Scholarly article on topic 'A Firefly-inspired Micro and Macro Clustering Approach for Wireless Sensor Networks'

A Firefly-inspired Micro and Macro Clustering Approach for Wireless Sensor Networks Academic research paper on "Computer and information sciences"

Share paper
Academic journal
Procedia Computer Science
OECD Field of science
{"Firefly approach" / "Bio-insppired algorithm" / "Wireless Sensor Network" / "Micro clustering" / "Macro clustering" / "Sensor attractiveness"}

Abstract of research paper on Computer and information sciences, author of scientific article — Nafaâ Jabeur

Abstract Bio-inspired algorithms have been widely used to solve Wireless Sensor Network (WSN) challenges. In several studies, they have demonstrated effective capabilities to fulfil the expected goals while adapting to contextual changes and using limited resources. In this paper, we propose a new firefly-based approach for WSN clustering. Our approach includes a micro clustering phase during which sensors self-organize into clusters. These clusters are polished during a macro-clustering phase where they compete to integrate small neighboring clusters. Our simulations show promising results where the number of clusters tend to stabilize independently from the density of the network and the various communication ranges of sensors.

Academic research paper on topic "A Firefly-inspired Micro and Macro Clustering Approach for Wireless Sensor Networks"



Available online at


Procedia Computer Science 98 (2016) 132 - 139

The 7th International Conference on Emerging Ubiquitous Systems and Pervasive Networks

(EUSPN 2016)

A Firefly-Inspired Micro and Macro Clustering Approach for

Wireless Sensor Networks

Nafaâ Jabeur^*

aGerman University of Technology in Oman (Gutech) - P.O Box 1816, PC 130, Muscat, Oman


Bio-inspired algorithms have been widely used to solve Wireless Sensor Network (WSN) challenges. In several studies, they have demonstrated effective capabilities to fulfil the expected goals while adapting to contextual changes and using limited resources. In this paper, we propose a new firefly-based approach for WSN clustering. Our approach includes a micro clustering phase during which sensors self-organize into clusters. These clusters are polished during a macro-clustering phase where they compete to integrate small neighboring clusters. Our simulations show promising results where the number of clusters tend to stabilize independently from the density of the network and the various communication ranges of sensors. © 2016 The Authors.PublishedbyElsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the Program Chairs

Keywords: Firefly approach; Bio-insppired algorithm; Wireless Sensor Network; Micro clustering; Macro clustering; Sensor attractiveness

1. Introduction

Wireless Sensor Networks (WSNs) are being broadly studied because of their ability to collect in-situ, real-time data about a wide range of objects and events of interest. Thanks to continuous research and development advances, the spatially distributed sensors are now able to adapt to new environmental changes and reason on their own actions. Nevertheless, these abilities are not yet performing well enough because of the limited power and processing capabilities of sensors as well as the commonly changing contextual conditions, particularly in some harsh environments. Due to these constraints, sensors are highly recommended to collaborate in order to achieve goals exceeding their own competencies. This collaboration is usually achieved within clusters, which are created based on

* Corresponding author. Tel.: +968 96275005; fax: +968 22061000. E-mail address:

1877-0509 © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license


Peer-review under responsibility of the Program Chairs


a variety of criteria, including the distance between cluster-heads (CHs) and base stations, the uniform distribution and size of CHs, the residual energy of sensors, and the number of CHs1. Clusters can also be created based on the spatial location of CHs, their connectivity degrees, and their semantics.

The research and development communities have recognized the multiple advantages of clustering in WSN. For instance, clustering allows for data aggregation, topology control, support of quality of service (QoS), and energy efficient routing by limiting participating nodes in route establishment2. It also allows for enhancing the network scalability and increasing the network lifetime in large-scale WSNs. To meet these goals, it is important to create the appropriate clusters. To address this issue, numerous approaches have essentially focused on selecting minutely the Cluster Heads (CHs) as they are commonly expected to bear higher processing and communication loads. Some of these approaches have not used any specific metric to elect CHs. They have, rather, used sensors' IDs (e.g., LCA3), connectivity, or distance from CHs (e.g., Max-Min D-Cluster4, EEUC5). Other works have proposed to perform several iterations for the selection of CHs based on weights assigned to sensors desiring to become CHs. These weights are generally derived from the residual energy of sensors (e.g., EECA6) and/or intra-cluster communication cost (e.g., HEED7).

