Department of Computer Science and Engineering
Suresh Gyan Vihar University, Jagatpura
Abstract:- Wireless Sensor
Networks forms the core of the infrastructural facilities and amenities that
constitutes a major part of modern living. Wireless Sensor Networks founds
tremendous applications in domains such as theft alarms, wildlife monitoring,
radiation/pressure/light/heat sensor networks and the list is endless. It
constitutes the core part of the modern Internet of Things (IoT) that will
revolutionize the modern living. The Iot specifies a scenario in which the
devices can communicate with each other using the internet over a flexible
framework and can be programmed to perform specific actions based on the
programming customization made by the users. For example, a refrigerator is
runs out of milk or bread can email the requirement to the dairy that can
entertain the mail and ship a delivery of the same to the location of the
refrigerator. As sensor nodes are battery powered, there is a critical aspect
to same battery power. This is possible only by avoiding the in-network
communication as much as possible. A fraction of communication overhead can be
reduced through clustering. In this paper, an approach for dynamic clustering
is proposed based on the varying traffic loads to various PAN coordinators so
as to maximize the battery life and therefore the network lifetime.
Keywords:- Wireless sensor network, Clustering
Protocols, Battery Life etc.
I. Clustering in Wireless Sensor Networks
Clustering forms, the backbone towards the
persistence of sensor nodes towards sensing data in such a way that a single
lithium ion battery can work even for one and a half year continuously. This is
because of the reduction in in-network communication to the central node
through the creation of clusters in such a way that all the node in the cluster
transmit the data to the cluster head and the cluster head is responsible to
transmit the data to the central node. The senario is expressed in the
following figures.
Fig. 1: Wireless Sensor Network without clustering
Fig. 2: Wireless Sensor Network with clustering and
Data Aggregation
The individual collections shown in figure 1.2 are known as clusters
and the nodes that belongs to a particular cluster sends the data only to the
cluster head. Thus, reducing the data transmission over long distance from the
individual nodes to the central computer. In the clustered approach, the nodes
transmit the data to the cluster head over a relatively very short distance,
thus, conserving the battery life and enhancing the network lifetime.
II. Dynamic Clustering over the WirelessSensor Network
Consider
a network of N nodes and a static number set initially k as the total number of
cluster over the network. Thus, on an average, there are N/k in each cluster.
Also, consider a rectangular plane of dimension aXa over which the sensor nodes
are (approximately evenly) speeded.
As state previously, there are k clusters each having (N/k)-1 nodes as
ordinary sensing nodes and a Cluster head that hold the responsibility of
aggregating data from each of the (N/k)-1 nodes. Also assume that each packet
senses the medium and sends the data packet to the cluster head in specified
TDMA frame.
Considering the first order radio energy dissipation model, let the
energy consumption per bit in the transmission circuitry be Et and
the energy consumption per bit in the processing circuitry be Ep.
Let there be B bits in a TDMA packet. Considering the initial energy level in
the battery be E, one can approximate the residual battery life after N rounds.
Let Me be the number of rounds after which the leader election takes
place and a message is broadcasted to all the other nodes in the cluster
regarding the node which is elected as the leader so that all the nodes may
transmit the data to the specific node. The specified node then aggregates the
data from all the nodes in its cluster and transmit the data to the central
computer.
It is important to
note that the leader election process is an overhead and is incurred only to
manage the network traffic. Rapidly electing new heads and consequently
broadcasting the message to all other nodes in the network induce an overhead
which is to be avoided. On the other hand, it is also important to note that
the node which is elected as the cluster head depletes its energy very
frequently as it has to perform all the data aggregation processing all be
itself for all the nodes in the network. Thus, frequent leader election leads
to an evenly consumption of battery power in all the nodes of the cluster. If
no election of leader takes place, then the node which handles the task of
leader will soon run out of the battery.
In addition to the depltion of the battery in the normal rounds during
the data gathering, the leader will deplete the energy
E = Ebroad*n*[(N/k)-1]
in view of broadcasting the message, where n
is the number of bits in the broadcasted message, and all the nodes depletes an
amount of energy equals to
E = n*Ep
in view of the reception of the message
regarding the leader of the cluster.
Let p be the average number of packets that
are transmitted by any node and let the length of each packet be l. For
implementation, the case study of Zigbee radio sensors is considered in which
the underlying operating system is tiny OS having packet size of l=114 bytes.
The important points to analyze in the scenario is:
For More Information Click Here
