Choosing Cluster Head in Artificially Intelligent Wireless Sensor Network


Dr. A. Narmada, Dr. P. Sudhakara Rao

ECE, Vignan’s Institute of Management and Technology for Women, Kondapur, Telangana




Reduced energy consumption and extended lifetime are basic requirements of Wireless Sensor network with distributed nature and dynamic topological changes. The motes are arranged in clusters and only one mote is chosen as cluster head to synchronize and data routing. The proposed work introduces an innovative approach of choosing cluster head in artificially intelligent wireless sensor network. In the proposed work the residual energy consumption plays the major role in choosing the cluster head and the Radial basis function based network model is used. The performance of the proposed algorithm is evaluated based on several factors such as dead nodes, energy consumption, cluster head formation, number of packets transferred to base station and cluster head. The performance of the proposed algorithm is compared with existing protocols such as LEACH and LEACH-C.


KEYWORDS:LEACH, LEACH-C, Artificial Neural Networks and Wireless Sensor Networks.



Wireless sensor network is the interconnection of sensor nodes or motes with limited processing abilities. The motes are generally battery operated in remoteterrains of harsh radio environments distributed over wide geographical regions. The mobility of nodes may be fixed or dynamic and are randomly operated tomonitor and communicate about environmental parameters like temperature,humidity, light etc where in these wirelesssensor networks are deployed. See Figure 1.




Figure 1: Architecture of Wireless Sensor Network                  Figure 2: Basic components of sensor node


The WSN architecture is interconnection of sensors, internet and remote controller as shown in figure 1. Sensors, transceiver and processor are main blocks of each sensor node. Environmental parameters like temperature, humidity, light etc are sensed by these sensors. The sensed data is processed by the processor with limited processing abilities and is communicated to central controller with the help of transceiver which is used to transmit and receive.


These nodes are powered by non-rechargeable batteries for several years from the deployment phase. The proposed work aims at designing a system that makes the wireless sensor network to consume least and optimum energy besides extending the network lifetime. The basic components of sensor mote are as shown in figure 2.


Routing the data is very important in WSN as it follows data centric routing rather than conventional address centric routing. The protocols of WSN are categorized into four types viz., based on structure of the network, based on communication model, based on topology and based on reliable routing scheme.


Network based routing protocols are further classified into two types viz., flat routing protocols and hierarchical routing protocols. All the nodes play similar role in flat routing type of protocols and nodes are organized into clusters. Each cluster is associated with one cluster head that is responsible for routing in hierarchical routing protocols.


The communication model based routing is further divided into three types viz., query based protocols, coherent and non-coherent based protocols, negotiation based protocols. Query for data is broadcasted in query based routing and reply is sent by the node whichhas got the requested data. After preliminary processing, the data is passed to aggregators in coherent routing where as raw data is processed by each node in non-coherent routing. Data is routed based on negotiations in negotiation based routing to avoid loss of data.


Location based routing and mobile agent based routing are two variants of topology based routing. Data is routed to appropriate location rather than the node in location based routing. Data is directly forwarded to the intended destination with help of mobile agent in mobile agent based routing.


QOS based routing and multipath routing are two variants of reliable routing protocol. Data quality and minimum energy consumption in QOS based routing. Load balancing is projected in case of multipath routing.


The proposed work concentrates on hierarchical based routing where LEACH(Low Energy Adaptive Clustering Hierarchy) is chosen to experiment new algorithms as it is very familiar and efficient among the existing protocols. LEACH randomizes cluster head selection in order to share equal energy to all the nodes of the network based on probability. Each iteration of LEACH operation is functional with two phases viz., setup phase and steady state phase.


The clusters are created in setup phase and decision of becoming cluster head is done by the nodes based on the suggested probability in each iteration. In order to become cluster head a random number is chosen from 0 to 1 and compared with the threshold value as per equation 1.




