[IEEE 2011 IEEE Colloquium on Humanities, Science and Engineering (CHUSER) - Penang, Malaysia...

Energy Efficient Clustering Algorithm in Wireless Sensor Networks using Fuzzy Logic Control

Z.W. Siew, A. Kiring, H.T. Yew, P. Neelakantan and K.T.K. Teo Modelling, Simulation and Computing Laboratory

School of Engineering and Information Technology Universiti Malaysia Sabah Kota Kinabalu, Malaysia

[email protected], [email protected]

Abstract — In general, environment monitoring cluster based hierarchical routing protocol is among the most common protocol being opted due to the load balancing among each other sensor. Sensors are randomly deployed in a specific area to collect useful information periodically for a few months or even a few years. Therefore, battery power limitation becomes a challenging issue. It is also impractical to maintain the network lifetime by changing the battery frequently. Low energy adaptive cluster hierarchical (LEACH) is one of the common clustering protocols that will elect the cluster head based on the probability model which will possibly lead to a reduce in network lifetime due to election of cluster head with a least desired location in the network. For wireless sensor networks, the distribution of cluster head selection directly influences the network’s lifetime. This paper presents factors which will affect the network lifetime and apply fuzzy logic based cluster head selection conducted in base station. The base station considers two selection criteria from sensor nodes which are energy level and distance to the base station to select the suitable cluster head that will prolong the first node die (FND) time, data stream guaranteed for every round and also increase the throughput received by the base station before FND.

Keywords – wireless sensor network; clustering hierarchical; first node die; fuzzy logic

I. INTRODUCTION Wireless sensor networks contain hundreds or thousands of

sensor nodes equipped with sensing, computing and communication abilities. The sensor nodes can be equipped with different sensors such as temperature sensor, humidity sensor, gas sensor et cetera. When it sense elements of the surrounding environment, the processing units inside the sensor nodes will convert the analog signal to digital data by using the analog to digital converter module and further sends the data to the base station for processing. Wireless sensor nodes not only communicate directly to base station, it can also communicate among its peers. In the coming future it is possible to implement smart adaptive sensor nodes by loading artificial intelligence algorithm embedded into the sensor nodes.

It is challenging to design the wireless sensor nodes due to the purpose of cost saving and the size of the wireless sensor nodes has to be compact for convenience purpose. The trade-

offs for being compact are limited computation capability, limited power source and small memory storage. The nodes energy is the most important issue because the nodes are small in size and it may be deployed in hazardous areas, thus making battery replacement unpractical and impossible. Therefore, it is more practical to save energy and prolong the network lifetime by improving the routing algorithm. Cluster based hierarchical routing protocol is an energy efficient routing protocol. In the cluster routing, the sensor nodes will be divided into a few groups with one cluster head elected for each group. The cluster head collects data from member nodes in the same cluster and aggregates the collected data so that it can be transmitted to the base station. Implementing this protocol will significantly reduce the overall energy used and reduce the network congestion by only allowing the cluster head to communicate with the base station [1, 2].

LEACH protocol is one of the most common cluster routing protocols which aims to achieve the load balancing in sensor nodes so it can prolong the network lifetime. In cluster routing protocol, energy consumption is concentrated on cluster heads which collect and aggregate the sensed data from member nodes and forward the aggregated information to the base station [3, 4]. In LEACH protocol each sensor nodes elects itself as a cluster head based on the probability model. Each sensor node will become cluster head in every cycle to evenly distribute the work loads. The main problem of LEACH protocol is it only depends on the probability model to elect the cluster head and therefore it is possible that no cluster heads or too many are selected in a single round [5]. Furthermore, the selected cluster heads may be located near to each other or even near to the edge of the networks which leads to inefficient energy distribution [6].

LEACH-C uses a centralized algorithm to elect the cluster heads where each node sends information about its current location and energy level to the base station and elects the cluster head using the simulated annealing algorithm [7]. Gupta proposed a cluster head election method using fuzzy logic based centralized algorithm [8]. By considering three parameters which are energy, concentration and centrality of sensor nodes, the base station will elect the suitable cluster head among all the nodes which show the improvement of

2011 IEEE Colloquium on Humanities, Science and Engineering Research (CHUSER 2011), Dec 5-6 2011, Penang

978-1-4673-0020-9/11/$26.00 ©2011 IEEE 392

network lifetime compared to LEACH. These cluster head election mechanism needs the location information of the sensor node (possible determined using a GPS receiver). To use the algorithm, the sensor nodes need to know the location information. At the first round, location information will be transmitted to the base station and the algorithm assumes all sensor nodes under minimal mobility constraint. GPS receivers consume large amounts of power since GPS is one of the communication radios. At current stage, it is too expensive to include GPS in the sensor nodes for wireless sensor networks.

