Energy Efficient Multipath Data Fusion Technique for Wireless Sensor Networks

ACEEE Int. J. on Network Security , Vol. 03, No. 02, April 2012

Energy Efficient Multipath Data Fusion Technique for
Wireless Sensor Networks
Sasikala V1 and C. Chandrasekar2
1
Professor, DRBCCC Hindu College, Chennai, Tamil Nadu
vsasikalaphd@gmail.com
2
Associate Professor, Dept. of Computer Science,
Periyar University, Salem – 636 011, Tamil Nadu, India.

Abstract—In wireless sensor networks (WSN), data fusion effectively. [4]
should be energy efficient. But, determining the optimal
number of aggregators in an energy efficient manner is a B. Existing Data Fusion Techniques in Wireless Sensor
challenging task. Moreover, the existing data fusion Networks
techniques mostly use the same path for transmitting  Witness based data fusion: In the witness based data
aggregated data to the sink which reduces the nodes lifetime.
fusion, the data fusion node doesn’t forward its result to the
In this paper, we propose a technique which combines energy
efficiency and multiple path selection for data fusion in WSN.
base station but will compute the Message Authentication
The network is partitioned into various clusters and the node Code (MAC) of the result (they call the MAC a proof). After
with highest residual energy is selected as the cluster head. receiving this information the data fusion node forwards the
The sink computes multiple paths to each cluster head for proofs to the base station. The data fusion node has to create
data transmission. The distributed source coding and the false proofs on the invalid result if the node has to be
lifting scheme wavelet transform are used for compressing compromised and wanted to send an invalid fusion result to
the data at the CH. During each round of transmission, the the base station. [5]
path is changed in a round robin manner, to conserve the
Dynamic data fusion: In sensor networks, the dynamic
energy. This process is repeated for each cluster. From our
simulation results we show that this data fusion technique application specified data fusion is supported by an
has less energy consumption with increased packet delivery architectural framework known as DFuse. The dynamic nature
ratio, when compared with the existing schemes. of applications in sensor networks is considered by the
advanced fusion applications. This dynamic data fusion
Index Terms— wireless sensor networks (WSN), Data Fusion bridges an important abstraction gap for developing these
Techniques, distributed source coding (DSC), Modified Unary applications.
Coding (MUC), Energy Efficient Fusion Techniques.
 Multi-sensor data fusion: Diverse sensors like temperature,
humidity light, and Carbon Monoxide are set in each sensor
I. INTRODUCTION node. The information about the environmental condition
A. Data Fusion in Wireless Sensor Networks can be provided by using more than one sensor. The fuzzy
In sensor networks, data fusion is an essential service. In rule based system helps in processing and fusion of these
order to compute useful information, such as average of all diverse sensor signals. [6]
the sensor readings, the maximum value among the sensor  Single Mobile agent (MA)-based autonomic data fusion:
readings or the number of sensors that detect an event the Here for autonomic data fusion, only one MA is used. In
data fusion process aggregates the data from independent small scale WSNs this approach is effective but in the
sensors. [1] By eliminating redundancy and power networks comprising hundreds or thousands of sensor nodes
consumption in data fusion, the performance of a network these solutions doesn’t scale acceptably.
can be increased. This is because the fault-tolerance between Multiple MA-based autonomic data fusion: In order to
the sensors is ensured and the available communication fuse the data from WSN sensors, in this approach number of
bandwidth between network components is effectively MAs are working parallel. Large scale WSNs also supports
managed. [2] this fusion technique. The itineraries of individual MAs are
For the purpose of energy-efficient information flow from derived using the relatively complex algorithms. [7]
several sensors to a central server or sink, in-network data Mobile agent based clustering data fusion: In this data
fusion is needed in wireless sensor networks. To effectively fusion, two cluster head models are used to control the size
fuse the data, the data fusion techniques should be of the clusters. All the sensor nodes in the detection region
synchronized at various levels. The credibility of the are divided into several clusters and the fault nodes are
aggregated report can be increased, by fusing the information removed through the partial results of data integration. The
from as many sensors as possible. [3] On considering the mobile agents are used in between the cluster heads for data
measurements of multiple sensors, the system performance fusion, and the path of the mobile agent is optimized. [8]
can be improved in data fusion which is a widely adopted
signal processing technique. Enabling collaboration among
the sensors improves the system sensing performance
© 2012 ACEEE 34
DOI: 01.IJNS.03.02.14


