Optimizing the Data Collection in Wireless Sensor Network

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 03 Issue: 02 | Feb-2016 www.irjet.net p-ISSN: 2395-0072
© 2016, IRJET | Impact Factor value: 4.45 | ISO 9001:2008 Certified Journal | Page 701
Optimizing the Data Collection in Wireless Sensor Network
R.Latha1,Valarmathi.M2
1 Assistant Professor, 2 PG Scholar
1,2 Computer Application
1,2 Vel Tech High Tech DR.Rangarajan DR.Sakunthala Engineering College
--------------------------------------------------------------------***-----------------------------------------------------------------------
Abstract - In Wireless Sensor Network (WSN), data
collection is an important task. How to efficiently collect
sensing data from all sensor nodes is important to the
performance of WSN. Sensor nodes will send the sensed data
to sink for further processing. Sink node will be at root level
and the Sensor nodes will act as intermediate and leaf nodes
(forms a Tree). The study of data collection had done with
Breadth First Search (BFS) and has been proved that BFS
data collection have achieved a maximum data collection. In
BFS, nodes will send the packets (its own and its child nodes)
one after the other to its parent node. This takes more time
to transfer the packets. If Parallel data collection algorithm
in Breadth First Search (PDC-BFS) is incorporated, the sink
can collect the data parallely from its child nodes. Besides,
nodes can aggregate the packets (its own and its child
nodes) and send to its parent node. This takes less time to
transfer the packets.
Key Words: Wireless Sensor Networks, Breadth First
Search, Parallel data collection, Binary tree.
1. INTRODUCTION
Sensor node is a small device to sense, receive and
forward packets to its neighbor node. A tree rooted at a
sink node is called data gathering tree, if every internal
collects the data from sensors that are its child nodes and
send this packet to its parent node. Finally the whole data
is send to the sink for further processing. The sensor
nodes and the sink are deployed in a geographical area
and a binary tree structure is constructed. This is many to
one data collection, so data collection scheduling is very
important. So that there is a possibility of maximizing the
data collection at sink.
Consider n sensor nodes, these n nodes are randomly
deployed in a geographical area of m*m. n sensor nodes
and a sink as root forms a Tree. At regular intervals, sink
node collects sensed data from these n sensor nodes. After
collecting, analyze the data collection capacity of the sink.
As the main aim of this project is to minimize the delay
and maximize the capacity. To achieve this, follow BFS
data collection following three models. The three models
are Disk Graph Model, Generalized Quasi Disk Graph
Model and Protocol Interference Model.
Following these three models, collect data and send data
to the sink node following BFS [1] approach. Capacity
analysis is done in the sink. The capacity analysis is the
amount of data sink can collect for a snapshot.
Implementing BFS, the data collected and the delay
required collecting data is analyzed. Each sensor node
measures independent field values at regular time
intervals and sends these values to sink node. The union of
all these sensing values from n sensors at a particular time
is called snapshot. Delay of Data collection D is the time
used by the sink to successfully receive a snapshot, i.e., the
time needed between completely receiving one snapshot
and next snapshot at the sink.
Distance between the sensor nodes will be different and
the amount of data to send will differ, so the energy
utilized to send the data will differ for different sensor
nodes. So each sensor node calculates the energy on its
own. Each sensor node calculates the distance and energy
to transfer data. Now three models mentioned above are
implemented and traverse the nodes parallel and collect
data parallely from nodes. This is Parallel Data Collection
in BFS (PDCBFS). In BFS, individual packets (its own and
its child nodes) are sent separately, but in PDCBFS every
node aggregates the packets (its own and its child node)
into a single packet and send to parent node. The amount
of data collected and the delay required collecting data is
analyzed.
Now performance analysis of data collected and delay
required to collect data for both BFS and PDCBFS is done
which shows the result of best of two algorithms (BFS and
PDCBFS).
2. RELEVANT WORK
This paper focuses on data collection in WSN using
parallel data collection in BFS. This section gives an idea of
related data collection algorithms.
It has been proved that the upper bound of delay rate or
capacity of data collection is W [7], [8]. It is obvious that
the sink cannot receive data at a rate faster than W since
W is the fixed transmission rate of individual link.
Therefore, the delay rate of this collection scheme
achieves the order of the upper bound, and the delay rate