For instance, weights reflect the obvious differences between sensors in terms of their capabilities and locations. They could, therefore, be effectively used to elect the right candidates for CHs, at the right locations, at the right time. In order to reach this goal, biologically inspired algorithms can be adopted, especially since the WSN could be effectively modeled based on observations of living systems8,9,10. In this regards, several works have been particularly inspired from Particle Swarm Optimization (PSO) and Artificial Bee Colony. Few other works have proposed firefly-inspired algorithms where sensors carry out intense competition to be CHs while trying to attract peers to their clusters. We argue in this paper that this approach is well suited for WSN clustering and propose a firefly-inspired approach to cluster the WSN based on sensors' residual energies, cluster sizes, and contextual spatial information. Our approach includes four phases, namely Initialization, Fetching, Intimidation, and Polishing where sensors self-organize into clusters and then neighboring clusters, especially small ones, self-organize to enhance the balance of clusters' sizes.

In the reminder of this paper, Section 2 will outline the current literature on bio-inspired WSN clustering. Section 3 will focus on firefly-based clustering. Section 4 will present our clustering solution called FiCA. Section 5 will extend FiCA to SFiCA where the spatial context is taken into consideration. Section 6 will outline the results of our simulations.

2. Related work on bio-inspired clustering

Self-organizing clustering algorithms for large-scale networks have been inspired long time ago from the collective behavior of small living organisms. Based on simple rules, purely local decisions, and limited amount of broadcasted messages, most of the nodes in a network can determine their roles as either cluster heads or cluster members11. In this context, several authors have proposed approaches inspired from Particle Swarm Optimization (PSO) for network clustering. For example, Tillett et al.12 have proposed a PSO algorithm where a 'swarm' of test solutions, comparable to a natural swarm of bees, is allowed to interact and cooperate to find the best solution to a given problem. For optimization reasons, the authors have proposed to match the number of nodes and CH candidates in each cluster with the objective of minimizing the energy consumed by the nodes while maximizing the total data gathered. Dong and Qi13 have used a PSO-based clustering algorithm with an enhanced search ability. The authors have proposed an algorithm that solves the clustering problem by using the fast search ability of the particle swarm optimization. Each particle is composed of a cluster center vector and represents a possible solution of the clustering problem. Latiff et al.14 have presented an energy-aware clustering for WSNs using PSO. The performance of the protocol was compared with conventional WSN clustering protocols such as LEACH (Low-Energy Adaptive Clustering Hierarchy). Guru et al.15 have described a number of extensions of the PSO algorithm by using clustering techniques to reduce the total communication distance of the WSN and hereby decreasing the energy cost. Charalambous and Cui16 have used bio-inspired intercellular communication to achieve a compact cluster via a lateral induction model in a purely distributed and energy-efficient manner. Initially, the sensor nodes collaborate to construct a functional cluster via lateral induction followed by a lateral inhibition phase. Once clusters are formed, a competition process is carried out between nodes to decide which ones should remain active and which ones should go to sleep.

