2014 IEEE 12th International Conference on Dependable, Autonomic and Secure Computing

A Spanning Tree based Data Collection for

Real-Time Streaming Sensor Data

Kyung Tae Kim, Jong Chang Park, Manyun Kim, Ung Mo Kim, Hee Yong Youn
College of Information and Communication Engineering
Sungkyunkwan University
Suwon, Korea
{harisu, jongchang, benimaru82, ukim, youn7147}@skku.edu

Abstract—Wireless sensor networks are widely used for protocols have been developed to organize and operate the
gathering data in the distributed fashion. Since the network network with energy efficiency as the primary criterion [6,7].
consists of low-cost nodes of limited battery power, it is a
challenging task to design an energy efficient routing scheme. In In this paper we propose a novel clustering algorithm for
this paper we propose a novel clustering algorithm based on the sensors of WSN based on bounded degree spanning tree for
spanning tree of bounded degree in each cluster for energy energy efficiency. In the proposed scheme each node elects
efficient WSNs. The proposed scheme selects the cluster-head itself as cluster-head (CH) with a probability, which is decided
considering the residual energy of each node, and partitions each by its energy level and operation time. Consequently, the
cluster for constructing efficient routing paths to the base station. possibility that the node of low energy to become CH is
Computer simulation shows that the proposed scheme effectively lowered, and as a result energy consumption of the nodes can
reduces and balances the energy consumption among the nodes, be balanced. After the clusters are formed, a tree of nodes is
and thus significantly extends the network lifetime compared to constructed within each cluster with the CH as the root of it.
the existing schemes such as LEACH, PEGASIS, and TREEPSI. Inside the tree, the routes from the leaf nodes to the root are
decided by the newly developed algorithm of spanning tree of
Keywords—Wireless Sensor Networks; Spanning Tree; Real- bounded degree. The resulting routes allow balanced data
time Streaming Data; Transmission Delay; Clustering. transmissions and energy consumption among the nodes. Each
node in a cluster sends the sensed data to their parent node,
I. INTRODUCTION while aggregates them to reduce the amount of data transferred.
Advanced integrated circuit technologies have led to the The CH fuses the data received from the member nodes, and
development of small sensor nodes equipped with sensing, data then transmits them to the BS. The CH selection and tree
processing and communication capability. Distributed in the construction occur in each round of operation. The proposed
target area, they form a network which can sense the scheme effectively distributes the load among the member
environmental data and react to the surrounding condition. This nodes so that their energy consumption can be balanced. Also,
makes the wireless sensor network (WSN) find a wide range of it extends the lifetime of the entire network by reducing the
civil and military applications including target field monitoring, frequency of CH selection and using the minimum diameter
intrusion detection, weather monitoring, tactical surveillance spanning tree of bounded degree in each cluster. Simulation
and disaster management, etc [1,2]. A WSN is composed of a results show that it achieves a significant improvement on
large number of sensor nodes and one base station (BS). The network lifetime compared to the representative routing
nodes are usually deployed in the region of interest in random schemes such as LEACH [8], PEGASIS [9], and TREEPSI
fashion, and the BS is engaged to provide commands to the [10].
sensor nodes and gather data sent from them [3]. The The remainder of the paper is organized as follows. Section
dissemination of commands and data collection are usually II presents the related work and Section III introduces the
made through the sensor nodes. With its capabilities of proposed scheme. Section IV evaluates the performance of the
monitoring and control, the WSN perceives the condition of schemes by computer simulation. Finally, Section V concludes
target area by integrating the local data collected from the the paper and outlines future research direction.
sensors [4,5].
Since sensor nodes have limited power supply which II. RELATED WORKS
cannot be recharged or replaced, their operation needs to be
energy efficient. If some sensor nodes deplete their energy, the A. Routing in WSN
WSN may not guarantee reliable operation due to partition of In general, the WSN gathers the data using hundreds or
the network. The limited energy in each node affects the even thousands of sensing nodes, where all data are collected
lifetime of the entire network, and thus energy efficiency has in the BS [11]. All the sensor nodes can directly transmit the
been a critical design issue for the protocols and algorithms sensed data to the BS. This single-hop transmission is simple,
developed for WSN. In the past various architectures and

