![International Journal of Research in Computer Science
eISSN 2249-8265 Volume 2 Issue 4 (2012) pp. 51-57
© White Globe Publications
www.ijorcs.org
A SURVEY ON CLUSTERING TECHNIQUES FOR
WIRELESS SENSOR NETWORK
Rudranath Mitra1, Diya Nandy2
1
Asst. Professor, Dept. of IT, Heritage Institute of Technology, Kolkata.
Email: mitra.rudra@gmail.com
2
Student M.Tech, Dept. of IT, Heritage Institute of Technology, Kolkata.
Email: diya1nandy@gmail.com
Abstract: Wireless sensor networks have been used in high reliability of the sensor networks. To make sensor
various fields like battle feilds, surveillance, schools, networks more reliable, the attention to research on
colleges, etc. It has been used in our day-to-day life. Its heterogeneous wireless sensor networks has been
growth increases day by day. Sensor node normally increasing in recent past [6-9].
senses the physical event from the environment such as
temperature, sound, vibration, pressure etc. Sensor In order to support data aggregation through
nodes are connected with each other through wireless efficient network organization, nodes can be
medium such as infrared or radio waves it depends on partitioned into a number of small groups called
applications. Each node has its internal memory to clusters. Each cluster has a coordinator, referred to as a
store the information regarding the event packets. In cluster head, and a number of member nodes.
this paper we will come to know the various Clustering results in a two-tier hierarchy in which
algorithms in clustering techniques for wireless sensor cluster heads (CHs) form the higher tier while member
networks and discuss them. Clustering is a key nodes form the lower tier. Figure 1 illustrates data flow
technique used to extend the lifetime of a sensor in a clustered network. The member nodes report their
network by reducing energy consumption .It can also data to the respective CHs. The CHs aggregate the data
increase network scalability. Sensor nodes are and send them to the central base through other CHs.
considered to be homogeneous since the researches in Because CHs often transmit data over longer distances,
the feild of WSNs have been evolved but in reality they lose more energy compared to member nodes.
homogeneous sensor networks hardly exist. Here we The network may be reclustered periodically in order
will discuss some of the impact of heterogeneous to select energy-abundant nodes to serve as CHs, thus
sensor networks on WSN and various clustering distributing the load uniformly on all the nodes.
algorithms used in HWSN. Besides achieving energy efficiency, clustering
reduces channel contention and packet collisions,
Keywords: Base Station (BS), Cell Header (CH), inter- resulting in better network throughput under high load.
cluster communication, intra- cluster communication. Clustering has been shown to improve network
lifetime, a primary metric for evaluating the
I. INTRODUCTION performance of a sensor network. Although there is no
With the advances in the technology of micro unified definition of “network lifetime,” as this
electromechanical system (MEMS) and developments concept depends on the objective of an application,
in wireless communications, wireless sensor networks common definitions include the time until the first/last
have emerged [1]. In past few years, wireless sensor node in the network depletes its energy and the time
networks (WSNs) have become one of the most until a node is disconnected from the base station.
interesting areas of research. A WSN is composed of a
number of wireless sensor nodes which form a sensor
field and a sink. These large numbers of nodes with
low-cost, low-power, and capable of communication at
short distances perform limited computation and
communicate wirelessly form the WSNs [2]. Specific
functions such as sensing, tracking and alerting [3] can
be obtained through cooperation among these nodes.
These functions make wireless sensors very useful for
monitoring natural phenomena, environmental changes
[4], controlling security, estimating traffic flows,
monitoring military application [5], and tracking
friendly forces in the battlefields. These tasks require Figure 1. Illustration of data flow in a clustered network
www.ijorcs.org](https://image.slidesharecdn.com/vol2-iss4-09-a-survey-on-clustering-techniques-for-wireless-sensor-network-120714062535-phpapp02/85/A-Survey-on-Clustering-Techniques-for-Wireless-Sensor-Network-1-320.jpg)
![52 Rudranath Mitra, Diya Nandy
The important components of wireless sensor algorithm (DEEC) [7], a probability based clustering
network are discussed below: algorithm was proposed. DEEC elects cluster heads
based on the knowledge of the ratio between residual
• Sensor Node: A sensor node is the core component
energy of each nodes and the average energy of the
of a WSN. Sensor nodes can take on multiple roles
network. This knowledge however requires additional
in a network, such as simple sensing; data storage;
energy consumption to share the information among
routing; and data processing.
the sensor nodes. Stable Election Protocol (SEP) [8] is
• Clusters: Clusters are the organizational unit for another heterogeneity-aware protocol. It does not
WSNs. The dense nature of these networks requires require energy knowledge sharing but is based on
the need for them to be broken down into clusters to assigning weighted election probabilities of each node
simplify tasks such a communication. to be elected cluster head according to their respective
• Clusterheads: Clusterheads are the organization energy. This approach ensures that the cluster head
leader of a cluster. They often are required to election is randomly selected and distributed based on
organize activities in the cluster. These tasks the fraction of energy of each node therefore assuring a
include but are not limited to data-aggregation and uniform use of the nodes energy. In SEP, two types of
organizating the communication schedule of a nodes (two tier in-clustering) and two level hierarchies
cluster. were considered. SEP is based on weighted election
• Base Station: The base station is at the upper level probabilities of each node to become cluster head
of the hierarchical WSN. It provides the according to the remaining energy in each node. A
communication link between the sensor network survey of clustering algorithm was presented in Ref.
and the end-user. [9]; the even distribution of sensors in clusters is
• End User: The data in a sensor network can be used another primary objective of clustering called load
for a wide-range of applications. [1] Therefore, a balancing that needs to be considered when designing
particular application may make use of the network a robust protocol for WSNs [7]. The clustering issue
data over the internet, using a PDA, or even a was also discussed in a review on wireless multimedia
desktop computer. In a queried sensor network sensor networks [1]. The contribution of this work is a
(where the required data is gathered from a query SEP extension called SEP-E, by considering a three-
sent through the network). This query is generated tier node classification in a two-level hierarchical
by the end user. network. The HEED protocol considers multi-hop
network and assumes that all nodes are equally
II. RELATED WORKS important. A node who has the highest residual energy
is considered as the cluster-head. In HEED, each nodes
Many researchers have proposed different execute a constant number of iterations.
techniques related to clustering in wireless sensor
network. The LEACH protocol [2] is an application- The new node type for the purpose of this study is
specific clustering protocol, which has been shown to referred to as “intermediate nodes”, which serves as a
significantly improve the network lifetime. It assumes bridge between the advanced nodes and the normal
that every node is reachable in a single hop and that nodes. The intermediate nodes can take on the role of
load distribution is uniform among all nodes. LEACH information fusion and filtering depending on the
assigns a fixed probability to every node so as to elect application settings, which we intend to study further.
itself as a CH. The clustering process involves only Our goal is to achieve a robust self-configured WSN
one iteration, after which a node decides whether to that maximizes its lifetime.
become a CH or not. Nodes take turns in carrying the
role of a CH. However, the LEACH protocol is not III. OVERVIEW OF PROPOSED ALGORITHMS
heterogeneity-aware, in the sense that when there is an
A. Heuristic Algorithms
energy difference to some threshold between these
nodes in the network, the sensors die out faster than a An heuristic algorithm is an algorithm that usually
more uniform energy setting [2]. In real life situation it has one or both of the following goals in solving a
is difficult for the sensors to maintain their energy problem:
uniformly, this makes energy imbalance between
• Finding an algorithm with reasonable run-time
nodes to occur easily. LEACH assumes that the energy
(time needed to set up clusters is affordable); and/or
usage of each node with respect to the overall energy
• With finding the optimal solution. This means that a
of the system or network is homogeneous.
heuristic algorithm leads to reasonable performance
Conventional protocols such as Minimum
and is not based on particular metrics.