Furthermore, Zhang et al.17 have proposed a distributed self-organizing Low-Complexity Clustering (B-LCC) algorithm for large-scale, dense WSNs inspired by the collective behavior of flocks and schools. The B-LCC algorithm does not require sensor locations, time synchronization or any prior knowledge of the network. Selvakennedy et al.18 have proposed an approach whereby the network is clustered around certain nodes estimated biologically fit. According to this approach, when a node possesses a special agent, it elects itself to become a CH. Such election prevents the need to maintain many state variables. A fixed number of such agents are used to ensure that a certain number of clusters are formed throughout the network useful life. This number is calculated in a way to minimize energy dissipation through data aggregation. Krishnaveni and Arumugam19 have presented a clustering algorithm based on Harmony Search (HS) and including Artificial Bee Colony (ABC) features. The HS is a stochastic meta-heuristic optimization algorithm inspired from musicians' improvisation process. The food source exploitation feature of ABC algorithm is applied to enhance the convergence rate of the HS method by considering the fitness values of cluster members. Charalambous and Cui16 have proposed a distributed clustering algorithm for WSNs based on the biological lateral induction model for further node activation control. Hasnat et al.20 have presented a Distributed Energy Efficient Clustering (B-DEEC) protocol based on an Artificial Bee Colony (ABC) algorithm. The authors have simulated the social interactions of bees to optimize the CH selection. This selection was optimized with a probabilistic model by performing neighborhood search and nominating a node as CH with maximum energy. Karaboga et al.21 have proposed a centralized energy efficient mechanism (working on the base station), based on fast searching features of ABC to cluster the WSN and ultimately prolong its life-time. Sarobin and Ganesan22 have presented a bio-inspired cluster-based deployment algorithm for energy optimization of the WSN and ultimately for improving the network lifetime. In the cluster initialization phase, a single cluster is formed with a single CH at the center of the sensing terrain. The second phase is for optimum cluster formation surrounding the inner cluster, based on swarming bees and a piping technique.

3. Firefly-based clustering

When sensors are competing for CHs, they are generally following a bully approach that gives priority to sensors with higher fitness value, such as residual energy (e.g., HEED7) or distance from sinks (e.g., EEUC5). Naturally, this fitness value should fade with distance as the distance highly affects communication costs and the complexity of controlling large clusters. It is, thus, important that the clustering approach reflects this fading as well as prevents the creation of clusters with large number of sensors. This could be modeled by a social behavior where individuals try to team up to reduce the number of competitors. In order to endow our algorithm with these abilities, we are proposing in this paper an approach inspired from the social life of firefly insects.

The Firefly Algorithm (FA) is a bio-inspired technique that has been used for solving nonlinear optimization problems. It is based on observations from the social insect colonies, where each individual (for instance firefly glowing through bioluminescence) appears to operate for its own benefit and yet the group as a whole performs to be highly organized. This characteristic actually fits well to the context of WSN, where sensor nodes collectively operate for the benefit of the network while potentially exhibiting competitive, selfish behaviors.

Senthilnath et al.23 have compared the performance of several bio-inspired algorithms, including FAs, ABC, and PSO and concluded that FAs can be efficiently used for data clustering. In the context of WSNs, few research works only have proposed FA-based approaches to cluster the network. In this regards, Sarma and Gopi24 have presented an energy-aware FA for WSNs which takes into account the maximum distance between the CH and the cluster members as well as the remaining energy of the CH candidates to determine the best k CHs. The algorithm is centralized as it is implemented on the base station. Sandeep et al.25 have applied a two-phases firefly-based approach to the basic LEACH protocol to enhance the energy efficiency. In the first phase (initialization), the base station broadcasts the percentage of CHs requirements for the entire network. Each node calculates then a random intensity number and declares its interest to be a CH if this number is less than a given threshold. In the second-phase (cluster formation), nodes continue to exchange their intensity and update their affiliations to CHs until the network is organized into clusters. It is not, however, clear when the clustering process stops.

4. FiCA: Firefly-based clustering approach

Based on the findings of Senthilnath et al.23 (see Section 3), we are proposing an approach inspired from fireflies behaviors, where we assume that the WSN (also the swarm) includes n sensors (also fireflies). Each sensor s has a set of solutions {xsi: i=0, ..., m} where each xsi is a neighboring CH candidate with a fitness value f(xsi). Every sensor has a probability to become a CH that matches the brightness of the firefly it represents. The attractiveness fis reflects the strength of the sensor s in attracting other sensors. It depends on its probability (brightness) as well as on its distance to each nearby sensors. Since we are assuming that sensors do not have location-aware facilities, we calculate this distance in terms of hops. In this case, the attractiveness is calculated as follows:

where /?so is the attractiveness of sensor s at r = 0 and y is the coefficient of brightness fading (corresponds to fireflies' light absorption coefficient).