of data collection is θ(W). Note that even for individual
sensors, the lowest achievable delay rate of this method is
θ(W=n) which also meets the upper bound.
Siyuan Chen, Shaojie Tang, Minsu Huang, Yu Wang [1]
studied the capacity of data collection in an arbitrary
network. They derived the upper and constructive lower
bounds of data collection capacity in arbitrary networks.
The BFS data collection method and path scheduling
algorithm [1] achieved an upper bound of θ (W). They also
examined the design of data collection under a general
graph model and discuss performance implications.
Mahajan and J. Malhotra [2] used BFS approach for path
determination. Breadth-First Search (BFS) is a node
search algorithm that begins at the root node and explores
all the neighboring nodes. Then for each of those nearest
nodes in graph, it explores their unexplored neighbor
nodes, and so on, until it finds the goal. It has been proved
by induction that the BFS tree is a shortest path tree
starting from its root. Every vertex has a path to the root,
with it's path length equal to its level (just follow the tree
itself), and no path can skip a level so this really is a
shortest path. In both [1] and [2], they proved that BFS
data collection have achieved a maximum data collection.
Barton and Rong [4] investigated the capacity of data
collection under general physical layer models for
example, cooperative time reversal communication model
where the data rate of an individual link is not ﬁxed as a
constant W but dependent on the transmitting powers and
distances of all simultaneous transmissions.
P. Balamurugan, K. Duraiswamy [9] gave a random rank to
the vertices of the nodes between 0 and 1. A link is formed
between any two nodes if they are within their
transmission range. If a sensor node has the highest rank
among its neighbors, then it is considered an associate
node, else it forms into the leaf node. Next, the associate
nodes form a complete graph and later form a Rooted
Directed Tree (RDT) after an implementation of Kruskal’s
Minimum Spanning Tree algorithm and the Breadth First
Search algorithm. The energy consumption model is
followed.
Firstly, a virtual tree topology is constructed based on
Grid-based WSNs [12]. Then two node-placement
techniques, namely Distance-based and Density-based
deployment schemes are followed to balance the power
consumption of sensor nodes. Finally, a collision-free MAC
scheduling protocol [12] is proposed to prevent the packet
transmissions from collision. In addition, extension of the
proposed protocols is made from a Grid-based WSN to a
randomly deployed WSN, enabling the energy-balanced
schemes as they are generally applied to randomly
deployed WSNs.
3. NETWORK MODEL
In [12], a N * N Grid-based WSN is considered, which
consists of a sink node and -1 sensor nodes. Fig.1
shows a 3 * 3 Grid-based topology composed of a sink
node and = 8 sensor nodes.
Figure 1: Grid based topology
In Fig 1.the Sink is placed in the coordinate (1,1) top right
position and the coordinate of each sensor node is
increased by 1 when moves left in x-axis and increased by
1 when moves down in y-axis. Fig. 1 shows the
constructed root-balanced tree which is rooted by sink.
3.1 Grid based WSN
For an example, Grid based topology is formed with a sink
node and six sensor nodes. Then a tree structure of this
Grid based topology is identified. In [12], the tree
constructed guarantees that the numbers of nodes in left
and right sub-trees of the sink node differ one at most had
been proved. This Grid based WSN is identified as tree
structure.
Figure 2: Sample Grid based WSN
Numbers in the sensor node denotes how many packets a
node would have to send to the sink. This will be
calculated by the sink. Based on the number of packets,
power required by the sensor node differs. This will be
discussed in section 4. The coordinates shown in fig. 2 are
logical coordinates. In this paper, the logical coordinates of
the Grid based WSN are not converted to physical
coordinates. The physical coordinates of the nodes are
initially given as input to geographically deploy the nodes.
3.2 Tree construction of WSN
Consider a Graph G with vertices V and Edges E. Vertices
are the sensor nodes and edges are the link between the