Transmission Energy (MTE) and Direct Transmission
(DT) do not also assure a balanced and uniformly use There are many types of heuristic algorithms that
of the sensor’s respective energy as the network exist in choosing cluster-heads. We will see each of
evolves. In Distributed Energy-Efficient Clustering these algorithms one by one.
www.ijorcs.org](https://image.slidesharecdn.com/vol2-iss4-09-a-survey-on-clustering-techniques-for-wireless-sensor-network-120714062535-phpapp02/85/A-Survey-on-Clustering-Techniques-for-Wireless-Sensor-Network-2-320.jpg)
![A Survey on Clustering Techniques for Wireless Sensor Network 53
Linked Cluster Algorithm (LCA) [4]: follows a simple set of rules to determine their
respective cluster-head. The 1st d rounds are called
LCA, was one of the very first clustering
flood-max, used to propagate the largest node ids.
algorithms developed. It was initially developed for
After this is complete, the 2nd d rounds of flooding
wired sensors, but later implemented in wireless sensor
occur. This round is called flood-min, used to allow
networks. In LCA, each node is assigned a unique ID
the smaller node id to reclaim some of their territory.
number and has two ways of becoming a cluster-head.
Then each node evaluates the logged entries following
The first way is if the node has the highest ID number
the rules listed below [1]:
in the set including all neighbor nodes and the node
itself. The second way, assuming none of its neighbors • Rule 1: Each node checks to see if it has received its
are clusterheads, then it becomes a clusterhead. own id in the 2nd d rounds of flooding. If it has, then it
can declare itself the cluster-head and skip the other
rules. Otherwise it proceeds to Rule 2.
Linked Cluster Algorithm 2 (LCA2) [5]:
• Rule 2: Each node looks for node pairs. Once this is
LCA2 was proposed to eliminate the election of an complete, it selects the minimum node pair to be the
unnecessary number of clusterheads, as in LCA. In cluster-head. If a node pair does not exist, they proceed
LCA2, they introduce the concept of a node being to Rule 3.
covered and non-covered. A node is considered
• Rule 3: Elects the maximum node id in the 1st d
covered if one of its neighbours is a cluster-head.
rounds of flooding as the cluster-head for this node.
Clusterheads are elected starting with the node having
the lowest ID among non-covered neighbours. This algorithm is valid only if the following two
assumptions are made:
• Assumption 1: During the flooding, no node id will
propagate further than d-hops from the originating
node.
• Assumption 2: All nodes that survive the flood-max
elect themselves cluster-heads.
B. Weighted Schemes
Weighted Clustering Algorithm (WCA) [6]:
The algorithm explained in this section is a non-
periodic procedure to the cluster-head election,
invoked on demand every time a reconfiguration of the
networks topology is unavoidable.[6].This clustering
algorithm tries to find a long-lasting architecture
during the first cluster-head election. When a sensor
Figure 2: Classification of proposed clustering schemes loses the connection with any cluster-head, the election
procedure is invoked to find a new clustering topology.
Highest-Connectivity Cluster Algorithm [6]: This is an important feature in power saving, as the re-
This algorithm is similar to LCA. In this scheme the election procedure, which consumes energy, occurs
number of node neighbours is broadcast to the less frequently. This algorithm is based on a
surrounding nodes. The result is that instead of looking combination of metrics that takes into account several
at the ID number, the connectivity of a node is system parameters such as: the ideal node degree;
considered. The node with the highest connectivity transmission power; mobility; and the remaining
(connected to the most number of nodes) is elected energy of the nodes. Depending on the specific
cluster-head, but in the case of a tie, the node with the application, any or all of these parameters can be used
lowest ID prevails. as a metric to elect cluster-heads. Another important
aspect of the algorithm is that it is fully distributed;
Max-Min D-Cluster Algorithm: meaning that all the nodes in the mobile network share
With Max-Min D cluster, the authors[11] propose a the same responsibility acting as cluster-heads.
new distributed cluster-head election procedure, where Cluster-head election procedure:
no node is more than d (d is a value selected for the
heuristic) hops away from the cluster-head. This The election procedure is based upon a global
algorithm provides load balancing among cluster- parameter , that is called combined weight, which is
heads. The cluster-head selection criteria is developed described by[6]:
by having each node initiate 2d rounds of flooding,
Wv = w1∆v + w2Dv + w3Mv + w4Pv
from which the results are logged. Then each node
www.ijorcs.org](https://image.slidesharecdn.com/vol2-iss4-09-a-survey-on-clustering-techniques-for-wireless-sensor-network-120714062535-phpapp02/85/A-Survey-on-Clustering-Techniques-for-Wireless-Sensor-Network-3-320.jpg)
![54 Rudranath Mitra, Diya Nandy
Where, w1, w2, w3, w4 are the weighing factors to clusterhead is made dynamically at each interval.
for the corresponding system parameters. The The elevation decision is made solely by each node
weighting factors can be chosen based upon the independent of other nodes to minimize overhead in
specific application. The combined weight is clusterhead establishment. This decision is a function
calculated by each node and broadcast across the of the percentage of optimal clusterheads in a network
network. The node with smallest Wv is chosen as the (determined a priori on application), in combination
cluster-head. The first component, w1∆v, helps in with how often and the last time a given node has been
efficient MAC functionality, as it is always important a clusterhead in the past. The threshold function is
to have a bound on the maximum number of nodes in a defined as:
cluster. The second component, Dv, is the average
distance from the neighbours and is strictly related to T(n) ={P/1−P(r mod 1/P ), if n € G
power consumption. It is known [6] that more power is Otherwise 0
required for long range transmission. The third where n is the given node, P is the a priori
component is due to mobility of the nodes. It is probability of a node being elected as a cluster-head, r
desirable that a cluster-head moves very slow, in order is the current round number and G is the set of nodes
to have a more stable cluster architecture. From this that have not been elected as cluster-heads in the last
point of view a node that moves slowly is always a 1/P rounds. Each node during cluster-head selection
better choice to be a cluster-head [6]. The last will generate a random number between 0 and 1. If the
component is directly related to the available energy in number is less than the threshold (T(n)), the node will
a node: if a node was already a cluster-head it may become a cluster-head. Following elevation to cluster-
have consumed a large amount of energy and should head, the new cluster-head will broadcast its status to
not be considered for the next cluster-head election. neighbouring nodes. These nodes will then determine
The weighing factors (w1,w2,w3,w4) can be chosen the optimal cluster-head (in terms of minimum energy
according to the system needs. For example, power required for transmission) and relay their desire to be
control is very important in CDMA networks [6], thus in that particular cluster. The broadcast messages as
the weight of the corresponding parameter. The well as cluster establishment messages are transmitted
flexibility of changing the weight factors helps in the using CSMA (Carrier Sense Multiple Access) to
application of this algorithm for different minimize collisions. Following cluster establishment,
implementations. cluster-heads will create a transmission schedule and
broadcast the schedule to all nodes in their respective
Complexity due to distributiveness:
cluster. The schedule consists of TDMA slots for each
The time required for the selection of the node with neighbouring node. This scheduling scheme allows for
minimum Wv depends on the implementation of the energy minimization as nodes can turn off their radio
algorithm. As it is not possible [6] to have a during all but their scheduled time-slot.