978-1-4799-5079-9/14 $31.00 © 2014 IEEE 202

DOI 10.1109/DASC.2014.44
but not energy efficient for transmitting data to a node locating communicate with only their closest neighbors and take turns
far. It is desirable to let the nodes transmit data via multi-hop to communicate with the BS. Once all the nodes have assumed
short range communication. Moreover, the routing protocol the role of communication with the BS, a new round starts.
needs to work in the presence of the failure of some nodes. This allows power consumption to be uniformly spread over all
During the routing, local processing inside each node is nodes. Here communication occurs between only adjacent
required to reduce the bandwidth requirement due to the nodes, and thus the transmission power can be reduced. Unlike
limited wireless channel bandwidth shared by many sensor LEACH, PEGASIS avoids cluster formation and lets only one
nodes in the network. There exist many routing schemes node in the chain transmit data to the BS. To locate the closest
proposed for WSNs [12,13]. neighbor node for each node, a node uses the signal strength to
measure the distance to all neighboring nodes, and then adjusts
Diffusion-based routing. This approach employs the the signal strength so that only one node can be heard. The
flooding mechanism, where the nodes need to continuously chain construction is performed in a greedy fashion..
broadcast the data [14,15].
Cluster-based routing. The sensor nodes are grouped into D. Tree-based Routing
clusters, and the member nodes send the data to their CH. After In [10], the authors proposed a Tree-based Efficient
receiving all data in the cluster, the CH directly transmits them Protocol for Sensor Information (TREEPSI). Here a root node
to the BS [8,16,17]. is selected before the data transmission phase begins. There are
Chain-based routing. Here one or more chains are formed, two ways for building the tree path; decided centrally by the
and a chain head is selected in each chain. Each sensor node BS and then broadcast to the nodes, or decided by each node
sends the data to two neighbor nodes in the chain, and the using a common algorithm. At the initial phase of this scheme,
neighbor nodes forward it to the chain head after fusing it with the root executes the tree formation process with the nodes
their data. Finally, the chain heads directly transmit the following a tree traversal algorithm. The data transmission
aggregated data to the BS [9,18,19]. phase begins after the tree is built. All the leaf nodes send the
sensed data to their parent nodes. The parent nodes combine
Tree-based routing. At the initial phase, the BS builds a the received data with their own data and send them to their
tree covering entire sensor nodes. The root node of the tree parents. This process continues until the root node aggregates
aggregates the data and then transmits them to the BS [10]. the entire data, and directly sends them to the BS. The BS then
re-selects a new root node. The tree path is not changed until
B. Cluster-based Routing the root node becomes dead. TREEPSI and PEGASIS use the
Low-Energy Adaptive clustering Hierarchy (LEACH) [8] is same method to transmit data from the leaf node to the chain
one of representative clustering schemes. In LEACH sensors head or root, respectively. However, the length of the path
are organized into clusters and one node in each cluster acting from the leaf node to the root node in TREEPSI is usually
as CH takes the responsibility of data collection, aggregation, shorter than that to the chain head of PEGASIS. As a result,
and finally transmission to the BS. [8] showed that the network TREEPSI requires smaller data transmission power than
is optimal in the sense of energy dissipation when around five PEGASIS. Since it is based on binary tree topology, however,
percent of the total nodes act as CH. The operation of LEACH the path in some network might become inadvertently long.
is divided into rounds, where a round begins with set-up phase
followed by steady-state phase. During the setup phase each III. THE PROPOSED SCHEME
node decides by itself whether it will be the CH or not. After
We first discuss the system and energy model adopted in
the CHs are decided, they broadcast an advertisement message
the proposed routing scheme.
to the rest of the nodes. Depending on the received signal
strengths, each node decides the CH to which it belongs for
that round. Each CH then creates a TDMA schedule for all the A. System and Energy Model
member nodes in its cluster and sends it to them. During the We consider a WSN of a number of sensor nodes
steady state phase, the member nodes start sensing and distributed randomly in the target area. The sensing nodes
transmitting data to the CH according to the TDMA schedule, periodically form the clusters and they have enough
and the information extracted by the CHs are sent to the BS. At transmission power to reach the BS. The following
the end of the current round, a new set of nodes become the assumptions on the sensor nodes and the underlying network
CHs for the subsequent round and the whole process repeats. are made, which are common in the routing of WSN.
As the CH heavily drains the energy, LEACH facilitated
randomized rotation of CH in order to evenly distribute energy • All the sensor nodes are randomly distributed in a 2-
consumption among the nodes. dimensional plane according to a homogeneous spatial
Poisson process with Ȝ intensity.
C. Chain-based Routing • All nodes are homogeneous and have the same
capabilities.
In [9] an enhancement of LEACH protocol was proposed, • Each node is assigned a unique identifier.
called Power-Efficient Gathering in Sensor Information • All sensor nodes have the same initial energy.
Systems (PEGASIS). PEGASIS is a chain-based protocol
trying to extend the network lifetime by letting the nodes