= 0 otherwise


Where P is the probability of cluster head, i represents an existing iteration, S represents a set of nodes that have become cluster heads in last 1/p iterations


If random number in less than threshold value then the node can become cluster head. Thus each node will become cluster head in 1/p rounds. Thus after 1/p rounds all the nodes are again suitable of becoming a cluster head. The nodes that are chosen as cluster head will transmit a message to other nodes to advertise. The non cluster head nodes now decide to which cluster the node must belong.The nodes must inform the cluster head that it is the member of cluster.

Steady – State Phase - In this phase all the non-cluster head nodes transmit the data in the allocated time slot to their corresponding cluster and cluster head performs some functioning over the data and passes it to base station.


A node can become cluster if the random number is less than threshold value. Within 1/p iterations each node can become clusterhead. Initially advertisements are broadcasted by the cluster heads to all the nodes. A node can become a member of a cluster by informing to its cluster head.


In steady-state phase, data is sent by all other nodes of cluster to cluster head for further processing and routing to base station.


Artificial neurons that resemble biological neurons of human brain are essential part of artificial neural network. The functions of human brain are imitated by artificial neurons to create intelligent behavior. The process of learning is implemented in artificial neurons to store the information and re-use it. The measure of success oftarget is defined as per the learning methodology of the system. The data set is divided into two sections viz., training set and testing set to measure the quantum of success. The training set represents the output and accuracy of the system performance is measured by the test set. Two methods of learning viz., learning with teacher and learning without teacher are identified and further classification is shown in figure 3.



Figure 3: Classification of learning


There are some special kinds of functions used for attaining efficiency in cluster head selection and one such function is radial basis function which is used in the proposed work whose output increases or decreases monotonically from the centre. Two parameters are chosen for the model viz., Distance scale and centre shape.


In linear RBF these parameters are fixed where as these parameters are variable in case of non linear RBF.


RBF exhibits Gaussian properties as per the equation 2




RBF is as shown in figure 4.


Figure 4: Radial Basis Function


Remainder of the paper is organized into four sections viz., section 2- reviews the related work, section 3- discusses about the proposed algorithms of cluster head selection, section 4- publishes experimental results and section 5- concludes about the work



The existing methodologies of how ANNs are implemented in WSNs are studied. The efficiency of WSN depends on efficiency of the protocols used particularly energy based. Network structure, communication model, topology and reliable routing are main characteristics on which the efficiency of the energy based routing protocols depend [1]. The factors effecting the efficiency of WSNs are discussed [2]. ANNs are introduced briefly [10]. ANN structure and model is elaborately illustrated [11] and Mc Culloch Pitts model is discussed. Different types of learning viz., learning with teacher, learning without teacher, supervised learning and unsupervised learning are discussed [12]. RBF based network is introduced and performance of RBF is compared with other kinds of networks.


It is very important to find exact node location for attaining efficiency in delivery of data [14]. Time difference of arrival (TDOA) is used with ANN to find exact location of nodes preferably the distance between base station and sensor nodes. The ANN is trained and tested with this information. The WSN is implemented with ANN with main focus in coverage and connectivity [15]. A two layer feed forward network is created with competitive learning method. The ANN is created in such a way that it takes inputs from sensors of the cluster and creates the outputs as cluster heads which consumes minimum energy.


The majority part of energy is spent in communication and hence data is optimized in order to reduce energy consumption during communication [16]. The energy consumption can be optimized based on various localization techniques [17]. Various methods of localization techniques and base station positioning are explored. The anchor free localization technique performed best among all the other techniques. SIR Protocol is developed [18], where intermediate node is selected based on the running AI algorithm to avoid node failures in datarouting. Neural network is implemented in WSN so as to find efficient path and to minimize sensor power consumption [19].



All the nodes in WSN are supplied with initial energy Eo before the start of the data transmission. Minimizing the energy consumption and maximizing network life time are main challenges in WSN.LEACH algorithm is based on random numbers and probability. Any node can become cluster head without sufficient energy for transmission and reception which is a serious drawback of the algorithm. The efficiency of the algorithm could be improved if a cluster head is selected based on the sufficient energy level rather than on the random number generation and probability.


LEACH-C algorithm is proposed in this paper so as to maximize the network life time and minimize energy consumption by choosing the cluster head with maximum energy level to avoid node failures.