centralized server in ad hoc sensor networks, the
algorithm proposes a distributed solution in which all TL-LEACH [5]:
nodes broadcast their ids along with Wv values. Each Two-Level Hierarchy LEACH (or TLLEACH) is a
node receives the broadcast from its neighbors and proposed extension to the LEACH algorithm. It
stores the information. The stored information is again utilizes two levels of cluster-heads (primary and
exchanged with the immediate neighbors and the secondary) in addition to the other simple sensing
process continues until all the nodes become aware of nodes. In this algorithm, the primary clusterhead in
the node with the smallest Wv. The time required will each cluster communicates with the secondaries, and
depend on the diameter of the underlying network. the corresponding secondaries communicate with the
nodes in their sub-cluster. Data-fusion can also be
C. Hierarchical Schemes performed as in LEACH. In addition, communication
LEACH [5]: within a cluster is still scheduled using TDMA time-
slots. The organization of a round will consist of first
Low-Energy Adaptive Clustering Hierarchy (or selecting the primary and secondary clusterheads using
LEACH) was one of the first major improvements on the same mechanism as LEACH, with the a priori
conventional clustering approaches in wireless sensor probability of being elevated to a primary cluster-head
networks. Conventional approaches algorithms such as less than that of a secondary node. Communication of
MTE (Minimum-Transmission-Energy) or direct- data from source node to sink is achieved in two steps :
transmission do not lead to even energy dissipation
throughout a network. LEACH provides a balancing of 1. Secondary nodes collect data from nodes in their
energy usage by random rotation of clusterheads. The respective clusters. Data-fusion can be performed at
algorithm is also organized in such a manner that data- this level.
fusion can be used to reduce the amount of data 2. Primary nodes collect data from their respective
transmission. The decision of whether a node elevates secondary clusters. Data-fusion can also be
www.ijorcs.org](https://image.slidesharecdn.com/vol2-iss4-09-a-survey-on-clustering-techniques-for-wireless-sensor-network-120714062535-phpapp02/85/A-Survey-on-Clustering-Techniques-for-Wireless-Sensor-Network-4-320.jpg)
![A Survey on Clustering Techniques for Wireless Sensor Network 55
implemented at the primary cluster-head level. The transmission time-frame will be transmitting to the
two-level structure of TL-LEACH reduces the base station. Data-fusion occurs at every node in the
amount of nodes that need to transmit to the base sensor network allowing for all relevant information to
station, effectively reducing the total energy usage. permeate across the network [5]. In addition, the
average transmission range required by a node to relay
EECS:
information can be much less than in LEACH,
An Energy Efficient Clustering Scheme (or EECS) resulting in an energy improvement versus the
is a clustering algorithm in which cluster-head hierarchical clustering approach.
candidates compete for the ability to elevate to cluster-
head for a given round. This competition involves B. GROUP
candidates broadcasting their residual energy to The Group algorithm is a grid-based clustering
neighbouring candidates. If a given node does not find algorithm. In this algorithm one of the sinks (called the
a node with more residual energy, it becomes a cluster- primary sink), dynamically, and randomly builds the
head. Cluster formation is different than that of cluster grid . The cluster-heads are arranged in a grid-
LEACH. LEACH forms clusters based on the like manner as in Fig. 3. Forwarding of data queries
minimum distance of nodes to their corresponding from the sink to source node are propagated from the
cluster-head . EECS extends this algorithm by Grid Seed (GS) to its cluster-heads, and so on. The GS
dynamic sizing of clusters based on cluster distance is a node within a given radius from the primary sink.
from the base station[4] . The result is an algorithm In terms of cluster-head selection, on a given round the
that addresses the problem that clusters at a greater primary sink selects a GS based on residual energy.
range from the base station require more energy for Once the GS has been selected, the GS selects cluster-
transmission than those that are closer. Ultimately, this heads along the corners of the grid at a range R. Each
improves the distribution of energy throughout the new cluster-head will then select more cluster-heads
network, resulting in better resource usage and along the grid until all cluster-heads have been
extended network lifetime. selected. These selections are based on the residual
HEED: energy of nodes near the corners of the grid. Data
transmission in GROUP is dependent on the type of
Hybrid Energy-Efficient Distributed Clustering (or data being collected. In the case of a location unaware
HEED) is a multi-hop clustering algorithm for data query (data that is not dependant on the location
wireless sensor networks, with a focus on efficient of the sensing node), the query is passed from the
clustering by proper selection of cluster-heads based central most sink in the network to its nearest cluster-
on the physical distance between nodes. The main head. That cluster-head will then broadcast the
objectives of HEED are to: message to neighbouring cluster-heads. If the data is
• Distribute energy consumption to prolong network location aware, then the requests are sent down the
lifetime; chain of cluster-heads towards the specified region
• Minimize energy during the cluster-head selection using unicast packets. For both data queries, data is
phase; transmitted upstream through the chain of cluster-
• Minimize the control overhead of the network. heads established during cluster formation. Energy
conservation is achieved due to the lower transmission
Clusterheads are determined based on two distance for upstream data. In LEACH, a cluster-head
important parameters: must transmit data to the base station directly, while in
GROUP, the data is transmitted across short ranges
1. The residual energy of each node is used to
through the upstream path[6].
probabilistically choose the initial set of
clusterheads. This parameter is commonly used in
many other clustering schemes.
2. Intra-Cluster Communication Cost is used by nodes
to determine the cluster to join.
IV. GRID SCHEMES
A. PEGASIS
Power-Efficient GAthering in Sensor Information
Systems (or PEGASIS) is a data-gathering algorithm
that establishes the concept that energy savings can
result from nodes not directly forming clusters. The
algorithm presents the idea that if nodes form a chain Figure 3: GROUP Example of Cluster Grid
from source to sink, only 1 node in any given
www.ijorcs.org](https://image.slidesharecdn.com/vol2-iss4-09-a-survey-on-clustering-techniques-for-wireless-sensor-network-120714062535-phpapp02/85/A-Survey-on-Clustering-Techniques-for-Wireless-Sensor-Network-5-320.jpg)
![56 Rudranath Mitra, Diya Nandy
V. CLUSTERING ALGORITHMS FOR composed of three types of nodes including Type_0,
HETEROGENEOUS WIRELESS NETWORK Type_1 and some management nodes as shown in Fig.