203
 • When two communicating sensor nodes are not within ­ p Eresidual
°° ( r div s ) , if i ∈ G
each other’s radio range, the data are forwarded 1 Einit (3)
through other nodes. T (i ) = ®1 − p × ( r mod )
p
°
• The routing and MAC infrastructure are in place, and °̄0 otherwise
the communication environment is contention and
error-free. with p as the probability of being CH, r as the number of the
current round, s is the number of consecutive rounds in which
• Data fusion or aggregation is used to reduce the the node has not been CH, G is the set of nodes that have not
number of messages in the network, and The been CH in the last 1/p rounds. Eresidual is the residual energy of
communication channel is symmetric. node, and Einit is the initial energy. If the threshold is reduced,
We employ the radio model of the previous schemes [8- the probability of the node of high energy to be CH will
10,16], which is the first order radio model. It consists of increase. Therefore, the nodes of large residual energy have
transmit mode and receiving mode: in the transmit mode, the high probability to be CH due to the energy factor in the
energy consumption of a node mainly due to the energy threshold. To ensure even load distribution over the whole
consumption of signal amplifier and transmission circuit; in the network, this additional factor is included in the process of CH
receiving mode, it is mainly due to reception circuit. ETx and selection. This meliorates the robustness of the network,
ERx denote the energy consumed for sending and receiving extending the lifetime of the network. When a node is selected
messages, respectively. To radio expense required to transmit l as CH, it broadcasts an advertisement message (ADV) to the
-bit message to a distance, d, ETx(l, d) is expressed as: rest of the nodes. This is a small message containing the node’s
ID and a header identifying it as an announcement message.
Each non-CH node receiving the messages compares the signal
­lE + lİ d2 , d <d0 strengths of them, and selects the CH showing the largest
° elec fs strength. After that, it transmits a join-request message (Join-
ETx (l,d) = ETx-elec (l) +ETx-amp (l,d) = ®
4
°lEelec + lİ fsd , d ≥ d0 (1) REQ) to the chosen CH to inform that it will be a member node.
¯ After the clusters are formed, the configuration of spanning
tree begins in each cluster..
When a node receives l-bit data, the energy consumed is 2) Formation of Spanning Tree
When a tree is constructed with the member nodes inside
each cluster, the CH becomes the root. Refer to Fig. 1 where
ERx (l ) = ERx-elec (l ) = lEelec (2) the nodes form a cluster with the CH at the center, node-x.
Among the member nodes, node-y is farthest from the CH. If a
Here, ETx-elec, ETx-amp, and ERx-elec are the energy cost of the circle passing node-y is drawn with node-x as the center, thus,
transmitter circuit, transmitter amplifier circuit, and receiver all the member nodes in the cluster will be inside the circle, S.
circuit, respectively. Eelec is the energy taken to transmit or
receive a 1-bit message with the transmitter or receiver circuit;
İfs is the energy taken to send a 1-bit message with the
transmitter amplifier circuit. If the distance between two nodes
is smaller than the threshold, d0, the free space (d2 energy loss)
model is used. Otherwise, the multi-path (d4 energy loss)
model is used. For simplicity of calculation, the transmission
range of each node is assumed to be the same. Also, all data
packets contain the same number of bits.