C. Need for Power Efficient Data Fusion in Wireless Sensor is changed in a round robin manner, to conserve the energy.
Networks This process is repeated for each cluster.
 The redundancy among sensed data and the network Thus this data fusion technique is energy efficient and
load can be reduced by exploring data correlation and involves multiple paths for transmission of data.
employing in-network processing. [9]
 Though the wireless sensor networks have infinite scopes, II. RELATED WORK
they have limited node battery lifetime. When there is Frank Yeong-Sung Lin et al [14] have proposed this paper
sufficient battery power, the network can be operated after which considers the energy consumption tradeoffs between
deployment. Once a battery is deployed over an inaccessible data aggregation and retransmissions in a wireless sensor
area it is impossible to replace the node battery and this network. By using the existing CSMA/CA (Carrier Sense
problem needs to be considered mainly. [10] Multiple Access with Collision Avoidance) MAC protocol,
Failures often occur in wireless sensor networks, since the retransmission energy consumption function is well
they are not considered for long periods of time in the field. formulated. This paper proposes a novel non-linear
Sensors running out of energy, ageing or harsh environmental mathematical formulation, whose function is to minimize the
conditions surrounding them are the reasons for these total energy consumption of data transmission subject to
failures. [11] data aggregation trees and data retransmissions.
 The sensor nodes must rely on small, usually non- Rabindra Bista et al [15] have proposed a new energy
renewable batteries, energy efficiency is considered as a most balanced and efficient approach for data aggregation in
important design concern in sensor networks. Consumption wireless sensor networks, called Designated Path (DP)
of energy in a sensor node is processed in data acquisition, scheme. In DP scheme, a set of paths is pre-determined and
processing and communication. The data transmission among run them in round-robin fashion so that all the nodes can
the sensor nodes is reduced to a minimal level since the participate in the workload of gathering data form the network
transmit power is governed in many of the applications. [12] and transferring the data to the sink node. They use Semantic
 Due to limited detection range and reliability of each Routing Tree (SRT) for disseminating any kind of aggregation
node, we need to make the monitoring range overlap each query to get aggregated value.
other, so that the accuracy and robustness of the network V. Bhoopathy et al [16] have proposed Energy Efficient
can be enhanced. Therefore, the data in the sensor nodes Secure Data Aggregation Protocol. First the network is
maintains certain redundancy. The redundant information divided into clusters, each cluster is headed by an aggregator
should be reduced and the energy has to be saved to prolong and the aggregators are connected to sink either directly or
the network lifetime and so each node transmits its detection through other aggregators. The aggregator is selected based
data to the sink node in the routing. [8] on the nearest distance to a set of sensor nodes and its
 The cost and continual power consumption of sensors energy level. Separate keys are distributed to the two levels
are high since extra circuits are required in hardware based i.e., sensor node to the aggregator and aggregator to the
approach in order to detect or frustrate the compromised sink. Whenever a sensor node wants to send data to another
node. [13] node; first the sensor node encrypts the data using a key
D. Problem Identification and Solution and sends it to the aggregator. For integrity of the data packet,
a MAC based authentication code is used.
In Wireless Sensor Networks (WSN), the data fusion
Yuanzhu Peter Chen et al [17] have proposed an Energy-
should be energy efficient. But the existing data fusion
Efficient Protocol for Aggregator Selection (EPAS). Then,
techniques mostly use the same path for transmitting
they generalize it to an aggregation hierarchy and extend
aggregated data to the sink, which results in reduced lifetime
EPAS to a Hierarchical Energy-Efficient Protocol for
of the nodes along the paths. Moreover, determining the
Aggregator Selection (hEPAS). They derive the optimal
optimal number of aggregators in an energy efficient manner
number of aggregators with generalized compression and
is a challenging task.
power consumption models, and present fully distributed
Here, we propose a technique which combines the energy
algorithms for aggregator selection.
efficiency and the multiple path selection for data fusion in
M. Umashankar et al [18] have proposed a novel power-
WSN. We assume multiple paths from each cluster to the
efficient data fusion assurance scheme using silent negative
sink. Initially, the nodes form a cluster and the number of
voting mechanism and data fusion assurance with random
aggregators that minimizes the total energy consumed by
key pre-distribution scheme has been proposed. The
transmitting and aggregating data is determined. Each sensor
proposed schemes has been compared and evaluated their
selects the closest aggregator as its cluster head. Then the
efficiency with a simple MAC based fusion assurance scheme
sensors send packets to their respective aggregator. Each
as well as the direct voting based fusion assurance scheme.
aggregator compresses the data it receives from the sensors
Huseyin Ozgur Tan et al [22] have proposed localized
of its cluster and finally forwards the data to the sink.
power efficient data aggregation protocols (L-PEDAPs) for
In the initial round, from the aggregators, the aggregated
wireless sensor networks. Their protocol is based on
data is transmitted to the sink using one of the established
topologies such as local minimum spanning tree (LMST) and
multiple paths. During each round of transmission, the path
© 2012 ACEEE 35
DOI: 01.IJNS.03.02.14