sensor nodes. Consider n nodes in the graph. For example
assume that the sensor nodes and sink are deployed over a
geographical area of 2000*2000 and there are 100 Sensor
Nodes spread in the geographical area of 2000*2000. At
regular intervals, sensor nodes generate b bits and want to
send it to the sink. The union of data from sensor nodes at
this regular interval is called snapshot. The data size of
one snapshot is n*b.
Figure 3: Tree structure of sensor nodes
In Fig. 3, A, B, C, D, E, F are the sensor nodes. Figure 3
shows the tree structure of six sensor nodes and one sink
node formed from the Grid based WSN in Fig 2 and they
are placed at the coordinates mentioned in the table 1.
Node A B C D E F Sink
X-
coordinate
Y-
Coordinate
50
90
120
80
20
30
70
10
80
20
180
40
100
190
Table 1: Coordinates of Sensor Nodes
This Coordinates are used to find distance between the
sensor nodes. The distance is calculated based on the
coordinates given in table 1.
Nodes Distance (meters)
C to A
D to C
E to B
F to B
A to Sink
B to Sink
67.1
53.9
72.1
72.1
50
22.7
Table 2: Distance between the Sensor Nodes in fig. 3
Based on the distance calculated in table 2, the energy
required by the sensor nodes is calculated. The energy
calculation formula is given in table 3.
4. MODELS
4.1 Disk Graph Model
A sensor node sends or receives data from or to another
sensor node if that sensor node is within a range. This is
Disk Graph Model. Here fix a range r for all sensor nodes.
A sensor node can transfer data to another sensor node
if
| - |<=r
Cannot transfer data to if
| - |>r
and are the locations of the sensor nodes.
Algorithm for Disk Graph Model:
1. INPUT: Graph G (V, E), root as sink s and range r.
2. Dg=0
3. For each sensor node
4. Find the neighbor nodes of the sensor nodes
5. For each neighbor node
6. Find distance d between the sensor nodes
7. If d<=r then
8. Dg=1
9. End if
10. End For
11. End For
4.2 Protocol Interference Model
A sensor node can send to sensor node if no other sensor
node is sending data simultaneously within the
interference range. This is Protocol Interference Model. A
sensor node can successfully transfer data to a sensor
node if
| - | ≤ R
and for any other simultaneously transmitting node ,
| - | ≥ (1 + Δ)R
where Xi, Xj and Xk are the locations of , and , R is
the interference range of all nodes, and Δ models the
interference zone around in which any simultaneous
transmission causes collision at . In the range of r/2,
there can be at most two nodes[1]. This interference range
R and range r is illustrated in the fig. 4.