B. The Proposed Approach

In this subsection we introduce the proposed scheme which
employs the tree-based clustering approach to evenly distribute
the energy load among the sensors in the network.
1) The Selection of Cluster-Head
The CH selection of the proposed scheme is similar to
Fig. 1. A cluster of nodes forming spanning tree.
LEACH. Here a fixed portion of sensors stochastically select
themselves as CHs as described in [8,16]. [8] analyzed the With the two semicircles of Fig. 1, it is explained how the
problem of the number of CHs for WSN, and they showed that spanning tree, T, is constructed with the nodes in the lower
5% of the nodes working as CHs allow good performance in circle, C. The spanning tree for the nodes of upper semicircle
homogeneous network with various parameter settings. In can be constructed similarly. Note that a spanning tree can be
order to select the CHs, each node determines a random constructed with an arbitrary node of S as the root by the same
number between 0 and 1. If the number is smaller than a way.
threshold, the node becomes CH for that round. The threshold
of node-i, T(i), is set as follows: First, C is partitioned into m equal circular sectors, Ki(i=1,
…, m), as shown in Fig. 2. In Fig. 2, for example, the area is

204
partitioned into three sectors, K1, K2, and K3. After that, the «¬ n / 2 i »¼ communication pairs can be generated in the ith slot.
node closest to the root in each sector is selected which is
called ‘sector root’. The sector roots become the children nodes Hence the minimum latency for completing packet
of node-x. It plays the role of root node of sub-tree formed with transmissions during the data collection phase is ( ǻ log 2 n ȼ + 1 )
the nodes in that sector. In the example of Fig. 2, node-p is the slots.
root of the tree in K2.

∧ ∧ ∧
K1 K 2 K3

Fig. 3. The operation of token passing mechanism.

