Вы находитесь на странице: 1из 4

(IJCNS) International Journal of Computer and Network Security, 41

Vol. 2, No. 2, February 2010

A New Clustering Algorithm for Increasing of


Lifetime in sensor Networks
Abbas Mirzaei Somarin1, Habib Motee Ghader2, Amir Masoud Rahmani 3 and Ali Ghafari4
1
Islamic Azad University-Ardabil Branch,Ardabil, Iran
a.mirzaei@iaut.ac.ir
2
Islamic Azad University- Tabriz Branch- Young Research Club, Tabriz, Iran
Habib_Moti@yahoo.com

3 Islamic Azad University- Oloom Tahgigat Branch-, Tehran, Iran


Rahmani74@yahoo.com

4 Islamic Azad University- Tabriz Branch , Tabriz, Iran


jane.doe@email.com

Abstract: One of the crucial problems in sensor networks is can be recounting according to active nodes because in this
limitation of energy, which affects in network lifetime. There are case networks lifetime will be increased. One of the effective
many algorithms for increasing network lifetime. One of the ways in increasing of networks lifetime is clustering. In
methods in this case, is network clustering. A good clustering sensor network clustering, sensor network divided into some
increases the lifetime of network. In this paper a new clustering branches and one of the nodes have been chosen as a top
algorithm based on Learning Automata for increasing network
branch. Duty of top branch is receiving information from
lifetime is proposed. The proposed algorithm based on LEACH
other branches and sends them to the Sink. In dynamic
Protocol. Performance of proposed algorithm is compared with
previous methods by implementing simulation. The obtained clustering network can be cluster just once and nodes of
results from proposed algorithm is better than others. Results are branches never transfer to others but network nodes can be
indication of high efficiency of suggestion algorithm. the members of other branches. In this paper by using of
learning automata we will change the LEACH algorithm
Keywords: Network Clustering, Algorithm. that increases network lifetime. In the rest of this paper we
will talk about learning automata in section 2. Then in
1. Introduction section 3 LEACH algorithm will be explained. In section 4
Wireless networks are networks, which are, consist of many suggested algorithm will be introduced and in section 5 the
little nodes by low facility. These nodes, which are called a rest of dramatization had been shown.
sensor, can feel special feature such as wetness, temperature, 2. Learning Automata
and pressure in their environment and send this for their
neighbors. In other words two major facilities of these Learning automata is an abstract model that chooses
sensors are the feeling of special parameter around randomly an operation from a set of finite operations and
environment is ability of connection. In some operations then applies it on the environment to the selected operation
these nodes might be join to each other by connective cables, environment is evaluated by learning automata and then
but in most cases a wireless network is completely wireless. informs the evaluated result by the help of a reinforcement
In this network nodes are generally fixed or fixed limited signal to the learning automata. The learning automaton
motions. updates its interior situation by utilizing selected operation
Despite of fixed networks and other wireless networks, and reinforcement signal and selects the next operation
which the quality of service is completely, clear in sensor afterwards. Interconnection between learning automata and
networks there is no fixed recounting. Some of these the environment is shown in Fig.1 [1].
recounting are network proper covering, amount of active
nodes in per time, accuracy of received information in Sink Random
Environment
(central node) and the time of transferring information to
Sink. Some of these recounting such as proper covering and α (n) Learning
β (n )
amount of active nodes in per time are depended to Automata

operation and others like accuracy of received information Figure 1. Learning automata connection with
and time of transferring information Sink intended the environment [2].
feature of network. One of the sensor networks important
features is probability of devastation in some nodes specially The environment shown by where α = {α1, α 2 ,...,α r } is a set
because of losing energy. For this reason there are so many of inputs β = {β1 , β 2 ,..., β r } is a set of outputs and
nodes in wireless networks. In this case if some of these
nodes has been destroyed others can replace them so some of c = {c1 , c 2 ,..., c r } is penalty probabilities. When β is a set
these nodes should be active but others should be inactive in of binary, so the environment is a P type. In this kind of
order not consume their energy. Therefore, quality of service environment β1 = 1 is considered as penalty and β 2 = 0 as
42 (IJCNS) International Journal of Computer and Network Security,
Vol. 2, No. 2, February 2010