Figure 4: Illustrates range r and Interference range R
In the fig. 4, dots denote the sensor nodes. r is the range
within which a sensor node can communicate with other
sensor nodes. R is the interference range in which two
sensor nodes in the same interference range R cannot
send data to its neighbor node simultaneously to avoid
collision. That is only one node in range R will be in active
state.
4.3 Generalized Quasi Disk Graph Model
One extension of the DG model is to consider nodes in R3.
Moreover, distances between nodes could be modeled
using the Manhattan norm (L1 norm). In the Manhattan
norm, the distance between two points u = ( , ) and v =
( , ) in the plane is given by d(u, v) = | − | + | −
|, while in the Euclidean norm (L2 norm), the distance is
d(u, v) = + . Alternatively, the
maximum norm (L∞ norm) is also popular, where d(u, v) =
max | − |, | − |. Distance can be calculated using
any of the above mentioned norms.
5. ENERGY CALCULATION
The lifetime of a sensor node depends on its power
(energy). Power required by the sensor nodes differs from
sensor node to sensor node. It depends on its position that
is how close the sensor node is from the sink.
This work uses the following energy consumption model
for all the sensor-network related projects [9].
Table 3: Energy calculation
The transmitter amplifier energy cost is to achieve an
acceptable Signal-to-noise ratio so that the signal received
can be processed
Let data size be k = 512 bits
Energy lost at C = Energy lost to receive + Energy lost to
transmit =230523.4 nJ
Energy lost at D = 374080.0 nJ
Energy lost at E = 291758.6 nJ
Energy lost at F = 291758.6 nJ
Energy lost at A = Energy lost to receive + Energy lost to
transmit = 179200.0 nJ
Energy lost at B = Energy lost to receive + Energy lost to
transmit = 77582.9 nJ
Total Energy Lost in the Network = 1444903.5 nJ =
0.00144J
Nodes Energy lost at nodes (nJ)
C
D
E
F
A
B
230523.4
374080.0
291758.6
291758.6
179200.0
77582.9
Table 4: Energy required by the Sensor Nodes in Fig. 1
Using the X and Y coordinates, the distance between the
sensor nodes is calculated and then the energy required by
each sensor node is calculated, based on its activities. If a
sensor node is a child node, then it consumes energy only
to transfer the packets. If a sensor node is a parent node,
then it requires energy to receive the packets from its
child nodes, energy to integrate the packets and energy to
transfer the packets to its parent. Based on these
conditions each sensor node calculates the distance from
its neighbor node and the energy it require.
6. DATA COLLECTION
6.1 BFS Data Collection
Breadth-first search (BFS) which is a Graph search
algorithm in graph theory that begins at the root node and
explores all the neighbor nodes. For each of those nearest
nodes, it explores their unexplored neighbor nodes, and so
on, until it finds the goal node. It has been proved by
induction that the breadth first search tree is a shortest
path tree starting from its root. Every vertex in tree has a
path to the root, with path length equal to its level, and no
path can skip a level so this really is a shortest path.
Operation Energy Dissipated
Transmitter Electronics(ETx-
elec)
Receiver Electronics(ERx-
elec)
(ETx-elec = ERx-elec = Eelec)
Transmitter Amplifier(ɛamp)
To transmit a message of size
k-bits, over distance d m
To receive a k-bit message
Cost for data fusion
50 nJ/ bit
100 pJ/ bit/
ETx(k, d) = Eelec * k +
ɛamp * k *
ERx(k) = Eelec*k
5 nJ/ bit/ message

Figure 5: Steps in BFS data collection
In the fig. 5, it shows the data collection based on BFS
(Breadth first Search). Dotted line represents the
exploring nodes; plain lines represent the link between the
sensor nodes. The sink starts searching its first child and
explores its neighbor nodes in step 1 and step 2. Node D
sends data to node C and node C tags its data and node D’s
data and send to node A. In step 3, node A tags its data to
the received data and send to the sink. In step 4, it
searches its next child node and explores its own child
nodes E and F in step 4 and step 5. This data is buffered in
node B and in step 6 this data is send to the sink. This
completes a snapshot. The network link has a capacity to
send or receive t bits/sec. The snapshot is divided into
time slots.
Siyuan Chen, Shaojie Tang [1] studied that the BFS-based
data collection method can achieve data collection
capacity of θ(W) at the sink. They proved that data
collection capacity is θ(W).
Figure 6: Time slots required for BFS data collection
The fig. 6 shows the time slots to complete a snapshot.
Numbers on arrow denotes the time slots it requires to
send the data to its parent node. The worst case is
considered that is every node sends t bit/sec to its parent
that is, it utilizes the complete data rate of a link. That is
why nodes D, E, F require one time slot to send the data to
its parent. Node C requires two time slots to send data to A
because it has buffered t data from D and its own data t
bits so it requires two time slots. Node A requires three
time slots to send the data to the sink because it has
buffered 2t data from node C, D and its own data t bits so it
requires three time slots. This is the case for node B. Total
time slots required to complete a snapshot is
1+2+3+1+1+3=11 time slots. A network with six nodes
requires eleven time slots to complete a snapshot.
6.2 Parallel Data collection in BFS
In this work, the sensor nodes is traversed in Breadth First
Search but traversed parallely. The sink starts exploring
its child nodes simultaneously following the TDMA
scheduling. Fig. 7 shows the traverse path followed in
parallel data collection in BFS.