Fig. 2. The partition of C. In order to employ the similar approach for data collection
within a cluster in the proposed scheme, the token passing
In the next step, each sector is partitioned again into m mechanism is used. As shown in Fig. 3, if node-a is elected as
subsectors. In Fig. 2, for example, the outer part of K2 is CH (root node), it sends a token to node-b which returns the
∧ ∧ ∧
partitioned into three subsectors, K 1 , K 2 , K 3 . Even though it token back to node-a. Here the numbers associated with the
links represent the order of token passing. Since node-a has
is not shown in the figure, the other sectors are also partitioned
other children nodes, it sends the token to node-c which then
into subsectors. A node closest to the sector root in each
returns the token back to node-a. Finally, node-a sends the
subsector is selected, and they are connected to the sector root.
token to node-d to collect the information. Since node-d has
They are called the ‘subsector root’, and node-p, q, r are
children nodes, it forwards the token to node-f which again
subsector roots of K2. Note here that a sector root can also be a
forwards the token to node-g. After node-g receives the token,
subsector root, if it is the closest node to the root node in the
it returns the token back to its parent, node-f, and finally node-
subsector such as node-p.
d. Node-d sends the token to node-e which then returns it to
The nodes in each subsector form a chain with the node-d. Since node-d has no more child node, it returns the
subsector root as head node of the chain. The root of entire tree token to node-a. The process continues until the token reaches
is at the center, and the sector roots compose the next level of the CH. Then, it executes the scheduling algorithm to generate
the tree. From the sector roots the nodes form a chain. Except the communication pairs. For the example of Fig. 3, three time
the root (actually the CH) the degree of sector roots is bounded slots are required to collect data, say S1 = {(g, f), (e, d), (c, a)},
by m+1 (m connections to the subsector roots and one to the S2 = {(f, d), (b, a)} and S3 = {(d, a)}. The direction of
root). The degree of subsector root is 2 (one connection to transmission for each communication pair is also identified..
sector root and the other to the neighboring node in the chain).
This property of bounded degree allows even distribution of IV. PERFORMANCE EVALUATION
the load among the nodes, and eventually maximized network
lifetime. In this section we evaluate the performance of the proposed
scheme using a simulator implemented in C++, and compare
3) Data Collection and Transmission the performance with the existing protocols. The simulation
After the selection of CHs and tree construction are over, program evaluates the amount of energy consumption, network
the data collection phase begins. At the beginning of this phase lifetime, number of messages received, and number of live
each CH accumulates the data sent from the member nodes of nodes with the protocols compared. Table I provides the
its cluster. This phase may take long time if the cluster size is parameters used in the simulation.
large. Data collection latency is defined as the number of time
slots taken for delivering data packets from all the nodes to the
TABLE I. THE PARAMETERS USED IN THE SIMULATION
BS. It is especially important to time-critical applications such
as battlefield surveillance and fire detection. For such Parameter Value
applications, it is crucial to minimize the data collection Network model
latency. In this regards, a scheduling approach was proposed in Network size 100m × 100m, 200m × 200m
SHORT [20]. Here as many communication pairs (parallel Location of BS Center of target area
packet transmissions) as possible are generated in each time Number of nodes (100 / 200)
slot. For a network of n sensor nodes and one CH, at most Data packet size 500 byte
Energy model

205
 Initial energy of each sensor 1.0J 100 sensor nodes distributed in 100m×100m area where each
Eelec 50 nJ/bit node begins with an initial energy of 1.0J. The figure clearly
fs
10 pJ/bit/ m2 illustrates the effectiveness of the proposed scheme in
mp
0.0013 pJ/bit/ m4 delivering significantly more messages than its counterparts.
EDA 5 nJ/bit/signal
Ω͑͢͡ᒝ ͽͶͲʹ͹ ͽͶͲʹ͹͞ʹ ΅΃ͶͶ΁΄ͺ ΁ΣΠΡΠΤ ΖΕ
ͥ͢

In the simulation the BS is assumed to be fixed and located

ͿΦΞΓΖΣ͑ΠΗ͑ΞΖΤΤΒΘΖΤ͑ΣΖΔΖΚΧΖΕ
ͣ͢

at the center of the target area, and it has unlimited power and ͢͡

other resources. In order to study how the proposed scheme

ΒΥ͑ΥΙΖ͑ͳ΄
ͩ
works with the network of different densities, the simulations
are conducted with two different sizes of network. 100 ͧ

simulation runs are executed, and the results are averaged. In ͥ

the simulations the radio model presented in Section III.A is ͣ

used [8-10,16]. For the simplicity, an ideal MAC layer and ͡

error-free communication links are assumed. ͡ ͢͡͡ ͣ͡͡ ͤ͡͡ ͥ͡͡ ͦ͡͡ ͧ͡͡ ͨ͡͡ ͩ͡͡ ͪ͡͡ ͢͡͡͡ ͢͢͡͡ ͣ͢͡͡
΅ΚΞΖ͑ΤΥΖΡΤ͙ΣΠΦΟΕΤ͚

ͽͶͲʹ͹ ΁Ͷ͸Ͳ΄ͺ΄ ΅΃ͶͶ΁΄ͺ ΁ΣΠΡΠΤ ΖΕ

͢͡͡
Fig. 5. The total number of messages received by BS.

