Computer Science Review 40 (2021) 100376

Contents lists available at ScienceDirect

Computer Science Review

journal homepage: www.elsevier.com/locate/cosrev

Review article

Strategies based on various aspects of clustering in wireless sensor

networks using classical, optimization and machine learning
techniques: Review, taxonomy, research findings, challenges and
future directions
∗
J. Amutha, Sandeep Sharma , Sanjay Kumar Sharma
University School of ICT, Gautam Buddha University, Greater Noida, Uttar Pradesh, India

article info a b s t r a c t

Article history: Wireless Sensor Networks (WSNs) have attracted various academic researchers, engineers, science,
Received 26 July 2020 and technology communities. This attraction is due to their broad research areas such as energy
Received in revised form 19 December 2020 efficiency, data communication, coverage, connectivity, load balancing, security, reliability, scalability,
Accepted 27 January 2021
and network lifetime. Researchers are looking towards cost-effective approaches to improve the
Available online xxxx
existing solutions that reveal novel schemes, methods, concepts, protocols, and algorithms in the
Keywords: desired domain. Generally, review studies provide complete, easy access or solution to these concepts.
Wireless Sensor Networks Considering this as a driving force and the impact of clustering on the deterioration of energy
Optimization consumption in wireless sensor networks, this review focus on clustering methods based on different
Machine learning aspects. This study’s significant contribution is to provide a brief review in the field of clustering
Routing
in wireless sensor networks based on three different categories, such as classical, optimization,
Security
and machine learning techniques. For each of these categories, various performance metrics and
Reliability
parameters are provided, and a comparative assessment of the corresponding aspects like cluster
head selection, routing protocols, reliability, security, and unequal clustering are discussed. Various
advantages, limitations, applications of each method, research gaps, challenges, and research directions
are considered in this study, motivating the researchers to carry out further research by providing
relevant information in cluster-based wireless sensor networks.
© 2021 Elsevier Inc. All rights reserved.

Contents

1. Introduction......................................................................................................................................................................................................................... 2
2. Related work ....................................................................................................................................................................................................................... 7
3. Taxonomy of clustering in wireless sensor networks ................................................................................................................................................... 10
3.1. Clustering environment......................................................................................................................................................................................... 10
3.2. Cluster characteristics ........................................................................................................................................................................................... 10
3.3. Clustering methods................................................................................................................................................................................................ 11
3.4. Clustering process.................................................................................................................................................................................................. 11
3.4.1. Cluster formation ................................................................................................................................................................................... 11
3.4.2. Cluster head selection............................................................................................................................................................................ 11
3.5. Clustering objectives ............................................................................................................................................................................................. 12
3.6. Cluster routing ....................................................................................................................................................................................................... 12
3.6.1. Network structure .................................................................................................................................................................................. 12
3.6.2. Routing topology .................................................................................................................................................................................... 12
3.6.3. Path establishment................................................................................................................................................................................. 12
3.6.4. Protocol operation.................................................................................................................................................................................. 12
4. Strategies used in various aspects of clustering process............................................................................................................................................... 12
4.1. Classical methods................................................................................................................................................................................................... 12

∗ Corresponding author.
E-mail address: sandeepsvce@gmail.com (S. Sharma).

https://doi.org/10.1016/j.cosrev.2021.100376
1574-0137/© 2021 Elsevier Inc. All rights reserved.
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

4.1.1. Applications ............................................................................................................................................................................................ 13

4.1.2. Advantages .............................................................................................................................................................................................. 13
4.1.3. Limitations .............................................................................................................................................................................................. 13
4.2. Optimization approaches ...................................................................................................................................................................................... 13
4.2.1. Evolutionary algorithm .......................................................................................................................................................................... 14
4.2.2. Swarm intelligence based clustering approaches............................................................................................................................... 14
4.2.3. Approximation algorithms based on load balanced clustering ........................................................................................................ 14
4.2.4. Applications ............................................................................................................................................................................................ 14
4.2.5. Advantages .............................................................................................................................................................................................. 15
4.2.6. Limitations .............................................................................................................................................................................................. 15
4.3. Machine learning techniques ............................................................................................................................................................................... 15
4.3.1. Supervised learning................................................................................................................................................................................ 15
4.3.2. Unsupervised learning ........................................................................................................................................................................... 16
4.3.3. Reinforcement learning ......................................................................................................................................................................... 16
4.3.4. Applications ............................................................................................................................................................................................ 16
4.3.5. Advantages .............................................................................................................................................................................................. 16
4.3.6. Limitations .............................................................................................................................................................................................. 17
5. Clustering aspects based on various approaches ........................................................................................................................................................... 17
5.1. Cluster formation and cluster head selection .................................................................................................................................................... 17
5.1.1. Classical methods ................................................................................................................................................................................... 17
5.1.2. Optimization approaches....................................................................................................................................................................... 17
5.1.3. Machine learning techniques................................................................................................................................................................ 19
5.2. Routing protocols in clustering ............................................................................................................................................................................ 19
5.2.1. Classical methods ................................................................................................................................................................................... 19
5.2.2. Optimization approaches....................................................................................................................................................................... 19
5.2.3. Machine learning techniques................................................................................................................................................................ 24
5.3. Reliability in clustering ......................................................................................................................................................................................... 24
5.3.1. Classical methods ................................................................................................................................................................................... 24
5.3.2. Optimization approaches....................................................................................................................................................................... 24
5.3.3. Machine learning techniques................................................................................................................................................................ 27
5.4. Security in clustering ............................................................................................................................................................................................ 27
5.4.1. Classical methods ................................................................................................................................................................................... 27
5.4.2. Optimization approaches....................................................................................................................................................................... 27
5.4.3. Machine learning techniques................................................................................................................................................................ 27
6. Unequal clustering in wireless sensor networks............................................................................................................................................................ 29
6.1. Classical methods................................................................................................................................................................................................... 29
6.2. Optimization approaches ...................................................................................................................................................................................... 30
6.3. Machine learning techniques ............................................................................................................................................................................... 32
7. Research findings, challenges, open issues, and future directions ............................................................................................................................... 32
7.1. Research findings ................................................................................................................................................................................................... 32
7.2. Challenges and open issues .................................................................................................................................................................................. 32
7.3. Future directions.................................................................................................................................................................................................... 33
7.4. Summary................................................................................................................................................................................................................. 36
8. Conclusion ........................................................................................................................................................................................................................... 36
CRediT authorship contribution statement ..................................................................................................................................................................... 37
Declaration of competing interest.................................................................................................................................................................................... 37
Acknowledgments .............................................................................................................................................................................................................. 37
References ........................................................................................................................................................................................................................... 37

During the clustering process, groups of sensor nodes form

1. Introduction clusters, and a Cluster Head (CH) is identified within every cluster.
This CH then gathers and disseminates the collected information
Wireless Sensor Network (WSN) is considered significant and through single-hop or multi-hop data transfer to the BS [10]. In
promising due to its cost-effectiveness, size, simplicity, and easy single-hop data transfer, CHs transfer the obtained data directly
deployment. It comprises sensor nodes that are deployed in a to the BS, whereas in multi-hop data transfer, CHs transfer the
sensing field, used for examining and recording environmental obtained data on to higher-level CHs and thereby transfers to the
information such as the light, temperature, vibration, sound, and BS. Besides, the network minimizes the centralized requirements
motions, thereby transmits them to the base station [1–3]. Thus, and inspires the local decision, which enhances the scalability.
these sets of nodes have the advantage of ensuring proper sensing Clustering reduces the communication overhead and power con-
and wireless communication [4]. Due to these specific character- sumption, minimizes the interference, and hence the network
istics, WSNs are utilized in various applications like healthcare, lifetime can be enhanced [11,12].
environmental monitoring, transportation system, military sys- Moreover, external causes and dynamic changes in the wire-
tems, industry, and public safety [5–7]. During data distribution, less sensor networks affect cluster head selection, routing, delay,
each node communicates with the Base Station (BS) through a localization, quality of service, fault detection, coverage, relia-
single-hop or multi-hop data transfer [8,9]. As the continuous bility, and security [46–51]. Hence, the network may require a
process of data transmission, the nodes having more significant redesign, but the classical approaches for wireless sensor net-
distance consume their resources quickly than other nodes, and works are programmed explicitly. Due to this effect, for a dynamic
hence to solve this issue, the clustering method is used in WSNs and complex environment, the network does not function prop-
with numerous nodes. erly. Machine Learning (ML) and various optimization approaches
2
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Table 1
List of survey papers based on clustering for WSNs.
Ref. Year Major contributions Schemes/Categories Approach Algorithms/Protocols Challenges/Issues Proposed solutions Limitation
specified
[13] 2006 Classification of Election of CHs and Classical LEACH, HEED, ACE, Connectivity, optimal Cross-layer architecture The paper does not
clustering approaches Execution of a GAF, DCA CH selection, based on MAC and focus on many existing
based on their clustering algorithm computing optimal network layers clustering strategies
objectives and design (probabilistic or clusters, node duty and fails to address
principles iterative) cycle and node strategies based on
synchronization unequal clustering,
optimization and
machine learning
techniques.
[14] 2006 Comparative analysis of Heuristic, Weighted, Classical & LCA, WCA, LEACH, Energy efficiency, Restriction on the The protocols
different schemes with Hierarchical and Grid Heuristic TL-LEACH, EECS, HEED, Lifetime of the number of nodes in mentioned in this
respect to power and Schemes PEGASIS, GROUP network, Reliability and the network guarantees article offers promising
quality Quality of service the balancing of load improvement over
among CHs conventional clustering;
but still there is a
massive amount of
work that should focus
on machine learning
approaches
[15] 2007 Taxonomy of clustering Clustering attributes Classical LCA, RCC, EEHC, Energy efficiency, Load Re-clustering recovers Many clustering
attributes; classification and Clustering LEACH, FLOC, ACE, balancing, Delay, failure of CH; approaches based on
in terms of objectives, algorithms HEED, DWEHC, MOCA Connectivity and Integrated clustering, the unequal clustering
cluster properties and Reliability dynamic clustering, and fuzzy-logic- based
clustering process route configuration are clustering have not
the viable solutions to been explored
extend the network
lifetime
[16] 2007 Considering the cluster Cluster-head election Classical ACE-C, ACE-L, CM, Designing an The network lifetime Distributed clustering
architecture as well as and Clustering with LEACH energy-efficient data maximizes as the strategies was
providing distributed mobility (CM) aggregation protocol in clustering parameter addressed, but failed to
clustering algorithms the mobile and cluster member fit these strategies in
for mobile sensor environment due to function increases various mobility
nodes the dynamic topology models
[17] 2008 An extensive survey is Deterministic, adaptive, Classical MECH, ACE-C, ACE-L, In hybrid approaches, To reduce the The need for a more
provided for data combined and hybrid LMSSC, LEACH, the energy utilization of energy, scalable, energy
aggregation in WSNs clustering schemes LEACH-C, LEACH-F, consumption of nodes stable clusters are efficient and effective
based on CH selection BCDCP, DMSTRP, EMCA, nearer to the BS maintained and clustering system are
and CH role rotation ERA, RRCH, HEED, restricts the scalability; number of CH highlighted only in
approaches CMRP, LEACH-ET, EECS, The maintenance of transmissions and classical methods of
LNCA, DWBCA, ECR, the network becomes receptions during the WSNs
EECR, PEGASIS, TCDGP difficult when there is cluster creation are
rise in the sensor minimized
nodes
[18] 2009 The location-based Schemes based on Classical and ZRR, DRR, LAR, CBR, Redundancy detection, To extend the potential While routing
schemes are assessed query, flooding grid, ACO-based GLS, TTDD, GBRA, secure routing, QoS applications of algorithms focus on
and the algorithms are ant-algorithm and GPSR, Ant-based routing, intelligent position-based routing position-based WSNs
evaluated in terms of curve-based intelligent routing routing, node mobility, algorithms, data and have presented
metrics like energy algorithms [19–24] data fusion aggregation is some major
consumption, reliability, necessary advantages, there are
scalability and data still issues to be
aggregation addressed like unequal
clustering and security
[25] 2010 The survey focuses Hierarchical, data Classical EEHC, LEACH, E-LEACH, The energy constrained Each protocol haves Even though feasibility
primarily on WSN centric, location-based LEACH-C, M-LEACH, nature of nodes creates their own advantages of the discussed
energy-efficient and QoS-aware LEACH-F, PEGASIS, specific challenges of and limitations. Based protocols are in terms
hierarchical clustering EB-PEGASIS, HEED, developing WSN on the topology and of only classical
protocols TEEN, APTEEN applications; The routing strategies, the approaches, more
metrics that affect the routing protocols must investigation is
creation of cluster be considered for necessary based on
process and CH specific applications machine learning and
communication are the optimization strategies
open issues to be
addressed
[26] 2011 An comprehensive Properties of CI CI : NN, FL, GA, Coarse Grain Wireless ad-hoc nature, For security and QoS Many CI techniques
survey of CI Paradigms, WSN PSO, ACO, RL Decomposition GA, Mobility, Topology issues, fuzzy logic have outperformed
applications to Applications in design Sequential PSO, changes, Energy based schemes are the traditional schemes
different WSNs and deployment, PSO-SinkPath, limitations and Physical best choice; SI is a against unpredictable
problem are presented; Localization, Data Traffic-PSO, COORD, distribution very prominent circumstances and
Besides, discussion of aggregation, Energy SSDP, GA-Loc, technique for MANET constraints in energy
CI approaches over Aware Routing and Two-Phase GA-Loc, routing schemes; RL is efficiency, network
classical schemes are Clustering PSO-Loc, Iterative PSO, well-suited when connectivity and
offered MLPMAC, GNMAC, dealing with computational
FS-MAC, SIR, distributed and capabilities. However,
GA-Routing, complex issues like the works that are
TwoTier-GA, Energy routing and clustering discussed are not the
Efficient GA Clustering, for WSNs best potential solutions
Multi-objective DE, and have not been
FROMS, Clique, compared with
Q-Routing, traditional or other CI
DRQ-Routing, TPOT-RL, approaches
QRC, QPR, AdaR,
RL-Flooding, Q-Fusion,
RLGR, FHNN, GA
Thermal Effect,
Dynamic Alliance-GA,
Immune Based
Scheduling, RL-MAC,
GA-PSO, Stochastic
Gradient GA, Q-RC,
Fuzzy-QoS, FuzzyMARS
[27] 2011 The effects of node Clustering attributes, Classical SEP, EDFCM, REECR, If CH is not selected in Heterogeneous WSNs In WSNs, eventhough
heterogeneity are Clustering algorithms, ZREECR, LEACH, DEEC, some rounds, the are more suited than classical methods based
examined on the WSNs Energy-Efficient SDEEC, DEBC, C4SD sensed data are not homogeneous schemes clustering are popular
performance and clustering Protocol transmitted to the BS, for real-life for minimizing energy
address various which impacts the applications; consumption and to
heterogeneous effective heterogeneous nodes maintain stability,
clustering algorithms communication of the support complex data other optimization
for WSNs based on network processing and longer based approaches and
different clustering storage machine learning
attributes techniques must also
be addressed.

(continued on next page)

