INTRODUCTION
Wireless sensor networks are the interconnected tiny sensor nodes which sense
the external physical phenomenon and report its data to the central repository.
Wireless Sensor Network (WSN) found its wide applicability ranging from the
military field to the civilian field. A sensor node is made up of following
three subsystems sensing subsystem, processing subsystem and radio subsystem
(Bajaber and Awan, 2011). The sensing subsystem may
be having one or more sensor for sensing the physical world phenomenon like
temperature, light, seismic signals. The processing subsystem is having the
dedicated processor according to the need of the application. The radio subsystem
is used to give connectivity to the sensor node with other nodes or base station.
Collaborative signal and information processing approach (Zhao
et al., 2003) is used in the WSN, where the localization and tracking
of the target is made by the groups of sensor nodes which organizes among itself
and exchange their data. Some of the major issues in the CSIP are network issues
like suitable routing protocols, database issues like data abstraction and query
optimization and security problems. In this routing is the major problem which
affects the lifetime of the network. In WSN sending a bit would be taking much
more energy than processing a bit. In many practical applications like battle
field monitoring, agricultural practices the sensor network should run unattended
for more number of days. It becomes possible only when we are designing an exclusive
suitable protocol for WSN. The clustering methods (Akkaya
and Younis, 2005) are suitable for sensor networks due to its scalability
factor and eliminating the redundancy among the data collected. In some applications
of WSN it needs to run without any human assistance and report it to the base
station. Here power consumption becomes the major issue as we can’t replace
the batteries in the node. The other possible solutions to this problem is concentrating
on MAC layer of the sensor, designing energy efficient routing protocol for
WSN.
The WSN can exhibit continual reporting of events or intermittent reporting
depending upon its application need. In both the cases clustering proves to
be the suitable strategy than the direct and sequential transmission. The clustering
techniques have the advantages like balancing the network load by distributing
the energy consumption all over the network, fault tolerant network, increasing
connectivity within the network and increased scalability factor. The clustering
protocol is divided into variable convergence time and constant convergence
time algorithms (Abbasi and Younis, 2007). Many clustering
algorithms have been proposed till date and it varies on its CH node selection,
data aggregation points.
LEACH (Heinzelman et al., 2002) is a popular
clustering protocol for the WSN. It supports the distributed cluster formation.
Each node selects its CH based on the received signal strength from all the
eligible CH nodes. Before each data collection round, the CH nodes will advertise
itself with other nodes in the network. The non-CH node will check the highest
Received Signal Strength value from the CH node and attaches itself with the
particular cluster. Once the cluster is finalized, the data from each cluster
member is given to its respective CH nodes. The responsibility of a CH node
is to aggregate these data and send it to the base station. Due to its additional
work energy consumption will be more for the CH nodes than the other nodes.
To avoid this rapid depletion of rare resource like energy in CH nodes, this
CH node position became rotational.
LEACH follows the proactive strategy but for certain application we need to
follow the reactive strategy like reporting to the base station when a particular
have occurred in the environment. In that case, TEEN (Manjeshwar
and Agrawal, 2001) protocol suits our need. This reactive protocol uses
two threshold value called as Hard Threshold (HT) and Soft Threshold (ST). When
a node senses a value higher than the HT and differs equal or greater than ST,
it transmits its value to the CH node. By using HT and ST number of transmission
is greatly reduced which makes it suitable for time critical events.
In APTEEN (Manjeshwar and Agarwal, 2002), both Proactive
and Reactive strategies are implemented. It gives both periodic reading and
time critical readings. The users have been given the choice to set time interval
for periodic data collection.
Yang and Zhang (2009) used unequal clusters to solve
the hot spots in the sensor networks. Hierarchical clustering algorithm UDCA
was proposed by (Abidoye et al., 2011). The sensor
network is divided into different layers. In each the nodes with stronger probability
value will be elected as Cluster Heads (CH). The node which is having stronger
probability value than all other nodes will be elected as CH node at the highest
layer, it compress the data given to it by its lower layers and pass it to the
base station.
GRESS (Shan et al., 2011) works in two phases.
In Initialization phase, all the nodes will find its communication cost with
the BS. Then the network will be divided into static clusters. In a cluster,
the node which is having low cost will be elected as cluster Head (CH). All
non-CH nodes have to send its ID, cost and energy residual to CH node. The CH
node will compute the sleep scheduling for each node in its clusters based on
this information. The sleep schedule differs for nodes based on its residual
energy. In State transformation phase, the non CH nodes after its stipulated
sleep time interval will go to listen state and check whether it have received
any query packet from the current CH node. If receives any packet it will enter
active state and becomes new CH node. The old CH node will become a non-CH member
of the cluster and enter into sleep state.
ADRP (Bajaber and Awan, 2011) provides us with the
adaptive routing strategy. It follows the centralized approach. All the sensor
nodes will be transmitting its residual energy level to the base station. The
base station will select the CH node with a set of next eligible CH nodes and
send it to each sensor nodes. Due to any environmental factors if a CH node
dies, the next CH node will be selected from the list without informing BS and
reclustering. Because of these factors it comes for more number of data collection
rounds than LEACH.