1. Type_0 and Type_1 nodes vary in capabilities of
A WSN is composed of hundreds of sensor
sensing, energy and software. They have the
networks distributed randomly. Clustering is one of the
responsibility of sensing events, while the management
best technique to increase network heterogeneity.
nodes perform management of both types of nodes
Heterogeneous network should be energy efficient to
during cluster formation. EDFCM is specially
take the advantages of node heterogeneity. So now we
proposed for heterogeneous networks to provide the
will discuss some of the clustering prtocols for
longer lifetime and more reliable transmission service.
heterogeneous wireless sensor network[1].
Unlike the other energy efficient protocols, the process
A. Stability-oriented clustering protocols for HWSNs of cluster head selection in EDFCM is based on a
method of one-step energy consumption forecast. It
Here we will discuss the stability period of wireless uses the average energy consumptions of the two types
sensor network. Stability period is actually the time of cluster heads in previous round for this purpose.
interval before the death of the first node. It is very The more remaining energy in a node after the
important for the applications where the response from operation of next round, higher the chances of node to
the response nodes must be reliable. be selected as a cluster head [3].
B. Stable Election Protocol for Clustered HWSNs
Smaragdakis G. et al. describe the impact of
heterogeneity on the heterogeneous-oblivious
protocols and instability of the protocols like LEACH,
in the presence of heterogeneity, once some nodes die
[2]. So he introduced a heterogeneity protocol named,
Stable Election Protocol (SEP).It Ensures that the
cluster-head is selected based the fraction of energy of
each node, this assures each nodes energy is properly
used. In SEP, two types of nodes (normal and
advanced) are considered. It is based on weighted
election probabilities of each node to become cluster
head according to the remaining energy in each node.
This prolongs the stability period i.e. the time interval Figure 4: Type_1 and Type_2 nodes are shown by circle
before the death of the first node. Virtually, there are and triangle respectively and management nodes by star
n·(1+α·m) nodes with energy equal to the initial
energy of a normal node. In the heterogeneous
scenario, the average number of cluster heads per D. Base Station Initiated Dynamic Routing
round per epoch is equal to n·(1 + α·m)·Pnrm. The Protocol:
weighed probabilities for normal and advanced nodes S. Verma et al. propose a routing protocol that is
are respectively: based on clustering and uses heterogeneity in nodes to
increase the network lifetime. In this scheme, some
Pnrm = Popt / (1+α m) , Padv = Popt (1+α)/(1+α m) nodes which are stronger than other nodes in terms of
In most rounds, no cluster head is selected by SEP. power, computational capability and location
In such rounds where no CH is selected, the data awareness, work as the cluster heads. They forward
packets cannot be transmitted to the base station. This information to their parents, towards the base station
is a great disadvantage to the reliable transmission in by aggregating all the information from their clusters
the networks, especially for some important real-time members. Following assumptions are considered in
transmission tasks. this scheme: all nodes are deployed uniformly in the
field and CHs will be assumed dead only when their
C. Novel Stable Selection and Reliable energy is very less. There is no collision between inter
Transmission Protocol for Clustered HWSN cluster and intra cluster communication. Transmission
H. Zhou et al propose a model of energy and power of the CH is adjusted in such a way that only
computation heterogeneity for heterogeneous wireless single hop broadcast is possible. In this algorithm, how
sensor networks[3]. They also propose a protocol far a CH is from the BS, is defined as level. Low level
named Energy Dissipation Forecast and Clustering means that CH is near to the BS and if level is high it
Management (EDFCM) for HWSNs. This algorithm means CH is away from BS accordingly. Data flow
balances the energy consumption round by round, will be always from higher level to lower level.
which will provide the longest stability period for Decision of levels by base station is based on the range
network. The heterogeneous model they consider is of the CH and normal node. Ranges of all the nodes
www.ijorcs.org](https://image.slidesharecdn.com/vol2-iss4-09-a-survey-on-clustering-techniques-for-wireless-sensor-network-120714062535-phpapp02/85/A-Survey-on-Clustering-Techniques-for-Wireless-Sensor-Network-6-320.jpg)
![A Survey on Clustering Techniques for Wireless Sensor Network 57
are enough to ensure the connectivity and coverage. much work to be done. Many energy improvements
BS sets its level to zero and broadcasts a packet to thus far have focused with minimization of energy
initiate the scheme. Base station mentions that this associated in the cluster-head selection process.
packet is only for CHs. Since the CHs have different Wireless sensor networks are not always
signal strength from normal nodes, they receive homogeneous, they may be heterogeneous too. This
paper surveys research protocols for clustering in
heterogeneous wireless sensor networks. Clustering is
a good technique to reduce energy consumption and to
provide stability in wireless sensor networks. We
classified all protocols according to stability and
energy efficiency of network. We summarize a number
of schemes, stating their strengths and limitations.
Finally on the basis of survey work, we conclude that
the heterogeneous wireless sensor networks are more
Figure 5: Cluster hierarchy in sensing field suitable for real life applications as compared to the
homogeneous counterpart.
different signal strength from normal nodes, they
receive the packet and set their levels accordingly. VIII. REFERENCES
When the CHs of first level are selected, they
[1] I.F. Akyildiz, W. Su, Y. Sankarasubramaniam and E.
broadcast their level. CHs at lower level receive the Cayirci, .Wireless sensor networks: a survey.,Computer
packet according to the signal strength[2]. They Networks 38 (4), 2002, 393.422.
choose their parent from upper level CHs only. This [2] K. Romer, O. Kastin, and F. Mattern, "Middleware
process is repeated again and again until all CHs are challenges for wireless sensor networks," ACM
connected. CH now broadcast a message that all sensor SIGMOBILE Mobile Computing and Communications
nodes should join the CH according to the RSS (Radio Review, Vol. 6, No. 4, 2002, 59-61.
Signal Strength). Communication between CH and [3] R. Shorey, A. Ananda, and W. T. Ooi, "Mobile,wireless,
sensing nodes is single hop, while communication and sensor networks," 1st edition, IEEE Press, John
between different CH is multiple hops. All CHs sends Wiley & Sons, 2006.
their position, level and energy consumption to the BS [4] I. F. Akyildiz,W. Su, Y. Sankarasubramaniam, and E.
at the end of the round. BS then analyzes the energy Cayirci, “A Survey on Sensor Netowrks,” IEEE
consumption of different CH at the same level. Communications Magazine, vol. 40, no. 8,pp. 102–114,
Aug 2002
VI. COMPARATIVE STUDY [5] S. Meyer and A. Rakotonirainy, “A Survey of Research
on Context-Aware Homes,” Workshop on Wearable,
Table1: Comparisons
Invisible, Context-Aware, Ambient, Pervasive and
Power Ubiquitous Computing, Adelaide Australia, 2003.
Classification Mobility Scalability
usage
[6] B. Warneke, M. Last, B. Liebowitz, Kristofer, and S.
LCA Heuristics CH Max Good Pister, “Smart Dust: Communicating with a Cubic-
mobile Millimeter Computer,” Computer Magazine, vol. 34, no.
WCA Weighted CH fixed Limited Limited 1, pp. 44–51, Jan 2001.