reward. In Q type environment, β includes numerous cluster. And the responsibility of each cluster head is
finite members and in S type β has a number of infinite gatherings the information’s from the other nodes of the
same cluster and sending of to the Sink. In LEACH protocol
members. ci is a penalty probability of operation. Learning the cluster head never be changed. And because this reason
automata is divided into two groups as stable and unstable the node that is considered as cluster head, consumes more
structure. Learning automata can be utilized in many energy. And this itself causes the declining of the life long
different ways such as: routing in communication network, of the network. In order to overcome this problem we can
face recognize, programming processes in a computer exchange dynamically the cluster head node between the
network. queue theory, access control in asynchronous members of cluster nodes. In suggested algorithm, we
transmit network, partitioning objects and finding an change the LEACH node in a way that the node of cluster
optimal structure for nerve systems [2, 7-11]. head in each period exchange dynamically between the
clusters nodes. The way of selecting the cluster head node
3. LEACH Protocol between the nodes of a cluster forms by the learning
One of the first and famous presented hierarchical protocols automata. In suggested algorithm there is considered a
to the sensor networks is LEACH protocol [3]. In this learning automata for each cluster the number of the actions
protocol, the time span of the nodes activity divides into of automata equals with the number of clusters nodes. Each
some periods. At the beginning of each period some of the action by automata corresponds with a node from the
nodes be selected haphazardly as a cluster head (CH). To do cluster. In figure 3 a cluster and its corresponded automata
this, each node produces a haphazard number between 0, 1. are showed.
Where as this number from the amount of T (n) that results
in by the use of formula (1), be less, the stated node be N2
presented as cluster head. In formula 1, p is the correlation N1
N3 N4
of the number of clusters to the total nodes of the networks, N3
r is the number of period and G the number nodes that in a1 a2
the last period of 1/p are not selected as cluster head. N5 N4
a5 N8
LA

 P a3 a4
, if n ∈ G
 N5 N6 N7
1 N7
T (n) =  1 − P × ( r mod ) N8

 p N9
0, otherwise N10 N11
Relation 1. Random number generation between zero and one
Example of Sensor network
and cluster
After determining the nodes of cluster head, the other nodes Equivalence Learning
on the basis of the power of the received signal from each
cluster, decide to be accepted as a member of which cluster. automata for cluster (1)
The cluster head node divides its responsibility glen in to
some time slots (Figure 2). This time slots on the basis of
TDMA mechanism are cooperated between the members of Figure 3. The assigning the cluster nodes to correspond
cluster. In each time slot, the cluster head connect with on learning automata
of the members of the cluster and receives the information The Way of choosing the clustering
packs of that member. The cluster head in every some slots
After clustering in LEACH protocol, turns to the selection of
sends its received information from its members to the Sink.
cluster head to each clusters. As it mentioned, for each of
In order to the distribution of the load on different nodes
the clusters there is considered a learning automata that the
after finishing of a period, to start a new period, the cluster
number of its actions equals with the number the nodes of
head by the declared mechanism above are exchange.
the some cluster. The learning automata between its actions
choose an action that its number is more than the other.
After choosing a node as a cluster head, in order to transfer
the information of its clusters nodes by the TDMA
mechanism connects with them and again the learning
automata choose another node as the cluster. (It chooses the
node as a cluster that the corresponding action that has more
amount than the other actions). The amounts of automata in
each time are exchange by the formula 2.
eni
Figure 2. Period and Time Slice in LEACH protocol Pni =
4. Improved Protocol (LALEACH)
d ni
Relation 2. action value Relation
As mentioned in the previous section LEACH algorithm do
the act of clustering and is selected a cluster head to each
(IJCNS) International Journal of Computer and Network Security, 43
Vol. 2, No. 2, February 2010

In this formula: transferring their outputs to the Sink, turns to the


exchange of cluster head node.
n : The number of cluster nodes. The way of exchanging the clusters nodes of number (1)
ni : The i th node. as the cluster head is occurs like this:
First stage: in this stage by the information of the
e ni : The remain energy of ni node. number (1) cluster the exists in figure of table 6, the
d ni : number of actions of the cluster (1) automata is
The average of distance of ni node with the other
appeared.
nodes of the same cluster.
p ni : The amount of i th action of automata. Node Remain Average Distance
In figure 4, period and the way of time slot in suggested No Energy AUTOMATA
protocol is showed. By comparing the figure 2 and 4 the way ACTIONS VALUES
of exchanging the node of cluster head for a cluster is n1 en1 = 3 J 2+6+8 3
d1 = = 5.34 p n1 = = 0.55
showed (clustering in each period is done and the exchange 3 5.43
of the cluster head node is done inside the clusters). en2 = 5 J 2 + 4 + 12 5
n2 d2 = =6 p n2 = = 0.83
3 6
n3 en3 = 1J 8 + 12 + 3 1
d3 = = 7.67 pn3 = = 0.13
3 7.67
n4 en4 = 2 J 6+4+3 p n4 =
2
= 0.46
d4 = = 4.3 4.3
3
Figure 6. The figure (5) network cluster number (1) nodes
information.