This project aims to maximize the FND period, data collection is guaranteed for each round and the proposed method does not require the location information of the nodes. A cluster head selection mechanism in base station using fuzzy logic is proposed. Fuzzy logic system can manipulate the linguistic rules into mathematical form; it can make real time decision due to the algorithm’s simplicity. Moreover extra parameters such as battery charging rate, data queue length are among the parameters that can be added into fuzzy logic system to produce an optimized solution [9, 10].

The paper is organized as follows. In the next section, overview of cluster based hierarchical routing protocol LEACH and fuzzy logic cluster based are demonstrated. Section III explains the system model followed by the analysis of the simulation results in section IV. Lastly, section V concludes the findings.

II. RELATED WORK A. Cluster Based Hierarchical Routing Protocol

In cluster routing, nodes with more energy will become the cluster head in the networks. By forming few clusters, the surrounding node only need to communicate to cluster head in the same cluster to reduce the energy consumption used to communicate among other nodes at different clusters.

LEACH [2] is one of the clustering mechanisms aim to distribute the energy concentration in the cluster head. This protocol operates in rounds; each round consists of two phases which are set-up phase and steady phase. Set-up phase include advertisement phase and cluster set-up phase while steady phase include schedule creation and data transmission. In each round, each node independently generates the random number between 0 and 1. If the generated number is less than the threshold value )(nT which defines by (1), the node wills self elect to become the cluster head for the current round.

⎪⎪⎩

⎪⎪⎨

⎧ ∈×−=

otherwise

Gnif

prp

p

nT

, 0

, )1mod(1)( (1)

r is the round which already ended;

p is the proportion of the nodes to cluster heads;

G is a set of nodes which have never been cluster head in the last p1 rounds.

Although LEACH protocol distributes the energy concentration in the cluster head but it has some disadvantages.

• LEACH depends on only the probability model to elect the cluster head, elected cluster head may be very close to each other.

• In each round the number of the cluster head is dynamic and it cannot guarantee the data stream received at the base station.

• Cluster head will appear in the edge of the network or in the place where the node density is very low.

It can be observed that the disadvantages of LEACH protocol are mainly due to only using the local information [3, 4]. By using centralized algorithm in the base station with global information, it will possibly select the suitable cluster head in the network.

B. Cluster-Head Election using Fuzzy-Logic In the centralized algorithm, base station has the global

view of the overall networks. The base station is more powerful than the sensor nodes in term of computation power, sufficient memory, unlimited power and storage. Gupta proved that, by considering three fuzzy parameters which are energy, concentration and centrality the network lifetime can be improved. Energy level is available in each node, concentration is the number of neighbor nodes and centrality is a value based on how central the node to the cluster.

In this project, expert knowledge is represented based on the following two descriptors:

• Node Energy – remaining energy level in the node.

• Distance – the distance between the node and the base station.

The remaining battery level can be obtained from each node via analog to digital converter (ADC) module which is built in the microcontroller. The distance information can be acquired when the base station broadcasts the elected cluster head information to the whole network and each node calculates the distance to the base station based on the received signal strength indicator (RSSI).

III. SYSTEM MODEL Fig. 1 shows the basic linear topology used to study the

effect of various parameters on the networks lifetime. BS represents the base station whereas node A , B and C are the sensor nodes. The relative distance between node A and BS is set as 7R whereas node B and C are located 6R and 5R from the BS. The inputs of fuzzy inference system (FIS) are energy and distance of a particular node to the base station. Based on the rule base, FIS will compute and decide the output which is the fitness of a node to become a cluster head. Base station will select ×p total number of nodes to become cluster heads based on their fitness. Sensor nodes will be listed in a table from highest to lowest fitness. The table restores every single round to keep the information up to date. In short, the selected cluster head should be the best leader in every round.


393

Figure 1. Basic linear model.

Below are some assumptions used in the system modelling.

• All the sensor nodes can transmit data with enough power to reach the base station and the base station have a wide transmission range which is capable of using a single broadcast to reach all the sensor nodes.

• The distance between sensor nodes and base station can be measured based on the RSSI.

• All the sensor nodes are homogenous, with the similar nodes model and same energy resources.

