CM and LEACH in
terms of energy consumption. This indicates that the mecha-
nism we provide for cluster-head election indeed improves the
performance because our mechanism can avoid the case that
all the sensor nodes send their data directly to the BS.
Table I lists the average system time in terms of the number
of rounds with different T values for a mobile sensor network
in an area of 200m × 200m having 100 sensor nodes. For
protocol CM and LEACH, the first dead node occurs at almost
the same round. However, the last node dies at a later round
for protocol CM compared with LEACH. Protocol CM-IR not
only has the first dead node occur at a much later round but
also extends the whole system lifetime. This trend holds in all
of the experiments we made. Generally, protocol CM makes
the system lifetime 5%-10% longer than LEACH. Protocol
CM-IR performs much better and has 40% - 55% longer
system lifetime than LEACH.
Clustering Factor
To study the impact of the clustering factor, we set the
clustering factor, T , from 0 to 5. Recall that when T = 0,
CM is equivalent to LEACH. Figure 6 shows the effects of
different T values in a wireless mobile sensor network system
where each node has initial energy 1.0J in an coverage area
of 200m × 200m. For each value of T , we run 50 different
cases and take the average number of rounds among these
50 cases. In general, as the value of T increases, the system
 T Protocol Round (first Number
(sec) dead node) of rounds VII. C ONCLUSIONS
350 CM-IR 411 831 In this paper, we provide different data-gathering protocols
CM 310 656
LEACH 297 604 for wireless mobile sensor networks where the sensor nodes
360 CM-IR 419 833 are capable of moving. Our protocols lead to a 40% - 55%
CM 313 656 longer system lifetime compared with LEACH; therefore,
LEACH 297 604
370 CM-IR 419 841
are more efficient in terms of energy. The experiments also
CM 309 659 indicates a better selection for the clustering factor. In this
LEACH 297 604 work, we explore the energy efficiency problem on data-
380 CM-IR 413 833 gathering protocol in wireless mobile sensor networks where
CM 309 657
LEACH 297 604 the sensor nodes can move. The future work includes:
390 CM-IR 420 843 • Considering the mobility when electing the cluster-heads.
CM 308 661
• The further analysis on the clustering factor T .
LEACH 297 604
400 CM-IR 420 856 • The study on the case that each sensor node has different
CM 310 658 speed and direction to move.
LEACH 297 604
R EFERENCES
TABLE I
[1] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, “A survey
S YSTEM LIFETIMES ( IN TERMS OF THE NUMBER OF ROUNDS ) USING on sensor networks,” IEEE Communications Magazine, vol. 40, no. 8,
DIFFERENT VALUES OF CLUSTERING FACTOR T FOR EACH PROTOCOLS . pp. 102–114, 2002.
[2] W. R. Heinzelman, A. Chandrakasan, and H. Balakrishnan, “Adaptive
protocols for information dissemination in wireless sensor networks,”
in Proceedings of ACM International Conference on Mobile Computing
and Networking (MobiCom ’99), 1999, pp. 174–185.
Effects of different T values
[3] ——, “Energy-efficient communication protocol for wireless microsen-
950
sor networks,” in Proceedings of the 33rd Hawaii International Confer-
: LEACH ence on System Sciences, 2000, pp. 1–10.
900 : CM
: CM−IR [4] C. Intanagonwiwat, R. Govindan, and D. Estrin, “Directed diffusion: a
850
scalable and robust communication paradigm for sensor networks,” in
Proceedings of ACM International Conference on Mobile Computing
800 and Networking (MobiCom ’00), 2000, pp. 56–67.
[5] K. Kalpakis, K. Dasgupta, and P. Namjoshi, “Maximum lifetime data
round

750 gathering and aggregation in wireless sensor networks,” in Proceedings

700
650
600
550
0 50 100 150 200 250 300
clustering factor;T(0~500)
Fig. 6. The effects of different T values on the total number of rounds in
a wireless mobile sensor network which has 100 sensor nodes in an area of
200m×200m; each sensor node has initial energy 1.0J and constant speed of
0.5 m/sec.
lifetime increases. When T is small, in some cases, LEACH
is better than CM. However, CM-IR is always better than
LEACH. Since we are exploring the impact of the value of
T , we focus on protocol CM and LEACH. For the plot, we
observe that when T is larger than 3.75, protocol CM leads to
a longer system lifetime. Recall that the message transmission
step is the last step in a round and consumes most of the
energy. As T is larger than 3.75, T is closer to the time when
the message transmission step starts and CM protocol can have
better prediction on the location; therefore, leads to less energy
consumption for transmission in the system. The results accord
to the expectation discussed in Section IV. Suppose a round
starts at t1 and ends at t2 and δt = t2 − t1 . The clustering
factor T should be greater than or equal to t1 + 34 δt.
of the 2002 IEEE International Conference on Networking (ICN’02),
2002, pp. 685–696.
[6] L. Li and J. Y. Halpern, “Minimum-energy mobile wireless networks
revisited,” in Proceedings of the IEEE International Conference on
Communications (ICC ’01), 2001.
[7] S. Lindesy, C. Raghavendra, and K. M. Sivalingam, “Data gathering
350 400 450 500
algorithms in sensor networks using energy metrics,” IEEE Transactions
on Parallel and Distributed Systems, vol. 13, no. 9, pp. 924–935, 2002.
[8] D. Petrovic, R. C. Shah, K. Ramchandran, and J. Rabaey, “Data
funneling: Routing with aggregation and compression for wireless sensor
networks,” in Proceedings of the First IEEE International Workshop on
Sensor Network Protocols and Application, 2003, pp. 156–162.
[9] R. C. Shah and J. M. Rabaey, “Energy aware routing for low energy ad
hoc sensor networks,” in Proceedings of IEEE Wireless Communications
and Networking Conference (WCNC), 2002, pp. 17–21.
[10] Y.-C. Tseng, S.-Y. Ni, and E.-Y. Shih, “Adaptive approaches to relieving
broadcast storms in a wireless multi-hop mobile ad hoc network,” IEEE
Transactions on Computers, vol. 52, no. 5, pp. 545–557, 2003.
[11] S. Hedetniemi and A. Liestman, “A survey of gossiping and broadcasting
in communication networks,” Networks, vol. 18, no. 4, pp. 319–349,
1988.
[12] K. Sohrabi, J. Gao, V. Aliawadhi, and G. J. Pottie, “Protocols for self-
organization of a wireless sensor network,” IEEE Personal Communica-
tions, pp. 16–27, 2000.
[13] A. Savvides, C. Han, and M. Srivastava, “Dynamic fine-grained localiza-
tion in ad hoc network of sensors,” in Proceedings of ACM International
Conference on Mobile Computing and Networking (MOBICOM ’01),
2001.
[14] A. Cerpa, J.Elson, D. Estrin, L. Girad, M. Hamilton, and J. Zhao,
“Habitat monitoring: Application driver for wireless communication
technology,” in Proceedings of ACM SIGCOMM Workshop on Data
Communications in Latin America and the Caribbean, 2001, pp. 3–5.

0-7803-8521-7/04/$20.00 © 2004 IEEE 4639

0-7803-8521-7/04/$20.00 (C) 2004 IEEE