3
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Table 1 (continued).
Ref. Year Major contributions Schemes/Categories Approach Algorithms/Protocols Challenges/Issues Proposed solutions Limitation
specified
[28] 2011 Taxonomy for WSN Foraging mechanisms; Swarm Intelligence SC, DCR, SDG, FF, Low memory, SI based algorithms Various routing
routing protocols are ACO metaheuristic; PZSWiD, FP, MO-IAR, Aggregation of provide computational protocols have been
provided and the Bee-inspired schemes EEABR, ACO-QoSR, in-network data, design concepts presented that focus
principles of swarm ASAR, Ant-Agg, T-ANT Energy consumption, inspired by adaptive on SI principles, mainly
intelligence as well as [29], AntChain, Scalability, biological behavior that focused on ant and
its usage to routing BeeSensor, PEADD, Self-organization and suits the challenges bees foraging behavior.
are explored JARA, E-DANTS Traffic patterns and constraints of However, this work
WSNs failed to address other
SI based and machine
learning based
clustering approaches
[30] 2012 Classification of Self-organized, Assisted Classical & LEACH, TB-LEACH, Selectivity, Even The complexity of A summary of different
cluster-oriented and Multi-factor Fuzzy-logic ACE-C, ACE-L, CCRP, distribution of CHs, clustering protocols clustering phases were
protocols that focused evaluation schemes WST-LEACH, Failure recovery, towards dynamic addressed only with
on their goals and MELEACH-L, EEDMC, Re-clustering, Data topology has to be classical and
strategies MECH, DDCHS, CES, aggregation, Application examined to provide fuzzy-based schemes
KMMDA, CHEF, FSCA, dependency, Balanced solutions concerning
CHLBC, CMATO, clustering and Network cluster inconsistency;
CLUDDA, APTEEN, connectivity Rotation of the CH role
BCDCP should be recognized
as a factor in
determining the
process of CH selection
[31] 2012 Review on clustering Cluster characteristics, Classical LEACH, TL-LEACH, Mobility of clusters, Classical WSN This study relies only
routing protocols; CH characteristics and HEED, TEEN, APTEEN, Energy Efficiency, clustering routing on a few of the
Summarize the benefits clustering process PEGASIS, EECS, EEUC, Scalability, algorithms rely on the important and
and goals of clustering; DWEHC, PANEL, ACE, Communication delay, classification and well-known clustering
Taxonomy of WSN BCDCP, TTDD, CCS Load Balancing, emphasis on different techniques. The fuzzy,
clustering that focus Algorithm Complexity algorithm phases evolutionary and
on the attributes of objectives in this area
cluster; Comparison of are not reviewed.
various taxonomy
approaches and other
relevant metrics
[32] 2012 Examines WSNs Data-centric, Location Classical and F & G, DD, SPIN, GBR, Limited energy To address overhead Most protocols
routing protocols from based, Hierarchical, Swarm intelligence EAR, RR, CADR, capacity, fault mobility and conclude with analysis
classical to swarm Network flow and ACQUIRE, EAD, GAF, tolerance, scalability, configuration variations and evaluation, with
intelligence based QoSaware GEAR, MECN, SMECN, productive cost, quality in energy-restricted no findings compared
protocols that focus on TBF, EAGRP, LEACH, of service, data areas, new algorithms to any state-of-the-art
structure of the PEGASIS, SOP, TEEN, aggregation and fusion, are needed traditional routing
network, energy APTEEN, HEED, node mobility, algorithms from WSNs
efficiency, EAR-CSN, BCEE, EAQSR, connectivity, security,
computational MCBR, MLDG, SAR, congestion, latency
complexity and path MLER, SPEED, AODV,
establishment PEADD, CRP, SC, SDG,
EBAB, ACO-C,
ACALEACH, MACS,
PZSWiD, ACMRA,
ACMT, ACLR, MSRP,
JARA, ACOLBR, EEABR,
FF, FP, E-DANTS, AR
and IAR, ACO-QoSR,
ASAR, BABR,
ACO-EAMRA, EAQR,
QDV, AntSensNet
[33,34] 2013 Taxonomy of clustering CH selection, Multi-hop Classical LEACH, HEED, PEGASIS, The various A lightweight Clustering and routing
and routing techniques; data transmission, PEACH, EECFP, performance metrics authentication protocols based on
Comparison and Heterogeneous LEACH-C, MCR, such as latency, delay mechanism ensures LEACH along with their
discussion of existing networks and Chain DE-LEACH, ID-LEACH, and fault tolerance secure network variants were
protocols based routing protocols LEACH-ER, TB-LEACH, play an important role communications, and described, but failed to
MEPA, IEEPB, OECM, for multimedia sensor furthermore consumes address the schemes in
RPCR, KMMDA, SEP, network based less resources machine learning and
EBC, BARC, MR-LEACH, applications, due to the optimization
W-LEACH, LEACH-SC, sensors that capture approaches
T-LEACH, DEECIC, the image in the
LEACH-H, LEACH-ME, region where they are
LEACH-TM, SHRP, CCRP, deployed; Owing to
BCDCP, ECCRA, MLCRA, the high mobility of
BCEE, SRMCF, EHEED, the nodes in Vehicular
EECHE, CCBRP, Ad Hoc Networks
EBRAMS, MBC, (VANETs), the way
SEC-LEACH, CCCR, ECR, data aggregation to be
ECHR, FTTT carried out remains a
major challenge; Since
sensor networks
function in an open
environment, security
remains a key issue
[35] 2013 The strategies analyzed Structure of the Classical LEACH, LEACH-C, TEEN, Node Deployment, Hierarchical protocols Though, energy-efficient
for each classification network, PEGASIS, APTEEN, GEM, Node Heterogeneity, are suitable for high protocols are utilized
techniques are Communication GEAR, IGF, SELAR, Data Reporting, Energy load and wide coverage for WSNs, still a lot of
discussed based on the schemes, Routing based MERR, GDSTR, MCBCR, Consumption Without sensor networks research to be done,
benefits, drawbacks, on topology and WRP, SSMCF, TBRPF, Losing Accuracy, not only in the field of
scalability, flexibility, reliability TORA, RR, E-TORA, ZRP, Scalability, Network energy conservation,
and reliability VGA, MIMO, TTDD, Dynamics, Fault but in other fields like
IEMF, IEMA, HPAR, Tolerance, Connectivity, security, QoS, real-time
ELCH, BCDCP, DD, SPIN, Transmission Media, application and
GBDD, ACQUIRE, Coverage, Quality of mobility
NHRPA, DHAC, HMRP, Service, Data
SHPER, SWE, DREAM, Aggregation
MWE, SPIN-PP,
SPIN-EC, DGR, SPIN-BC,
SPIN-RL, GRAB,
PAGER-M, RPL, OGF,
HGR, MIP, ROAM,
SPEED, LMR, CBMPR,
DCF, SAR, MGR

(continued on next page)

are used to overcome this problem. So far, different reviews However, an overall comprehensive review of cluster head sel-
based on several aspects of the clustering process in WSNs were ection, routing protocols, unequal clustering, reliability, and secu-
reported by various researchers [15,26,52–64]. rity for WSNs based on different mechanisms and optimization
4
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Table 1 (continued).
Ref. Year Major contributions Schemes/Categories Approach Algorithms/Protocols Challenges/Issues Proposed solutions Limitation
specified
[36] 2014 Clustering approach Clustering objectives, Classical, Heuristic, LEACH, TEEN, APTEEN, Many random-based Deterministic Although clustering
review; Comparison clustering Fuzzy, PEGASIS, CSS, EEHC, classical clustering approaches are more approaches have been
focusing on different characteristics, and Unequal-based FLOC, TL-LEACH, MOCA, strategies employs efficient than extensively explored in
metrics such as cluster classification of CCN, HEED, SEP, EECS, direct communication probabilistic approaches various areas, certain
size, cluster count and clustering algorithms EDACH, TCCA, CAWT, with the BS, rather due to the size of the aspects of clustering
complexity of ACDA, CMEER, EEMC, than multi-hop cluster, cluster count, like dynamic nature of
algorithm LEACH-DT, DCA, communication; The energy level, position clustering, clustering
DWEHC, TASC, EECABN, slow and centralized of the nodes and CHs; methods for reactive
CHEF, LEACH-FL, ECPF, nature lets heuristic the use of networks, optimization
GCA, MOGA, BEES, algorithms unpractical heuristic-based algorithms,
PSO-C, HCC, ACE, for many applications; strategies may be the machine-learning
BCDCP, PEACH, MECH, Examining the impact best option for techniques are still not
EEDC, EACLE, LCM, of network mobility; small-scale applications investigated
PANEL, EEDCF, UCS, Another open challenge
EEUC, EEDUC, EB-UCP, is achieving the QoS
LUCA, EADUC, EBCAG, criteria of a WSN
MRPUC, PEBECS, EAUCF
[37] 2014 Intelligent routing Ant Colony Intelligent-based Q-Routing, AdaR, ATP, Effective methods are RL and ACO are the Intelligent
protocols that Optimization, Genetic routing FROMS, QELAR, BAR, required to deal with best choices for WSNs algorithm-based routing
contribute to network Algorithm, Fuzzy Logic, SC, FF, FP, EEABR, uncertainty, promote to deal with routing protocols look
lifetime improvement Reinforcement Learning ACORC, FCH, FMO, the development of issues, however, ACO promising as they have
in the perspective of and Neural Networks GA-Routing, GA- EECP, reliable mechanisms involves high success over uncertain
protocol specification, SIR and maximize communication environments.
results and inference effectiveness overhead; FL is Moreover many of
implementation are appropriate for them lack comparisons
studied achieving multiple with conventional,
objectives, routing optimization, machine
optimization and learning or other
clustering heuristics; swarm-intelligent based
GA and NNs are more algorithms.
suited for clustering
strategies that are
pre-deployed
[38] 2015 Various number of Virtual forces-based, Classical, Fuzzy VFA - Virtual Force Sensor coverage, Sensor The self-deployment of Providing rather a
algorithms for pattern-based, logic, Metaheuristic Algorithm, IVFA - communication, application-specific comprehensive view of
movement-assisted grid-quorum based, Improved VFA, EVFA - Self-deployment in WSNs could provide the state-of-the-art, it
sensor deployment computational Energy-considered VFA, three-dimensional RoI, valuable data to also offers valuable
have been enumerated geometry based and CPVF - Connectivity Path planning and maximize resources in perspectives for
and in addition fuzzy logic-based Preserving VFA, ESD - Application-specific a sensing area in choosing the most
highlighted their schemes Enhanced WSNs self-deployment compliance with the suitable
benefits and drawbacks Self-deployment needs of the self-deployment
by considering different Algorithm, BODVFA - framework and algorithm in spite of
concepts, assumptions, Back-Off Delay based contextual information only some of the fuzzy
goals and constraints Virtual Force Algorithm, logic and
VOR - VORonoi based meta-heuristics
algorithm, FBPC - approaches
Fuzzy Based Priority
Coverage algorithm,
ACO, GA, PSO, SA, ABC,
LODICO - LOcal
DIstributed
COevolutionary
computation, FGA -
Force based GA,
SMART - Scan-based
Movement-Assisted
sensoR deploymenT,
ZONER - ZONE-based
Sensor Relocation
[39] 2016 Provides a Clustering based on: Classical, Heuristic, Algorithms : Max–Min, During clustering Actuators can be used Only ACO-based swarm
retrospective overview Mobility-aware, K-hop, ACO Lowest-ID, process, the nodes are in clustering algorithms intelligence algorithm
of clustering strategies, Load-balanced, Highest-Degree, Least not adaptable for owing to the task of have been used as a
with extensive insights Energy-aware, Cluster Change; MBC, dynamic algorithms; gathering sensing data learning technique for
about the origin and Multi-metric, SCP, WCA, KCMBC, The key challenges of and in addition serves WSNs; Optimization
the research efforts Access-based and Data CEIF, DDCA, LBKC, VANET are the mobility as the medium to the algorithms based on
over the past decades; aggregation DiLoC, LEACH, HEED, and the delay sensitive data collected by the models for biological
A concise summary of TL-LEACH, COGMESH communications sensors behavior is not
data aggregation in highlighted
clustering; Directions
for future research are
outlined
[40] 2017 Survey of latest Optimization Strategies, Multi-Objective MOEAs [41], SIOAs Intensify MOO research A feasible alternative Various solutions are
research and MOO Algorithms, Optimization [42], MOGA, NSGA, in the sense of for Cognitive Radio explored based on
development initiatives Software Tools, Existing NPGA, SPEA, DE, AIS, multi-hop WSNs; In Assisted WSN both classical and
that solve the different WSN literatures using ICA, ACO, PSO, ABC, WSNs, security is (CR-WSN) can be used optimization
problems of energy MOO ANN, RL, FL, FRMOO, indeed one of the for some WSN approaches. However,
efficiency, coverage and FA significant challenges applications, such as research directions
network lifetime using as the nodes are indoor sensing, does not focus on
the multi-objective deployed in an heterogeneous multi-hop
optimization technique accessible and harsh multi-class sensing and transmissions, dynamic
(MOO) were addressed environments; real-time surveillance deployment of nodes
Trade-offs between the and solving
security and QoS optimization problems
specifications results in in 3D networks
additional new
challenges
[43] 2018 Homogeneous and Homogeneous and Classical, PSO UMBIC, MOIA, EECA, Selection of CH; In Heterogeneous Several clustering
heterogeneous heterogeneous SEED, LEACH-SF, distributed sensors, networks have strategies are specified
clustering techniques clustering based DSM-NbC, EEAOC, inaccessible and improved performance for the homogeneous
are primarily explored schemes WCA, EEWCA, LEACH, unattended nodes that over homogeneous networks, whereas
and compared to each TL-LEACH, TEEN, are prone to failure networks as they some are for the
other. Moreover PEGASIS, DEEC, DDEEC, results in topology choose the CH from heterogeneous
comparison is provided EDEEC, EDDEEC, MAPP, maintenance the higher power networks, but does not
for centralized and M2 CM, EEHC, HCTE, challengeable nodes, while all nodes focus on unequal
distributed clustering LEACH-C, MHRPUC, have full functional clustering, optimization
approaches ACHTH-LEACH, and processing capacity methods and machine
LEACH-SWDN, ECDS, in homogeneous learning algorithms
DWEHC, EEUC, EECS, networks
LACP, SBMC, EECR-PSO,
LEACH-DT, CRABDP,
C4SD, SEP

(continued on next page)

5
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Table 1 (continued).
Ref. Year Major contributions Schemes/Categories Approach Algorithms/Protocols Challenges/Issues Proposed solutions Limitation
specified
[44] 2019 A systematic analysis Probabilistic, Classical, Fuzzy and PRODUCE, EDUC, LUCA, Load balancing is a Probabilistic strategies Although the existing
of various unequal deterministic, heuristic based EEUC, EEDUC, UCR, primary issue in which are simple and works strategies are
clustering strategies is hierarchical and preset Unequal LEACH, EC, CHs are chosen from well in large-scale well-reviewed, certain
provided with their clustering schemes EB-UCP, UHEED, COCA, the entire nodes in the WSNs; Deterministic aspects of clustering
objectives, MRPUC, PEBECS, network algorithms can be used like dynamic clustering,
characteristics, EADUC, EBCAG, EUCS, for more accurate and mobility of the
classifications, UCMR, EPUC, CUCA, stable applications; network have not
advantages and Improved EADUC, Heuristic approaches investigated in terms
disadvantages. In EAUCF, IFUC, FUCP, are the best choice for of swarm-intelligence
addition, the FBUC, DUCF, EBUC, attaining an optimal and machine learning
classifications of these GAEEP, Improved PSO, solution in a specific techniques
approaches are SMEBUC, FAMACROW, application
compared on the basis nCRO-UCRA, UMBIC, environment
of different cluster EDDUCA, UCCGRA, UCS
properties, CH
properties and
clustering process
[45] 2020 A systematic survey of Basic concepts of Conventional LEACH, HEED, LCA, Topology management, Trust level method for Although, distributed
clustering strategies clustering; Clustering TEEN, PEGASIS, load balancing, network all WSN nodes is approaches have low
based on WSNs that in WSN; Statistical LEACH-C, EEHC, ACE, connectivity, security determined on the latency, selected CHs
focus on properties of evaluation of clustering DWEHC, TL-LEACH, basis of the node do not always satisfy
network and objectives schemes UCS, EEUC, BCDCP, behavior in a cluster to all network needs
of clustering TTDD, MOCA, DEEC, detect malicious nodes owing to the restricted
DWCA, EECS, EEDC, network information
PRODUCE, MHP,
EEDUC, UCR, FUCA,
DFCR (215 protocols
specified)