relative neighborhood graph (RNG). Their solution involves When the energy of the cluster head goes down, the network
route maintenance procedures and it is also adapted to performance degrades. In order to balance the network energy
consider the remaining power levels of nodes in order to consumption, the clustering needs to ensure that energy
increase the network lifetime. dissipation across the network should be balanced and cluster
D. Kumar et al [23] have proposed a novel Energy Efficient head should keep on changing.
Clustering and Data Aggregation (EECDA) protocol for the  The initiator I broadcasts a request message for
heterogeneous WSNs. This protocol includes a novel cluster energy ( E REQ ) with its own energy level ( RLini )
head election technique and a path would be selected with
maximum sum of energy residues for data transmission instead information to its surrounding nodes.
of the path with minimum energy consumption.  The sensor node S i compares its own energy level
Hasan Cama et al [24] have proposed energy-efficient
( RLi ) with the initiator.
.
secure pattern based data aggregation (ESPDA) for wireless
sensor networks. ESPDA prevents the redundant data  If RLi  RLini , then,
transmission from sensor nodes to cluster-heads. They have sends a reply message for energy (EREP).
also presented a security protocol and NOVSF block-hopping Else
technique that provides data communication security.
waits for cluster head advertisement messages. (CH ADV ) .
Mohamed Watfa et al [25] have proposed an energy-
efficient approach to query processing by implementing new The initiator selects the cluster head with maximum residual
optimization techniques applied to in-network aggregation. energy and the next initiator node is the node having the
They did not concentrate on the data routing strategies and second maximum residual energy.
loss tolerance mechanism. The initiator node is changed every time when the energy
Xu Li et al [26] have proposed a novel two-stage delay level of the node decreases.
model based on IEEE 802.11 CSMA/CA MAC layer. This After CH is selected by the initiator, clusters are formed in
model uses hop count to measure end-to-end delay when the network.
the network traffic is low, and degree sum along the routing The nodes in the cluster broadcasts a CHADV to the CH and
path when the traffic is high. They introduced the novel CH sends it to the sink along with the cluster ID.
concept of DEsired Progress (DEP) for hop selection and A join request message JREQ is transmitted by the member
devised a localized delay-bounded and energy-efficient data node along with CHADV.
aggregation (DEDA) protocol accordingly. The transmission range gets minimized since the initiators
Siddhartha Chauhan et al [27] have proposed an energy collect the energy information about the nearest sensors.
efficient data gathering protocol (EEDGP) for wireless sensor The nodes having energy greater than the energy level of
network. This protocol reduces the transmission of the data the initiator ensure minimization of E REP message
packets thereby reducing the energy consumption of sensor transmission.
nodes. Once the selected cluster head node receives the J REP
Mohammad Mostafizur Rahman Mozumdar et al [28] have
message from member nodes, it sends a joint reply message
proposed an efficient data aggregation algorithm for
JREP back to the nodes. Then the CH transmits data to the
Cluster-based sensor network. The proposed algorithm
sink node.
selects a cluster leader that will perform data aggregation in a
partially connected sensor network. The algorithm reduces
the traffic flow inside the network by adaptively selecting
the shortest route for packet routing to the cluster leader.