Figure 7: Steps in BFS parallel data collection
In step 1, the sink starts searching all its child nodes in
parallel. It explores its child node, this child nodes
explores its child nodes simultaneously. That is from the
fig. 7 in step 1, node D sends the data to its parent node.
Node C aggregates its own data with the node D’s data and
node E sends data to node B. In step 2, node C and node F
send data to its parent node. The data is buffered at their
parent nodes. In step 3, node A sends data to the sink. In
step 4, node B sends data to sink.
Data buffered and this data is aggregated and send this as
a single packet. There is not much work on capacity of
data aggregation in wireless sensor networks, except in
recent papers [10] and [11]. Giridhar and Kumar [10]
investigated a more general aggregation problem in
random sensor network where a symmetric function of
sensor measurements is used for data aggregation in WSN.
It was shown that for random planar multi-hop network,
the maximum rate for computing divisible functions (i.e., a
subset of symmetric functions) is θ ( W). Notice that they
defined the maximum rate as the data rate per sensor. In
addition, using a technique called block coding, they
further showed that type-threshold functions can be
computed at a rate of θ ( W). Moscibroda [11] then
further studied the aggregation capacity for arbitrarily
deployed networks (he called this as worst-case capacity)
under both protocol interference model and physical
interference model. He showed that the worst-case
capacities of data aggregation are θ ( W) and Ω( )
respectively.
Figure 8: Time slots required for parallel data
collection in BFS
By following parallel data collection in BFS, time slots less
than the time slots required for normal BFS is achieved.
Fig. 8 shows that node C and node E together require one
time slot to send data to its parent node. Node A and node
B each requires one time slots to send data to the sink
because it aggregates the packets into a single packet
during transmission. Total time slots require are
1+1+1+1+1=5 time slots to complete a snapshot.
Previously BFS achieved eleven time slots to complete a
snapshot. With six nodes in a network, it got six time slots
difference. If the network grows, then there will be a
drastic difference between the time slots utilized.
Here comes the conclusion that, parallel data collection in
BFS gives the optimal data collection in WSN.
Barton and Rong [4] investigated the capacity of data
collection under general physical layer models (e.g.
cooperative time reversal communication model) where
the data rate of an individual link is not ﬁxed as a constant
W but dependent on the transmitting powers and
distances of all simultaneous transmissions. To calculate
the transmitting power, the energy required for each and
every node is calculated. Based on this study, the data rate
of the links is not fixed as a constant W. Then the time
slots required by all sensor nodes will become one time
slot even though they are having many packets to send.
Following this aggregation of packets into single packet, in
above example it takes only five time slots to complete a
snapshot.
Node T[N] Child T[C] Parent T[P]
Sink T[R]
A
B
C
A,B
C
E,F
D
-
Sink
Sink
A
Table 5: Relationship between nodes
Table 5 shows the relationship between the nodes. T[N]
represents the Nodes in the network, T[C] represents the
child nodes of T[N], T[P] represents the parent node of
T[N].