After determining the amount of the automata’s actions


related to the cluster (1) as table of figure 6, now it is
selected a node as a cluster head that its amount of action is
more than the others, so it is selected as the cluster head
node. After selecting the node of number (2) as a cluster
Figure 4. Period and Time slice in LELEACH protocol. head, it declares its being as cluster head to the all nodes
exist in its cluster. The other cluster head contents of the
Figure 5 shows a sensor network. At first this network by table of figure 6m which are being up-to-date in selecting of
the LEACH algorithm is divided into three clusters. By the cluster head are being selected.
next period that a new algorithm will be occurs, the
After connecting the node ( n2 ) by all nodes of its cluster
suggested algorithm will exchange the cluster head
dynamically. and transferring its information to the cluster head node, it
Cluster 4 Cluster 5 makes up-to-date its remained energy in table (6). After
N20 N21 N22
N23
getting up-to-date the information of the table, the node,
CH which its amount of action is more than the others, is
CH
Cluster 1 N18 N19
selected as cluster head. This process continues until the
N2
N17
beginning of next stage.
2m
CH
Cluster 2 N16
N1
12m N6 5. The Result of Simulations
N15
4m In this section the result of suggested algorithms simulation
6m
N5 is comparing and assessing with the result of previous
Cluster 6
8m CH
algorithm. The results of suggested algorithm is compared
N4 N7 and assessed in respect of lifetime by the algorithm of
N14
3m LEACH, HEED, Extended HEED clustering. In the tested
N8 CH
N13
simulation of the sensor environment is considered
N12 N25
150×150 and the radio range of the sensor environment is
N3
CH considered 30 meters. The first energy of the nodes is
N11
N9 considered 2J. Trials to some of the different sensor nodes is
N10
N24 done by N={100, 150, 200, 250, 300, 350, 400, 450, 500}
Cluster 3 the shown results is the result of min of results gained from
10 performing different simulations. The results from
Figure 5. an Example of clustered sensor network different algorithms are compared with each other respects
of lifetime that is one of the main standards of the quality of
At first after clustering by the use of LEACH algorithm, service in sensor networks. The results of assessing are
the node of cluster head is appear. After connecting the showed in Figure (7) it is clear that the lifetime of network
cluster head node with the all of its clusters node and in the suggested algorithm is more than the other ways.
44 (IJCNS) International Journal of Computer and Network Security,
Vol. 2, No. 2, February 2010

[6] M. Esnaashari, M. R. Meybodi1, “A novel clustering


algorithm for wireless sensor networks using Irregular
Cellular Learning Automata”, IEEE, 27-28 Aug. 2008.
[7] Meybodi, M. R. and Beigy, H., New Class of Learning
Automata Based Scheme for Adaptation of
Backpropagation Algorithm Parameters, Proc. Of
EUFIT-98, Sep. 7-10, Achen, Germany, pp. 339-344,
1998.
[8] Oommen, B. J. and Ma, D. C. Y., Deterministic
Learning Automata Solution to the Keyboard
Optimization Problem, IEEE Trans. On Computers, Vol.
37, No. 1, pp. 2-3, 1988.
[9] Beigy, H. and Meybodi, M. R., Optimization of Topology
of neural Networks Using Learning Automata, Proc. Of
Figure 7. Comparison of suggested algorithm with others. 3th Annual Int. Computer Society of Iran Computer
Conf. CSICC-98, Tehran, Iran, pp. 417-428, 1999.
6. Conclusion [10] Beigy, H. and Meybodi, M. R."Optimization of
In this paper a new algorithm based on LEACH algorithm is Topology of Neural Networks Using Learning Automata,
suggested. The goal of suggested algorithm is the increasing Proc. Of 3th Annual Int. Computer Society of Iran
the lifetime of the sensor network. In LEACH algorithm the Computer Conf. CSICC-98, Tehran, Iran, pp. 417-428,
nodes of cluster head are selecting with out considering their 1999.
remained energy and because the nodes of cluster head [11] Hashim, A.A., Amir, S.and Mars, p. Application of
consume more energy than the other node, so LEACH Learning Automata to Data Compression, In Adaptive
algorithm in this respect that chooses haphazardly the nodes and Learning Systems, K. S. Narendra (Ed), New York:
of cluster head causes decreasing the lifetime of the Plenum Press, pp. 229-234, 1986.
network. In suggested algorithm instead of selecting
haphazardly the node of cluster head, it is selecting between
the nodes their remained energy is more than the other
nodes and also near to its neighbor nodes.
So, by this way nodes with high energy are selecting as
cluster head and increase the lifetime of network. The
effectiveness of suggested algorithm compared with cleared
clustering LEACH, HEED, Extended HEED algorithms has
better results.

Reference

[1] D. Chen and P. K. Varshney, "QoS support in wireless


sensor networks: a survey", in Proc. of International
Conference on Wireless Networks (ICWN '04), pp. 227-
233, Las Vegas, Nev., USA, June 2004.
[2] H.MotieGhader, S..Parsa, M.Hossein Nejad,
“ Application of Learning Automata for DAG Scheduling
on Homogeneous Networks", 17th Iranian Conference on
Electrical Engineering, Iran University of Science and
Technology, ICEE2009.
[3] S. M. Abolhasani, M. Meybodi, “Usage of Learning
Automata for Routing, Fault Tolerance and Topology
Control in Wireless Sensor Network”, MSC Thesis,
March 2008.
[4] W. R. Heinzelman, A. Chandrakasan, and H.
Balakrishnan, “An Application-Specific Protocol
Architecture for Wireless MicrosensorNetworks,” IEEE
Transactions on Wireless Communications, vol.1 ,no.4 ,
pp.660 –670 , October2002 .
[5] O. Younis and S. Fahmy, "Distributed Clustering in Ad-
hoc Sensor Networks: A Hybrid, Energy-Efficient
Approach", In Proc. of IEEE INFOCOM, March 2004 .

Вам также может понравиться