III. PROPOSED WORK
A. Cluster Head Selection and Data Transmission
In this technique, we select the node with highest residual
energy as the cluster head in order to prolong the lifetime of
the network. We randomly consider few nodes as the initiator
nodes (I) to collect the information of the nearest sensor
nodes and to select the cluster heads depending upon the
energy information.
We assume that in this sensor network, the sink node has
the knowledge about all other node’s location. The sensor
nodes are assumed to be immobile and have limited energy.
The initiator node (I), determines the CH based upon the
residual energy of the nodes. In a homogeneous network,
cluster head uses more energy than non cluster head nodes. Figure 1. Selection of Cluster Head
© 2012 ACEEE 36
DOI: 01.IJNS.03.02.14


supported in LSWT.
The individual redundancy can be reduced by the scalar
quantization after the LSWT process. The distributed
redundancy is reduced when low frequency component of
the sensor data is processed by DSC.
In the first step, we do the LSWT and it gets repeated by
iteration on the S (even part of original data). A multi-level or
multi resolution decomposition is created due to this.
In the next step, quantization, we reduce the individual
redundancy by choosing a scalar quantization in our
applications. For data compression combined with WT,
quantization is widely studied and successfully implemented.
The data values which have higher frequencies are set to
zero after the quantization. The individual redundancy can
be reduced significantly in the quantization process.
The high frequency data sets are encoded using the
Figure 2. Data transmission from CH to Sink Modified Unary Coding (MUC). The nonzero values {Ai}
In figure 1, the Initiator I1, I2, and I3 sends energy request are only encoded here.
EREQ to all the surrounding nodes. For example, in this figure, Ai  0 :- encoding with 2  Ai bits 1 plus 10 bits relative
when the node I1 sends E REQ to the nodes, the energy position value.
consumption of the nodes is compared with I1. Since node 2 Ai  0 :- encoding with 2 Ai  1 bits 1 and 10 bits relative
has a higher energy level than I1 and so node 2 is selected as position value.
the cluster head and node 4 which has the next highest energy H(Y) is a sample value with n bits used for encoding the
level is selected as the next initiator for cluster head selection. low frequency part when each value is mapped to the source
Similarly, when the node I2 sends E REQ to the nodes, codebook ranging from, 0 to2n-1. This is known as base data.
The compress process ends and base data is sent to the base
the energy consumption of the nodes is compared with I2.
station when the data is collected from the sensor node else
Here, node 9 has the highest energy level and it is selected as
it will continue to the next step. The distributed redundancy
the Cluster head and the node 11 which has the next highest
is still more reduced by DSC after mapping. Portioning is
energy level is selected as the next initiator for cluster head
done in the data set to split it into different cosets as the
selection.
correlation of the data from neighbor layers. The fully
Then when the node I3 sends to the surrounding nodes,
compressed data is the original data value which is replaced
the energy consumption of the nodes is compared with I3.
with coset index represented by H(X|Y).
Since node 22 has the highest energy level, it is selected as
After compressing the data the base station collects these
the cluster head and no other surrounding node has the next
data. Decompression of all the data is done when the base
highest energy level other than initiator I3. So, we select
station receives the collected data.
node I3 as the next initiator node.
In figure 2, the data transmission from CH to the sink is C. Multi path Routing from Cluster Head to Sink
described. After the selection of CH, the clusters are formed. For data transmission from CHs towards the sink node,
The nodes in the cluster transmit CH ADV with J REQ to the multiple paths are created in the multi path discovery phase.
CH. The CH then transmits this data to the sink. These multi-paths are node disjoint. Due to the utilization of
most available network resources, the multi-path routing
B. Data Compression usually prefers the node disjoint paths and thus they are
The distributed source coding (DSC) is explained in the fault tolerant. There is a minimum impact to the diversity of
paper [19]. the routes since when an intermediate node in a set of node
In WSNs we construct an algorithmic framework which disjoint paths is failed, only the path containing that node
supports DSC for high and low frequency signal gets affected.
compression. To preprocess the original data for signal The path discovery procedure is executed according to
decomposition and noise deduction, we use a lifting scheme the following phases:
wavelet transform (LSWT) in order to separate the low Initialization phase: The information about the neighbors
frequency component from the high frequency component, having highest quality data is maintained by each CH since it
and strength the correlation among distributed sensor data. broadcasts a HELLO message through the clusters. The
Compared to the traditional transforms, LSWT is better neighboring table is maintained and updated in this phase.
suitable for WSNs due to the following reasons: The knowledge about the list of neighboring nodes of the
1) Efficiency is higher compared to the FFT or DCT. cluster head are maintained in the neighboring table. Hop
2) Time domain analysis is supported by transformed data. count which represents the distance in hops for message
3) Multi-scale analysis and integer to integer mapping are from its originator is present in the HELLO message.
© 2012 ACEEE 37
DOI: 01.IJNS.03.02.14