FA - Firefly Algorithm, FRMOO - Fuzzy Random MultiObjective Optimization, EMOCA - Evolutionary MultiObjective Crowding Algorithm, MOEAs - Multi-objective Evolutionary Algorithms, SIOAs - Swarm
Intelligence based Optimization Algorithms, ICA - Imperialist Competitive Algorithm, AIS - Artificial Immune System, NSGA - Non-dominated Sorting GA, NPGA - Niched Pareto GA, SPEA - Strength Pareto
Evolutionary Algorithm, MOGA - Multi-objective Genetic Algorithm, GPSR - Greedy Perimeter Stateless Routing, GBRA - Grid-Based Routing Algorithms, GLS - Grid Location Service, CBR - Curve-Based
Routing, LAR - Location-Aided Routing, ZRR - Zonal Rumor Routing, DRR - Directed Rumor Routing, MO-IAR - Many-to-One Improved Ant Routing, Ant-Agg - Ant-aggregation, FHNN - Fuzzy Hopfield Neural
Network, RLGR - Geographic-based Routing protocol for WSNs based on RL, QPR - Q-Probabilistic Routing, QRC - Q-Routing with Compression, TPOT-RL - Team-partitioned Opaque-Transition Reinforcement
Learning, DRQ-Routing - Dual Reinforcement Q-routing, SIR - Sensor Intelligence Routing, FS-MAC - Fuzzy logic based MAC, GNMAC - Generalized Neuron MAC, MLPMAC - MultiLayer Perceptron based
Media Access Control (MAC), GA-Loc - GA based node localization algorithm, SSDP - Service Directory Placement Protocol, COORD - Coordinated algorithm, EB-PEGASIS - Energy Balancing PEGASIS, LEACH-F
- LEACH with Fixed Cluster, M-LEACH - Multi-hop LEACH, E-LEACH - Enhanced-LEACH, DEBC - Distributed Energy-Balance Clustering, SDEEC - Stochastic Energy-Efficient Clustering, REECR - Residual Energy
and Energy Consumption Rate, ZREECR - Zone-Based REECR, EDFCM - Energy Dissipation Forecast and Clustering Management, C4SD - Cluster-Based Service Discovery Protocol, CRABDP - Clustering Routing
Algorithm Based on Distance Probability, LEACH-DT - LEACH-Distance-based Thresholds, SBMC - Sector Based Multi-hop Clustering, LACP - Layered Architecture Clustering Protocol, ECDS - Energy-Constrained
minimum Dominating Set, LEACH-SWDN - LEACH - Sliding Window and Dynamic Number of Nodes, ACHTH-LEACH - Adaptive Cluster Head Election and Two-hop LEACH, MHRPUC - Multi-Hop Routing
Protocol with Unequal Clustering, HCTE - Hierarchical Clustering Based Routing Algorithm, M2 CM - Multi-Mode Clustering Model, MAPP - Mobile Anchor nodes Path Planning, DDEEC - Developed Distributed
Energy Efficiency Clustering Protocol, EEAOC - Energy-Efficient Adaptive Overlapping Clustering, DSM-NbC - Need-Based Clustering (NbC) with Dynamic Sink Mobility, LEACH-SF - LEACH-Sugeno fuzzy
inference system, SEED - Sleep-awake Energy - Efficient Distributed, MOIA - Multi-Objective Immune Algorithm, UMBIC - Unequal Multi-hop Balanced Immune Clustering, UCCGRA - Unequal Clustering
and Connected Graph, EDDUCA - Energy Degree Distance Unequal Clustering Algorithm, UMBIC - Unequal Multi-hop Balanced Immune Clustering, nCRO-UCRA - Novel Chemical reaction optimization based
unequal clustering and routing algorithm for Wireless Sensor Networks, FAMACROW - Fuzzy and ant colony optimization based combined MAC, routing and unequal clustering cross-layer protocol for
wireless sensor networks, SMEBUC - Sink Mobility based Energy Balancing Unequal Clustering, IPSO - Improved PSO, GAEEP - Genetic Algorithm based Energy-Efficient Adaptive Clustering Hierarchical
Protocol, EBUC - Energy Balanced Unequal Clustering, DUCF - Distributed Load Balancing Unequal Clustering in Wireless Sensor Networks using Fuzzy approach, FBUC - Fuzzy logic Based Unequal Clustering,
FUCP-Fuzzy based Unequal Clustering Protocol, IFUC - Improved Fuzzy Unequal Clustering, EADUC - Energy Aware Distributed Unequal Clustering, CUCA - Coverage aware and Unequal Clustering Algorithm,
EPUC - Energy and proximity based unequal clustering algorithm, UCMR - Unequally Clustered Multi-hop Routing protocol, EUCS - Enhanced Unequal Clustering Scheme, EB-UCP - Energy Balancing Unequal
Clustering Protocol for WSN, EC - Energy-Efficient Clustering, EDUC - Energy Driven Unequal Clustering, PRODUCE - Probability Driven Unequal Clustering Mechanism for WSN, Q-Routing - Q-learning based
Routing, AdaR - Adaptive Routing, ATP - Adaptive Tree Protocol, FROMS - Feedback Routing for Optimizing Multiple Sinks, QELAR - Q-learning-based Energy-efficient and Lifetime-Aware Routing, BAR - Basic
Ant Routing, SC - Sensor-driven Cost-aware ant routing, FF - Flooded Forward ant routing, FP - Flooded piggy backed ant routing, EEABR - Energy-efficient Ant Based Routing, ACORC - Routing using Ant
Colony Optimization Router Chip, FCH - Cluster-Head election using Fuzzy logic, FMO - Fuzzy Multi-Objective routing, GA-Routing - Genetic Algorithm based routing, GA- EECP - GA based Energy-Efficient
Clustering Protocol, SIR - Sensor Intelligence Routing, MGR - Multimedia Geographic Routing, RPL - Routing Protocol for Low power and Lossy Networks, DCF - Directional Controlled Fusion, DGR -
Directional Geographical Routing, CBMPR - Cluster-Based Multi-Path Routing, HMRP - Hierarchy-Based Multipath Routing Protocol, GRAB - GRAdient Broadcast, LMR - Label-based Multipath Routing, ROAM
- Routing On-demand Acyclic Multipath, IEMF and IEMA - Itinerary Energy Minimum for First-source-selection and IEM Algorithm, MIP - Multi-agent based Itinerary Planning, HGR - Hybrid Geographic
Routing, PAGER-M - Partial-partition Avoiding Geographic Routing-Mobile, OGF - On-demand Geographic Forwarding, MERR - Minimum Energy Relay Routing, GDSTR - Greedy Distributed Spanning Tree
Routing, SELAR - Scalable Energy-efficient Location Aided Routing, IGF - Implicit Geographic Forwarding, GEM - Graph Embedding for Routing, DREAM - Distance Routing Effect Algorithm for Mobility,
SPIN-RL - SPIN with Reliability, SPIN-BC - SPIN for Broadcast Networks, SPIN-EC - SPIN with Energy Conservation, SPIN-PP - SPIN for Point to Point Communication, MWE - Multiple Winner Algorithm,
SWE - Single Winner Algorithm, DHAC - Distributed Hierarchical Agglomerative Clustering, SHPER - Scaling Hierarchical Power Efficient Routing, NHRPA - Novel Hierarchical Routing Protocol Algorithm,
ELCH - Extending Lifetime of Cluster Head, GBDD - Grid Based Data Dissemination, HPAR - Hierarchical Power Aware Routing, MIMO - Multihop Virtual Multiple Input Multiple Output, VGA - Virtual Grid
Architecture Routing, ZRP - Zone Routing Protocol, TORA - Temporarily Ordered Routing Algorithm, E-TORA - Energy-aware TORA, WRP - Wireless Routing Protocol, TBRPF - Topology Dissemination Based
on Reverse-Path Forwarding Protocol, SSMCF - Scalable Solution to Minimum-Cost Forwarding, MCBCR - Minimum-Cost Bandwidth-Constrained Routing, EEABR - Energy Efficient Ant Based Routing, FF -
Flooded Forward ant routing, FP - Flooded Piggy back ant routing, E-DANTS - Energy-Delayant-based routing, AR and IAR - Ant colony based Reinforcement learning algorithm, ACO-QoSR - ACO-based
Quality-of-Service Routing, ASAR - Ant based Service-Aware Routing algorithm, BABR - Basic Ant Based Routing, ACO-EAMRA - Ant COlony-based Energy-Aware Multipath Routing Algorithm, EAQR - Energy
efficient ACO based QoS Routing, QDV - Quality of service based Distance Vector routing protocol, AntSensNet - Ant-based routing for wireless multimedia Sensor Networks, PEADD - Pheromone based Energy
Aware Directed Diffusion, CRP - Comprehensive Routing Protocol, SC - Sensor driven and cost-aware ant routing, SDG - Self-organizing Data Gathering scheme, EBAB - Energy Balanced Ant-Based routing
protocol, ACO-C - Adaptive clustering for energy efficient WSN based on ACO, ACALEACH - Ant colony Clustering Algorithm, MACS - Multipath routing based on Ant Colony System, PZSWiD - Probabilistic,
Zonal and Swarm-inspired system for Wildfire Detection, ACMRA - Ant Colony based Multipath Routing Algorithm, ACMT - Ant Colony Multicast Trees, ACLR - Ant Colony optimization based Location aware
Routing, MSRP - Multi-sink Swarm-based Routing Protocol, JARA - Jumping Ant Routing Algorithm, ACOLBR - Ant Colony Optimization-based Load Balancing routing Algorithm, MLDG-Maximum Lifetime Data
Gathering, SAR - Sequential Assignment Routing, MLER - Maximum Lifetime Energy Routing, SPEED - Stateless Protocol for real-time communication in sensor networks, EAQSR - Energy-Aware QoS Routing
protocol, MCBR - Message-initiated Constraint- Based Routing, AODV - Ad-hoc On-Demand distance Vector routing protocol, EAR-CSN - EAR for Cluster-based Sensor Networks, BCEE - Balanced-Clustering
Energy-Efficient protocol, GEAR - Geographic and Energy-Aware Routing, MECN - Minimum Energy Communication Network, SMECN - Small MECN, TBF - Trajectory-Based Forwarding, EAGRP - Energy-Aware
WSN Geographic Routing Protocol, EAD - Energy-Aware Data-centric routing, ACQUIRE - ACtive Query forwarding In sensor nEtworks, CADR - Constrained Anisotropic Diffusion Routing, RR - Rumor
Routing, EAR - Energy-Aware Routing, SPIN - Sensor Protocol for Information via Negotiation, DD - Directed Diffusion, F & G - Flooding and gossiping, TCDGP - Tree-Clustered Data Gathering Protocol,
EECR - Energy Efficient Clustering Routing algorithm, DWBCA - Distributed Weight-Based Clustering Algorithm, LNCA - Local Negotiated Clustering Algorithm, LEACH-ET - LEACH-Energy Threshold, CMRP -
Clustering and Multi-hop Routing with Power control in WSNs, RRCH - Round-Robin Cluster Header, ERA - Energy Residue Aware, EMCA - Enhanced Multihop Clustering Algorithm, DMSTRP - Dynamic
Minimum Spanning Tree Routing Protocol, LEACH-F - LEACH with Fixed cluster, LMSSC - Least Mean Squared Subtractive Clustering, COGMESH - COGnitive wireless MESH network, DiLoC - Distributed
and Location-aware Clustering, LBKC - Load Balance K-hop Clustering algorithm, DDCA - Distributed Dynamic Clustering Algorithm, CEIF - Cluster-head Election Irresponsible Forwarding, KCMBC - K-hop
Compound Metric Based Clustering, SCP - Stable Clustering Protocol, EAUCF - Energy Aware Fuzzy Unequal Clustering, PEBECS - Partition Energy Balanced and Efficient Clustering Scheme, MRPUC -
Multi-hop Routing Protocol with Unequal Clustering, EBCAG - Energy-Balancing unequal Clustering Approach for Gradient-based routing, EADUC - Energy-Aware Distributed Unequal Clustering, LUCA -
Location-based Unequal Clustering Algorithm, EB-UCP - Energy-Balancing Unequal Clustering Protocol, EEUC - Energy-efficient Unequal Clustering, EEDUC - Energy-Efficient Distributed Unequal Clustering,
UCS - Unequal Clustering Scheme, EEDCF - Energy Efficient Deployment and Cluster Formation, LCM - Link-aware Clustering Mechanism, EACLE - Energy-Aware CLustering scheme with transmission power
control for sEnsor networks, EEDC - Energy-efficient and Dynamic Clustering, HCC - Hierarchical Control Clustering, PSO-C - Centralized PSO, BEES - BEE-colony-inspired backbone Selection algorithm, MOGA
- Multi-objective GA-based clustering algorithm, GCA - Genetic Clustering Algorithm, ECPF - Energy-aware Clustering Protocol using Fuzzy-logic, EECABN - Energy Efficient Clustering Algorithm Based on
Neighbors, TASC - Topology Adaptive Spatial Clustering, EEMC-Energy-Efficient Multi-level Clustering, CMEER - Clustering Method for Energy Efficient Routing, ACDA - Autonomous Clustering via Directional
Antenna, CAWT - Clustering Algorithm via Waiting Timer, TCCA - Time Controlled Clustering Algorithm, EDACH - Energy-Driven Adaptive Clustering Hierarchy, CCN - Clustering Communication Based on
number of Neighbors, FTTT - Fault Tolerant Target Tracking Protocol, ECHR-Energy-Aware Coverage- Preserving Hierarchical Routing Protocol, ECR - Energy-Centric optimized Recovery algorithm, CCCR -
Coverage Conscious Connectivity Restoration, SEC-LEACH - Security Based LEACH, MBC - Mobility Based Clustering, EBRAMS - Energy Balanced Routing Algorithm Based on Mobile BS, CCBRP-Chain-Chain
Based Routing Protocol, EECHE - Energy Efficient Cluster Head Election, EHEED - Enhanced HEED, SRMCF - Source Routing for Minimum Cost Forwarding, BCEE- Balanced Clustering Energy-Efficient,
MLCRA-MultiLayer Clustering Routing Algorithm, ECCRA-Energy-Efficient Coverage and Connectivity Preserving Routing Algorithm, SHRP - Simple Hierarchical Routing Protocol, LEACH-TM - LEACH-Trust
Transmission Mechanism, LEACH-ME - LEACH-Mobile-Enhanced Protocol, LEACH-H - Hybrid Cluster Head Selection LEACH, DEECIC - Distributed Energy-Efficient Clustering with Improved Coverage, T-LEACH
- Threshold-LEACH, LEACH-SC - LEACH-Selective Cluster, W-LEACH - Weighted LEACH, MR-LEACH - Multi-hop Routing LEACH, BARC-Battery Aware Reliable Clustering, SEP-Stable Election Protocol, KMMDA-
K-Means Minimum Mean Distance Algorithm, RPCR-Regional Partitioned Clustering Routing Algorithm, OECM-Optimal Energy Consumption Model, IEEPB-Improved Energy-Efficient PEGASIS-Based Protocol,
MEPA-Message Passing Algorithm, TB-LEACH-TimeBased LEACH, LEACH-ER-LEACH-Ensuring Reliable Data Delivery, DE-LEACH-Differential Evolution, ID-LEACH-Identification Based LEACH, EECFP-Energy Efficient
Cluster Formation Protocol, PEACH-Power-Efficient and Adaptive Clustering Hierarchy Protocol, HGMR-Hierarchical Geographic Multicast Routing, CCS - Concentric Clustering Scheme, TTDD-Two-Tier Data
Dissemination, PANEL-Position-based Aggregator Node Election protocol, BCDCP-Base-station Controlled Dynamics Clustering Protocol, APTEEN - Adaptive Periodic Threshold-sensitive Energy Efficient Sensor
Network Protocol, CLUDDA - Clustered Diffusion with Dynamic Data Aggregation, CMATO - Cluster-Member-based fAult-Tolerant scheme, CHLBC - Cluster Head Load Balanced Clustering, FSCA - Fuzzy Self
Clustering Algorithm, CHEF - Cluster Head Election mechanism using Fuzzy logic, KMMDA - K-Means Like Minimum Mean Distance Algorithm, CES - Cluster-based Energy-efficient Scheme, DDCHS - Density
and Distance based Cluster Head Selection, MECH - Maximum Energy Cluster-head, EEDMC- Energy-Efficient Distributed Multi-level Clustering, MELEACH-L - More Energy-efficient LEACH for Large-scale
sensor networks, WST-LEACH - Weighted Spanning Tree for LEACH, CCRP - Cluster-Chain Routing Protocol, ACE-C - Algorithm of Cluster-head Election by Counting, ACE-L - ACE-Location, LCA - Linked
cluster algorithm, RCC- Random competition based clustering, EEHC- Energy Efficient Hierarchical Clustering, LEACH - Low Energy Adaptive Clustering Hierarchy, FLOC - Fast Local Clustering service, ACE
- Algorithm for Cluster Establishment, HEED - Hybrid Energy-Efficient Distributed Clustering, DWEHC - Distributed Weight-Based Energy-Efficient Hierarchical Clustering, MOCA - Multi-hop Overlapping
Clustering Algorithm, GAF - Geographical Adaptive Fidelity, DCA - Distributed Clustering Algorithm, TB-LEACH - Time-Based-LEACH, WCA - Weighted Clustering Algorithm, EECS-Energy Efficient Clustering
Scheme, TL-LEACH - Two-Level Hierarchy LEACH, PEGASIS-Power-Efficient GAthering in Sensor Information Systems, GROUP- GRid-clustering ROUting Protocol, ACE-C-Algorithm of Cluster-head Election by
Counting, ACE-L-Algorithm of Cluster-head Election by Location, EEUC- Energy-Efficient Uneven Clustering, LEACH-C LEACH-Centralized, MCR-Multi-hop Communication Routing.

6
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Fig. 1. Taxonomy of clustering in wireless sensor networks.

criteria has not been proposed. This hitherto unexplored area lack in-depth analysis, detailed explanation and fails to address
of investigation is the objective of this work. In this article, a clustering strategies based on unequal clustering, optimization
comprehensive, meticulous study of these aspects in classical, approaches and machine learning techniques.
machine learning, and optimization approaches based on energy Hence, several works related to classical methods [89–98],
efficiency and improved network lifetime are presented. More- Optimization approaches [99–115] and Machine Learning tech-
over, various challenges and research directions that help the niques [116–124] with respect to cluster formation and CH se-
researchers provide insight into the wireless sensor network field lection, routing protocols, reliability and security are compre-
are also analyzed. hensively discussed and analyzed with various parameters in
The rest of this paper is organized as follows: Section 2 dis- Section 5 and Unequal clustering in Section 6. Research findings,
cusses the related work and lists several survey papers for a challenges, open issues, and future directions are also addressed
period from 2006 to 2020. Section 3 describes the taxonomy of based on all these aspects. Moreover, the comprehensive solution
clustering in wireless sensor networks based on environment, is by providing a state-of-the-art classification of clustering for
characteristics, methods, process, objectives, and routing. Sec- wireless sensor networks based on different dimensions, such as
tion 4 depicts strategies such as classical methods, optimization the taxonomy of clustering, strategies, clustering aspects, chal-
approaches, and machine learning techniques that are used to lenges, open issues, and research findings, drive this survey dif-
study the various aspects of clustering. A detailed literature re- ferent from the existing other surveys and is considered as the
view based on cluster formation, cluster head selection, rout- novelty of this work.
ing, reliability, and security is provided in Section 5. Section 6 Literature studies [96–98,125–127] show that the clustering
provides an analysis of relevant literature, relying on unequal process began with Low-Energy Adaptive Clustering Hierarchy
clusters in wireless sensor networks. Section 7 addresses a few (LEACH) [89], a distributive approach for randomly selecting CH.
of the research findings, challenges, issues, future directions, and In WSNs, LEACH is still considered the basis for several other
summary. Finally, Section 8 discusses the conclusion. progressive clustering schemes. However, the limitation of LEACH
protocol is that the sensors transmit information to the CH only
2. Related work during their specified TDMA slot, and all nodes are assigned
within the transmission range of each other. There are numerous
The existing survey papers that focus on clustering for WSNs LEACH variants proposed in the literature [90–92] since these
is presented in Table 1. It analyzes the major contributions, var- schemes focus on CH selection and broadcast information to
ious schemes/categories, approaches, protocols/algorithms spec- the BS by employing single-hop transmission. The CH selection
ified, challenges/issues, proposed solutions and limitations. The process in clustering is essential since the appropriate selection
overall taxonomy of clustering in wireless sensor networks that of CHs reduces long-term energy consumption.
comprises of clustering environment, characteristics, methods, The existing LEACH clustering protocol is improved by im-
process, objectives and cluster routing are briefly analyzed in posing a threshold limit for the CH selection and switching the
Section 3 and presented in the form of diagram in Fig. 1. There energy level among the nodes simultaneously [90]. This improved
are several other survey papers in WSN that focus on lifetime protocol minimizes the energy consumption in a distributed way
optimization of multi-hop WSNs [65–67], WSNs lifetime maxi- and is best suited for the different applications that change the
mization [68] multi-path routing [69], quality of service [56,70– energy, nodes, and area. The limitation is that this routing proto-
73], node placement [74], energy consumption models [75], bio- col failed to address in a heterogeneous environment.
mimic optimization [76], distributed topology [77], congestion A distributed LEACH-based CH selection algorithm named
control [78,79], privacy [57,80–84], security [85], coverage [86], LEACH with Distance-based Thresholds (LEACH-DT) consider the
reliable clustering [87] and in-network aggregation [88]. There distances from the nodes to the sink node, which equalizes
are several overlapping surveys and investigations. Most of them the energy consumption between the sensors efficiently [91]. In
7
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