For simplicity reasons, we use the following three idealized rules in describing our firefly algorithm: (i) all fireflies are unisex so that one firefly will be attracted to other fireflies regardless of their sex; (ii) an important and interesting behavior of fireflies is to glow brighter mainly to attract prey and to share food with others; (iii) attractiveness is proportional to their brightness, thus each agent firstly moves toward a neighbor that glows brighter. In contrast with any existing firefly-based solution, our proposed clustering algorithm includes four main steps (Figure 1-right): Initialization, Fetching, Intimidation, and Polishing. In the initialization step, all sensors have to compute their initial brightness. To this end, we define the initial brightness p0 of each sensors as follows:

Pso ~ Eresidual / Emax (2)

where Eresiduai is the current energy and Emax is the maximum energy (of the sensor battery). At any moment of the clustering process, the brightness fis of every sensor s must not fall below a given fimin. This reflects the fact that this brightness will fade continuously until it becomes unattractive.

During the Fetching step, each sensor will iteratively look for the appropriate Cluster Head Candidate (CHC) to which it will belong or declare itself a CHC if it fails to find a sensor with higher attractiveness. In each iteration, the sensor will select the CHC to which belongs its neighbor k with the highest attractiveness. For instance, this attractiveness is not necessary the actual attractiveness of k. Indeed, because not all sensors are mobile as well as because of increasing energy consumption incurred from mobility, we are adopting an approach where every sensor promotes its currently selected CHC. In every iteration of the Fetching phase, we consider the coefficient y of brightness fading (light absorption coefficient in Equation (1)) of a given sensor as the energy consumption calculated according to the following Equation6:

where ed is the energy dissipated, per bit per m2 and et is the energy spent by transmission circuitry per bit. Like in other works6, we chose the values 100 x 10-12 and 50 x 10-9 for ed and et respectively. Furthermore, b is the number of bits to transmit or receive, d is the distance from transmitter to receiver and a is a constant > 2 which depends on the attenuation the signal will suffer in that environment. In this paper, we will consider the common values of a = 2 and a = 4. The Fetching phase of every sensor will finish once the current attractiveness falls below the threshold pmin.

In order to balance the size of clusters, our first mechanism is to allow sensors who already decided on their CHs to intimidate neighboring peers from joining their clusters. Intimidation happens when sensors belonging to a given cluster promote a low attractiveness (zero or a value below a given threshold) to any neighboring sensors willing to join the cluster and located at a hop distance greater than a predefined value. For instance, intimidation reflects a natural behavior demonstrated by some animals that become aggressive and fight other animals to prevent them from joining their group.

The Fetching and Intimidation phases may result in clusters with a variety of sizes, either because of the competition between neighboring CHCs or because of low sensor density in some spatial areas. We, thus, carry out a Polishing phase which is actually a macro firefly algorithm that allows small clusters (i.e., which sizes do not exceeding a predefined threshold) to be aggregated with other neighboring clusters. In this phase, which is representing our second mechanism for balancing cluster sizes, leaf sensors of every cluster will set up their attractiveness based on the number of hops they are far from their CHs (Figure 1 -left). These sensors must not be among those who already showed earlier intimidating behaviors. The calculation of the attractiveness of any cluster A toward a neighboring cluster B is derived from the attractiveness of all leaf sensors of A toward neighboring leaf sensors of B and will be calculated as follows:

Ps = Ps0 e-^

Y = edbda + etb

Attr(A,B) =

where n refers to the number of pairs of neighboring sensors, one belonging to A and one to B, and xi refers to the number of hops that sensor i is away from the CH of A. In the example of Figure 1 -left, three clusters (namely A, B, and D) are competing to integrate the smallest cluster C. Our Equation (4) leads to the following results: Attr(A,C) = —x 2 = 2/3 « 0.67, Attr(B,C) = -^x 3 = 3/7 « 0.43, and Attr(D,C) = — x 2 = 2/6 « 0.34

* 1+2 \ ' / 2+2+3 \ ' / 3+3

The cluster C will thus be merged with cluster A. This result is reasonable since merging C with D will lead to sensors located relatively far from the CH of D. Merging cluster C with cluster B has the advantage of securing more connectivity. However, it will also lead to sensors a bit far from the CH of B compared to the solution of merging cluster C with cluster A. The implementation of our proposed micro and macro FA, that we call FiCA, will follow the activity diagram depicted in Figure 1 -right. This implementation is executed in a distributed manner by each sensor.