Primary path discovery phase: The information on D. Combined Algorithm
computing the cost function for CHs neighboring nodes is 1. Among the N sensor nodes, randomly consider k
contained in the CH after initialization phase. After the nodes as the initiator nodes to collect the information.
preferred next hop CH is computed by the sink node an RREQ 2. The initiator I1 selects the cluster head based upon the
message is sent to the most preferred next hop. In the same energy level information.
way, the preferred next hop CH of the sink computes its most For each neighbor Ni of I1 , i  1, 2  r
preferred next hop in the source node’s direction. An RREQ
If RLi  RLi  RL ini , then,
message is sent to the next hop and the operation continues
until the source node. S i sends a reply message for energy (EREP).
Alternative Paths discovery phase: The next most Else
preferred neighbor is considered as the alternative path and waits for cluster head advertisement messages .
the sink sends an alternate RREQ message to that neighbor. End For
Each node accepts only one RREQ message in order to avoid 3. I1 select the node St as CH1 such that RLt = max{RLi},
having paths with shared nodes. When two or more nodes 4. CH1 broadcast a to Ni
receive one RREQ message, the first RREQ message is only 5. Each node Ni sends request to CH1.
accepted rejecting the remaining messages. 6. On accepting from the CH, nodes join the cluster.
For example in this figure 3, we compute multiple paths 7. Each CH send its cluster details to the sink.
from the cluster head to the sink. The source node transmits 8. The sink establishes multiple paths towards each CH and
the data to the cluster head CH4 and the CH4 wants to transmit designates one path as the primary path.
the data to the sink. So, it computes four paths: 9. At first cycle R1, sends the sensed data to its CH.
Path 1: Source-CH4-CH2-Sink 10. At CH, The data from its members is compressed using
Path 2: Source-CH4-CH3-Sink DSC.
Path 3: Source-CH4-CH3-CH1-Sink 11. CH sends the compressed data to the sink, using the
Path 4: Source-CH4-CH5-CH3-CH1-Sink primary path.
The best path is the path with minimum number of hops. 12. After the sink receives the compressed data, it
We select the path 1 (CH4-CH2-Sink) for transmission decompresses all the data.
When a failure occurs in that primary path the alternative 13. During the next cycle R2, CH chooses the next alternative
path is selected. Path 2 (CH4-CH3-CH1-Sink) is chosen as path from the multi path set, to transmit the data.
the alternative path for further transmission. The path is 14. If all the alternate paths are used, then in the next cycle,
changed in a round robin manner in order to conserve energy. again the primary path is selected.