LEACH Hierarchical BS fixed Max Good [7] A. A. Abbasi and M. Younis. A survey on clustering
PEGASIS Grid BS fixed Max Good algorithms for wireless sensor networks. Computer
Communication, 30(14-15):2826–2841, 2007.
VII. CONCLUSION AND FUTURE SCOPE [8] I. F. Akyildiz, T. Melodia, and K. R. Chowdhury. A
survey on wireless multimedia sensor networks.
In this paper we have examined the current state of Computer Networks, 51(4):921–960, 2007.
proposed clustering protocols, specifically with respect [9] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E.
to their power and reliability requirements. In wireless Cayirci. A survey on sensor networks. IEEE
sensor networks, the energy limitations of nodes play a Communications Magazine, 2002.
crucial role in designing any protocol for
implementation [1]. In addition, Quality of Service
metrics such as delay, data loss tolerance, and network
lifetime expose reliability issues when designing
recovery mechanisms for clustering schemes. These
important characteristics are often opposed, as one
often has a negative impact on the other. Protocols
presented in this paper offer a promising improvement
over conventional clustering; however there is still
www.ijorcs.org](https://image.slidesharecdn.com/vol2-iss4-09-a-survey-on-clustering-techniques-for-wireless-sensor-network-120714062535-phpapp02/85/A-Survey-on-Clustering-Techniques-for-Wireless-Sensor-Network-7-320.jpg)
A Survey on Clustering Techniques for Wireless Sensor Network
- 1. International Journal of Research in Computer Science eISSN 2249-8265 Volume 2 Issue 4 (2012) pp. 51-57 © White Globe Publications www.ijorcs.org A SURVEY ON CLUSTERING TECHNIQUES FOR WIRELESS SENSOR NETWORK Rudranath Mitra1, Diya Nandy2 1 Asst. Professor, Dept. of IT, Heritage Institute of Technology, Kolkata. Email: mitra.rudra@gmail.com 2 Student M.Tech, Dept. of IT, Heritage Institute of Technology, Kolkata. Email: diya1nandy@gmail.com Abstract: Wireless sensor networks have been used in high reliability of the sensor networks. To make sensor various fields like battle feilds, surveillance, schools, networks more reliable, the attention to research on colleges, etc. It has been used in our day-to-day life. Its heterogeneous wireless sensor networks has been growth increases day by day. Sensor node normally increasing in recent past [6-9]. senses the physical event from the environment such as temperature, sound, vibration, pressure etc. Sensor In order to support data aggregation through nodes are connected with each other through wireless efficient network organization, nodes can be medium such as infrared or radio waves it depends on partitioned into a number of small groups called applications. Each node has its internal memory to clusters. Each cluster has a coordinator, referred to as a store the information regarding the event packets. In cluster head, and a number of member nodes. this paper we will come to know the various Clustering results in a two-tier hierarchy in which algorithms in clustering techniques for wireless sensor cluster heads (CHs) form the higher tier while member networks and discuss them. Clustering is a key nodes form the lower tier. Figure 1 illustrates data flow technique used to extend the lifetime of a sensor in a clustered network. The member nodes report their network by reducing energy consumption .It can also data to the respective CHs. The CHs aggregate the data increase network scalability. Sensor nodes are and send them to the central base through other CHs. considered to be homogeneous since the researches in Because CHs often transmit data over longer distances, the feild of WSNs have been evolved but in reality they lose more energy compared to member nodes. homogeneous sensor networks hardly exist. Here we The network may be reclustered periodically in order will discuss some of the impact of heterogeneous to select energy-abundant nodes to serve as CHs, thus sensor networks on WSN and various clustering distributing the load uniformly on all the nodes. algorithms used in HWSN. Besides achieving energy efficiency, clustering reduces channel contention and packet collisions, Keywords: Base Station (BS), Cell Header (CH), inter- resulting in better network throughput under high load. cluster communication, intra- cluster communication. Clustering has been shown to improve network lifetime, a primary metric for evaluating the I. INTRODUCTION performance of a sensor network. Although there is no With the advances in the technology of micro unified definition of “network lifetime,” as this electromechanical system (MEMS) and developments concept depends on the objective of an application, in wireless communications, wireless sensor networks common definitions include the time until the first/last have emerged [1]. In past few years, wireless sensor node in the network depletes its energy and the time networks (WSNs) have become one of the most until a node is disconnected from the base station. interesting areas of research. A WSN is composed of a number of wireless sensor nodes which form a sensor field and a sink. These large numbers of nodes with low-cost, low-power, and capable of communication at short distances perform limited computation and communicate wirelessly form the WSNs [2]. Specific functions such as sensing, tracking and alerting [3] can be obtained through cooperation among these nodes. These functions make wireless sensors very useful for monitoring natural phenomena, environmental changes [4], controlling security, estimating traffic flows, monitoring military application [5], and tracking friendly forces in the battlefields. These tasks require Figure 1. Illustration of data flow in a clustered network www.ijorcs.org
- 2. 52 Rudranath Mitra, Diya Nandy The important components of wireless sensor algorithm (DEEC) [7], a probability based clustering network are discussed below: algorithm was proposed. DEEC elects cluster heads based on the knowledge of the ratio between residual • Sensor Node: A sensor node is the core component energy of each nodes and the average energy of the of a WSN. Sensor nodes can take on multiple roles network. This knowledge however requires additional in a network, such as simple sensing; data storage; energy consumption to share the information among routing; and data processing. the sensor nodes. Stable Election Protocol (SEP) [8] is • Clusters: Clusters are the organizational unit for another heterogeneity-aware protocol. It does not WSNs. The dense nature of these networks requires require energy knowledge sharing but is based on the need for them to be broken down into clusters to assigning weighted election probabilities of each node simplify tasks such a communication. to be elected cluster head according to their respective • Clusterheads: Clusterheads are the organization energy. This approach ensures that the cluster head leader of a cluster. They often are required to election is randomly selected and distributed based on organize activities in the cluster. These tasks the fraction of energy of each node therefore assuring a include but are not limited to data-aggregation and uniform use of the nodes energy. In SEP, two types of organizating the communication schedule of a nodes (two tier in-clustering) and two level hierarchies cluster. were considered. SEP is based on weighted election • Base Station: The base station is at the upper level probabilities of each node to become cluster head of the hierarchical WSN. It provides the according to the remaining energy in each node. A communication link between the sensor network survey of clustering algorithm was presented in Ref. and the end-user. [9]; the even distribution of sensors in clusters is • End User: The data in a sensor network can be used another primary objective of clustering called load for a wide-range of applications. [1] Therefore, a balancing that needs to be considered when designing particular application may make use of the network a robust protocol for WSNs [7]. The clustering issue data over the internet, using a PDA, or even a was also discussed in a review on wireless multimedia desktop computer. In a queried sensor network sensor networks [1]. The contribution of this work is a (where the required data is gathered from a query SEP extension called SEP-E, by considering a three- sent through the network). This query is generated tier node classification in a two-level hierarchical by the end user. network. The HEED protocol considers multi-hop network and assumes that all nodes are equally II. RELATED WORKS