ͩ͡
In WSNs the ultimate goal of energy saving is to prolong
΃ ΖΤΚΕΦΒΝ͑ͶΟΖΣΘΪ͙͑ͻ͚

ͧ͡
the network lifetime. In other words, ‘reduced energy
consumption’ means ‘prolonged lifetime’ in WSNs. The
ͥ͡ network lifetime is indicated by the number of live sensor
nodes. The two graphs of Fig. 6 show the network lifetime for
ͣ͡ two different network sizes of 100m×100m and 200m ×200m.
͡ ͽͶͲʹ͹ ΁Ͷ͸Ͳ΄ͺ΄ ΅΃ͶͶ΁΄ͺ ΁ΣΠΡΠΤΖΕ

͡ ͢͡͡ ͣ͡͡ ͤ͡͡ ͥ͡͡ ͦ͡͡ ͧ͡͡ ͨ͡͡ ͩ͡͡ ͪ͡͡ ͢͡͡͡ ͢͢͡͡ ͣ͢͡͡ ͢͡͡

΅ΚΞΖ͑ΤΥΖΡΤ͙ΣΠΦΟΕΤ͚ ͪ͡

ͩ͡
(a) 100 nodes
ͿΦΞΓΖΣ͑ΠΗ͑ΝΚΧΖ͑ΟΠΕΖΤ

ͨ͡

ͧ͡
ͽͶͲʹ͹ ΁Ͷ͸Ͳ΄ͺ΄ ΅΃ͶͶ΁΄ͺ ΁ΣΠΡΠΤ ΖΕ
ͦ͡
ͣ͡͡
ͥ͡

ͤ͡
ͧ͢͡
΃ ΖΤΚΕΦΒΝ͑ͶΟΖΣΘΪ͙͑ͻ͚

ͣ͡

͢͡
ͣ͢͡
͡
͡ ͢͡͡ ͣ͡͡ ͤ͡͡ ͥ͡͡ ͦ͡͡ ͧ͡͡ ͨ͡͡ ͩ͡͡ ͪ͡͡ ͢͡͡͡ ͢͢͡͡
ͩ͡ ΅ΚΞΖ͑ΤΥΖΡΤ͙ΣΠΦΟΕΤ͚

(a) 100m×100m
ͥ͡

ͽͶͲʹ͹ ΁Ͷ͸Ͳ΄ͺ΄ ΅΃ͶͶ΁΄ͺ ΁ΣΠΡΠΤ ΖΕ

͡ ͢͡͡
͡ ͢͡͡ ͣ͡͡ ͤ͡͡ ͥ͡͡ ͦ͡͡ ͧ͡͡ ͨ͡͡ ͩ͡͡ ͪ͡͡ ͢͡͡͡ ͢͢͡͡ ͣ͢͡͡ ͤ͢͡͡ ͥ͢͡͡ ͦ͢͡͡ ͧ͢͡͡
ͪ͡
΅ΚΞΖ͑ΤΥΖΡΤ͙ΣΠΦΟΕΤ͚
ͩ͡
ͿΦΞΓΖΣ͑ΠΗ͑ΝΚΧΖ͑ΤΖΟΤΠΣΤ

(b) 200 nodes ͨ͡

ͧ͡

Fig. 4. The total remaining energy of the network. ͦ͡

ͥ͡

Fig. 4 depicts the total remaining energy of the network ͤ͡

with LEACH, PEGASIS, TREEPSI, and the proposed scheme

ͣ͡

͢͡
for two different node densities. The results indicate that the ͡
residual energy of the proposed scheme is significantly larger ͡ ͢͡͡ ͣ͡͡ ͤ͡͡ ͥ͡͡ ͦ͡͡
΅ΚΞΖ͑ΤΥΖΡΤ͙ΣΠΦΟΕΤ͚
ͧ͡͡ ͨ͡͡ ͩ͡͡ ͪ͡͡ ͢͡͡͡

than the other three schemes, particularly in the early rounds.