Algorithm for Time slot calculation in Parallel Data
collection in BFS
1. INPUT: Graph G (V,E), range r
2. Data = 0
3. At time = 0 For each T[R]->T[C]
4. time slot TS=0
5. For each T[N]
6. For each T[C]
7. TS++
8. Data = Data + data of T[C]
9. End For
10. End For
11. End For
7. DATA COLLECTION ANALYSIS AT SINK
(for parallel data collection)
Let total time slot required to collect a snapshot be T. T=
α.τ + ( C – 1 ) is the at most time slot required. Here α
denotes level of the tree, τ is the max of the time slot
required by a node. Time slot required by a node t = b/W ,
b is the bits to transfer, W is the maximum transfer rate of
a node, τ = max(t1,t2………tn). C is the number of child
nodes a Sink have. Delay of data collection D = (α + (C-1))
τ. Delay of data collection is minimized by using parallel
data collection in Network. After data collection for one
snapshot, data collection capacity analysis is done at the
sink node. Data collection Capacity C = n*b/D, here n is the
number of Sensor Nodes, b is number of bits each Sensor
Node can transfer, D is the delay between the snapshots.
So C = = . Here α, C are
constants. So the upper bound of θ (W) is obtained, by this
derivation the upper bound is achieved. Thus the data
collection in sink is maximized.
8. PERFORMANCE ANALYSIS
Nodes are placed randomly and uniformly in given
geographical area. The energy required to transfer the
data for each and every node is calculated. The distance
between the nodes is calculated using any of the three
norms (GQDG model) formula. After setting up this, data is
collected using BFS and PDCBFS algorithm.
In BFS algorithm, parent nodes receives the packet
individually and when a node sends data no other node in
the network transmits the data which takes more time
(delay) for sink to collect the data. But in PDCBFS
algorithm, parent node combines packets to one packet
and transmits to its parent and nodes transmit data
parallely without collision. To avoid the collision Protocol
Interference Model is implemented.
Data is collected using both these algorithms (BFS and
PDCBFS). Analysis is done for a snapshot. Data collected
for one snapshot and delay taken to collect a snapshot is
calculated in sink. This calculation is done separately for
both these algorithms. Graph is plotted for both data
collected and delay taken. This analysis is done deploying
the nodes uniformly and randomly. First two graphs for
uniform placement of nodes and second two graphs for
random placement of nodes in geographical area.
Figure 9: Graph to compare the delay for uniformly
placed nodes
Fig. 9 shows the comparison of delay taken by BFS and
PDCBFS for uniformly placed nodes. This graph shows that
PDCBFS takes much less delay than BFS and there is only a
moderate increase in the delay for PDCBFS. But in BFS,
after 40 nodes very fast increase in the delay is found. This
shows that delay taken for PDCBFS to collect data is much
less than BFS algorithm.
Figure 10: Graph to compare data collected for
uniformly placed nodes
Fig. 10 shows the comparison of data collected by BFS and
PDCBFS for uniformly placed nodes. This graph shows that
PDCBFS has collected more data than BFS. In BFS, after 60
nodes there is no increase in the amount of data collected
but in PDCBFS, only after 90 nodes there is no increase in
the amount of data collected. Graph 10 shows that the
PDCBFS is much better than BFS in data collection.