LEACH-DT, the sensor nodes are self-selected to become CHs with of EAMR is that instead of direct communication between the CH
different probabilities. Furthermore, this algorithm considers only and BS, this protocol utilizes inter-cluster communication using
direct communication from the CHs to the sink node and failed the relay nodes.
to consider the multi-hop communication. Multi-Level Route-aware Clustering (MLRC) algorithm is used
A sliding window is defined, and the predicted number of CHs to overcome the above limitation [98] since a considerable
are balanced in the LEACH protocol with the Sliding Window Dy- amount of energy loss arises during inter-cluster data transfer
namic number of Nodes (LEACH-SWDN) [92]. The sliding window because of the large distances between the relay nodes (CHs) and
dynamically adjusts the nodes’ interval during the threshold cal- the BS. A routing tree for inter-cluster transmission is constructed
culation to generate the random number. However, the limitation to provide connectivity between CHs. Thus every CH performs the
of LEACH-SWDN is that the CHs are always kept at the optimal role of relay that accepts and transfers the information to other
range and gets compressed with the decrease in the number of CHs. The tree structure acquires the advantage of optimal trans-
nodes. Moreover, the clustering enhances the network efficiency mission path information to the BS. However, MLRC optimizes
by various protocols, as discussed below. energy usage and maximizes the network lifetime, and it is a
A Hybrid Energy-Efficient Distributed clustering (HEED) pro- decentralized method in which every node communicates only
tocol chooses CHs by considering the metrics like the residual with its adjacent nodes.
energy and the degree of the node [93]. HEED’s primary objec- Genetic algorithm (GA) is one of the most common optimiza-
tives are the network lifetime enhancement, incurs low message tion techniques that are population-based, selecting CH by con-
overhead, attains uniform distribution of CH, and guarantees con- verging into the most viable method, and considering the most
nectivity of clustered networks. Moreover, HEED does not make appropriate CH node. The optimized CH selection is a primary
awareness about the node’s location and is distributed only based concern to attain energy efficiency in WSN. Genetic Algorithm-
on local information. The network is also viewed as a grid-like based Optimized Clustering (GAOC) protocol and Multiple data
structure due to the cluster range limitation, where routing tables Sinks based GAOC (MS-GAOC) are the two reactive protocols
or reactive routing schemes determine the optimal routes. that are used to accomplish this task [99]. To deal with the
A Flow-Balanced Routing (FBR) protocol attains energy ef- energy-hole challenge and reduce the distance between the nodes
ficiency by using four algorithms which are used for network and the BS, MS-GAOC is used. The limitation is that the sensor
clustering, construction of multi-hop backbone, flow-balanced nodes are not aware of the location and do not consider the
transmission, and re-routing [94]. In FBR, the formation of a clus- environmental factors.
ter is defined in terms of the overlapping degrees of sensors and Similarly, and Optimal Multi-hop Path Finding Method
is initiated only once during the network’s initialization. During (OMPFM) [100] identify the optimal multi-hop route between
the construction of the multi-hop backbone, the CHs are con- the CH and BS to reduce the energy utilization and to prolong
structed with the sink node at the top. A flow-balanced routing the lifetime of the network. To find the optimal path, GA is used
algorithm is used to distribute the power consumption of sensors by a fitness function. However, OMPFM does not consider the
uniformly, and the re-routing algorithm redefines the network multi-hop communication in the mobile WSNs.
topology when any topological change occurs. Furthermore, FBR’s Various schemes involve the GA process and the integration
limitation is that the complexity of network design in every round of fitness functions for CH selection. In a self-organizing network
is increased as the number of sensors rises, consuming available based on GA, [101], sensor nodes are placed at random and
resources. optimize the network to prolong the energy efficiency. Moreover,
Hierarchical Distributed Management Clustering (HDMC) pro- it is less effective in a rapid environment due to the overhead
tocol prolongs the network lifetime, and network coverage [95]. traffic from the BS. The issue of transmission overhead is resolved
Each node becomes a CH based on its coverage area, available by a Genetically Derived Secure cluster-based Data Aggregation
resources, and its activity history. Generally, there are two meth- (GDSDA) scheme [102], moreover ensures data security and mini-
ods used for cluster formation. Initially, the CH will be chosen, mizes energy consumption. However, the hot-spot problem arises
and then the member node gets joined in the first method. In the in the GDSDA scheme. The GA based load balanced clustering
second method, the cluster and the member nodes are initially algorithm [103] balances the equal and unequal load of the sensor
defined, and later CH will be chosen from one among the member nodes, thereby converges very fast in generating efficient results
nodes. In HDMC, the nodes in a region are compared, and CHs will but failed to consider CH selection in its approach.
be selected, which selects the clustering process from the BS and A Genetic Algorithm based, Self-Organizing Network Cluster-
aims to reach the maximum coverage threshold. The limitations ing (GASONeC) scheme [104] dynamically optimize the clusters
of HDMC are that the nodes do not support mobility and fail to and assign a surrogate node as CH. Furthermore, GASONeC failed
consider CH’s dynamic behavior. to analyze the performance in a heterogeneous environment. In
In a Low Energy Fixed Clustering Algorithm (LEFCA) [96], the hybrid Fuzzy-Genetic scheme [105], the fuzzy is used due to
through the entire lifetime of the network, the sensor node be- the uncertainty of input data and the elimination of encountering
comes the cluster member that remains in the same cluster, obstacles. Moreover, the optimal local solution was acquired in
thereby maximizes the lifetime of the network. In LEFCA, the CHs fuzzy logic, and the GA is used to deal with it. This scheme has
are determined by the BS, followed by the set-up phase. Data led only to some convergence between the time consumed and
transmission takes place in the steady-state phase, and if neces- the distance traveled.
sary, the changing mechanism of CH is performed in the CH deci- To solve the issue of scheduling addressed by the target cov-
sion phase. However, LEFCA selects the CH by choosing a random erage problem, a target-oriented GA-based algorithm [106] forms
cluster member instead of selecting the nearest neighbor’s re- cover sets consisting of sensors with multiple sensing ranges and
maining energy. Besides, it also decreases the energy dissipation directions that monitors all targets in the network and maxi-
significantly. The above limitation is resolved by an Energy-Aware mizes the lifetime of the network. The limitation is that it needs
Multi-hop Routing (EAMR) protocol, which improves the network to determine the target’s priority as per their vulnerability. A
energy’s dissipation using the relay nodes [97]. The relay nodes Gravitational Search Algorithm (GSA) finds the optimal CHs, bal-
broadcast the cluster information through inter-cluster commu- ances energy consumption, and maximizes stability and scala-
nications, thereby increasing the network’s scalability, and the bility [107]. GSA does not focus on the heterogeneity of sensor
re-distributor nodes will perform the new CH role. The limitation nodes. A Harmony Search Algorithm (HSA) [108], and Optics
8
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

inspired optimization (OIO) [109] solves the issues related to A LEACH-SF uses the Fuzzy C-Means (FCM) clustering algo-
clustering and routing in WSNs. For clustering, an HSA-based rithm to form balanced clusters. To enhance the network lifetime,
CH selection algorithm is used along with the objective function an optimized Sugeno fuzzy system determines appropriate clus-
and encoding scheme. A routing algorithm computes a near- ter heads [120]. The LEACH-SF was designed only for static sensor
optimized route by using mapping and a fitness function. How- nodes, and CHs communicates with the BS using single-hop rout-
ever, fault tolerance and delay are not taken into account for ing that failed to deal with large-scale networks. An energy-based
transmitting the information to the BS. Clustering with Fuzzified Updates (ECFU) algorithm [121] uses
A clustering algorithm based on Simulated Annealing (SA) and a machine learning technique to address the issues regarding
Particle Swarm Optimization (PSO) finds optimal CHs and pro- message overhead and the data transmission complexity. ECFU
longs the energy efficiency in WSNs [110]. However, only a min- contains different phases like Cluster Construction, Fuzzy based
imal number of performance parameters are analyzed. An Evolu- re-clustering, Machine Learning based Data Transmission, and
tionary Multipath Energy-Efficient Routing (EMEER) protocol uses Cluster Head Role Change. The cluster construction build clusters
the cuckoo search algorithm to determine the optimal CH [111] and are adjusted by the Fuzzy Inference System (FIS) in the Fuzzy
but failed to investigate the performance in terms of the alive
based re-clustering process. However, ECFU considers only static
node during communication. Another energy-restricted cluster-
sensor nodes. The CH preserves dynamic signal strength for every
ing strategy based on Particle Swarm Optimization [112] min-
node in a cluster that impacts its propagation distance.
imizes energy consumption but failed to use in heterogeneous
A Deep Learning-based Defense Mechanism (DLDM) [122] is
WSNs. While a Type-2 Mamdani Fuzzy Logic System (T2MFLS)
used for energy-efficient cluster-based protective measures along
handle uncertainties, and imprecise inputs, another Clustering
with deep learning to detect DoS attacks effectively. The CH
Routing protocol called CRT2FLACO based on the Type-2 Fuzzy
preserves dynamic signal strength for every node in a cluster that
Logic and Ant Colony Optimization (ACO) [113] select the CHs
and used for multi-hop routing. Moreover, the node density in- impacts its propagation distance. A multicast routing protocol
fluences the reliability of the WSN. that utilizes reinforcement learning called Feedback ROuting to
Another protocol based on the fuzzy logic known as a Fuzzy Multiple Sinks (FROMS) [123] discovers efficient multicast routes
logic based Unequal Clustering, and Ant Colony Optimization globally. FROMS fails to investigate its potential in clustering,
(ACO) based Routing, Hybrid (FUCARH) [114] eliminates the medium access, and neighborhood management. A Neuro-Fuzzy
energy-hole issue and prolongs the network lifetime. For efficient Rule-Based Cluster Formation and Routing Protocol [124] per-
inter-cluster routing, the ACO-based routing strategy is used. A forms reliable routing and maximizes the network lifetime. The
threshold principle is used to detect rapid environmental vari- deep learning approach is combined with the Neuro-Fuzzy In-
ations during the data transfer. Even though FUCARH operates ference System (NFIS) by considering large inputs to provide a
both proactively and reactively, it failed to focus on mobile single qualitative output, thereby used for the decision making
nodes. A Butterfly Optimization Algorithm (BOA) [115] finds an process. The BS performs the training process, and the rules are
optimal CH, and the optimal route is predicted by Ant Colony propagated to the sensor nodes. The limitation lies in the sensor
Optimization (ACO). Therefore, a combination of BOA and ACO nodes that all the nodes are not always secure. Therefore, trust
minimizes the overall consumption of energy and prolongs the or security-based mechanisms can be implemented to provide
lifetime of the network. However, the BOA algorithm re-initializes efficient and secure routing in WSNs.
the clustering process if the remaining energy of the CH is greater The following observations are drawn from the extensive lit-
than the threshold level. erature work: The improper formation of the cluster leads to the
To solve the hot-spot issue in WSN, two algorithms based on CH over-burden that causes extension of communication latency,
reinforcement learning are used. The first algorithm, Reinforce- consumes high CH energy, and deteriorates the sensor networks’
ment Learning Based Clustering Algorithm (RLBCA) form CHs, overall performance. Consequently, routing and load balancing
and the second algorithm, On-Demand Mobile Sink Traversal using the sensor nodes and the cluster heads are the critical issues
(ODMST), aggregate data from CHs by using mobile sink [116]. for clustering sensor nodes. Grid-based routing protocols can be
The advantage of these two algorithms is that they avoid ex- used to optimize routing efficiency, thus ensuring the enhanced
tra overhead on mobile sinks, thereby minimize energy con- performance of the network on the QoS parameters [128]. Despite
sumption. The limitation is that these algorithms can only be most of these works related to cluster-based schemes, almost
implemented in small-scale networks, with the static and ho-
all of the existing schemes used only distance-based clustering
mogeneous nature of nodes. Another clustering algorithm which
techniques. Moreover, the challenge in the designing of routing
was based on the fuzzy logic known as Distributed Fuzzy Logic-
algorithms is the decision making process.
Based Clustering Algorithm (DFLC), operates in a distributed man-
Most of the literature studies focus on the fuzzy rules which
ner [117]. Every node executes the fuzzy logic engine, thereby
make efficient decisions that provide an optimal value in the
inferencing the sensor node’s properties that best fit the root
route discovery process, rather than other intelligent-based tech-
node to detect rapid environmental variations during the data
niques. Also, it is observed that the energy-efficient schemes
transfer. Even though DFLC avoids the network crash caused by
energy depletion, it cannot be executed on mobile networks. for clustering do not provide optimal results in terms of QoS
The Shuffled Frog Leaping Algorithm, PS-SFLA determines the parameters. Concerning unequal clustering, most of the existing
fuzzy parameters [118] and comprises three phases: pre-process techniques aim to solve the energy-hole challenge in addition
clustering and routing. During pre-processing, all the input pa- to the role of CH selection. However, no definite scheme exists
rameters are extracted. In the second phase, SFLA optimizes and for data transfer among CHs despite cluster-based routing. The
selects the parameters with the corresponding effectiveness co- discerning facts about GA from the above literature are: (i) While
efficients. The third phase performs more precise tuning of the the use of GA has proposed numerous different methods for CH
fuzzy rules, thereby enhances the performance factors of WSNs in selection, neither strategy has efficiently been used to conserve
the context of the application. Also, a Fuzzy Shuffled Frog Leaping energy optimally. (ii) CH must be chosen such that multi-hop
Algorithm (FSFLA) [119], uses the SFLA by employing the Mam- communication, and the other factors are also taken into account.
dani fuzzy rule-base table. Moreover, to reduce the total energy (iii) GA based optimized CH selection in a heterogeneous environ-
consumption, the network levels that focus on the distance needs ment failed to use the node density parameter as well as other
to be adjusted, which is the limitation of PS-SFLA. energy-related parameters in its fitness function.
9
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Fig. 2. Cluster size (a) Equal (b) Unequal.

Fig. 3. Taxonomy of clustering process in WSN.

3. Taxonomy of clustering in wireless sensor networks minimize energy consumption [129]. In centralized clustering,
the cluster node or base station requires global information about
The taxonomy of cluster analysis based on various strategies the network to control the entire process. These approaches
such as environment, characteristics, methods, process, objec- systematize the nodes into clusters and assign CH for every
tives, and routing are shown in Fig. 1. Various attributes, such cluster. Centralized clustering limits the network performance
as the nature of nodes, deployment, mobility pattern, criteria, based on the transmission of reliable data, scalability and min-
performance metrics, the position of sink nodes, are used to imizes network efficiency [45]. In Hybrid clustering, the features
analyze and categorize the examined clustering strategies. of distributed and centralized clustering are combined. In such
a scenario, distributed schemes can be utilized for cluster head
3.1. Clustering environment coordination, and centralized schemes can be used for cluster
heads to form individual clusters [31].
Based on the clustering environment, WSNs can be distributed,
centralized, or hybrid in nature. In distributed clustering, the sen- 3.2. Cluster characteristics
sor nodes work together to build routes or join a formed cluster
without global information like geographical position and the Clustering characteristics correlate with the cluster’s inter-
energy level of clusters. Each sensor node can accomplish its tasks nal structure having related attributes such as cluster count,
by executing its algorithm and determine of becoming a cluster cluster size, and cluster communication. The two categories of
head or not. WSNs utilize distributed clustering approaches to cluster counts, viz. fixed and variable, centered on clustering
10
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Fig. 4. Taxonomy of routing protocols.

schemes’ variability. The cluster size is of two classes, viz. equal 3.4.1. Cluster formation
and unequal, having the same and different sizes of the clusters This phase reduces the workload of CHs near the sink node
respectively, as shown in Fig. 2. The unequal clustering avoids the by having fewer members of clusters. Each member node joins
energy-hole problem and achieves a more uniform consumption the nearest cluster head concerning the Received Signal Strength
of energy in the network [130,131]. The communication between Indication (RSSI). Based on the local data received, each node is
sensor nodes in clustering takes two forms: Intra and inter-cluster qualified to become a cluster head by computing its radius [98,
communication. In intra-cluster communication, communication 139].
is between the sensor nodes and the CH. However, in inter-cluster
communication, the communication is between the CHs and the 3.4.2. Cluster head selection
BS or sink node [132]. As CH is primarily used for aggregating and distributing the
information to the sink node, CH selection plays a very crucial
3.3. Clustering methods part in optimizing energy consumption [140]. However, the ap-
propriate CH selection enhances the WSN lifetime [141]. In a
The clustering process can be performed with various meth- cluster-based network, CH near the base station quickly exhausts
ods such as hierarchical, partition-based, spectral, grid-based, its energy, which leads to hotspot problems. Unequal cluster-
and density-based clustering. In hierarchical clustering, a tree ing algorithms are utilized [142] to overcome this problem. The
structure is used, which comprises of top-down or bottom-up appropriate CH selection focus on different criteria such as CH
approaches. It is highly flexible and suitable for point-to-point selection, communication, cluster head mobility, and role of CH,
communication [15]. In partition-based clustering, the clusters as shown in Fig. 3. The CH selection criteria focus on different
are partitioned into different divisions, where each division rep- metrics such as distance between nodes and cluster head, clus-
resents a cluster, which is appropriate for a low number of ter head and sink and nodes and sink; remaining energy; RSSI;
nodes [133]. The spectral type of clustering utilizes a similar- cluster density; node degree; position metrics, and node weight.
ity matrix. Moreover, it is convenient for a small number of The data transfer between the cluster member nodes and
nodes, used in image processing, and requires low processing CH and, in turn, to the base station decides the CH selection’s
time [134]. In grid-based clustering, the region is partitioned into communication criteria. The nodes or CHs convey their infor-
more regions, and based on certain features, sensor nodes are mation directly to their corresponding CHs or sink nodes in a
added to these regions. It requires low processing time, low com- single-hop communication or intermediate node in multi-hop
putational complexity, and shows high-performance data distri- communication.
butions [135]. In density-based clustering, the higher density is CH mobility is another critical factor in evaluating the cluster
the area containing more number of clusters. The higher density head selection technique in WSN since a large number of WSN
regions are separated from the regions with lower density. It is applications are mobile. When the CH is fixed, stable clusters
used in dynamic clustering and shows excellent performance in are formed and facilitate inter-cluster and intra-cluster network
harsh environments [136–138]. management. On the other hand, when the CH becomes mobile,
the sensors’ membership keeps on changing, and the clusters
3.4. Clustering process must be continuously monitored. Sometimes, CHs move to lim-
ited distances for better efficiency of the network by repositioning
The clustering process in WSN maximizes the network life- itself.
time, and it mainly contains two phases, cluster formation, and The cluster head plays the role of relay for the generated traffic
cluster head selection, as shown in Fig. 3. or perform data aggregation of sensor nodes. It can also act as
11
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Fig. 5. Strategies used for various aspects in clustering process.