important. A node who has the highest residual energy is considered as the cluster-head. In HEED, each nodes Many researchers have proposed different execute a constant number of iterations. techniques related to clustering in wireless sensor network. The LEACH protocol [2] is an application- The new node type for the purpose of this study is specific clustering protocol, which has been shown to referred to as “intermediate nodes”, which serves as a significantly improve the network lifetime. It assumes bridge between the advanced nodes and the normal that every node is reachable in a single hop and that nodes. The intermediate nodes can take on the role of load distribution is uniform among all nodes. LEACH information fusion and filtering depending on the assigns a fixed probability to every node so as to elect application settings, which we intend to study further. itself as a CH. The clustering process involves only Our goal is to achieve a robust self-configured WSN one iteration, after which a node decides whether to that maximizes its lifetime. become a CH or not. Nodes take turns in carrying the role of a CH. However, the LEACH protocol is not III. OVERVIEW OF PROPOSED ALGORITHMS heterogeneity-aware, in the sense that when there is an A. Heuristic Algorithms energy difference to some threshold between these nodes in the network, the sensors die out faster than a An heuristic algorithm is an algorithm that usually more uniform energy setting [2]. In real life situation it has one or both of the following goals in solving a is difficult for the sensors to maintain their energy problem: uniformly, this makes energy imbalance between • Finding an algorithm with reasonable run-time nodes to occur easily. LEACH assumes that the energy (time needed to set up clusters is affordable); and/or usage of each node with respect to the overall energy • With finding the optimal solution. This means that a of the system or network is homogeneous. heuristic algorithm leads to reasonable performance Conventional protocols such as Minimum and is not based on particular metrics. Transmission Energy (MTE) and Direct Transmission (DT) do not also assure a balanced and uniformly use There are many types of heuristic algorithms that of the sensor’s respective energy as the network exist in choosing cluster-heads. We will see each of evolves. In Distributed Energy-Efficient Clustering these algorithms one by one. www.ijorcs.org
- 3. A Survey on Clustering Techniques for Wireless Sensor Network 53 Linked Cluster Algorithm (LCA) [4]: follows a simple set of rules to determine their respective cluster-head. The 1st d rounds are called LCA, was one of the very first clustering flood-max, used to propagate the largest node ids. algorithms developed. It was initially developed for After this is complete, the 2nd d rounds of flooding wired sensors, but later implemented in wireless sensor occur. This round is called flood-min, used to allow networks. In LCA, each node is assigned a unique ID the smaller node id to reclaim some of their territory. number and has two ways of becoming a cluster-head. Then each node evaluates the logged entries following The first way is if the node has the highest ID number the rules listed below [1]: in the set including all neighbor nodes and the node itself. The second way, assuming none of its neighbors • Rule 1: Each node checks to see if it has received its are clusterheads, then it becomes a clusterhead. own id in the 2nd d rounds of flooding. If it has, then it can declare itself the cluster-head and skip the other rules. Otherwise it proceeds to Rule 2. Linked Cluster Algorithm 2 (LCA2) [5]: • Rule 2: Each node looks for node pairs. Once this is LCA2 was proposed to eliminate the election of an complete, it selects the minimum node pair to be the unnecessary number of clusterheads, as in LCA. In cluster-head. If a node pair does not exist, they proceed LCA2, they introduce the concept of a node being to Rule 3. covered and non-covered. A node is considered • Rule 3: Elects the maximum node id in the 1st d covered if one of its neighbours is a cluster-head. rounds of flooding as the cluster-head for this node. Clusterheads are elected starting with the node having the lowest ID among non-covered neighbours. This algorithm is valid only if the following two assumptions are made: • Assumption 1: During the flooding, no node id will propagate further than d-hops from the originating node. • Assumption 2: All nodes that survive the flood-max elect themselves cluster-heads. B. Weighted Schemes Weighted Clustering Algorithm (WCA) [6]: The algorithm explained in this section is a non- periodic procedure to the cluster-head election, invoked on demand every time a reconfiguration of the networks topology is unavoidable.[6].This clustering algorithm tries to find a long-lasting architecture during the first cluster-head election. When a sensor Figure 2: Classification of proposed clustering schemes loses the connection with any cluster-head, the election procedure is invoked to find a new clustering topology. Highest-Connectivity Cluster Algorithm [6]: This is an important feature in power saving, as the re- This algorithm is similar to LCA. In this scheme the election procedure, which consumes energy, occurs number of node neighbours is broadcast to the less frequently. This algorithm is based on a surrounding nodes. The result is that instead of looking combination of metrics that takes into account several at the ID number, the connectivity of a node is system parameters such as: the ideal node degree; considered. The node with the highest connectivity transmission power; mobility; and the remaining (connected to the most number of nodes) is elected energy of the nodes. Depending on the specific cluster-head, but in the case of a tie, the node with the application, any or all of these parameters can be used lowest ID prevails. as a metric to elect cluster-heads. Another important aspect of the algorithm is that it is fully distributed; Max-Min D-Cluster Algorithm: meaning that all the nodes in the mobile network share With Max-Min D cluster, the authors[11] propose a the same responsibility acting as cluster-heads. new distributed cluster-head election procedure, where Cluster-head election procedure: no node is more than d (d is a value selected for the heuristic) hops away from the cluster-head. This The election procedure is based upon a global algorithm provides load balancing among cluster- parameter , that is called combined weight, which is heads. The cluster-head selection criteria is developed described by[6]: by having each node initiate 2d rounds of flooding, Wv = w1∆v + w2Dv + w3Mv + w4Pv from which the results are logged. Then each node www.ijorcs.org
- 4. 54 Rudranath Mitra, Diya Nandy Where, w1, w2, w3, w4 are the weighing factors to clusterhead is made dynamically at each interval. for the corresponding system parameters. The The elevation decision is made solely by each node weighting factors can be chosen based upon the independent of other nodes to minimize overhead in specific application. The combined weight is clusterhead establishment. This decision is a function calculated by each node and broadcast across the of the percentage of optimal clusterheads in a network network. The node with smallest Wv is chosen as the (determined a priori on application), in combination cluster-head. The first component, w1∆v, helps in with how often and the last time a given node has been efficient MAC functionality, as it is always important a clusterhead in the past. The threshold function is to have a bound on the maximum number of nodes in a defined as: cluster. The second component, Dv, is the average distance from the neighbours and is strictly related to T(n) ={P/1−P(r mod 1/P ), if n € G power consumption. It is known [6] that more power is Otherwise 0 required for long range transmission. The third where n is the given node, P is the a priori component is due to mobility of the nodes. It is probability of a node being elected as a cluster-head, r desirable that a cluster-head moves very slow, in order is the current round number and G is the set of nodes to have a more stable cluster architecture. From this that have not been elected as cluster-heads in the last point of view a node that moves slowly is always a 1/P rounds. Each node during cluster-head selection better choice to be a cluster-head [6]. The last will generate a random number between 0 and 1. If the component is directly related to the available energy in number