This is because it allows better cluster formation and short data (b) 200m×200m
transmission path, with appropriate number of clusters and
Fig. 6. The comparison of network lifetime.
configuration of spanning tree of bounded degree in each
cluster. In LEACH, the data packets from some member nodes
The improvement gained through the proposed protocol is
may need to be transmitted to the CHs over long distance. It
further identified by the network lifetime shown in Fig. 6(a),
leads to large energy consumption during data transmission.
which shows the number of nodes remaining alive over the
Fig. 5 shows the total number of messages received by the number of rounds. Notice from the figure that with the
BS over time. For this experiment, we consider a network of proposed protocol, all nodes remain alive until 682 rounds,

206
while the corresponding number for LEACH, PAGASIS, and Program 2014 (1391105003). Corresponding author: Hee
TREEPSI is 413, 496, and 579, respectively. Also the proposed Yong Youn.
protocol always allows a larger number of live nodes than
other protocols. This is due to the conspicuous selection of REFERENCES
CHs considering the residual energy and configuration of
[1] I. F. Akyildiz et al., “Wireless sensor networks: a survey,”
spanning tree with bounded degree in each cluster. Computer Networks, vol. 38, no. 4, March 2002, pp. 393–422.
In Fig. 6(b), 100 nodes are placed in a 200m ×200m area [2] Y. A. Obaisat and R. Braun, “On Wireless Sensor Networks:
with the BS located at (100,100). Observe that the proposed Architectures, Protocols, Applications, and Management,” IICT,
2007, pp. 1–11.
protocol consistently outperforms other protocols even for
[3] D. Culler, D. Estrin, and M. Srivastava, “Overview of sensor
larger size network. In LEACH, as data are transmitted directly networks,” IEEE Computer, vol. 37, no. 8, 2004, pp. 41–49.
from the member nodes to the CH, many long distance [4] M. A. M. Vieira, C. N. Coelho Jr., D. C. da Silva Jr., and J. M. da
transmissions are required inside a cluster. The number of Matam, “Survey on wireless sensor network devices,” in Proc. of
transmissions of long distance will increase as the network size IEEE ETFA, vol. 1, 2004, pp. 537–544.
grows, and consequently does the overall energy consumption [5] S. J. Choi, K. T. Kim, and H. Y. Youn, “An Energy-Efficient Key
of the network. In PEGASIS energy is saved since the number Predistribution Scheme for Secure Wireless Sensor Networks Using
of nodes delivering the packets to the BS is reduced to one. Eigenvector”, International Journal of Distributed Sensor Networks,
However, the chain formed in PEGASIS may have long edges, Vol.2013, pp.1-13, June 2013.
especially the link to the BS. As a result, the nodes connected [6] P. Kumar, M. P. Singh, and U. S. Triar, “A Review of Routing
to these long links quickly deplete the energy and thus limit the Protocols in Wireless Sensor Network,” International Journal of
lifetime of the entire network. Although TREEPSI can alleviate Engineering Research & Technology, vol. 1, 2012, pp. 1–14.
this problem, the path may contain a detour in the topology. [7] K. T. Kim and H. Y. Youn, “An Energy Efficient MAC Protocol
Employing Dynamic Threshold for Wireless Sensor Networks”,
Also, the tree path does not change until the root node dies International Journal of Distributed Sensor Networks, Vol.2012,
which leads to imbalance of energy consumption among the pp.1-12, October 2012..
nodes. Fig. 6 confirms that the proposed scheme significantly [8] W. R. Heinzelman, A. Chandrakasan, and H. Balakrishnan,
extends the network lifetime by avoiding these problems. “Energy-Efficient Communication Protocol for Wireless Micro-