a sink node that acquires actions based on the intended targets. node. Data aggregation is performed in every node in the tree-
Besides, the role of the cluster head performs energy efficiency based topology. Grid-based routing topology divides the network
and load balancing. into different grids. Here, the routing operation is performed
without the need for a routing table. Area-based routing topology
3.5. Clustering objectives is considered an up-to-date topology, in which some sensor nodes
designate a specific area and finds application in mobile WSNs. In
Clustering objectives are used to facilitate the needs of vari- Cluster-based routing topology, cluster formation is performed by
ous applications used in wireless sensor networks. The different collecting sensor nodes, and a CH is elected from each cluster that
objectives used in network clustering are: Data Aggregation [143– serves as an interface between the members of the cluster and the
145], Load balancing [93,146], Fault Tolerance [147], Network BS.
lifetime [148,149], Energy Efficiency [150,151], Scalability [152],
Reliability [153], QoS support [154] and Security [155]. 3.6.3. Path establishment
It identifies routes from the source to the destination. There
3.6. Cluster routing are various categories of path establishment, such as proac-
tive, reactive, and hybrid-based mechanisms. The proactive-based
A routing protocol initiates cluster routing in WSNs for situ- routing is also known as table-driven routing, in which each
ations where a source node cannot transfer its data directly to node chooses an optimal path and transmits its relevant infor-
the destination. Routing is a very challenging issue because of mation to the routing table. Reactive-based routing does not
the following reasons: Firstly, it cannot create a global address- have a predetermined path for route establishment and causes
ing scheme; therefore, classical IP-based protocols could not be delays during route-discovery. The nodes immediately react to
used to deploy sensor nodes. Secondly, the generated data traffic the sudden changes in the sensing field. In hybrid routing, nodes
will suffer from redundancy because of the data generated by transmit their data periodically and respond to the rapid changes
multiple sensor nodes. Thirdly, resource management is exploited in sensing events.
carefully because sensor nodes are inhibited by the transmission
power, processing capacity, storage, and on-board energy. The 3.6.4. Protocol operation
taxonomy of routing protocols is shown in Fig. 4 The routing protocols are categorized into a negotiation,
query-based, multi-path, coherent, and QoS-based routing based
3.6.1. Network structure on the protocol operation [160]. Negotiation based routing takes
The routing protocols are categorized into flat, hierarchical, place before the real transmission of data. The descriptors are
and location-based routing based on the structure of the network. used to negotiate between the sensor nodes to reduce redundant
In flat routing, sensing and routing tasks are performed together information. Query-based routing depends on the queries from
due to all nodes’ similar functionality. In hierarchical routing, to the destination node, responded by the source node. Multipath-
achieve scalability and energy efficiency, the network is subdi- based routing constructs multi-route from source to destination
vided into clusters. In location-based routing, the routing path is to enhance the network performance. Coherent routing and non-
determined using node location information [156,157]. coherent routing are the methods used to minimize energy con-
sumption during data processing. QoS-based routing guarantees
3.6.2. Routing topology the quality of service parameters, such as reliability, delay, and
According to different topologies, routing can be categorized bandwidth of the data.
into a chain, tree, grid, area, and cluster-based routing proto-
cols [135,158]. In chain-based routing topology, chains are used to 4. Strategies used in various aspects of clustering process
construct nodes for data transmission. A CH is elected from each
cluster, which aggregates data, similar to a sink node. In turn, it The strategies such as classical methods, optimization ap-
is classified into data-centric and geographic-based protocols. In proaches, and machine learning techniques are used for various
data-centric routing protocols, the absence of global identifiers aspects of the clustering process, as shown in Fig. 5. The various
makes it challenging to select sensor nodes to interrogate. In aspects include cluster formation, cluster head selection, routing
contrast, the geographic protocol uses location information to protocols, reliability, security, and unequal clustering.
devise an efficient search that identifies the destination’s route.
The geographic protocol is very suitable for sensor networks 4.1. Classical methods
because of the data aggregation, which minimizes the number
of transmissions towards the base station by eliminating redun- Classical methods use the statistical technique for determining
dancy among the packets. Besides, it is used for large multi-hop the number of clusters. Besides, classical methods are used to ex-
wireless networks [159]. change the role of CH that calculates the residual energy of nodes
In tree-based routing, nodes are constructed in a tree-based or energy consumption based on the threshold, which triggers
structure. The information is transmitted to their parent nodes the network to re-cluster [161]. Classical methods focus on the
from the leaf nodes, which transfers the obtained data to the root way to select cluster heads in cluster-based routing protocols. The
12
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Fig. 6. Taxonomy of optimization approaches in clustering process of WSN.

cluster heads are randomly selected, which causes an irregular rely on the network’s size and topology. The two-level clustering
flow of traffic between them. Although the selection process is scheme results in a low range of transmission, and fewer nodes
suitable for WSNs applications, they have to overcome issues like are needed to communicate long distances, which minimizes
scalability, connectivity, coverage, and load balancing [127]. In the overall consumption of energy. Reducing the number of CHs
LEACH [89] protocol, each node has fair chances to become a CH, provides the overall network efficient coverage and decreases the
but it cannot be a CH in successive rounds, thereby managing the overlap of the cluster, thereby increasing the effectiveness of the
tasks among CHs. MECH considers the energy and the number algorithms functioning at the CHs level.
of hops during data transfer to the BS. In contrast, TL-LEACH
minimizes the data size due to the two aggregation phases in 4.1.3. Limitations
the CHs. T-LEACH does not trade-off CHs frequently concerning Most of the classical methods like LEACH perform single-hop
the predefined threshold. FBR [94] uses overlapping parameters inter-cluster communication that is inadequate for a large-scale
and avoids the re-clustering of nodes. PECRP considers the energy network environment, leads too much energy consumption. The
and distance to determine the best CH as well as considers the various issues related to coverage, energy holes, and hot spot
hot spot issue. EEHC [162] uses the heterogeneous nodes together arise because many nodes with relatively low energy act as CHs
with the distinct levels in three tiers. DS-LEACH uses the density for similar rounds, while higher energy nodes die prematurely.
of sensors to distribute the CHs in the network uniformly. MERA HEED results in a large overhead because the development of
uses the chain communication mechanism between each node clusters needs several evaluations. Determining the number of in-
stances for cluster creation while dealing with the cost of energy
and the closest neighbor within and between clusters. LCRPOCH
utilization and connectivity is challenging. A large-scale environ-
uses overlapping CHs between clusters on the boundary that en-
ment generates a large amount of energy consumption because
able multi-hop transmission of information. LEFCA selects the CH
of the inter-cluster routing from the source to the BS. For nodes
by the distributed method and uses the fixed cluster mechanism.
with varying initial energy levels, CH selection without energy
HEER [163] aggregates information and communicates it through
concerns does not guarantee effective load balancing.
a Hamiltonian path to the BS. The network is initially partitioned
in the Improving LEFCA algorithm [96], and the CHs are defined 4.2. Optimization approaches
only once by the base station.
The Optimization technique is used to resolve NP-hard prob-
4.1.1. Applications lems that are unable to be solved within a specified time by
Classical methods are better suited for event-detecting WSN, using traditional methods. Optimization can be of heuristic or
periodic reporting scenarios, high mobility scenarios [164], peri- meta-heuristic. Heuristic approaches work on one problem at a
odical data gathering applications [165], both proactive and reac- time, and they are greedy, which means that the solutions get
tive applications. It facilitates asynchronous sensor network ap- trapped in a local minimum. Also, the heuristic uses trial and
plications where BSs obtain the node interpretations and choose error to find solutions for complex problems. Meta-heuristic work
aggregators for efficient and persistent data storage applica- on the set of problems at a time and finds a global minimum.
tions [166]. Meta-heuristics may not provide us with the best solution, but
gives an optimal solution most of the time [175]. Meta-heuristic
4.1.2. Advantages approaches are combined with clustering techniques to attain an
It is impossible to select a node that acts as a CH in a par- energy-efficient network by identifying optimal solutions. Several
ticular round as the CH again so that every node can share the researchers use various meta-heuristics techniques like evolu-
same load; cluster members prevent additional dissipation of tionary algorithms (EAs), approximation algorithms, and swarm
energy following their allotted time frames. Using node’s location intelligence (SI) based approaches. The taxonomy of optimization
information, the classical protocol establishes balanced cluster algorithms used in the clustering process of WSNs is shown in
sizes and enhances intra-cluster communication, and does not Fig. 6.
13
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Table 2
Comparison between optimization approaches in WSN.
Approach Initially proposed Year Mimic Operator Feature Selection of CH Limitations
Genetic algorithm Holland et al. [167] 1992 Genes, Chromosome Useful for searching Based on more energy Management of mobile
from parents to optimal solution and less distance nodes are difficult
offspring
Ant colony optimization Dorigo et al. 1977 Ant, Pheromone Finds optimal path Based on residual Not good in terms of
[168] in seeking food source energy energy consumption
Firefly algorithm Yang [169] 2010 Fireflies, Distance Useful in Based on distance Discovery of impaired
in seeking preys on multi-objective path and transmission
basis of flashlight problems of data
Bacterial foraging Passino [170] 2002 Bacteria, in seeking Amount of chemotaxis Used in solving Based on the direction Not suitable for large
optimization nutrients optimization problems vector geographical area
Particle swarm Eberhart and Kennedy 1995 Birds, Position and velocity Finds optimal solution Based on high energy Increased network
optimization [171,172] in seeking food source on the basis of energy node overhead
consumption
Artificial bee colony Karaboga [173] 2005 Honeybees, in seeking Nectar amount Effective for low Based on intra-cluster Slow data-transmission
food source convergence problems communication efficiency
Grasshopper optimization Saremi et al. 2017 Grasshopper swarms, Position vector Able to solve real Based on the distance Requires all search
algorithm [174] in seeking food source problems with agents in defining the
unknown search spaces next position

4.2.1. Evolutionary algorithm the goodness of particles produced by using an objective function
The Genetic Algorithm (GA) is an evolutionary algorithm used that takes into account three goals [200]. The first goal tends
mainly for clustering and routing schemes in WSNs. GA is used to reduce the distance between CHs and the BS. The second
to extend the lifetime of cluster heads in the clustering pro- goal significantly reduces the relay nodes, and the third goal
cess, which prolongs the network [176] and improves the sys- minimizes the load of the network’s relay node. GSA, along with
tem efficiency [177,178]. A hybrid method based on GA and the PSO, can also be used in the CH selection process, enhancing
bee colony algorithm enhances the clustering process, improves stability and scalability and balancing energy consumption [201].
the optimal route through nodes [179] and improves the qual- A hybrid meta-heuristic approach using the Improved Ge-
ity of service efficiently [180,181]. The energy consumption of netic Algorithm (IGA) and Binary Ant Colony Algorithm (BACA)
data is also minimized in each round by reducing the total dis- ensure optimal coverage, minimizes the sensing of redundant
tance [182]. Although genetic algorithms solve non-continuous, information, and optimizes the multi-objective function by de-
multi-dimensional, non-differential, and non-parametrical prob- termining the minimum number of sensors [202]. Artificial Bee
lems, it has some disadvantages [76]. Some of them are: longer Colony (ABC) prolongs the network’s performance in WSNs and
running times cannot handle a large population and cannot as- achieves better exploitation and exploration rate at the time of
sure constant optimization response times. cluster head selection [203]. Moreover, an improved ABC (iABC)
approach obtains optimal CHs in WSNs [204] and a Bio-inspired
4.2.2. Swarm intelligence based clustering approaches Energy-Efficient Clustering Protocol (Bee-Cup) finds the shortest
Swarm Intelligence (SI) is another field that designs algo- path for routing in WSNs [205]. Fractional Grasshopper Opti-
rithms by the inspired behavior of animals or insect societies. mization Algorithm (Fractional-GOA) technique minimizes en-
Some of them include Ant Colony Optimization, Firefly Algorithm, ergy consumption by adapting the sleep/awake scheduling of
Artificial Bee Colony, Particle Swarm Optimization, Grasshopper the nodes [206]. The solution that yields the maximum value of
Optimization, and Bacterial Foraging Optimization. fitness is considered the optimal solution and is considered for
Ant Colony Optimization (ACO) approach helps to determine sensor activation in Fractional-GOA.
a safe and near-optimal route to the destination. ACO-based
routing scheme mitigates the energy usage and extends network 4.2.3. Approximation algorithms based on load balanced clustering
lifetime by updating the pheromone operator and equally divides In clustering, due to data processing, aggregation, and data
energy among all sensor nodes [183]. In addition, ACO-based communication to the sink node, the cluster heads utilize more
approaches identifies the shortest route [157], selects the cluster energy. Load balancing, therefore, plays a vital role in the func-
head [184,185], determine an efficient path [186,187], improves tioning of WSN. The two schemes used for load-balanced clus-
the selection process of the CH [188–190], find multiple routes tering based on approximation algorithms are the Load-Balanced
and used to optimize the search process in terms of exploration Clustering Algorithm (LBCA) and the Greedy Load Balanced Clus-
and exploration degree [191]. tering Algorithm (GLBCA). In LBCA, all sensor nodes are as-
Firefly algorithm (FA) is a recent technique inspired by ran- sumed to have equal traffic load. To form clusters, LBCA utilizes
domized solutions described as fireflies, and brightness is corre- a Breadth-First Search (BFS) scheme for every node to discover
lated to the fitness values. It is used with the harmony search the least CH by calculating the time complexity function. This
algorithm in WSN for CH selection and cluster formation, which scheme is not applicable in large WSN because the construction
in turn minimizes energy consumption [192] and shortens the of BFS for every sensor node requires large memory. Hence, an-
length of the route by using a technique named Mobile Data other approximation algorithm, known as GLBCA, is used, where
Transporter (MDT) [193]. Bacterial Foraging Optimization (BFO) is sensor nodes use unequal loads. GLBCA uses a bipartite graph of
one of the Swarm optimization approaches used to enhance the the nodes to form clusters by calculating the cardinality func-
lifespan of the network in WSNs. It can also be used for mobile tion [207]. The comparison between different optimization ap-
sensing operation, and routing in WSN [194]. proaches used in WSNs, in terms of mimic behavior, operator,
Particle Swarm Optimization (PSO) is an approach in which feature, criteria for selection of cluster head, and limitations, are
‘bird’ represents a unique solution and is also called a particle. given in Table 2.
This particle contains a fitness value that evaluates the quality
of the solution [195]. For the clustering and routing process, a 4.2.4. Applications
Binary PSO (BPSO) algorithm is used [196,197]. The PSO algo- Swarm Intelligence is suitable for applications where changes
rithm [198] and a two-tier PSO optimally selects the cluster head in mobility and topology are of utmost concern. Evolutionary
from sensor nodes by using a particle encoding scheme, and the Algorithms can be utilized for some centralized problems like
fitness value [199]. A PSO based routing algorithm also analyzes scheduling, localization, sensor fusion, energy-aware routing, and
14
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Fig. 7. Taxonomy of machine learning techniques in WSN.

clustering. Evolution Strategies are applied to continuous pa- achieve scalability, prolong the network lifetime, and reduced
rameter estimation problems associated with laboratory studies. latency during data collection. Load-Balanced Clustering increases
The GAs handle continuous or integer optimization problems, a the reliability of the system and facilitates connectivity between
chromosome offers the best solution to the specific issue in the the different nodes in WSN.
GA application, and fitness function defines the chromosome reli-
ability based on various needs. Also, to address clustering, classi- 4.2.6. Limitations
fication, and configuration management, GA based algorithm is Evolutionary Algorithms are not flexible when there is an
applied. BFO is used in many studies, such as quantization of environmental change. They need high memory and process-
color images, face recognition, address engineering-related issues, ing requirements [209]. In WSNs, GA is feasible for clustering
and MANETs. ABC technique is used in WSNs, mechanical and only there are pre-deployed clustering schemes and most of
civil engineering, image processing, software, and control engi- them are proposed for homogeneous networks. There are sev-
neering. Firefly algorithm is suited to different applications that eral limitations in ACO algorithms, such as the performance is
include flow shop and job shop scheduling, quadratic assignment highly dependent on previous cycles. Therefore, ACO algorithm
problem, and path planning for a crewless aerial vehicle. An involves intense overhead communication to handle the routes
approximation algorithm focuses on resolving the linear pro- independently. The limitations of the ABC algorithm are the low
gramming problem [208] and uses the data acquired from the convergence speed and can be quickly trapped into the local
solution for assigning jobs to machines. optimum. According to social insect capability, the forager will
stop hiring new swarms if the reliability of the path becomes
4.2.5. Advantages very poor. The firefly algorithm is a centralized method in which
Ant colony optimization is used for routing purposes and all the clustering operations are performed at the BS. Clustering
has been considered to be an effective and scalable algorithm. algorithms consider the load balancing only during the processing
The ACO-based routing protocols select the route as per the and communication of sensor nodes.
probability of enabling nodes to maintain good connectivity to
the BS. ACO is flexible, secure, provides routing, data aggre- 4.3. Machine learning techniques
gation, prevent network traffic, minimize energy consumption,
and enable multi-path information transfer in WSNs. The GA Machine Learning (ML) is a technique that learns from experi-
algorithm effectively analyzes the search space by simultaneous ence or study. The advantage of machine learning is its ability to
fitness evaluation function and merging partial solutions through deliver generalized solutions. Due to its interdisciplinary nature,
a crossover. Genetic algorithms analyze and manipulate a range machine learning plays an essential role in different fields like
of possible solutions. GA retains a search threshold and resolves computing, medicine, and engineering. Recent advancements in
issues with multi-criteria optimization. The more specific ad- machine learning aid to solve various issues in WSNs. In WSN, it is
vantage of GA is its capability to convey rule-based, practical not a simple process without ML to access vast amounts of sensor
solutions to several problems with ML strategies, and pattern data and obtain useful information from the data collected. ML
recognition. Particle swarm optimization is often used to search increases the efficiency and cost-effectiveness of computational
for optimal inter-cluster routing paths to prolong the network’s processes and analyzes complex data automatically. It is broadly
lifetime. BFO approach attains better performance than other categorized as supervised, unsupervised, and reinforced learning.
methods that are bio-inspired and traditional. Moreover, BFO is The taxonomy of ML techniques in WSN is shown in Fig. 7.
computationally useful in solving complex numerical problems.
ABC algorithm is developed to optimize multivariable and mul- 4.3.1. Supervised learning
timodal continuous functions. The firefly algorithm discovers an In Supervised learning, a collection of inputs and outputs
optimal global solution with quick convergence and under multi- are provided, and the relationship between them can be found
parameter evaluation function. Distributed load-balanced clusters during the training process. The main aim of supervised learning
15
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Table 3
Machine learning techniques to solve various issues in WSN.
ML Tech. Con. Cov. ACP AD Rou. EH Syn. TT DA Loc. RCO FD
Supervised learning
Regression Y [210] N N N N Y [211] N N Y [212] Y [213] N N
Decision tree Y [214] N N Y [215] N N N N Y [216] N N N
Random forest N N N N N N N N N N N N
ANN N N Y [217] N Y [218] N N N N Y [219] N Y [220]
Deep Learning N N N Y [221] Y [222] N N N N N N Y [221]
SVM Y [214] N Y [223] N N N N N N Y [224] N Y [225]
Bayesian N N N Y [226] Y [227] N Y [228] Y [229] Y [230] Y [231] N Y [232]
K-NN N N N Y [233] N N N N Y [234] N N N
Unsupervised learning
K-means N N N N N N N N N N N N
Hierarchical N N N N N N Y [235] N Y [236] N N N
Fuzzy C-means Y [237] N N N N N N N N Y [238] N N
SVD N N N N N N N N N N N N
PCA N N N N N N N N Y [239] N N N
ICA N N N N N N N N N N N N