is less than the threshold (T(n)), the node will a node: if a node was already a cluster-head it may become a cluster-head. Following elevation to cluster- have consumed a large amount of energy and should head, the new cluster-head will broadcast its status to not be considered for the next cluster-head election. neighbouring nodes. These nodes will then determine The weighing factors (w1,w2,w3,w4) can be chosen the optimal cluster-head (in terms of minimum energy according to the system needs. For example, power required for transmission) and relay their desire to be control is very important in CDMA networks [6], thus in that particular cluster. The broadcast messages as the weight of the corresponding parameter. The well as cluster establishment messages are transmitted flexibility of changing the weight factors helps in the using CSMA (Carrier Sense Multiple Access) to application of this algorithm for different minimize collisions. Following cluster establishment, implementations. cluster-heads will create a transmission schedule and broadcast the schedule to all nodes in their respective Complexity due to distributiveness: cluster. The schedule consists of TDMA slots for each The time required for the selection of the node with neighbouring node. This scheduling scheme allows for minimum Wv depends on the implementation of the energy minimization as nodes can turn off their radio algorithm. As it is not possible [6] to have a during all but their scheduled time-slot. centralized server in ad hoc sensor networks, the algorithm proposes a distributed solution in which all TL-LEACH [5]: nodes broadcast their ids along with Wv values. Each Two-Level Hierarchy LEACH (or TLLEACH) is a node receives the broadcast from its neighbors and proposed extension to the LEACH algorithm. It stores the information. The stored information is again utilizes two levels of cluster-heads (primary and exchanged with the immediate neighbors and the secondary) in addition to the other simple sensing process continues until all the nodes become aware of nodes. In this algorithm, the primary clusterhead in the node with the smallest Wv. The time required will each cluster communicates with the secondaries, and depend on the diameter of the underlying network. the corresponding secondaries communicate with the nodes in their sub-cluster. Data-fusion can also be C. Hierarchical Schemes performed as in LEACH. In addition, communication LEACH [5]: within a cluster is still scheduled using TDMA time- slots. The organization of a round will consist of first Low-Energy Adaptive Clustering Hierarchy (or selecting the primary and secondary clusterheads using LEACH) was one of the first major improvements on the same mechanism as LEACH, with the a priori conventional clustering approaches in wireless sensor probability of being elevated to a primary cluster-head networks. Conventional approaches algorithms such as less than that of a secondary node. Communication of MTE (Minimum-Transmission-Energy) or direct- data from source node to sink is achieved in two steps : transmission do not lead to even energy dissipation throughout a network. LEACH provides a balancing of 1. Secondary nodes collect data from nodes in their energy usage by random rotation of clusterheads. The respective clusters. Data-fusion can be performed at algorithm is also organized in such a manner that data- this level. fusion can be used to reduce the amount of data 2. Primary nodes collect data from their respective transmission. The decision of whether a node elevates secondary clusters. Data-fusion can also be www.ijorcs.org
- 5. A Survey on Clustering Techniques for Wireless Sensor Network 55 implemented at the primary cluster-head level. The transmission time-frame will be transmitting to the two-level structure of TL-LEACH reduces the base station. Data-fusion occurs at every node in the amount of nodes that need to transmit to the base sensor network allowing for all relevant information to station, effectively reducing the total energy usage. permeate across the network [5]. In addition, the average transmission range required by a node to relay EECS: information can be much less than in LEACH, An Energy Efficient Clustering Scheme (or EECS) resulting in an energy improvement versus the is a clustering algorithm in which cluster-head hierarchical clustering approach. candidates compete for the ability to elevate to cluster- head for a given round. This competition involves B. GROUP candidates broadcasting their residual energy to The Group algorithm is a grid-based clustering neighbouring candidates. If a given node does not find algorithm. In this algorithm one of the sinks (called the a node with more residual energy, it becomes a cluster- primary sink), dynamically, and randomly builds the head. Cluster formation is different than that of cluster grid . The cluster-heads are arranged in a grid- LEACH. LEACH forms clusters based on the like manner as in Fig. 3. Forwarding of data queries minimum distance of nodes to their corresponding from the sink to source node are propagated from the cluster-head . EECS extends this algorithm by Grid Seed (GS) to its cluster-heads, and so on. The GS dynamic sizing of clusters based on cluster distance is a node within a given radius from the primary sink. from the base station[4] . The result is an algorithm In terms of cluster-head selection, on a given round the that addresses the problem that clusters at a greater primary sink selects a GS based on residual energy. range from the base station require more energy for Once the GS has been selected, the GS selects cluster- transmission than those that are closer. Ultimately, this heads along the corners of the grid at a range R. Each improves the distribution of energy throughout the new cluster-head will then select more cluster-heads network, resulting in better resource usage and along the grid until all cluster-heads have been extended network lifetime. selected. These selections are based on the residual HEED: energy of nodes near the corners of the grid. Data transmission in GROUP is dependent on the type of Hybrid Energy-Efficient Distributed Clustering (or data being collected. In the case of a location unaware HEED) is a multi-hop clustering algorithm for data query (data that is not dependant on the location wireless sensor networks, with a focus on efficient of the sensing node), the query is passed from the clustering by proper selection of cluster-heads based central most sink in the network to its nearest cluster- on the physical distance between nodes. The main head. That cluster-head will then broadcast the objectives of HEED are to: message to neighbouring cluster-heads. If the data is • Distribute energy consumption to prolong network location aware, then the requests are sent down the lifetime; chain of cluster-heads towards the specified region • Minimize energy during the cluster-head selection using unicast packets. For both data queries, data is phase; transmitted upstream through the chain of cluster- • Minimize the control overhead of the network. heads established during cluster formation. Energy conservation is achieved due to the lower transmission Clusterheads are determined based on two distance for upstream data. In LEACH, a cluster-head important parameters: must transmit data to the base station directly, while in GROUP, the data is transmitted across short ranges 1. The residual energy of each node is used to through the upstream path[6]. probabilistically choose the initial set of clusterheads. This parameter is commonly used in many other clustering schemes. 2. Intra-Cluster Communication Cost is used by nodes to determine the cluster to join. IV. GRID SCHEMES A. PEGASIS Power-Efficient GAthering in Sensor Information Systems (or PEGASIS) is a data-gathering algorithm that establishes the concept that energy savings can result from nodes not directly forming clusters. The algorithm presents the idea that if nodes form a chain Figure 3: GROUP Example of Cluster Grid from source to sink, only 1 node in any given www.ijorcs.org