[initialization J

Micro Clustering

[CH not yet decided]


[hop distance from decided CH ^reached a predefined threshold]

Micro Intimidation

[micro clustering finished]

Macro Clustering

[polishing \ ^ [new cluster size after aggregation_

finished \Polis"hing)reached a predefined threshold(Macro intimidation) [macro clustering finished]

Fig. 1. (left) Macro Firefly Clustering, (right) FiCA (Firefly-based Clustering Approach) activity diagram

5. SFiCA: Spatial Firefly-based Clustering Approach

In several applications, including environment monitoring and safety, specific spatial locations are given high priority during related decision making process. This is the case, for example, of schools that should be given high priority during evacuation when a sudden disaster like earthquake happens. In such cases, the frequency of collecting data about events of interest around sensitive locations will become intensive. Consequently, the sizes of clusters should be reduced in order to balance processing load and prevent an accelerated depletion of CHs' energies. We thus propose to extend our FiCA algorithm with spatial capabilities that will allow sensors to self-organize into clusters based on the semantic of their current locations. The calculation of the attractiveness in our extended algorithm, called SFiCA (Spatial FiCA) is done according to the following Equation:

P's = <pps0 e~yrZ (5)

where ps0 is the initial brightness as defined in Equation (1) and ^ is coefficient factor assigned to the sensitivity of the location to the event of interest. The more the location is sensitive to the current event, the higher is the coefficient This coefficient will ultimately increase the chances of sensors in sensitive areas to become CHs. We will thus expect to have more clusters with smaller sizes around the important spatial areas with respect to the event of interest. More details about SFiCA and its implementation will be the subject of an upcoming publication.

6. Performance evaluation

In order to test the performance of our approaches FiCA and SFiCA, we created a java-based application from which we simulated a randomly deployed WSN. The WSN was modeled by a multiagent system and every sensor is represented by a software agent. Our multiagent system was created with the java-based platform Jade. We run simulations for FiCA only, for different densities of the network while varying the sensors' communication ranges. Figure 2, Figure 3, and Figure 4 depict the results of the micro FA clustering (left side) and the results of the macro FA polishing (right side). As shown is these figures, our FiCA algorithm was able to successfully cluster the WSN. The resulting clusters have relatively different sizes. This is particularly because we did not impose any restriction about the number of sensors in each cluster. These sizes may become further unbalanced, particularly when competition is low (for example in scares networks or when some sensors have much higher attractiveness compared to their neighboring).

Fig. 2. Simulation of FiCA for 800 randomly deployed nodes (left) micro FA, (right) macro FA

Fig. 3. Simulation of FiCA for 400 randomly deployed nodes (left) micro FA, (right) macro FA

Fig. 4. Simulation of FiCA for 100 randomly deployed nodes (left) micro FA, (right) macro FA

In Figure 2, 800 sensors have been deployed in 600x600 area. The right side of the figure shows the results of the polishing phase performed on the clustering results depicted on left side of the figure. Because we set the size of small sensors to 3, few merging operations have been performed. This is expected since most of the clusters have sizes more than 3 due to the density of the network. In the left sides of Figure 3 and Figure 4 we also depict the results of the polishing phases for network densities of 400 and 100 respectively.