ML Tech.- Machine Learning Techniques, ANN - Artificial Neural Network, SVM - Support Vector Machine, k-NN - k-Nearest Neighbor, SVD - Singular Value Decomposition, PCA - Principal Component
Analysis, ICA - Independent Component Analysis, Y -Yes, N - No, Con.-Connectivity, Cov.-Coverage, Rou.- Routing ACP-Avoids Congestion Problem, AD - Anomaly Detection, EH - Energy Harvesting, Syn. -
Synchronization, TT - Target Tracking, DA - Data Aggregation, Loc.- Localization, RCO - Reduces Communication Overhead, FD - Fault Detection.

algorithms is to forecast objective output and create a model 4.3.4. Applications

that signifies the relation and dependency links among the in- Fuzzy Logic is well adapted for defining and resolving complex
put features. Supervised learning algorithms are categorized into multi-objective tasks such as clustering, scheduling, and routing.
regression and classification. Regression estimates precision per- Fuzzy Logic (FL) is a discipline of mathematics that conveys
formance concerning minimum errors. Classification is catego- approximate human reasoning. Fuzzy Logic is used for pattern
rized into (i) Logic-based algorithms that use Decision Tree (DT) recognition, digital image processing, and various control systems
and Random Forest (RF) techniques (ii) Perceptron-based algo- like power systems, elevators, and home appliances. Fuzzy-c-
rithms that uses Neural Networks (NNs)/Artificial Neural Net- means clustering strategy is used in different fields like pattern
works (ANNs), Deep Learning (DL), and Gradient Descent (GD) identification, bioinformatics, and business intelligence. The cen-
techniques. (iii) Statistical Learning algorithms that use Support tralized issues such as fusion of sensors, clustering, data mining,
Vector Machine (SVM) and Bayesian techniques (iv) Instance- routing, and scheduling can be solved by using Neural Networks.
based algorithms that use k-Nearest Neighbor (k-NN) techniques Optimal sensor nodes, sink location, and localization are addi-
[240]. tional feasible application areas for neural networks. Many real-
time WSN applications use ANN due to its higher computational
4.3.2. Unsupervised learning specifications. In WSNs, reinforcement learning is extensively
In unsupervised learning, even the created model extracts the used for distributed issues such as medium access control, service
relationship, and there is no output related to inputs. This learn- positioning, and routing. The application of PCA minimizes the
ing approach classifies similar patterns into clusters. Unsuper- data complexity at either the sensor’s level or the level of the
vised learning overcome issues in WSNs like routing, data aggre- CH. ICA examines the data that was collected from different
gation, connectivity problems, and anomaly detection. It is further application areas, such as social networking and digital images.
divided into clustering and dimensionality reduction. Clustering Bayes Classifiers are used to identify outliers. Semi-supervised
techniques include Fuzzy-C-Means, K-Means, Hierarchical-based learning is suitable for many real-time applications, such as nat-
and K-Medoids algorithms, Dimensionality reduction includes ural language processing, filtration of spam, video surveillance,
various techniques like Singular Value Decomposition (SVD), Prin- and protein sequences classification. Thus, machine learning has
cipal Component Analysis (PCA) and Independent Component a wide variety of WSN applications, such as efficient routing,
Analysis (ICA). reducing overhead, and delay-aware communication.

4.3.3. Reinforcement learning 4.3.5. Advantages

The algorithm of reinforcement learning schemes continu- The membership function in fuzzy logic can be structured
ously learns with the interaction between the environments and dynamically to depict an objective’s behavior, based on different
gathers information. It determines the optimal result, which helps needs. To simultaneously accomplish several objectives, fuzzy
to maximize performance. Q-learning is indeed one of the rein- logic is ideal for implementing clustering heuristics and rout-
forcement learning approaches used in WSN. ing optimization. The critical advantage of Decision Tree (DT)
The Machine Learning algorithms are used in WSNs due to is reducing uncertainty in making decisions and facilitates sys-
the following facts: (i) Optimal sensor nodes in the target area tematic research. Random Forest (RF) classifier is the preferred
coverage problem can be obtained easily (ii) Improves the WSN technique for classifying hyperspectral data. ANN very quickly
performance by energy-harvesting and long-lasting maintenance categorizes non-linear and challenging data sets. Deep Learning
(iii) Improves the network efficiency by segregating the faulty strategies deal with techniques of data processing with layered
nodes from normal nodes (iv) Reduces the data transmission structures. Support Vector Machine (SVM) determines an optimal
overhead at the sensor or the CH level (v) Accurate localization hyperplane to classify the data. The K-Nearest Neighbor method
of the sensor node can be found out quickly, and even their loca- identifies the possible missing values and also minimizes the
tions are changed due to internal and external factors. Machine dimensionality. Extraction of information from high-level fea-
learning techniques are used to solve numerous challenges in tures and trained to achieve multiple objectives are the main
WSN such as connectivity, coverage, congestion problem, MAC advantages of deep learning. To minimize dimensionality, the
protocol, anomaly detection, routing, energy harvesting, synchro- PCA integrates all the data and eliminates the lowest priority
nization, target tracking, data aggregation, localization, communi- information. Regression predicts precise results with minimal
cation overhead, fault detection, mobile sink path selection, and errors. WSNs use the semi-supervised learning method to resolve
data quality estimation is given in Table 3. localization and fault detection problems.
16
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Fig. 8. Clustering aspects based on various approaches.

4.3.6. Limitations
Despite many advantages and various applications, there are
limitations of using ML in WSNs: The FL algorithm produces
a non-optimal solution, and topology variations need to be re-
learned by using fuzzy rules; The primary drawback of rout-
ing algorithms focused on reinforcement learning is the limited
awareness of future information. Therefore for highly dynamic
situations, the algorithms are not adequate as they require a Fig. 9. Formation and operation of clusters.
long time to discover optimal route [241]. ML algorithms do not
make reliable predictions automatically since they need to learn
from historical data; The performance is based on the historical
information; if the size of the information too large, then the
energy used for data processing is also considerable.

5. Clustering aspects based on various approaches

The different clustering aspects of WSNs based on classical

Fig. 10. Reservation phase in LBR.
methods, optimization approaches, and machine learning tech-
niques are shown in Fig. 8.

5.1. Cluster formation and cluster head selection In the reservation phase, a matrix is formed in each node
known as Reservation Matrix (R-matrix) in which assigns one (1)
5.1.1. Classical methods when it acts as a CH and zero (0) when it acts as a general node.
Sensor nodes are organized into clusters controlled by their Then a total-matrix is formed, indicating which node to become
CH in Low Energy Adaptive Clustering Hierarchy (LEACH) mech- a CH in each round. The clustering phase is followed by the end
anism. The function of CH in each cluster is shared between of the reservation phase.
cluster nodes to save path costs. Also, LEACH uses CH’s role to Moreover, the CH selection is based on various metrics such
distribute the load energy evenly [89,242]. In another protocol as highest trust value [246], two-level thresholds to detect and
named node ranked-LEACH, the CHs of each cluster are selected identify the malfunctioning of the cluster head [247], high en-
based on the node with the highest rank function, which is ergy [93], weight value [248], residual energy and RSSI of sensor
represented in Eq. (1). It overcomes the random election of CH, nodes [249,250]. The cluster formation and CH selection using
which improves the network lifetime [243]. classical methods according to various parameters are given in
j ji
Table 4.
∑ 1/dout
NR(ni ) = PO(ni ) ∗ α ∗ NR(nj ) ∑ jk
+ (1 − α ) (1)
5.1.2. Optimization approaches
0 k∈NH dout
Optimization approaches can be heuristic or meta-heuristic.
where, NR(ni ) represents node i rank score, PO(ni ) is the energy Heuristic approaches use trial and error to find solutions for
of node i, d is the distance, NH represents sets of neighbors k, and complex problems. The CH selection is performed on a ran-
α is the damping factor. dom basis in heuristic approaches. Due to the random selection
A LEACH-ERE (Expected Residual Energy) protocol focuses of nodes, the CH selection becomes unevenly distributed. Also,
on the appropriate selection of CHs to maximize the WSN life- the network’s performance gets affected because of homoge-
time [244]. The clusters are formed during the set-up stage. The neous based sensor networks [251]. Hence, to keep the system
steady-state stage is partitioned into frames that broadcast data stable, and to select the CH in a deterministic fashion, an Energy-
to the CH by the nodes during the allotted transmission slot, as Efficient Heterogeneous Clustered (EEHC) protocol [162] and a
shown in Fig. 9. The number of frames is obtained by Eq. (2). Stable Energy-Efficient Clustering Protocol (SEECP) are used [252].
tss Phase A Heuristic Algorithm for Clustering Hierarchy (HACH) uses a
Nframe = (2) heuristic crossover operator, which generates an improved solu-
n∗tslot + tCHtoBS
tion that coordinates energy consumption in WSN and maximizes
where, tss Phase represents the operation time to be a cluster
the distribution of CHs [253].
head, tslot is the slotted time, and tCHtoBS is the time required for
Meta-heuristic works on a set of problems at a time and finds a
the transmission from CH to the BS. Another mechanism known
global minimum. It may not provide us with the best solution, but
as LEACH Based on Reservation (LBR) determines cluster heads
it gives an optimal solution most of the time. These approaches
and highlights the reservation phase, which helps to reduce the
are combined with clustering techniques to achieve an energy-
control messages that are used in the clustering phase [245], as
efficient network in terms of selecting CHs by finding optimal
shown in Fig. 10.
17
 J. Amutha, S. Sharma and S.K. Sharma
Table 4
Cluster formation and Cluster Head (CH) selection using classical methods.
Ref. CH selection Clustering Cluster CH selection Role of CH Distance CH mobility Feature Experiment Simulation Deployment Nature & Area of Location of Future
method algorithm formation criteria between (or) environment of nodes number of interest (AOI) BS/sink node direction/Remarks
nodes Simulation performance nodes
parameters

[243] LEACH Node Set-up phase Residual Distribution Node rank Static Selection of CH is NS2 Throughput, Uniform Static, Square, Not specified Node Rank (NR)
Ranked– energy, RSSI of energy algorithm based on the weight End-to-end delay, distribution 100 nodes 600 × 600 affects every node
LEACH and number load among parameter Average energy ratio, m2
of links sensors Packet delivery ratio,
Number of alive nodes

[245] LEACH LBR Four-step Total-Matrix, Form TDMA Based on Dynamic Reservation phase MATLAB Number of live nodes, Random Static, Square, 133.5 m The memory usage
process time slot based time R-matrix reduces the wastage Energy consumption, 100 nodes 100 × 100 seems negligible
table slot table of energy in CH nodes Number of message m2 because of the usage
passing in clustering of R-Matrix
phase

[246] Multi-level trust Secure cluster Collection of Energy, Trustworthi- Euclidean Static Detect and eliminate MATLAB Trust level, Number of Random Static, Square, Center Communication and
evaluation method head election sensor nodes distance and ness distance malicious misbehavior nodes, 400 nodes 500 × 500 (250 × 250) m memory overhead of
algorithm trust degree CHs Case of compromised m2 the trust model can
CH, False positive and be considered
false negative alarms

[249] Two-tier DCHS Data commu- Residual Data RSSI Static Selection of frequent Real-time Total network energy, Random Static, Square, Primary tier Other parameters
18

architecture nication energy and aggregation, CH election can be experiment First Node Dead, Half 100 nodes 50 × 50 m2 rather than RSSI and
phase RSSI generation of avoided and of the Node Alive, residual
TDMA event-driven Data delivery at BS energy can be
schedule simulator considered

[250] Dynamic method DCHSM Voronoi Perceived Solve Euclidean Dynamic Solve the problem of MATLAB Network coverage, Random Static, Square, Center Re-division of the AOI
polygon probability overlapping distance unbalanced cluster Lifetime of nodes, 250 nodes 100 × 100 can be performed
method and survival coverage and head selection Total deployment m2 after the death of all
time unbalanced nodes and the active redundant nodes
estimation energy nodes, Average
Consumption residual energy of
nodes

[254] Protocol-based DECH Collection of Minimum Discards the RSSI Static Re-selection of optimal MATLAB Stability Period, Random Static, Square, Center DECH increases
method sensor nodes link cost redundancy CHs and balance the R2013a Network Lifetime, 100 nodes 100 × 100 (50 × 50) m network stability and
of data network load Throughput m2 network life span by
placing the triggered
nodes efficiently

[255] Classical method RaCH Collection of Remaining Minimizes Euclidean Static Minimizes the wastage NS3 Network Lifetime, Random Static, 100 Square, Center Performance measures
sensor nodes energy the energy distance of energy during Stability/Instability nodes 100 × 100 can also be
dissipation communications Period, Number of m2 considered based on
CHs the throughput in the

Computer Science Review 40 (2021) 100376

network.

Ref. - Reference, DCHS - Distributed Cluster Head Scheduling, TDMA - Time-Division Multiple Access, RSSI - Received Signal Strength Indicator, DCHSM - Dynamic Cluster Head Selection Method, LEACH - Low Energy Adaptive Clustering Hierarchy, LBR - LEACH Based on Reservation, DECH - Equally
Distributed Cluster Heads, RaCH - Radial Cluster Heads Selection.
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