EECPL (Bajaber and Awan, 2011) also follows the centralized
approach. The initial configuration of the sensor nodes will be given to the
BS. Based on its remaining energy level, new CH nodes are selected and informed
to the sensor nodes. This protocol reduces the responsibility of the CH nodes
by transferring the duty of data transmission to the cluster sender node. The
CH node only generates the TDMA schedule for its cluster member node. The cluster
member nodes will be transmitting data to their neighbor nodes and aggregates
there. At the end the aggregated data will be given to the cluster sender and
from there it will be sent to the base station.
All these protocol are functioning in the manner that the CH node will be sending its data directly to the base station. Those CH nodes located away from the base station will be spending a considerable amount of its energy during its transmission. This paper gives a new protocol where the distanced CH nodes send its data in multi hop fashion to conserve its energy.
PROPOSED MODEL
System model: The sensor network is randomly distributed over the sensing region. The cluster Head nodes are elected by the Base Station in each data round by its remaining energy levels. The cluster member nodes have to send its data to the CH nodes in its scheduled time and aggregation is done in CH node. This paper assumes that a unique identifier is given for each sensor node in the network and it knows its location by using GPS receiver. EEGTP protocol operates in two different phase.
Cluster formation phase: The sensor nodes will be communicating its
location data gathered through the use of GPS (Fig. 1) and
its remaining energy reserves to the central base station. The base station
will elect the CH nodes which is having its energy level more than the Threshold
level. And in each data collection round the Threshold level is reduced to elect
new CH nodes. Also the already elected CH nodes would not be participating in
next election of CH nodes to evenly distribute the energy consumption among
the nodes.
|
| Fig. 1: |
Sensor nodes is transmitting its residual energy and location
data to the central base station |
|
| Fig. 2: |
Taking the shortest energy path in choosing the next hop within
the intra-cluster communication |
|
| Fig. 3: |
Intra-cluster communication model |
The base station will be sending the new CH node ID for each sensor. If a sensor
node receives its ID it becomes CH nodes for current data collection round.
Data collection phase: When the cluster member nodes are sending its
data to its CH nodes there is a higher probability of collision to occur. In
order to avoid the collision the CH node is generating the TDMA schedule for
its member node. A minimum spanning tree (Fig. 2) is built
between the far by sensor node with the CH node. The member node have to wake
up its radio only during its time and go to the sleep state after sending its
data to its neighbor node which involves less transmission cost as shown in
Fig. 3, so as to conserve its radio consumption.
|
| Fig. 4: |
Acyclic graph for intra-cluster and shortest path for intra-cluster
communication |
The data aggregation takes place in every node and the final aggregated is
given to the CH node. The CH node sends the aggregated data to the base station
through the shortest path which involves multi hop transmission. The intermediate
nodes in the multi hop communication are the other CH nodes. The inter cluster
communication is avoided by using separate CDMA codes for each Clusters. Acyclic
graph for Intra-Cluster and Shortest path for Inter-Cluster Communication has
been shown in Fig. 4.
Radio model: The radio chosen for this EEGTP is same as (Bajaber
and Awan, 2011):
Where, n = Number of bits, d = Distance between two nodes, Eelec = Electronics energy and εfs = Power loss in free space. EEGTP assumes n = 4000 bits, Eelec = 50 nJ bit-1, εfs = 10 pJ bit-1 and initial of each sensor node is equal to 1 Joule. The aggregation cost for per bit is 50 nJ bit-1.
SIMULATION AND RESULTS
The sensor network is simulated in the MATLAB environment. The sensor nodes are randomly distributed as shown in the Fig. 5. EEGTP performance is compared with LEACH-C in intra cluster energy consumption in each round, average energy consumption for a node and total energy consumption of the CH nodes in each round. In all the three above criteria EEGTP outperforms LEACH-C. We compared EEGTP with LEACH-C since both uses centralized approach.
Intra cluster energy consumption: In LEACH-C all the cluster members
are directly sending its data to the CH nodes and the CH node is in responsibility
of aggregating the data and sending it to the base station.
|
| Fig. 5: |
Random distribution of 100 sensor nodes |
|
| Fig. 6: |
Intra-cluster energy consumption comparison |
|
| Fig. 7: |
Average energy consumption of the nodes |
In EEGTP each node is sending its data to its nearby neighbor node which involves
lesser transmission cost. The aggregation takes place at all the nodes except
the first sending node which reduces aggregation cost in CH nodes (Fig.
6).
Average energy consumption of nodes: In Fig. 7, average energy consumption of nodes in LEACH-C is compared with the EEGTP protocol. Since in EEGTP the nodes are only communicating with its neighbor nodes and even CH nodes itself is transmitting the aggregated data to the base station in the shortest path, there is a drastic reduction in the energy consumption of the nodes.
CONCLUSION
Energy efficient data gathering is a major constrained in wireless sensor networks. In this paper lifetime of the network is analyzed, it’s metrics differs based upon the sensing application. EEGTP proves to be more efficient than its counterpart. But when it goes for latency in the network it lag behind its rival protocols. In our future work, the latency of the network also will be considered.
Subscribe Today