IV. SIMULATION RESULTS
Energy Efficient Multipath Data Fusion (EEMD)
Technique is evaluated through NS2 [21] simulation. A
random network deployed in an area of 500 X 500 m is
considered. Initially 100 sensor nodes are placed in square
grid area by placing each sensor in a 50x50 grid cell. 10 cluster
heads are deployed in the grid region according to our
protocol. The sink is assumed to be situated 100 meters away
from the above specified area. In the simulation, the channel
capacity of mobile hosts is set to the same value: 2 Mbps.
The simulated traffic is CBR with UDP source and sink. The
number of sources is per cluster is varied fro 1 to 4.
Table 1 summarizes the simulation parameters used
TABLE 1: SIMULATION PARAMETERS

Figure 3. Multi-path clustering
© 2012 ACEEE 38
DOI: 01.IJNS.03.02.14


A. Performance Metrics Figure 4 gives the average end-to-end delay when the rate is
The performance of EEMD technique is compared with increased. It shows that our proposed EEMDF protocol has
the compressive data gathering (CDG) technique [20]. The lower delay when compared to CDG.
performance is evaluated mainly, according to the following Figure 5 gives the packet delivery ratio when the rate is
metrics. increased. It shows that our proposed EEMDF protocol
 Average end-to-end delay: The end-to-end-delay is achieves good delivery ratio when compared to CDG.
averaged over all surviving data packets from the Figure 6 gives the energy consumption when the rate is
sources to the destinations. increased. It shows that our proposed EEMDF protocol
 Average Packet Delivery Ratio: It is the ratio of the utilizes lower energy when compared to CDG.
number .of packets received successfully and the B. Based on Sources
total number of packets transmitted. In our second experiment, we vary the no. of sources per
 Energy: It is the average energy consumed for the cluster from 1 to 4, keeping the CBR sending rate as 100kb.
data transmission.
B. Results
A. Based on Rate
In our initial experiment, we vary the transmission sending
rate as 100,150,200, 250 and 300 Kb for CBR traffic. We keep
4 sources per cluster.

Figure 7. Sources Vs Delay

Figure 4. Rate Vs Delay

Figure 8. Sources Vs Delivery Ratio

Figure 5. Rate Vs Delivery Ratio

Figure 9. Sources Vs Energy
Figure 7 gives the average end-to-end delay when the
no. of source is increased. It shows that our proposed EEMDF
protocol has lower delay when compared to CDG.
Figure 8 gives the packet delivery ratio when the no. of
source is increased. It shows that our proposed EEMDF
Figure 6. Rate Vs Energy protocol achieves good delivery ratio when compared to CDG.
Figure 9 gives the energy consumption when the no. of
source is increased. It shows that our proposed EEMDF
protocol utilizes lower energy when compared to CDG.