sensor Networks,” in Proc. of HICSS, Maui, Hawaii, 2000.
V. CONCLUSION AND FUTURE WORK [9] S. Lindsey, C. S. Raghavendra, “PEGASIS: Power-Efficient
gathering in sensor information systems,” in Proceedings of the
In this paper we have proposed a novel clustering algorithm IEEE Aerospace Conf., Canada, March 2002, pp. 1–6.
based on spanning tree in each cluster for energy efficient [10] S. S. Satapathy and N. Sarma, “TREEPSI: tree based energy
wireless sensor networks. In the proposed scheme each node efficient protocol for sensor information,” in Proc. of 2006 IFIP
elects itself to be CH by a probability, which was decided WOCN, 2006.
considering the energy level of the nodes to optimize the [11] X. Liu, “A survey on clustering routing protocols in wireless sensor
selection of CH. Also, the proposed scheme introduces networks,” Sensors, vol. 12, 2012, pp. 11113–11153.
spanning tree of bounded degree for connecting the nodes in [12] S. Saranya and M. Princy, “Routing Techniques in Sensor
Network–A Survey,” Procedia Engineering, vol.38, 2012, pp. 2739–
each cluster. The proposed protocol achieves substantial 2747.
performance enhancement in terms of network lifetime by
[13] J. N. Al-Karaki and A. E. Kamal, “Routing Techniques in Wireless
balancing the energy load among the nodes. Also, it minimizes Sensor Networks: a Survey,” IEEE Wireless Communications, vol.
the energy consumption required for the communication 11, no. 6, 2004, pp. 6–28.
between the nodes. Simulation results show that the proposed [14] Zhao, S. Yu, F. Zhao, and Baohua, “An Energy Efficient Directed
scheme successfully balances the energy consumption among Diffusion Routing Technique,” in Proc. of CIS, 2008, pp. 1067–
the nodes and thus extends the network lifetime compared to 1072.
the existing schemes such as LEACH, PEGASIS, and [15] C. Intanagonwiwat, R. Govindan, and D. Estrin, “Directed
TREEPSI. Diffusion: a Scalable and Robust Communication Paradigm for
Sensor Networks,” in Proc. of ACM MobiCom, 2000, pp. 56–67.
There are several issues required to investigate in the future. [16] W. B. Heinzelman, A. Chandrakasan, and H. Balakrishanan, “An
First, the network lifetime could be extended further without Application-Specific Protocol Architecture for Wireless
affecting other performance metrics by clustering the nodes in Microsensor Networks,” IEEE Transactions on Wireless
a better way. Second, the tree construction can be improved Communications, vol. 1, 2002, pp. 660–670.
considering the distribution of the nodes inside the cluster. [17] K. T. Kim and H. Y. Youn, “A Stochastic and Optimized Energy
Finally, some constraints made with the proposed protocol Efficient Clustering Protocol for Wireless Sensor Networks",
International Journal of Distributed Sensor Networks, Vol.2014,
could be relieved to make the scheme more robust. March 20 2014 pp.1-12.
[18] S. Mahajan and J. Mahotra “A Novel Chain Based Wireless Data
ACKNOWLEDGMENT Sensor Network (ECBSN) Technique,” IJCST, vol. 2, 2011, pp. 83–
87.
This research was supported by Basic Science Research
[19] S. Mahajan, J. Malhotra, and S. Sharma, “Improved Enhanced
Program through the National Research Foundation of Korea Chain Based Energy Efficient Wireless Sensor Network,” Wireless
(NRF) funded by the Ministry of Education, Science and Sensor Network, vol.5, 2013, pp. 84–89.
Technology (2013R1A1A2040257 and 2013R1A1A2060398), [20] Y. Yang, H. Wu, and H. Chen, “SHORT: shortest hop routing tree
the second Brain Korea 21 PLUS project, and MSIP (Ministry for wireless sensor networks,” International Journal of Sensor
of Science, ICT & Future Planning), Korea in the ICT R&D Networks, vol. 2, 2007, pp. 368–374.

207