A. Radio Model Fig. 2 shows the block diagram of the radio model used in

this project. The model consists of the transmitter model and receiver model for energy calculation purpose. The radio model referred from [3]. The expended energy during transmission and reception for k bits message to a distance d between transmitter and receiver node is given by (2).

),()(),( dkEkEdkE ampTxelecTxTx −− += (2)

⎪⎩

⎪⎨⎧

+

+=

4

2

dkkE

dkkE

mpelec

fselec

ε

ε

, ;

0

0

ddifddif

≥<

Distance 0d is the threshold for swapping amplification model,

it can be calculated as mpfsd εε=0 . To receive k -bit message, the radio model expended as in (3).

kEkE elecRx =)( (3)

Figure 2. Radio energy model.

B. Simulation Parameters The simulation is carried out in MATLAB with the radio

model parameters shown in Table I. The network topology and base station are varied from each case which will be discussed later.

C. Cluter Head Selection Based on Distance Distance is one of the factors that may affect the energy

consumption as shown in (2). Distance d refers to the relative distance among the cluster head, member nodes and the base station. Equation (4) illustrates node A being elected as cluster head, using the modified linear topology as shown in Fig. 1. Distance 1d , 2d and 3d in (4) represent the total transmit distance by each sensor nodes.

))(()()(5

(4) ))(())(

())(()(2

23

22

21

23

22

21

kdndndkEkEnE

dkkEdk

kEndkkEnkEnE

ampelecelecTotal

ampelecamp

elecampelecelecTotal

ε

εε

ε

++++=

++

++++=

In the case of n equal to 1 and n represents numbers of

sensor nodes, where node C becomes the cluster head, the overall energy consumed is less than if node B and node A being selected as cluster head.. This is due to the fact that

node C is located nearer to the base station, and also the distance to base station is far larger by other member nodes. Therefore the distance to the base station is significantly affecting the energy consumption. The previous statements become invalid when n not equal to 1. It can be proved by the following cases where TotalE can be calculated based on (4).

Case 1: A become cluster head, with n numbers of sensor nodes in B and in C .

))(495())(15( 22 kRnRkEnE ampelecTotal ε+++=

Case 2: B become cluster head, with n numbers of sensor nodes in A and in C .


Case 3: C become cluster head, with n numbers of sensor nodes in A and in B .


TABLE I. PARAMETERS OF THE RADIO MODEL Parameter Value

Initial Energy, 0E 0.05J

Packet Size, k 4000bit

Energy of transceiver, elecE 50nJ/bit

Energy of data aggregate, DAE 5nJ/bit/message

fsε 10pJ/bit/m2

mpε 0.0013pJ/bit/m4

5R R R

A B C BS

packetbit k

packetbit k

sElectronicTransmit

Amplifier Tx

sElectronicReceive

)(dETx

kEelec2kdEamp

kEelec

RxE

d


394

Referring to Case 2 if =n 5, node B become cluster head, and the overall energy consumption is less than Case 3 and follow by Case 1. It can be observed that node B is more suitable to become cluster head compared to node C , since the n numbers of nodes from A and B spend more energy to transmit to node C . As a result, selecting cluster head in the middle of linear topologies will produce less energy consumption per rounds. According to equation above, cluster head B will shows less energy consumption per rounds compared to cluster head A and C , but the tradeoff of it is FND occurred faster than node C .

D. Balance in Energy Distribution Although proper cluster head election based on distance can

achieve minimum energy used per round, but it will drain energy faster which leads to FND faster because it overloads the cluster head. FND can be prolonged by distributing the workload to maintain the level of remaining energy in each node.

This paper examines three protocols, namely direct communication, random cluster head selection protocol and energy constraint cluster head selection protocol using the same radio model and same network topology (base station located at x = 100m and y = 160m) as shown in Fig. 3. Fig. 4 shows the simulation results of the three protocols with 25 randomly deployed sensor nodes for duration of 150 rounds.

Fig. 4 illustrates the network lifetime of the three protocols, it is observed that FND time for energy constraints cluster head selection protocol is longer than random cluster head selection protocol and followed by direct communication protocol. In direct communication protocol, each sensor sends its data directly to the base station. It will quickly drain the battery when the sensor nodes are far away from the base station because it requires a large amount of transmitting power. On the other hand, since only one randomly selected cluster head sends data (gathered from sensor nodes) to the base station for random cluster head selection protocol, therefore it outperforms the direct communication protocol.

0 20 40 60 80 100 120 140 160 180 2000

20

40

60

80

100

120

140

160

180

200

X-coordinate (200m)

Y-c

oord

inat

e (2

00m

)

Figure 3. Modified topology model.