© 2012 ACEEE 39
DOI: 01.IJNS.03.02.14


V. CONCLUSION [11] Maen Takruri, and Subhash Challa “Data Fusion Approach
for Error Correction in Wireless Sensor Networks” InTech
In this paper, we have proposed a technique which December 2010.
combines the energy efficiency and the multiple path selection [12] Leeger Yu and Anthony Ephremides “Enhancing Energy
for data fusion in WSN. We have selected the node with Efficiency For Data Fusion In Sensor Networks With A Detection
highest residual energy as the cluster head. Randomly Mission” The 11th International Symposium on Wireless Personal
selected initiator nodes collect the information of the nearest Multimedia Communications (WPMC) 2008.
sensor nodes and select the node with highest energy as the [13] Hung-Ta Pai, and Yunghsiang S. Han “Power-Efficient Direct-
cluster head. The node with the next higher energy level is Voting Assurance for Data Fusion in Wireless Sensor Networks”
taken as the initiator. The sink computes multiple paths to IEEE Transactions On Computers, Vol. 57, No. 2, February 2008.
[14] Frank Yeong-Sung Lin, Hong-Hsu Yen, and Shu-Ping Lin “A
each cluster head for data transmission. Each CH uses
Novel Energy-Efficient MAC Aware Data Aggregation Routing in
distributed source coding and the lifting scheme wavelet Wireless Sensor Networks” Sensors 2009.
transform for compressing the data and after the sink receives [15] Rabindra Bista and Jae-Woo Chang “Energy-Efficient Data
all the collected data, it will decompress all the data.. Initially Aggregation for Wireless Sensor Networks” Intech 2010.
the path with minimum number of hops is considered as the [16] V. Bhoopathy and R. M. S. Parvathi “Energy Efficient Secure
primary path and the compressed data is transmitted to the Data Aggregation Protocol for Wireless Sensor Networks”
sink through this path. During each round of transmission, EuroJournals Publishing, Inc. 2011.
the path is changed in a round robin manner, to conserve the [17] Yuanzhu Peter Chen, Arthur L. Liestman and Jiangchuan Liu
energy. This process is repeated for each cluster. From our “A Hierarchical Energy-Efficient Framework for Data Aggregation
in Wireless Sensor Networks” IEEE 2006.
simulation results we have shown that this data fusion
[18] M. Umashankar, and C. Chandrasekar “Energy Efficient
technique has less energy consumption with increased packet Secured Data Fusion Assurance Mechanism For Wireless Sensor
delivery ratio, when compared with the existing schemes. Networks” EuroJournals Publishing, Inc. 2011.
[19] Zixiang Xiong, Angelos D. Liveris, and Samuel Cheng
REFERENCES “Distributed Source Coding for Sensor Networks” IEEE 2004.
[20] Chong Luo, Feng Wu, Jun Sun and Chang Wen Chen,
[1] Jungmin So, Jintae Kim, and Indranil Gupta “Cushion:
“Compressive data gathering for large scale wireless sensor
Autonomically Adaptive Data Fusion in Wireless Sensor Networks”
networks” MobiCom’09, September 20–25, 2009, Beijing, China.
IEEE 2005.
[21] Network Simulator, http://www.isi.edu/nsnam/ns
[2] WEILIAN SU, and Theodoros C. Bougiouklis “Data Fusion
[22] Hu seyin Ozgur Tan, Ibrahim Korpeoglu, and Ivan Stojmenovic,
Algorithms In Cluster-Based Wireless Sensor Networks Using
“ Computing Localized Power-Efficient Data Aggregation Trees
Fuzzy Logic Theory” Proceed ings Of The 11th WSEAS
for Sensor Networks”, IEEE Transactions on parallel and distributed
International Conference On Communications, Agios Nikolaos,
systems, Vol 22, No 3, 2011.
Crete Island, Greece, July 26-28, 2007.
[23] D. Kumar, T.C. Aseri, R.B. Patel, “EECDA: Energy Efficient
[3] Wei Yuan, Srikanth V. Krishnamurthy, and Satish K. Tripathi
Clustering and Data Aggregation Protocol for Heterogeneous
“Synchronization of Multiple Levels of Data Fusion in Wireless
Wireless Sensor Networks”, Int. J. of Computers, Communications
Sensor Networks” IEEE 2004.
& Control, Vol. VI (2011), No. 1, pp. 113-124.
[4] Rui Tan, Zhaohui Yuan, Guoliang Xing, Xue Liu, and Jianguo
[24] Hasan Çama, Suat Özdemira, Prashant Nairb, Devasenapathy
Yao “System-level Calibration for Data Fusion in Wireless Sensor
Muthuavinashiappana and H. Ozgur Sanlia, “ Energy-Efficient
Networks” IEEE 2010.
Secure Pattern Based Data Aggregation for Wireless Sensor
[5] M.Umashankar, and C.Chandrasekar “Witness Based And
Networks”, Journal Computer Communications archive, Volume
Voting Based Data Fusion Assurance Mechanism In Sensor
29, Issue 4, February, 2006.
Networks” Ijrras 4 (2). August 2010.
[25] Mohamed Watfa, William Daher and Hisham Al Azar, “A
[6] P. Manjunatha, A.K. Verma and A. Srividya “Multi-Sensor
Sensor Network Data Aggregation Technique”, International Journal
Data Fusion in Cluster based Wireless Sensor Networks Using
of Computer Theory and Engineering, Vol. 1, No. 1, April 2009.
Fuzzy Logic Method” 2008 IEEE.
[26] Xu Li, Shuo Yan, Chendong Xu, Amiya Nayak and Ivan
[7] Aristides Mpitziopoulos, Damianos Gavalas, Charalampos
Stojmenovic, “ Localized Delay-bounded and Energy-efficient Data
Konstantopoulos, and Grammati Pantziou “Mobile Agent
Aggregation in Wireless Sensor and Actor Networks”, 7th
Middleware for Autonomic Data Fusion in Wireless Sensor
International Wireless Communications and Mobile Computing
Networks” Springer Science+Business Media, LLC 2009.
Conference (IWCMC), pp 7 - 12 2011
[8] Xiangbin Zhu, and Wenjuan Zhang “A Mobile Agent-based
[27] Siddhartha Chauhan, Dr. Lalit Kumar Awasthi, “ Energy
Clustering Data Fusion Algorithm in WSN” International Journal
Efficient Data Gathering Protocol for Wireless Sensor Networks”,
of Electrical and Computer Engineering 5:5 2010.
International Journal of Computer Science & Technology (IJCST),
[9] Hong Luo, Jun Luo, Yonghe Liu†, and Sajal K. Das “Energy
Vol. 2, Issue 4, 2011.
Efficient Routing with Adaptive Data Fusion in Sensor Networks”
[28] Mohammad Mostafizur Rahman Mozumdar, Nan Guofang,
DIALM-POMC ‘05 Proceedings of the 2005 joint workshop on
Francesco Gregoretti, Luciano Lavagno, “ An Efficient Data
Foundations of mobile computing.
Aggregation Algorithm for Cluster-based Sensor Network”, Journal
[10] Amir Sepasi Zahmati, Bahman Abolhassani, Ali Asghar
of Networks, Vol. 4, No. 7, 2009.
Beheshti Shirazi, and Ali Shojaee Bakhtiari “An Energy-Efficient
Protocol with Static Clustering for Wireless Sensor Networks”
World Academy of Science, Engineering and Technology 28 2007.