A. Proposed Cluster head selection algorithm based on ANN

The proposed work is explained accordingthe following steps of algorithm and flowchart as shown in figure 5.


a)       A random deployment of the sensor nodes in an area where, X – axis = 200m and Y – axis = 200m.

b)       The algorithm is broken into rounds and each round has two phases, setup phase and steady – state phase. For few initial rounds LEACH – C protocol is applied, as all the nodes are provided with the same initial energy E0.

c)       In the setup phase, based on a random number J, cluster are created and cluster heads are formed i.e. J represents the number of clusters to be formed for the existing round. The decision of becoming a cluster head depends on the amount of energy left in the node, i.e. the nodes having maximum energies will become cluster heads.

d)       We are using artificial neural networks (RBF Network) to find the maximum energy values amongst all the nodes as in equation (3) –

e)       0 otherwise

f)        In the steady - state phase non cluster head nodes transmit data to the corresponding cluster heads nodes, after some processing by the cluster heads the data is sent to the base station.

g)       When the maximum number of rounds reached the algorithm ends Where R represents the number of rounds that are carried out, n(E) represents energy of nodes and J represents Random integer that decides the number of clusters that is to be formed



The proposed cluster head selection approach is demonstrated in this section. LEACH-C routing protocol is enhanced and then it is used for data transmission, the proposed algorithm is based on single hop routing i.e. nodes can only transmit data to cluster heads or base-station [21]. The results of Enhanced LEACH-C with ANN, LEACH and LEACH-C Protocol are compared. The proposed scheme is simulated  inMatlab.  The  standard  values used in simulation are given in Table 1.



Figure 5: Flowchart of enhanced LEACH-C with ANN


Table 1: Simulation Parameters for Cluster Head Selection




200m X 200m

Number of Nodes


Initial Energy Per Node

0.5 J

Total Energy

50 J

Transmitting Energy, ETX


Receiving Energy, ERX


Data Aggregation Energy, EDA

5 nJ/b/message

Probability  of  Becoming  Cluster  Head  Per




Size of Data Packets

4000 bits

Threshold distance, d0


Transmit Amplifier Energy


Energy for Free Space Loss, EFS

0.0013 pJ/b/m4

Energy for Multi-path Loss, EMP

10 J/b/m2

A. Energy consumption for transmittingdata to Base Station:

a)         If the distance of the node to the base station is greater than d0 than the energy required to transmit and to receive the data is given by equation (4) and (5) -


Etrans(k,d) = Eelec * k +Emp * k * (d)p ------ (4)

Erecv(k,d) = Eelec * k(d)p              -------(5)


P is Path loss exponent taken to be 4, K is the Size of message being transmitted and received, Etrans is the amount of energy required to transmit the data packets and Erecv is the amount of energy required to receive the data packets


b)     If the distance of the node to the base station is greater than d0 then the energy required to transmit and to receive the data is given by equation (6) and (7) -


Etrans(k,d) = Eelec * k +Emp * k * (d)p ------ (6)


Erecv(k,d) = Eelec * k(d)p-------(7)



P is Path loss exponent taken to be 2, K is the Size of message being transmitted and received, Etrans is The amount of energy required to transmit the data packets and Erecv is The amount of energy required to receive the data packets[20,22]



Figure 6 : Energy model used for Cluster head selection


B.  ClusterHeadSelectionusingRBFNetworks:

The result of Enhanced LEACH-C is compared with LEACH and LEACH-C based on various factors such as number of nodes dead, energy consumption of the network, cluster head formation, and nodes dying etc. The key concern for every routing protocol is to decrease the energy consumption and to increase the network lifetime. The LEACH-C protocol is enhanced by applying maximum energy concept, in which few nodes havingmaximum energy amongst all the other nodes are selected as cluster head. Using the concept network lifetime of the sensor network is improved by around 200 rounds.