Figure 11: Graph to compare the delay for uniformly
placed nodes
Fig. 11 shows the comparison of delay taken by BFS and
PDCBFS for randomly placed nodes. This graph shows that
PDCBFS takes much less delay than BFS and there is only a
moderate increase in the delay for PDCBFS. But in BFS,
after 70 nodes very fast increase in the delay is found. This
shows that delay taken for PDCBFS to collect data is much
less than BFS algorithm.
Figure 12: Graph to compare data collected for
uniformly placed nodes
Fig. 12 shows the comparison of data collected by BFS and
PDCBFS for randomly placed nodes. This graph shows that
PDCBFS has collected more data than BFS. In BFS, after 40
nodes there is no increase in the amount of data collected
but in PDCBFS, only after 80 nodes there is no increase in
the amount of data collected. Graph 10 shows that the
PDCBFS is much better than BFS in data collection. From
these graphs, the conclusion is clear that parallel data
collection in BFS algorithm collects more data than in less
delay when compared to BFS algorithm. And it is already
proved that PDCBFS collects data with an upper bound of
θ (nW) which is exceeded the upper bound θ (W) of BFS
algorithm.
9. CONCLUSION
In this paper, a tree structure of Wireless Sensor Networks
is constructed. The distance between the sensor nodes and
energy needed by the sensor nodes to transfer and receive
the data is calculated. The distance and energy to achieve
the link data rate is calculated so that the data from
different sensor nodes can be buffered and aggregated
into a single packet and send through that link. First data
is collected using BFS and data collected in a snapshot and
the time slots required to collect a snapshot is calculated.
In [1], an upper bound of θ (W) for data collection is
achieved. In this paper, the data is collected parallely using
BFS approach and send to the sink and the data collected
in a snapshot and the time slots required for a snapshot is
calculated. This is implemented deploying the nodes
randomly and uniformly in a geographical area. This
statistics of data collected and delay taken to collect a
snapshot is analyzed using graph. From that graphs, it has
been proved that parallel data collection in BFS collects
more data in less delay than BFS algorithm. By this work,
the upper bound is exceeded which has been achieved
using BFS algorithm; this has been proved in previous
section.
As this work is based on TDMA and energy calculation, if
any node stops responding it can lead to total failure of the
network. So a secondary parent to every node is deployed
so that if any parent node stops working then the child
node can transfer the data to its secondary parent. But the
secondary parent does not simulate the parent. It does not
sense the data but receive the data from the child node
and transfer that data to its parent node. Then the
network works smoothly all the time. This will be our
future work.
REFERENCES
[1] Siyuan Chen, Shaojie Tang, Minsu Huang, Yu
Wang, “Capacity of Data Collection in Arbitrary
Wireless Sensor Networks”, IEEE Computer
Society, Jan 2012.
[2] Mahajan and J. Malhotra, "Energy Efficient Path
Determination in Wireless Sensor Network Using
BFS Approach“, Wireless Sensor Network, Vol. 3
No. 11, 2011, pp. 351-356. doi:
10.4236/wsn.2011.311040.
[3] P. Gupta and P.R. Kumar, “The Capacity of
Wireless Networks”, IEEE Transactions on
Information Theory, vol.46, no. 2, pp. 388–404,
March 2000.
[4] R. Zheng and R.J. Barton, “Toward optimal data
aggregation in random wireless sensor networks,”
in Proc. of IEEE Conference on Computer
Communications (Infocom), 2007.
[5] X.-Y. Li, Y. Wang, and Y. Wang, “Complexity of Data
Collection, Aggregation, and Selection for Wireless
Sensor Networks,” IEEE Trans. Computers, vol. 60,
no. 3, pp. 386-399, Mar. 2011.
[6] “Fault-tolerant Scheduling for Data Collection in
Wireless Sensor Networks”, Jitendra Pradhan,

Anil Sarje, 2012 International Conference, pg No.
382-386.
[7] Enrique J. Duarte-Melo and Mingyan Liu, “Data-
gathering wireless sensor
networks: Organization and capacity,” Computer
Networks, vol.43, pp. 519–537, 2003.
[8] Mingyan Liu David L. Neuhoff Daniel Marco,
Enrique J. Duarte-Melo, “On the many-to-one
transport capacity of a dense wireless sensor
network and the compressibility of its data,” in
Proc. of International Workshop on Information
Processing in Sensor Networks, 2003.
[9] P. Balamurugan, K. Duraiswamy, “Evaluation of
Energy and Delay Computing Algorithm for Data
Gathering in Wireless Sensor Networks”,
European Journal of Scientific Research, ISSN
1450-216X Vol.66 No.1 (2011), pp. 85-95.
[10] A. Giridhar and P.R. Kumar, “Computing and
communicating functions over sensor networks,”
IEEE Journal on Selected Areas in Communications,
vol. 23, no. 4, pp. 755–764, 2005.
[11] Thomas Moscibroda, “The worst-case capacity of
wireless sensor networks,” in IPSN ’07:
Proceedings of the 6th international conference on
Information processing in sensor networks, 2007.
[12] Chih-Yung Chang, Hsu-Ruey Chang, “Energy-
aware node placement, topology control and MAC
scheduling for wireless sensor networks” ,
Computer Networks 52 (2008) 2189–2204.

Optimizing the Data Collection in Wireless Sensor Network

Recommended

More Related Content

What's hot (16)

Similar to Optimizing the Data Collection in Wireless Sensor Network (20)

More from IRJET Journal (20)

Recently uploaded (20)

Optimizing the Data Collection in Wireless Sensor Network