In Figure 5-left, we estimated the effect of communication ranges and sensors' density on the number of CHs. As we can see on this figure, the more we increase the communication range, the more the number of CHs decreases. This result is actually expected since by increasing the communication ranges, several sensors and CHs would be attracted by other CHs. The result could also be interpreted by the fact that the WSN tends to reach an equilibrium where competitive forces between sensors does not allow any furtherer merging between neighboring clusters. Our results could be different if we limited the size of clusters to a given threshold. Furthermore, we studied the effects of communication ranges and density on the percent of initial and final CHs. To this end, we calculated the number of initial CHCs and the number of final CHs. Our results depicted in Figure 5 -right show that the percent of final CHs obtained from the initial number of CHs for different network densities and different communication ranges tends to grow for small densities slower than for higher densities. This percent tends to stabilize when the communication range


45 -40 35 30 g 25

15 10 5 0

30 40 50 60 70 Communication range

Numofnodes100 Num ofnodes600-

Num of nodes 200

-Num ofnodes 800

-- Num ofnodes 400

Fig. 5. Effect of sensors' densities and communication ranges on: (left) the final number of CHs (right) the percent of initial and final CHs

90 100

7. Conclusion

In this paper, we presented a new WSN clustering approach, called FiCA, based on a firefly algorithm. Our approach consists of four steps: Initialization, Fetching, Intimidation, and Polishing. During the first three steps, called micro clustering, our approach allows CH candidates to attract neighboring peers to their clusters based on initially random calculated attractiveness values. During the last step, also called macro clustering, our approach allows small neighboring clusters to be aggregated. Because of the fading of CHs' attractiveness, the clustering process will always stop and prevent the creation of large clusters, particularly when the density of the network is high. We also presented an approach called SFiCA where spatial contextual data could be taken into consideration to increase the spatial awareness of sensors and improve their decision-making process.

The implementation of our FiCA algorithm demonstrated promising results in terms of cluster distribution. Nevertheless, some performance issues still need to be fixed. In addition to this ongoing work, we are planning to implement the SFiCA algorithm and then compare the performance of both algorithms with some existing clustering algorithms, including LEACH. We are also working on modifying our firefly approach by allowing sensors to attract other sensors based on additional criteria, such as their energy, semantics, and quality of their services. This is actually going to reflect the idea that clusters will compete to attract the best (also strongest) peers, which are not necessarily the closest ones.


1. Boyinbode O, Le H, Mbogho A, Takizawa M, Poliah R. A Survey on Clustering Algorithms for Wireless Sensor Networks, Network-Based

Information Systems (NBiS), 2010 13th International Conference on, Takayama, 2010, pp. 358-364

2. Ye M, Li C, Chen G, Wu J. An Energy Efficient Clustering Scheme in Wireless Sensor Networks, Ad Hoc & Sensor Wireless Networks, 2006,

Vol.1, pp.1-21

3. Barker DJ, Ephremides A, Flynn JA. The design and simulation of a mobile radio network with distributed control, IEEE Journal on

Selected Areas in Communications (1984) 226-237

4. Amis A, Prakash R, Vuong T, Huynh D. Max-Min D-Cluster Formation in Wireless Ad Hoc Networks, IEEE INFOCOM, March 2000

5. Li C, Ye M, Chen G, Wu J. An energy efficient unequal clustering mechanism for wireless sensor networks, in: Proc. of2005 IEEE International

Conference on Mobile Adhoc and Sensor Systems Conference (MASS05), 2005, pp. 604-611

6. Heinzelman W, Chandrakasan A, Balakrishnan H. Energy-Efficient Routing Protocols for Wireless Microsensor Networks, Proceedings of the

33rd Hawaii International Conference on System Sciences HICSS, Maui, HI, USA, January 2000

7. Younis M, Fahmy S. HEED: A Hybrid Energy-Efficient Distributed Clustering Approach for Ad Hoc Sensor Networks, IEEE Transactions on

Mobile Computing, 2004, vol. 3, no. 4

8. Jones K.H, Lodding K.N, Olariu S, Wilson L, Chunsheng X. Communal Cooperation in Sensor Networks for Situation Management, Proc. 9th

Int. Conf. on Information Fusion, 2006, pp.1-8

9. Antoniou P., Pitsillides A.Congestion Control in Autonomous Decentralized Networks Based on the Lotka-Volterra Competition Model, J.