The plot shown in Figure 7 compares the total amount of energy consumed by Enhanced LEACH with ANN, LEACH and LEACH-C over a session of 1000 rounds. From graph the amount of energy consumed by Enhanced LEACH with ANN is less than energy consumed by LEACH and LEACH-C protocol thus prolonging the networks lifetime



Figure 7: Energy Consumption for Enhanced LEACH-C with ANN, LEACH-C, and LEACH

Table 2: Energy Consumption for Enhanced LEACH-C with ANN, LEACH-C, and LEACH

Enhanced LEACH-C with










































50 .1






As it is illustrated from Figure 8, the number of nodes dying to number of rounds which shows that Enhanced LEACH-C with ANN is improving the lifetime of the network than LEACH-C, and LEACH in terms of First Node Dies, Half Network Dies, and Full network diesSee Figure 8.


Also, Table 3 shows the number of nodes dying for Enhanced LEACH-C with ANN, LEACH-C, and LEACH. It is show that Enhanced LEACH-C with ANN haveincreased lifetime over the other two protocols. See Table 3.


The total Number of packets that are transmitted to the base-station and to the cluster heads by using LEACH with ANN, LEACH and LEACH-C protocol are given in Table 4.



Figure 8: Number of nodes dying with respect to number of rounds for Enhanced LEACH-C with ANN, LEACH-C, and LEACH.


Table 3: Number of nodes dying with respect to number of rounds for Enhanced LEACH-C with ANN, LEACH-C, and LEACH









First Node dies




Half Network dies




Full Network dies





Table 4: No: of Data Packets being transmitted to Base-Station and to the Cluster Heads in CH Selection




Enhanced LEACH-C



with ANN











The purpose of the study is to select the cluster heads that aggregates the data and pass it to base station, by the use of artificial neural networks in such a way that it improves the lifetime of the network. When working with artificial neural network, the type of learning becomes an important factor. The proposed method for Cluster head selection uses the concept of maximum energy left with the nodes to become cluster head i.e. a node is chosen as a cluster head when the amount of energy left with the node is maximum amongst all the other nodes. The type of artificial neural network chosen to select the cluster head amongst the node is radial basis network function. The proposed algorithm is analyzed on various factors including number of nodes dead to number of rounds, energy consumption of the protocols, cluster head formation,number of nodes dying with respect to number of rounds and the total number of packets sent to the base station and the cluster head. It has been observed that enhanced LEACH-C with Artificial Neural Network protocol provides better results in comparison to LEACH and LEACH-C protocols, i.e. the use of artificial neural network are improving the lifetime of the network to some extent.


Hierarchical routing protocols are focused in the proposed work, but ANN can be applied to other routing protocols like Location Based protocols, Query based protocols etc. Use of Self organizing Maps, Principal Component Analysis and Support Vector Machine can further enhance the lifetime of wireless sensor networks.



[1]       Nikolaos A. Pantazis, Stefanos A. Nikolidakis, Dimitrios D. Vergados, “Energy-Efficient Routing Protocols in Wireless Sensor Networks: A Survey,” IEEE CommunicationsSurveys & Tutorials, 2013, Vol. 15,Issue. 2, pp. 551-590.

[2]       Yick J., Mukherjee B., Ghosal D., “Wireless Sensor Network Survey,” Computer networks, 2008, Vol. 52,Issue. 12, pp. 2292 – 2330.

[3]       I.F. Akyildiz, W. Su, Y. Sankarasubramaniam, E. Cayirci, “Wireless sensor networks: a survey,” ELSEVIER ComputerNetworks (38), 2002, pp. 393–422.

[4]       Anastasi G., Conti M., Di Francesco M.,PassarellaA.,“Energy Conservation in Wireless Sensor Networks: A Survey,” Ad hocnetworks, 2009, Vol. 7, Issue. 3, pp.537 – 568.

[5]       Al – Karaki J. N., Kamal A. E., “Routing Techniques in Wireless Sensor Networks: A Survey,”Wireless communications, IEEE,2004, Vol. 11, Issue. 6, pp. 6 – 28.

[6]       W. Heinzelman, A. Chandrakasan, H. Balakrishnan, “Energy-Efficient Communication Protocol for Wireless Microsensor Networks,” InProc. 33rd Hawaii International Conference on System Sciences, HI,USA, 2000, Vol. 8, pp. 110.