0 50 100 1500

5

10

15

20

25

Rounds

Num

ber

of A

live

Nod

es

energy

randomdirect

Figure 4. Network lifetime 1.

In random cluster head selection protocol, the base station will randomly select one cluster head among the nodes. The random protocol will not consider the remaining energy in sensor nodes, therefore the energy constraint cluster head selection protocol which considers energy level will perform better than random protocols in terms of FND time due to the remains of energy for each sensor. However, energy constraint protocols will not guarantee minimum energy consume per round.

E. Fuzzy Logic There are pros and cons in using either energy or distance as the factor to elect the cluster head. Using fuzzy logic to combine and consider both factors will make the cluster head selection process simpler. The optimum results from both factors can be obtained based on the fuzzy rule base system. Fuzzy inference technique via Mamdani method is used due to its simple structure. There are four important steps in the Mamdani method:

• Fuzzification of the input variables, transforms the system inputs which are crisp values into fuzzy sets.

• Rule evaluation, taking the fuzzified inputs and evaluate them to the antecedents of the fuzzy rules.

• Aggregate conclusions, the process of unification of the outputs of all rules.

• Defuzzication, transforms the fuzzy set obtained by the inference engine into a single crisp value.

In the defuzzification module, the fuzzy set obtained from the inference engine needs to be transformed to the single crisp value. COA (centre of area) is used to calculate the crisp value from the aggregate output membership function using the (5) where )(xAμ is the membership function of the fuzzy sets:

))(.( ∫∫= xdxdxxuxfitness A (5)

Base station

Sensor nodes


395

The following membership functions are set for the simulation study. The linguistic variables used to represent the node’s energy and distance are divided into three levels; high, medium, and low for energy as in Fig. 5; then far, middle and near for distance as in Fig. 6. The outputs to represent the node cluster head selection is named as fitness which is divided into nine levels: very small, small, rather small, medium small, medium, medium large, rather large, large and very large as in Fig. 7. Table II shows the fuzzy rule base, consists of 932 = rules.

Figure 5. Membership functions of distance.

Figure 6. Membership functions of energy.

Figure 7. Membership functions of fitness.

TABLE II. FUZZY RULE BASE

No Energy Distance Fitness 1 low far very small 2 low near small 3 low middle rather small 4 medium far medium small 5 medium near medium 6 medium middle medium large 7 high far rather large 8 high near large 9 high middle very large

Table II shows the rule base set in the fuzzy inference system (FIS), the rule set as if the energy is ‘low’ and the distance is ‘far’ then the fitness to be cluster head is ‘very small’. On the other hand, if the energy is ‘high’ and the distance is ‘middle’, the fitness becomes ‘very large’. Referring to the distance rule base, ‘middle’ is set to higher priority than ‘near’ because the cluster head in the ‘middle’ zone will perform better than the ‘near’ zone which has been explained in section III.

In fuzzy based centralized algorithm, base station will calculate the fitness for every node to become cluster head via FIS where the base station will select the cluster heads according to ×p total number of nodes. The data transmission phase is similar to the LEACH steady state phase.

IV. RESULTS AND DISCUSSIONS The proposed method needs extra information to elect the

cluster heads. Therefore, an additional 24 bits of energy and distance information are added to the data transmitted from every member node to the cluster head. On the other hand, 480 bits is added to the transmission from the cluster head to the base station as extra energy and distance information for each cluster members and the cluster head. The simulation was carried out in MATLAB with the parameters show in Table III.

In Fig. 8, the simulation results show that the proposed fuzzy logic method performs better than the LEACH approach before FND in both 5% and 7% chosen as cluster head cases. However, the proposed method loses to LEACH after 60% of total alive nodes. In LEACH, cluster head is elected based on a random probability model causes the cluster head to exhaust the energy rapidly compared to the proposed method which considers critical factors in electing the cluster head.

TABLE III. SIMULATIONS PARAMETERS

Parameter Value Topology size 30×100m2

Base Station Position 15m, 150m Simulation Round 200

Number of Node, n 70

0 20 40 60 80 100 120 140 160 180 2000

10

20

30

40

50

60

70

Rounds

Num

ber

of A

live

Nod

es

proposed

LEACH

7%

5%



396

0 0.5 1 1.5 2 2.5

x 106

0

10

20

30

40

50

60

70

Total Data Receive by Base Station (bits)

Num

bers

of

Aliv

e N

odes

proposed

LEACH

7%

5%


This is because the proposed method uses energy level as one of the factor to be considered in electing the cluster head. All the sensor nodes distribute the energy consumption evenly according to the FIS. After FND, the remaining nodes die faster compared to the LEACH since the energy form the remaining nodes are almost equally less. Fig. 9 shows the data received by the base station for the LEACH and fuzzy approaches in the case of 5% and 7% of chosen cluster head. Received data by the base station can be further increased by increasing the number of cluster head from 5% to 7%. It can be observed that the data received by fuzzy approach are more than the LEACH in both cases.