© 2012 ACEEE 40
DOI: 01.IJNS.03.02.14


Sasikala V received the MCA., degree from Dr. C. Chandrasekar received his Ph.D.
University of Madras, Chennai, Tamilnadu, degree from Periyar University, Salem, TN,
the M.Phil., degree in Computer Science from India. He has been working as Associate
Manonmaniam Sundaranar University, Professor at Dept. of Computer Science, Periyar
Tirunelveli, Tamilnadu. She is pursuing her University, Salem – 636 011, Tamil Nadu, India.
PhD under the guidance of Dr. C. Chandrasekar The topic of his doctoral dissertation was “An
and currently working as Assistant Professor Optimization and Seamless Mobility for
& HOD of Computer Application Department Integration of Wireless LAN and Cellular
in DRBCCC Hindu College, Chennai, Tamil Nadu, India. Networks”. His research interest includes Wireless networking,
Mobile computing, Computer Communication and Networks. He
was a Research guide at various universities in India. He has been
published more than 50 research papers at various National /
International Journals.

© 2012 ACEEE 41
DOI: 01.IJNS.03.02.14

Energy Efficient Multipath Data Fusion Technique for Wireless Sensor Networks

Related slideshows

More Related Content

Energy Efficient Multipath Data Fusion Technique for Wireless Sensor Networks