solutions. Moreover, optimization techniques are used to achieve 5.2. Routing protocols in clustering
their objectives using different fitness functions [175].
CH selection is performed during the clustering process in 5.2.1. Classical methods
a Hybrid Artificial Bee Colony, Monarchy Butterfly Optimization An Energy Harvesting Wireless Sensor Network (EH-WSN) is
Algorithm (HABC-MBOA) [256] and during the initial clustering used for clustering and energy harvesting in WSN [282]. During
of nodes in an integrated PSO and Fuzzy-based CH selection ap- the clustering process, the node’s radius of each node is computed
proach [257]. However, the CH selection is based on geographical in a centralized manner, which identifies the node’s position.
location information, degree of CH [258], the distance between However, the node membership function is computed in a dis-
CH, base station, intra-cluster distances, remaining energy [259, tributed manner. The node with the smallest delay is selected
260] to enhance the dissipation of energy and to achieve data as CH, which further transfers the information to the BS. A Grid-
aggregation in an efficient manner [261]. Cycle Routing Protocol (GCRP) is used to select a Grid Cell Head
Based on the computation of multi-objective fitness value, the (GCH) for every grid, which updates the latest location of mobile
CH efficiency is being improved by an ABC-based algorithm [262, sink [283]. A global agent is used to report data to the mobile
263]. Similarly, the fitness value is derived based on the resid- sink [284] and to track the current sink location information, a
ual energy, Inter and intra-cluster distance [264,265] for veri- Virtual Grid-based Dynamic Routes Adjustment (VGDRA) method
fying evaluating the eligibility of CHs [266]. Through the effi-
is used. According to this sink node’s latest information, cell
cient process of CH selection, the consumption of energy in the
headers re-configure the paths between themselves. Also, by a
networks was significantly reduced [267] and stabilizes energy
set of communication rules, only a limited cell-header is used in
consumption [268].
the route reconstruction phase, minimizing the overall communi-
The GA performs the selection of CH in GADA-LEACH, which
cation cost [285]. A master node is selected for each grid, respon-
uses relay nodes that act as an interface between the BS and the
sible for data delivery and routing in the Grid-based Multipath
CH [269]. Due to the randomized selection of CHs and genetic
with Congestion Avoidance Routing (GMCAR) protocol [286].
algorithms, this method leads to inadequate load balancing. The
GA-based clustering technique creates energy-efficient clusters, Besides, to maximize the network lifespan and provide se-
thereby minimizing WSN energy consumption [270]. This scheme curity by resisting some specific attacks, a Secure and Energy-
also identifies optimal clusters by considering the delay fac- Efficient Multipath (SEEM) scheme is used [287]. A hybrid routing
tor and total dissipated transmission energy. An effective vector scheme known as Grid-Based Reliable Routing (GBRR) incorpo-
encoding strategy is used in the Differential Evolution (DE) algo- rates the cluster and grid-based strategies’ characteristics. The
rithm to resolve the load-balancing issue and choose CHs in the clusters are formed by combining two or more grid based on
GA-based clustering scheme [271]. The genetic algorithm’s stabil- their density. Virtual grid-based clusters are formed by using the
ity period is enhanced by a stable aware ERP (SAERP) [272]. The route reply strategy around the next-hop nodes of the specified
cluster formation and CH selection using optimization approaches routes. This scheme ensures reliable node-to-node data trans-
according to various parameters are given in Table 5. fer using a greedy and perimeter forwarding technique [288].
Another hybrid routing protocol based on the cluster and chain-
5.1.3. Machine learning techniques based scheme known as Cluster- Chain Mobile Agent Routing
Machine Learning (ML) makes the computing processes more (CCMAR) performs data collection within the clusters and data
reliable, cost-effective and analyzes the complex data automat- transmission [289].
ically [240]. An artificial neural network selects CH, and Chief An Energy Efficient and QoS aware multi-path Routing proto-
Node (CN) based on the backpropagation technique [218]. The col, known as EQSR, achieve load balancing and recovery from
adjacent nodes are monitored and marked to reduce the redun- node failure. Load balancing is achieved by splitting the traffic
dancy. The nodes with the maximum output value are identified across the set of node-disjoint paths. Also, it increases the relia-
as CN and the second-highest value as CH. A Naive Bayes is a bility of data delivery through the Forward Error Correction (FCC)
form of a classifier based on the Bayes theorem and improves scheme. EQSR employs a queuing model that manages traffic
the network’s existence by evaluating the optimal CHs [227, by differentiating service [290]. Another scheme that uses QoS
273]. Similarly, Fuzzy-based approaches are involved in CH se- is the SAR protocol [291]. Less residual energy and less quality
lection, and a fuzzy membership function is directly linked to of service are used during the building of paths. It also creates
every cluster to minimize the target function and improve energy multiple paths by constructing the source node at the root of the
efficiency. Some of them are Adaptive Neuro-Fuzzy Inference tree. A Multi-constrained QoS Multi-Path routing (MCMP) scheme
System (ANFIS) [141], Fuzzy Energy-Aware Unequal Clustering delivers packets according to the QoS metrics such as reliability
(EAUCF) [274], FUZZY-TOPSIS [275], C-Means-based Fuzzy Clus- and delays to the base station [292]. The protocol delivers the
tering (OCM-FCM) [276] and Distributed load balancing Unequal information with the minimum number of hops, thereby con-
Clustering using Fuzzy (DUCF) [277].
sumes more energy. The various routing protocols using classical
The optimal CHs are identified based on different parameters
methods, based on various parameters, are given in Table 7.
like the residual energy, distance, number of neighbors, and node
density. An Energy-efficient k-Means Technique (EKMT) finds the
5.2.2. Optimization approaches
optimal CHs by reducing the sum of squared distances between
To build low-power clusters, a meta-heuristic technique kn-
nodes [278]. In an Optimal K-means (OK-means) algorithm, the
own as Artificial Bee Colony- Standard Deviation (ABC-SD) al-
cluster formation is performed in the installation phase, and
gorithm is proposed [194]. An Improved version of Cluster-bas-
during the communication phase, every node transfers the data
to the BS. The chance of becoming CHs depends on the energy ed Wireless sensor network routing using Artificial bee colony
level, and location metrics [279]. To achieve classification accu- algorithm considering Quality of service (ICWAQ) algorithm opti-
racy, SVM is implemented with the Improved flower pollination mizes the node clustering during the selection process of CH and
algorithm (IFPA) [280]. A semi-supervised classification technique minimizes the energy cost of CH selection [263]. The Bee-Sensor-
in WSN clustering maximizes the network lifetime based on two C protocol scheme improves the scalability and reduces the rout-
objectives [281]. The first objective focuses on cluster formation, ing overhead [293]. It constructs a cluster scenario and elects the
and the second objective is to transfer data to the BS. The cluster cluster head when an event occurs. Bee-Sensor-C comprises three
formation and CH selection using machine learning techniques phases: formation of the cluster, multipath construction, and
according to various parameters are given in Table 6. data transmission, which is shown in Fig. 11. Before the cluster

19
 J. Amutha, S. Sharma and S.K. Sharma
Table 5
Cluster formation and Cluster Head (CH) selection using optimization approaches.
Ref. CH selection Clustering Cluster CH selection Role of CH Distance CH mobility Feature Experiment Simulation Deployment Nature & Area of Location of Future
Method algorithm formation criteria between (or) environment of nodes number of Interest (AOI) BS/sink node direction/Remarks
nodes Simulation performance nodes
parameters

[141] GA and GA-ANFIS Based on GA Energy level and Data Wireless Static Detect malicious MATLAB Number of live Uniform Static, Square, Center Detects the malicious
Fuzzy logic distance aggregation radio signal nodes, thereby sensors, longevity, distribution 100 nodes 100 × 100 nodes but failed to
and prolongs the WSN average energy, data m2 eliminate it.
broadcasts to lifespan packets received by
BS the BS

[150] HSA and PSO HSA-PSO Collection of Total energy and Energy- Euclidean Static Obtain global search MATLAB Residual energy, alive Random Quasi- Rectangle, (50,150) m Improves searching
sensor nodes distance efficient data distance with faster 2014b nodes, dead nodes, stationary, 100 × 200 efficiency at each
transmission convergence throughput 100 nodes m2 region with the help
of HSA

[175] SA and DE DESA Collection of Residual energy Prolonging Euclidean Static Improve the MATLAB Residual energy, alive Random Quasi- Square, Not specified DESA maximizes the
sensor nodes and distance the network distance performance of WSN 2015a nodes, dead nodes, stationary, 200 × 200 death of cluster heads.
lifetime by optimal CHs throughput 100 nodes m2

[251] MFO MFSTERP Nodes are Residual energy Load Euclidean Static MFO focus on the MATLAB Number of alive Random Static, Square, Center Cluster nodes support
divided to and relative Balancing distance stability of the cluster nodes, average energy, 100 sensor 100 × 100 m2 (50,50) m threshold function to
form clusters distance and cluster isolation round history of dead nodes forward data to CH
quality parameters nodes, energy variance
per round

[253] Heuristic- HACH HEECHS Based on the Relay Euclidean Rotating or Use heuristic crossover MATLAB Number of alive Random Static, Square, Variable Unable to deploy for
based protocol high energy distance re-clustering with GA nodes, average 100,400 100 × 100 positions GPS-free sensor
GA CH residual energy, nodes m2 applications
performance
comparison of
different WSNs
heterogeneity level
20

[256] ABC and HABC-MBOA- Initialization Dimension Enhance the RSS Static Identify an optimal MATLAB Residual energy, Random Random way Square, Center HABC-MBOA can be
MBOA CHSS phase network solution with faster R2015a throughput, alive Point 400 × 400 devised for
lifetime convergence nodes, dead nodes, mobility m2 energy stabilized CH
mean residual energy model, selection
and throughput, 1000 sensor
standard deviation in nodes
residual energy and
throughput

[257] Fuzzy and Based on Fuzzy Distance and total Minimize the Euclidean Static PSO algorithm Simulation Number of messages Uniformly Static, Square, Center More appropriate
PSO fuzzy Initialization consumption of size and select the CHs tool received at the BS, distributed 100 nodes 500 × 500 (250,250), & algorithms for initial
clustering phase energy reduce the Alive nodes, Node m2 edge clustering
and PSO time of mortality (500,575) m
computation

[258] Cuckoo and CHS-CHSS Clusters are Remaining energy, Routing the Euclidean Static Suitable method for Custom Total energy Uniform Static, Square, (100,100), CHS-CHSS can be
Harmony formed by node degree and data packet distance and transporting the Simulator consumption, number distribution 100 to 300 200 × 200 (150, 50), improved by
Search based joining nodes distance intra-cluster aggregated data in developed in of alive nodes nodes m2 (200, 200) m incorporating delay
method to nearer CH distance large scale WSNs C and sensitivity and
MATLAB network with
obstacles

[294] Firefly FCR Collection of Energy, distance Enhance the Euclidean Not specified Selection of best MATLAB Normalized energy for Random Static Square, Center The time required to

Computer Science Review 40 (2021) 100376

nodes and delay of WSN lifetime distance cluster head R2015a 100 nodes, normalized 100 × 100 determine the CH
nodes network energy with m2 should be minimized
distance

[295] GSO and GGWSO Collection of Distance, delay, Security and Arbitrary Static Select the best cluster MATLAB Initial energy, number Uniform Static, Square, Center Simulation is only
GWO sensor nodes energy, and the energy distance head based on hybrid R2015a of nodes and distribution M sensor 100 × 100 performed based on
risk factor awareness GSO and GWO performance analysis nodes m2 energy and security
on the range with
respect to the delay

Ref. - Reference, SA - Simulated Annealing, DE - Differential Evolution, HSA- Harmony Search Algorithm, PSO - Particle Swarm Optimization, DESA - Simulated annealing (SA) and differential evolution-based scheme, GA - Genetic Algorithm, GSO - Group Search Optimization, GWO - Grey Wolf Optimization,
MFO - Moth Flame Optimization, HACH - Heuristic Algorithm for Clustering Hierarchy, FCR - Firefly with Cyclic Randomization, GGWSO - Grouped Grey Wolf Search Optimization, MFSTERP - MFO (Moth Flame optimization) based Stable Threshold-sensitive Energy-efficient cluster-based Routing Protocol,
HEECHS - Heuristic-Crossover Enhanced Evolutionary Algorithm for Cluster Head Selection, CHS-CHSS - Cuckoo and Harmony Search-based Cluster Head Selection Scheme, ABC and MBOA - Artificial Bee Colony and Monarchy Butterfly Optimization Algorithm, HABC-MBOA-CHSS - Hybrid Artificial Bee Colony
and Monarchy Butterfly Optimization Algorithm Cluster Head Selection Scheme, GA-ANFIS - Genetic Algorithm (GA) and an Adaptive Neuro-Fuzzy Inference System.
 J. Amutha, S. Sharma and S.K. Sharma
Table 6
Cluster formation and Cluster Head (CH) selection using machine learning techniques.
Ref. Clustering Clustering Topology CH selection criteria Feature Number of Simulation / Distribution Nature & No. Area of Interest (AOI) Simulation environment Future direction / Remarks
algorithm method CHs Tools used of sensor of nodes performance parameters
nodes

[10] HQCA Distributed Grid-based Residual energy, cluster density, High reliability, better Not specified MATLAB Random Static, Square, Network lifetime, number of HQCA can be enhanced by
mean distance, cluster quality scalability, low error rate and 50, 100, 200, 100 × 100 m2 , 200 × 200 CHs incorporating more performance
good performance 300, 500 m2 , 300 × 300 m2 , parameters
nodes 500 × 500 m2

[140] FCHA Distributed Cluster-based Betweenness centrality and CH is dynamically formed by Not specified NS-2.35 Uniform Static, Square, Remaining energy, network FCHA does not support
residual energy the sink node distribution 200 nodes 100 × 100 m2 lifetime and total alive nodes mobility scenarios

[142] FBUC Distributed Unequal Distance and degree of node Extends the lifetime of CH Not specified MATLAB Deployed Static, Square, Total quantity of alive nodes, Inter-cluster energy
clustering nodes closer the sink node arbitrarily 100 nodes 200 × 200 m2 total clusters, network lifespan consumption is high

[218] ANN Distributed Hybrid Residual energy, distance and The chief node aggregates and 3 cluster OMNET Not specified Static, Square, Total alive nodes and energy This method does not focus on
traffic load computes the average energy heads 100 nodes 100 m × 100 mand500 × 500 consumption with respect to securing chief nodes
and the energy threshold m2 rounds

[225] SVM Centralized Hybrid Decision function Hausdorff distance calculation Not specified MATLAB Nonlinear Static, Not specified Detection accuracy, Fault Fault prediction & dynamic
and nonlinear classification distribution N number of probability classification can be used to
of data nodes predict the fault

[227] Naive Bayes Distributed Tree Residual energy and sum of Naive Bayes technique is used Not specified MATLAB Random Static, Square, Number of live nodes Flexibility of algorithm makes
scheme local distance to determine the CH node 100 nodes 100 × 100 m2 multiple networks as a model

[232] FDS based on Distributed Tree PJ and decision rules Identifies the faulty sensor 10 cluster TOSSIM tool Probability Static, Square, False alarm rate, detection FDS performs two-phase
21

Bayesian nodes heads distribution 100 nodes 500 × 500 m2 accuracy, and memory verification in deciding the
overhead faulty nodes

[278] EKMT Centralized Hybrid Remaining energy and the Finds the optimal CHs by Not specified MATLAB. Random Static, Square, Jitter and delay The factors affecting the
distance k-means algorithm 100 nodes 400 × 400 m2 process of CH is a major
challenge

[280] IFPA and Distributed Tree SVM Aggregation of data in cluster Not specified MATLAB Random Static, Square, Mean-square error and scatter Hausdorff metric determines
SVM heads by using SVM 2018b distribution 100,200, 300 200 × 200 m2 , 300 × 300 index the distance difference between
nodes m2 , 400 × 400 m2 two datasets.

[281] Semi- Distributed Hybrid Centroid based approach CH performs continuous or 3 cluster MATLAB Random Static Variable network size Number of sensor nodes, Twin rolling ball scheme is to
supervised periodic data transmission heads R2017a distribution Network size, End-to-end delay, overcome the issue of greedy
classification depending on different zones Number of rounds forwarding of data packets
technique

[296] K-means Hybrid Tree High-end sensor node acts as Offline phase is performed by 30 cluster OPNET14.5 Random Static, 720 Not specified Delay, Data sent and received The sequence of different
CH the BS and online phase by heads sensor nodes attacks are configured prior to
the CH node in each cluster (30 clusters every CH
with 24
nodes in
each cluster)

[297] ACSO and Distributed Hybrid ACSO Cluster head (CH) selection and Not specified PYTHON Random Static, Square, Bandwidth, BER, PDR, Buffer Effectiveness of this scheme

Computer Science Review 40 (2021) 100376

adaptive SVM malicious nodes are reported distribution 50 nodes 100 × 100 m2 occupancy, distance, channel depends on the set of
classification load and QoS classifiers which increases the
prediction accuracy

Ref. - Reference, ANN - Artificial Neural Network, SVM - Support Vector Machine, IFPA - Improvement Version for the Flower Pollination Algorithm, ACSO - Adaptive Chicken Swarm Optimization, FDS - Fault Detection Scheme, EKMT - Energy Efficient K-Means Clustering Technique, FBUC - Fuzzy Based
Unequal Clustering, HQCA - High-Quality Clustering Algorithm, FCHA - Fuzzy-based Cluster Head Amendment, BER- Bit Error Rate, PDR- Packet Delivery Ratio, QoS - Quality of Service.
 J. Amutha, S. Sharma and S.K. Sharma
Table 7
Routing protocols using classical methods.
Ref. Protocol Type Approach Cluster size CH selection Energy Phases Feature Simulation Simulation Nature & No. of Area of interest Type of Sink and Multi-sink Sink mobility Future
efficiency environment nodes & Deployment position support pattern direction/Remarks
performance
parameters

EH-WSN Hierarchical Hybrid Unequal Based on energy Yes Two phases: Setup Prolongs network Matlab Number of available Static, Square, Static, No Not specified Affects the rate of
[282] phase and Data lifetime, improved nodes, average energy homogeneous, 50, 1 × 1 Km2 , Center packets in the base
Transmission Phase throughput, reduces of nodes, number of 100 nodes Random (500,500) m station.
network overhead packets

[283] GCRP Grid-based Distributed Equal Remaining energy Yes Six phases: Grid Delivery of sensed NS-2.31 Throughput, Static, Square, Mobile, No Predefined Absence of random
and distance construction, data with minimal end-to-end delay, homogeneous, 200 × 200 m2 , Moves along the sink mobility pattern.
selection of cell delay packet delivery ratio 150, 200, 250 Random periphery of AOI
head, formation of nodes
cycle, sink location
update, data
transmission, grid
cell head re-election

[284] MSDD Grid-based Distributed Equal Not specified Yes Two phases: global MSDD is query-driven TOSSIM Packet delivery ratio Static, Square, Mobile, Yes Random Once the grid
agent relaying and and communicates to and delay homogeneous, 240 × 240 m2 , Fix the sink structure is
effective path the BS about disasters 100 nodes Random above at 2m/s constructed, MSDD
forwarding. immediately, a global does not change the
agent is introduced to two-tier grid structure
track the sink

[285] VGDRA Grid-based Distributed Equal Based on the set Yes Two phases: Find appropriate NS-2 Virtual backbone Static, Square, Mobile, No Predefined VGDRA scheme fails
of nodes close to reconstruction routes to the latest structure construction homogeneous, 200 × 200 m2 , Moving along the to evaluate the
the center of the phase, dissemination mobile sink location cost, per round routes 100-400 nodes Random periphery of the performance of sink
Cells phase reconstruction cost, sensor field node
network lifetime,
network convergence
time

[286] GMCAR Grid-based Distributed Equal Master node is Yes Three phases: Prolongs the overall Network Average residual Static, Square, Mobile, No Predefined To satisfy the delay
selected based on Formation of grid, lifetime by utilizing simulator energy, normalized heterogeneous, 350 × 350 m2 , BS is always requirements in
residual energy generation of the densely deployed remaining energy for 100 nodes Random located in corner routing, grid density
routing table and grids network grids, first of the sensor as well as hop count
data transmission grid death time, field factors must be
number of active considered.
22

nodes, delivery ratio,

average end-to-end
delay, average buffer
occupancy

[287] SEEM Density-based Distributed Equal Base station Yes Three phases: Reduces the additional NS-2 Average network Static, Square, Static No Stationary SEEM focus on
performs the role Formation of route maintenance throughput, control homogeneous, 150 × 150 m2 , multipath concept
of the server topology, overhead overhead, network 130 nodes Random which considers
maintenance of lifetime security and energy
route and data efficiency
transfer simultaneously