- 6. 56 Rudranath Mitra, Diya Nandy V. CLUSTERING ALGORITHMS FOR composed of three types of nodes including Type_0, HETEROGENEOUS WIRELESS NETWORK Type_1 and some management nodes as shown in Fig. 1. Type_0 and Type_1 nodes vary in capabilities of A WSN is composed of hundreds of sensor sensing, energy and software. They have the networks distributed randomly. Clustering is one of the responsibility of sensing events, while the management best technique to increase network heterogeneity. nodes perform management of both types of nodes Heterogeneous network should be energy efficient to during cluster formation. EDFCM is specially take the advantages of node heterogeneity. So now we proposed for heterogeneous networks to provide the will discuss some of the clustering prtocols for longer lifetime and more reliable transmission service. heterogeneous wireless sensor network[1]. Unlike the other energy efficient protocols, the process A. Stability-oriented clustering protocols for HWSNs of cluster head selection in EDFCM is based on a method of one-step energy consumption forecast. It Here we will discuss the stability period of wireless uses the average energy consumptions of the two types sensor network. Stability period is actually the time of cluster heads in previous round for this purpose. interval before the death of the first node. It is very The more remaining energy in a node after the important for the applications where the response from operation of next round, higher the chances of node to the response nodes must be reliable. be selected as a cluster head [3]. B. Stable Election Protocol for Clustered HWSNs Smaragdakis G. et al. describe the impact of heterogeneity on the heterogeneous-oblivious protocols and instability of the protocols like LEACH, in the presence of heterogeneity, once some nodes die [2]. So he introduced a heterogeneity protocol named, Stable Election Protocol (SEP).It Ensures that the cluster-head is selected based the fraction of energy of each node, this assures each nodes energy is properly used. In SEP, two types of nodes (normal and advanced) are considered. It is based on weighted election probabilities of each node to become cluster head according to the remaining energy in each node. This prolongs the stability period i.e. the time interval Figure 4: Type_1 and Type_2 nodes are shown by circle before the death of the first node. Virtually, there are and triangle respectively and management nodes by star n·(1+α·m) nodes with energy equal to the initial energy of a normal node. In the heterogeneous scenario, the average number of cluster heads per D. Base Station Initiated Dynamic Routing round per epoch is equal to n·(1 + α·m)·Pnrm. The Protocol: weighed probabilities for normal and advanced nodes S. Verma et al. propose a routing protocol that is are respectively: based on clustering and uses heterogeneity in nodes to increase the network lifetime. In this scheme, some Pnrm = Popt / (1+α m) , Padv = Popt (1+α)/(1+α m) nodes which are stronger than other nodes in terms of In most rounds, no cluster head is selected by SEP. power, computational capability and location In such rounds where no CH is selected, the data awareness, work as the cluster heads. They forward packets cannot be transmitted to the base station. This information to their parents, towards the base station is a great disadvantage to the reliable transmission in by aggregating all the information from their clusters the networks, especially for some important real-time members. Following assumptions are considered in transmission tasks. this scheme: all nodes are deployed uniformly in the field and CHs will be assumed dead only when their C. Novel Stable Selection and Reliable energy is very less. There is no collision between inter Transmission Protocol for Clustered HWSN cluster and intra cluster communication. Transmission H. Zhou et al propose a model of energy and power of the CH is adjusted in such a way that only computation heterogeneity for heterogeneous wireless single hop broadcast is possible. In this algorithm, how sensor networks[3]. They also propose a protocol far a CH is from the BS, is defined as level. Low level named Energy Dissipation Forecast and Clustering means that CH is near to the BS and if level is high it Management (EDFCM) for HWSNs. This algorithm means CH is away from BS accordingly. Data flow balances the energy consumption round by round, will be always from higher level to lower level. which will provide the longest stability period for Decision of levels by base station is based on the range network. The heterogeneous model they consider is of the CH and normal node. Ranges of all the nodes www.ijorcs.org
- 7. A Survey on Clustering Techniques for Wireless Sensor Network 57 are enough to ensure the connectivity and coverage. much work to be done. Many energy improvements BS sets its level to zero and broadcasts a packet to thus far have focused with minimization of energy initiate the scheme. Base station mentions that this associated in the cluster-head selection process. packet is only for CHs. Since the CHs have different Wireless sensor networks are not always signal strength from normal nodes, they receive homogeneous, they may be heterogeneous too. This paper surveys research protocols for clustering in heterogeneous wireless sensor networks. Clustering is a good technique to reduce energy consumption and to provide stability in wireless sensor networks. We classified all protocols according to stability and energy efficiency of network. We summarize a number of schemes, stating their strengths and limitations. Finally on the basis of survey work, we conclude that the heterogeneous wireless sensor networks are more Figure 5: Cluster hierarchy in sensing field suitable for real life applications as compared to the homogeneous counterpart. different signal strength from normal nodes, they receive the packet and set their levels accordingly. VIII. REFERENCES When the CHs of first level are selected, they [1] I.F. Akyildiz, W. Su, Y. Sankarasubramaniam and E. broadcast their level. CHs at lower level receive the Cayirci, .Wireless sensor networks: a survey.,Computer packet according to the signal strength[2]. They Networks 38 (4), 2002, 393.422. choose their parent from upper level CHs only. This [2] K. Romer, O. Kastin, and F. Mattern, "Middleware process is repeated again and again until all CHs are challenges for wireless sensor networks," ACM connected. CH now broadcast a message that all sensor SIGMOBILE Mobile Computing and Communications nodes should join the CH according to the RSS (Radio Review, Vol. 6, No. 4, 2002, 59-61. Signal Strength). Communication between CH and [3] R. Shorey, A. Ananda, and W. T. Ooi, "Mobile,wireless, sensing nodes is single hop, while communication and sensor networks," 1st edition, IEEE Press, John between different CH is multiple hops. All CHs sends Wiley & Sons, 2006. their position, level and energy consumption to the BS [4] I. F. Akyildiz,W. Su, Y. Sankarasubramaniam, and E. at the end of the round. BS then analyzes the energy Cayirci, “A Survey on Sensor Netowrks,” IEEE consumption of different CH at the same level. Communications Magazine, vol. 40, no. 8,pp. 102–114, Aug 2002 VI. COMPARATIVE STUDY [5] S. Meyer and A. Rakotonirainy, “A Survey of Research on Context-Aware Homes,” Workshop on Wearable, Table1: Comparisons Invisible, Context-Aware, Ambient, Pervasive and Power Ubiquitous Computing, Adelaide Australia, 2003. Classification Mobility Scalability usage [6] B. Warneke, M. Last, B. Liebowitz, Kristofer, and S. LCA Heuristics CH Max Good Pister, “Smart Dust: Communicating with a Cubic- mobile Millimeter Computer,” Computer Magazine, vol. 34, no. WCA Weighted CH fixed Limited Limited 1, pp. 44–51, Jan 2001. LEACH Hierarchical BS fixed Max Good [7] A. A. Abbasi and M. Younis. A survey on clustering PEGASIS Grid BS fixed Max Good algorithms for wireless sensor networks. Computer Communication, 30(14-15):2826–2841, 2007. VII. CONCLUSION AND FUTURE SCOPE [8] I. F. Akyildiz, T. Melodia, and K. R. Chowdhury. A survey on wireless multimedia sensor networks. In this paper we have examined the current state of Computer Networks, 51(4):921–960, 2007. proposed clustering protocols, specifically with respect [9] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. to their power and reliability requirements. In wireless Cayirci. A survey on sensor networks. IEEE sensor networks, the energy limitations of nodes play a Communications Magazine, 2002. crucial role in designing any protocol for implementation [1]. In addition, Quality of Service metrics such as delay, data loss tolerance, and network lifetime expose reliability issues when designing recovery mechanisms for clustering schemes. These important characteristics are often opposed, as one often has a negative impact on the other. Protocols presented in this paper offer a promising improvement over conventional clustering; however there is still www.ijorcs.org