[7]       Saraswat J., Rathi N., Bhattacharya P.P., “Techniques to Enhance the Lifetime of Wireless Sensor Network: A Survey,” Global Journal of Computer Science and Technology, 2012, Vol. 12, Issue. 14- E.

[8]       Sindhwani N., Vaid R., “VLEACH: An Energy Efficient Communication Protocol for WSN,” MechanicaConfab, 2013, Vol. 2, Issue. 2, pp. 79 – 84.

[9]       Nikolidakis S. A., Kandris D., Vergados D. D., Douligeris C., “Energy Efficient Routing in Wireless Sensor Networks through Balanced Clustering,” Algorithms, 2013, Vol. 6, Issue. 1, pp. 29 – 42.

[10]    Vidushi Sharma, Sachin Rai, Anurag Dev, “A Comprehensive Study of Artificial Neural Networks”, International Journal ofAdvanced Research in Computer Science and Software Engineering,2012, Vol. 2, Issue 10, pp. 278-284.

[11]    MaindSonali B., Wankar P., “Research Paper on Basic of Artificial Neural Networks,” International Journal on Recent and Innovation Trends in Computing and Communication, 2014, Vol. 2, Issue.1, pp. 96 – 100.

[12]    Haykin S, Network N., “Learning Process In: A Comprehensive Foundation Neural Network.”, second edition, pp. 63-66.

[13]    Haykin S, Network N. Radial Basis Function Network In: A Comprehensive Foundation Neural Network, second edition, pp. 256-280.

[14]    Sampat Kumar Satyamurti, Rakesh Joshi, “ANN Assisted Node Localization in WSN using TDOA,” International Journal of Innovative Research in Computer and Communication Engineering, 2014,Vol. 2, Issue 4, pp. 3871-3877.

[15]    Kumar N., Kumar M., Patel R. B., “Coverage and Connectivity Aware Neural Network Based Energy Efficient Routing in Wireless Sensor Networks,” Journal onApplication of Graph Theory in Wireless Ad-hoc Networks and Sensor Networks, 2010, Vol. 2, Issue1,pp. 45-60.

[17]    Akojwar S. G., Patrikar R. M., “Improving Life Time of Wireless Sensor Networks Using Neural Network Based Classification Techniques with Cooperative Routing,” International Journal ofCommunications, 2008, Vol.2, Issue1, pp. 75-86.

[18]    Tripathi R. K., “Base-Station Positioning, Node-Localization and Clustering Algorithm for Wireless Sensor Network”, 2012.

[19]    Barbancho. J., Leon C., Molina J., Barbancho A., “Using Artificial Intelligence in Wireless Sensor Routing Protocols,” In Knowledge –Based Intelligent Information and Engineering Systems, Springer Berlin Heidelberg,2006, pp. 475 – 482.

[19]    Hosseingholizadeh. A., Abari A., “A Neural Network Approach for Wireless Sensor Network Power Management,” In Proc. 28th IEEEInter. Symp on Reliable Distributed Systems, Niagara Fall, NY, USA,2009.

[20]    Enami N., Moghadam R. A., Dadashtabar K., Hoseini M., “Neural Network Based Energy Efficiency in Wireless Sensor Networks: A Survey,” International Journal ofComputer Science and Engineering Survey, 2010, Vol. 1, Issue. 1, pp. 39 – 53.

[21]    M. J. Handy, M. Haase, D. Timmermann, “Low Energy Adaptive Clustering Hierarchy with Deterministic Cluster-Head Selection,” In Proc. 4th InternationalWorkshop on Mobile and Wireless Communications Network, USA,2002, Vol. 1, pp. 368-372

[22]    Muruganathan Siva D., Ma D. C., Bhasin R. I., Fapojuwo A., “A Centralized Energy Efficient Routing Protocols for Wireless Sensor Networks,” CommunicationMagazine, IEEE, 2005, 43.3, pp. S8– 13.



Received on 19.04.2018            Accepted on 28.06.2018     

©A&V Publications all right reserved

Research J. Engineering and Tech. 2018;9(4): 335-342.

DOI: 10.5958/2321-581X.2018.00046.6