Artificial Neural Networks, 2009, vol. 5769, pp 986-996

10. Jabeur N, Sahli N, Zeadally S. ABAMA: An Agent-Based Architecture for Mapping Natural Ecosystems onto Wireless Sensor Networks, Invited Paper, in Proceedings of 9th International Conference on Future Networks and Communications (FNC-2014), Elsevier Procedia Computer Science, Volume 34, Niagara Falls, Ontario, Canada, August 2014 - doi:10.1016/j.procs.2014.07.020

11. Jacobsen RH, Zhang Q, Toftegaard TS. Bioinspired Principles for Large-Scale Networked Sensor Systems: An Overview. Sensors (Basel, Switzerland), 2011, 11(4), 4137-4151

12. Tillett, J.; Rao, R.; Sahin, F. Cluster-head identification in ad hoc sensor networks using particle swarm optimization. In IEEE International Conference on Personal Wireless Communications, New Delhi, India, 15-17 December 2002; pp. 201-205

13. Dong J, Qi M. A new clustering algorithm based on PSO with the jumping mechanism of SA. In Proceedings of the 3rd International Conference on Intelligent Information Technology Application, IITA'09; IEEE Press: Piscataway, NJ, USA, 21-22 Nov. 2009; pp. 61-64

14. Latiff N.M.A, Tsimenidis C.C, Sharif B.S. Energy-Aware Clustering for Wireless Sensor Networks using Particle Swarm Optimization. In IEEE 18th International Symposium on Personal, Indoor and Mobile Radio Communications, PIMRC 2007, Athens, Greece, 3-7 September 2007; pp. 1-5

15. Guru S.M, Halgamuge S.K, Fernando S. Particle Swarm Optimisers for Cluster formation in Wireless Sensor Networks. In Proceedings of the 2005 International Conference on Intelligent Sensors, Sensor Networks and Information Processing Conference, Melbourne, Australia, 5-8 December 2005; pp. 319-324

16. Charalambous C, Cui S. A biologically inspired networking model for wireless sensor networks. IEEE Netw. 2010, 24, 6-13

17. Zhang Q, Jacobsen RH, Toftegaard TS. Bio-inspired low-complexity clustering in large-scale dense wireless sensor networks, Global Communications Conference (GLOBECOM), 2012 IEEE, Anaheim, CA, 2012, pp. 658-663

18. Selvakennedy S, Sinnappan S, Shang Y. A biologically-inspired clustering protocol for wireless sensor networks, Comput. Commun, vol. 30, pp. 2786-2801, October 2007

19. Krishnaveni V, Arumugam G. A novel enhanced bio-inspired harmony search algorithm for clustering, Recent Advances in Computing and Software Systems (RACSS), 2012 International Conference on, Chennai, 2012, pp. 7-12

20. Hasnat MA, Akbar M, Iqbal Z, Khan ZA, Qasim U, Javaid N. Bio inspired distributed energy efficient clustering for Wireless Sensor Networks, Information Technology: Towards New Smart World (NSITNSW), 2015 5th National Symposium on, Riyadh, 2015, pp. 1-7

21. Karaboga D, Okdem S, Ozturk C. Cluster based wireless sensor network routing using artificial bee colony algorithm, International journal of Wireless Networks, 7(18), 2012, pp: 847-860

22. Sarobin V.R, Ganesan R. Bio-Inspired, Cluster-Based Deterministic Node Deployment in Wireless Sensor Networks, International Journal of Technology (2016) 4, pp: 673-682, ISSN 2086-9614

23. Senthilnath J, Omkar S.N, Mani V. Clustering using firefly algorithm: performance study, in Swarm and Evolutionary Computation, Vol. 1(3), 2011, pp.164 - 171

24. Sarma N.V.S.N, Gopi M. Implementation of Energy Efficient Clustering Using Firefly Algorithm in Wireless Sensor Networks, 1st International Congress on Computer, Electronics, Electrical, and Communication Engineering (ICCEECE2014), IPCSIT vol. 59 (2014), IACSIT Press, Singapore, DOI: 10.7763/IPCSIT.2014.V59.1

25. Sandeep K.E, Kusuma S.M., Kumar V.B.P. Fire-LEACH: A Novel Clustering Protocol for Wireless Sensor Networks Based on Firefly Algorithm, International Journal of Computer Science Theory and Application, 1(1), 2014, pp: 12-17