Increasing the number of cluster head will not much affect the network lifetime as shown in Fig. 8. However, it can significantly improve the throughput to the base station. Extra information for the proposed method is not considered into the throughput for fair comparison. Proper number of cluster heads should be carefully decided to achieve optimum in data collection. Using too many cluster heads will lead to additional overhead and cause the sensor nodes sacrifice faster without providing significant information.

V. CONCLUSION As a conclusion, the proposed method achieved its

objective to prolong the first node die (FND) time and also improve the throughput to the base station before FND. It shows improvement compared to the LEACH protocol because the base station have partial global knowledge of the networks

and the use of fuzzy logic to elect the suitable cluster head based on two factors which are energy and distance.

The proposed method can be further improved by adopting some intelligent algorithm to modify the shape of each fuzzy set to achieve optimum result in specific cases. With the global knowledge of the network, multi-hop routing can be implemented to reduce the energy consumption in the cluster head.

ACKNOWLEDGMENT The authors would like to acknowledge the financial

assistance of the Ministry of Higher Education of Malaysia (MoHE) under Fundamental Research Grant Schemes (FRGS), no. FRG0104-TK-1/2007.

REFERENCES [1] J.N. Al-Karak, and A.E. Kamal, “Routing techniques in wireless sensor

network: A survey,” IEEE Wireless Communication, vol. 11, issue 6, pp. 7-34, 2004.

[2] I. F. Akyildiz ,W. Su, Y. Sankarasubramaniam, and E. Cayirci, “Wireless sensor networks: a survey,” Computer Networks, vol. 38, pp. 393-422, 2002.

[3] W. Heinzelman, A. Chandrakasan, and H. Balakrishnan, “Energy-efficient communication protocol for wireless microsensor networks,” Proceedings of the 33rd Annual Hawaii International Conference on System Sciences (HICSS), Maui, Hawaii, pp. 3005-3014, 2000.

[4] I. Mahgoub, and M. Ilyas, Sensor Network Protocols, CRC Press, Taylor & Francis Group, pp. 3–11 - 3–13, 2006.

[5] J. Zheng, and A. Jamalipour, Wireless Sensor Networks A Networking Perspective. Wiley, USA, pp. 87-89, 2009.

[6] J.M. Kim, S. Park, Y. Han, and T. Chung, “CHEF: Cluster head election mechanism using fuzzy logic in wireless sensor networks,” Proceedings of 10th Internatinal Conference on Advanced Commmunication Technology (ICACT), Phoneix Park, Korea, pp. 654-659, 2008.

[7] W. Heinzelman, A. Chandraksan, and H. Balakrishnan, “An application-specific protocol architecture for wireless sensor networks,” IEEE Transcation on Wireless Communications, vol. 1, no. 4, pp. 660-670, 2002.

[8] I. Gupta, D. Riordan, and S. Sampalli, ”Cluster-head election using fuzzy logic for wireless sensor networks,” Proceedings of Communication Networks and Services Research Conference (CNSR2), Halifax, Nova Scotia, Canada, pp. 255-260, 2005.

[9] C.S. Chin, P. Neelakantan, H.P. Yoong, and K.T.K. Teo, “Optimisation of fuzzy based maximum power point tracking in PV system for rapidly changing solar irradiance,” Global Journal of Technology and Optimisation (GJTO), vol. 2, no. 2, pp. 130-137, 2011.

[10] C.S. Chin, P. Neelakantan, H.P. Yoong, S.S. Yang, and K.T.K. Teo, “Maximum power point tracking for PV array under partially shaded conditions.” Proceedings of 3rd International Conference on Computational Intelligence, Communication Systems and Networks (CICSyN) ,Bali, Indonesia, pp. 72-77. 2011.


397

[IEEE 2011 IEEE Colloquium on Humanities, Science and Engineering (CHUSER) - Penang, Malaysia...

Documents

Transcript of [IEEE 2011 IEEE Colloquium on Humanities, Science and Engineering (CHUSER) - Penang, Malaysia...