[288] GBRR Hybrid Distributed Equal Based on residual Yes Four phases: data GBRR effectively Matlab Network lifetime, Static, Square, Static, No Static The scheme uses a
energy and and cluster identifies the residual energy, homogeneous, 200 × 200 m2 , Location of BS decentralized
degree formation, grid head redundant nodes energy consumption 100 nodes Random (250, 100) m forwarding technique
stage and cluster
head competition
phase

CCMAR Hybrid Distributed Equal Remaining energy, Yes Two phases: Cluster Due to the hybrid NS-2 Energy consumption, Static, Square, Static, No Static Failed to address fault
[289] distance and bits formation and data network topology, transmission delay, homogeneous, 100 × 100 m2 , Center position tolerance and security
transferred aggregation data aggregation and network lifetime 100 nodes Random (50,50) m in CCMAR
transmission are

Computer Science Review 40 (2021) 100376

performed faster, thus
reduces the
transmission delay

[290] EQSR Tree-based Distributed Packets of Residual energy, Yes Five Phases: Path EQSR maximizes the NS-2 Energy consumption, Static, Square, Static, No Not specified EQSR can be analyzed
equal size buffer size and discovery; Path network lifetime, delivery ratio, average homogeneous, 500 × 500 m2 , Upper right by varying metrics
SNR ratio refreshment; Path increases the delay 300 nodes Random corner of the such as the size of
selection; Traffic throughput, uses simulation field the network, buffer
allocation and Data service differentiation, size and path length
transfer and minimizes the
delay

[298] IMP-MECA Area-based Distributed Equal Remaining energy Yes Three phases: Initial In IMP-MECA, nodes NS-2 Average distance, Static, Circle or Mobile, Yes Predefined To identify the
and node identity location of the consume less energy residual energy, homoge-nous, hexagonal Mobile sink along shortest route, nearest
sinks; Stoppage due to the change in number of active 100 nodes Random, the radius of the sink is necessary
Strategy for Sinks; location of sink nodes, total energy Radius of hexagon
energy-efficient gradually. consumption, hexagon 23 m
hierarchical routing
phase

(continued on next page)

 J. Amutha, S. Sharma and S.K. Sharma
Table 7 (continued).
Ref. Protocol Type Approach Cluster size CH selection Energy Phases Feature Simulation Simulation Nature & No. of Area of interest Type of Sink and Multi-sink Sink mobility Future
efficiency environment nodes & Deployment position support pattern direction/Remarks
performance
parameters

[299] Ring routing Area-based Distributed Equal Based on the Yes Five Steps: Ring Minimize the delay OPNET Average energy Static, Square, Mobile sink, No Not specified Extensions and
protocol anchor nodes construction, caused in consumption, delay homogeneous, 600 × 600 m2 , Inside or outside modifications can be
23

Advertisement of data reporting by 200 nodes Uniformly the ring performed based on
Sink Position, considering various deployed multiple mobile sinks
Obtaining BS mobility
location, Data parameters of the
Dissemination, Ring mobile sink
change

EH-WSN - Energy Harvesting Wireless Sensor Network, GCRP - Grid-cycle routing protocol, MSDD - Multiple mobile Sinks Data Dissemination, VGDRA - Virtual Grid based Dynamic Routes Adjustment, GMCAR - Grid-based Multipath with Congestion Avoidance Routing protocol, SEEM - Secure and
Energy-Efficient multipath Routing protocol, GBRR - Grid-Based Reliable Routing, IMP-MECA - Improved Mobile-sink based Energy-efficient Clustering Algorithm, CCMAR - Cluster-Chain Mobile Agent Routing protocol, EQSR - Energy Efficient and QoS aware multipath routing protocol.

Computer Science Review 40 (2021) 100376

J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

formation, each node is identified by its ID and corresponding

sink node ID. In the first phase, the cluster structure is being built.
In the second phase, multipath formation is initiated between the
CH and its corresponding base station. During the third phase, the
data transmission to the sink node takes place.
An Evolutionary based clustered routing protocol (ERP) scheme
is used to find a feasible solution by incorporating the fitness
function [300]. In LEACH-GA protocol, GA is used to determine
the optimal probability of nodes of becoming a CH, and in the
two-tiered architecture, GA is used to establish a suitable routing
strategy for WSNs and achieves double the network lifetime
compared to other traditional schemes [181,301].

5.2.3. Machine learning techniques

A Distributed Fuzzy logic-based unequal Clustering approach
and Routing algorithm (DFCR) is distributed in nature, employs
a scheme for CH selection, balances the load for data routing,
thus solves the hotspot issue [302]. Similarly, a Fuzzy and Ant
Colony Optimization Based Combined MAC, Routing, and Unequal
Clustering Cross-Layer Protocol for Wireless Sensor Networks Fig. 11. Phases of Bee-Sensor-C.
(FAMACROW)
avoids hotspots by unequal clustering mechanism, uses fuzzy
logic for selection of CH and ACO for inter-cluster communication
and transmission of data to the sink node [303]. In the machine
anteed delivery of information [307]. The narrow bandwidth and
learning technique, the Q-Routing scheme is used in earlier re-
limited capacity of the channel easily block the data transmission,
search [304]. To learn the best path, it uses the minimal delivery
which results in data loss.
time and allocates a Q-value to each neighbor’s nodes. Q-value
Moreover, the sensor nodes are prone to interference, which
considers the neighbor as the next hop to the sink.
affects the quality of data transmission and reliability of the
Reinforcement Learning (RL) is a technique where rules or
whole network system. The factors that affect reliability are the
patterns are generated by computer programs from the vast
environment, which is highly influenced by natural calamities,
amount of data. It also relies on agents that acquire knowledge
electromagnetic ratio, and other uncertain facts. Another signif-
by exploring actively in their environment. The knowledge gained
icant factor that affects reliability is that the sensor nodes are
decides the next action or activity, and a reward is received
battery-driven, which is not easily replaced, and this minimizes
from the deployed scenario for a particular state. RL is a Markov
the connectivity and affects the reliability of the network. Re-
Decision Process (MDP) technique that comprises of collection cent research reveals that numerous work focuses on routing
of states (s) and actions (A), set of transition probability (P) and protocols, network lifetime, localization, and energy efficiency in
reward (R) corresponding to actions. MDP finds an optimal policy clustering. However, only a few studies focus on reliability, but
that is termed as the mapping from states to actions. RL deals at present, research works that aim at reliability are increasing
with routing in WSNs because of the following reasons: improve every year [153,308].
network performance; fewer requirements for computation and
memory at each node; less time to converge; highly flexible; 5.3.1. Classical methods
easy to implement; achieve optimal results with no additional Distributed Sensor Networks (DSN), consider the problem of
costs [37,305]. measuring the reliability and expected delay in the network [309].
An adaptive tree protocol is a RL strategy used for restricted The DSN reliability is defined as the probability that there is a co-
routing and a Message-initiated Constraint Based Routing (MCBR) mmunication path between the BS and sensor, that is oper-
in RL strategy is used for achieving load balancing and con- ational in the target cluster. An efficient clustering algorithm
trol congestion in the network. Learning happens in all three based on mobile sink performs network clustering based on their
phases; in the initialization process, a spanning tree is built based energy level, thus enhancing the reliability of the network. Also,
on the RL technique. When one node broadcasts a packet to communication cost is reduced by using a mobile sink [310].
any sensor node, its current Q-value is also transmitted along
with the packet. In the forwarding phase, the node sent the 5.3.2. Optimization approaches
received packet to its parent. The tree architecture changes ac- Genetic algorithm and Bee colony optimization (BCO) ensures
cording to the neighbor Q-values. In the confirmation phase, if the reliability of the WSN in two stages. The genetic algorithm im-
the packet information is not processed within time, the neighbor proves the clustering process of nodes, and BCO finds an optimal
Q-value of the corresponding node updates and change occurs in route-transition across nodes [180]. Similarly, another algorithm
the relevant parent–child relationship [306]. The various routing for reliability evaluation in WSN uses randomly deployed sen-
protocols using optimization approaches and machine learning sor nodes, which investigate reliability analysis in two different
techniques based on various parameters are given in Table 8. ways [311]. First, an algorithm is based on the ordered binary de-
cision diagram, and second, another algorithm focused on simula-
5.3. Reliability in clustering tion in Monte Carlo. These two algorithms combined to compute
the reliability considering the failure of an individual component
Network reliability is an essential metric in evaluating the and improves the network efficiency with QoS parameters for the
performance of WSNs. In WSNs, various applications like surveil- reliability routing of WSNs [312].
lance, military, investigation operation, etc., reliability is consid-
ered as an essential factor because it ensures the quality and guar-
24
 J. Amutha, S. Sharma and S.K. Sharma
Table 8
Routing protocols using optimization approaches and machine learning techniques.
Ref. Protocol Type Approach Cluster size CH selection Energy model Phases Feature Protocol Location aware Simulation Simulation Nature, Type, No. Area of interest Type of sink Sink mobility Future
operation environment of nodes & Deployment & position pattern direction/Remarks
performance
parameters

[139] MRP-ACO Cluster-based Distributed Equal Residual First-order Three phases: Reduction of Multi-path No Matlab Average Static, Square, Static, Not specified The performance
energy and model Cluster Formation, energy utilization energy, homogeneous & 200 × 200 m2 , (0,200) m can be extended
path length, Constructing and prolongs the network heterogeneous Uniform with additional
Multipath, Data network lifetime, 100 to 500 nodes metrics like speed
transfer standard
deviation of
energy

[181] Genetic Cluster and Centralized Equal Higher power First Four Phases: Determines an Coherent Yes Simulation Achieved Static, Square, Static, Corner Random Genetic
Algorithm Location- relay nodes order radio Initial population, energy-efficient lifetime, heterogeneous, 480 × 480 m2 , of the algorithms evolve
based model Crossover, routing scheme execution 100 to 3000 Random network rapidly into the
Mutation, times, nodes optimal solution
Dynamic lifetimes in for local networks
reconfiguration lossy
environments

[194] ABC-SD Tree-based Centralized Unequal Energy CC2420 Four phases: Link quality, QoS No Castalia 3.2 Energy Static, Square, Static, Pre-defined Does not focus
efficiency and Bootstrapping, energy efficiency, simulator consumption, homogeneous, 100 100 × 100 m2 , (0,0) sink on the effects of
QoS Set-up, Clustering, throughput and first sensor to 400 sensors Random location mobile sensors
Steady-state scalability dead, energy and mobile BS
efficiency,
throughput,
network
coverage,
communica-
tion
overhead,
packet loss
25

[263] ICWAQ Cluster-based Centralized Equal Based on the Free space Three Steps: Energy efficiency, QoS No Matlab Total energy Static, Square, Static, Edge Not specified Future work focus
threshold radio model Initialization, throughput, dissipation, 100 nodes 200 × 200 m2 , (250, 575 m) on clustering and
level fitness calculation, employs QoS number of Random routing including
Update position alive nodes, mobile nodes
normalized
network
energy,

[269] GADA-LEACH Hierarchical Distributed Equal Based on the First-order Four Steps: Optimization of Multi-path No Matlab Network Static, Square, Not specified Network lifetime
cluster-based fitness radio model Initialize, fitness cluster head by lifetime by homogeneous, 100 × 100 m2 , Static, can be assured
function calculation, using a genetic varying 100 sensor nodes Random Outside AOI by direct
crossover, algorithm packet size communication of
communication nodes

[293] Bee-Sensor-C Area-based Distributed Unequal Based on the First-order Three phases: Energy efficiency, Multi-path No Java Control Static, Square, Static, Not specified Improved so that
detection of radio model Formation, reduces routing framework overhead, homogeneous, 200 × 200 m2 , (200, 100) m it may be
an event construction & overhead, energy variable nodes Uniform compatible with a
transmission scalability efficiency, dynamic scenario
latency,
energy
standard
deviation,
packet

Computer Science Review 40 (2021) 100376

delivery rate,
routing
building time

[300] ERP Cluster-based Distributed Equal Based on the Radio model Two phases: Prolongs the Multipath No Matlab Number of Static, Square Static, inside Not specified Future direction
optimal node cluster and network alive nodes, heterogeneous 100 × 100 m2 , the sensor focus on varying
steady-state lifetime, better advanced and 100 sensor nodes Uniform field sink location
energy normal dead
consumption nodes,
average
remaining
energy

(continued on next page)

 J. Amutha, S. Sharma and S.K. Sharma
Table 8 (continued).
Ref. Protocol Type Approach Cluster size CH selection Energy model Phases Feature Protocol Location aware Simulation Simulation Nature, Type, No. Area of interest Type of sink Sink mobility Future
operation environment of nodes & Deployment & position pattern direction/Remarks
performance
parameters

[302] DFCR Cluster-based Distributed Unequal Based on Free space or Four Steps: Energy-efficient Multipath No Matlab Network Static, Square, Static, Edge Not specified DFCR is designed
energy & multipath formation of and energy lifetime, homogeneous, 100 100 × 100 m2 , (50,0)m, for the stationary
sink node channel clusters & virtual balancing average sensor nodes Random Center sensor nodes
backbone, data remaining (50,50) m
sharing & data energy,
routing energy
population
26

[303] FAMA-CROW Hierarchical Distributed Unequal Based on the CC2500 Three phases: Reducing energy Link quality Yes Matlab and Residual Static, Square, Static, Not specified Use of LQI can
cluster-based highest construction, consumption and NS2 energy, dead homogeneous, 1000 × 1000 m2 , Center (0,0) maximize the
proficiency of identification & removing hot nodes, 1000 nodes m reliability of
the node steady-state phase spots problem latency, Random FAMACROW
delivery ratio

[306] Adaptive Tree and Distributed Equal Not specified Radio model Three phases: Retain the Not specified No Matlab Latency, Not specified 7 × 7 sensor Mobile Not specified Metrics analyzed
spanning tree grid-based initialization, maximum delivery ratio, grid based on energy
forwarding and connectivity to energy efficiency and
confirmation the BS efficiency delivery ratio

ABC-SD - Artificial Bee Colony- Standard Deviation, hom. - homogeneous, het.- heterogeneous, ICWAQ - Improved version of Cluster-based Wireless sensor network routings using Artificial bee colony algorithm considering Quality of service, ERP - Evolutionary based clustered Routing Protocol, GADA-LEACH
- Genetic Algorithm and Distance Aware DFCR - Distributed Fuzzy logic-based unequal Clustering approach and Routing algorithm, MRP-ACO - Multipath Routing Protocol based Ant Colony Optimization (ACO), FAMACROW -Fuzzy and Ant Colony Optimization Based Combined MAC, Routing, and Unequal
Clustering Cross-Layer Protocol for Wireless Sensor Networks, LQI - Link Quality Indicator.

Computer Science Review 40 (2021) 100376

J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

tags, and route construction request by the parent nodes. Hence,

an intruder cannot initialize the tree construction and cannot
forge route requests.
An Intrusion-tolerant routing protocol for Wireless Sensor
Networks (INSENS) minimize damage caused by the intruders
on modifying, injecting, or blocking the packets by incorporating
a distributed security mechanism [318]. It includes one-way
hash keys and validation codes that resist multipath routing and
wormhole attacks. Moreover, it integrates bi-direction verifica-
tion to overcome the rushing attack and secure communication
Fig. 12. Modules of secure routing protocol. through a critical configuration scheme focused on transitory
global keys. A Secure Threshold sensitive Energy Efficient sensor
Network (Sec-TEEN) technique enhances the network security
5.3.3. Machine learning techniques against external threats [319]. Sec-LEACH is used to design the
A fuzzy clustering ensemble approach based on the weighting Sec-TEEN protocol. The cluster head selection broadcasts its id
strategy is used to select high-quality fuzzy clusters [313]. The and nonce to generate the MAC address, which is used to prevent
fuzzy cluster quality is considered reliable, and entropy is ap- replay attacks.
plied over the similarity between fuzzy clusters. There are two
phases in the fuzzy clustering ensemble approach. The clusters 5.4.2. Optimization approaches
are formed during the first step by a single clustering algorithm A Fruit Fly Optimization Algorithm (FFOA) detects the node
and forwarded to the next stage. The fuzzy clustering ensemble capture attack and discovers the optimal node [320]. Initially, the
is the second stage, which is partitioned into four steps. Out cost of energy consumption, network information, and keys are
of the four steps, the first step computes each fuzzy cluster’s attained by the intruder. The intruder establishes and calculates
similarity with other clusters, whereas the second step calculates the matrix, which expresses the number of keys assigned to the
each fuzzy cluster’s reliability. The final cluster is obtained by nodes. Also, it calculates the correlation between the nodes and
the third and the fourth steps, comprises of a reliability-based the routes. A secured clustering approach based on bio-inspired
weighted graph or the weighted fuzzy co-association matrix. A technique creates a list of candidate nodes and selects the trusted
Reliability-based Enhanced Technique for the Ordering of Prefer- CH by the honey bee mating algorithm.
ence by Similarity Ideal Solution (RE-TOPSIS) protocol is used in By an Adaptive Chicken Swarm Optimization algorithm (ACSO),
CH selection, which is competent and reliable [314]. This protocol it is identified that CHs can be selected, network lifetime and
is used in combination with fuzzy logic, which uses a Multi- scalability can be improved. An adaptive SVM classification is
Criteria Decision Making (MCDM) approach to chose the best also used with an intrusion Detection System (IDS) where the
cluster head by using the rank index value and maximizes the malicious sensor nodes are reported based on the acknowledg-
network lifetime. A reliability estimation methodology focus on ment [297]. By the acknowledgment based method, the sensor
the k-means algorithm and analyzes the complex system perfor- node is recognized whether the node is attacked or not. Then
mance [315]. In the cluster-based system, reliability is analyzed in the malicious sensor nodes are processed based on the conditions
sub-systems and the division of family systems. The k-means al- to detect the various types of attacks. After detecting an attack,
gorithm implements the division through several iterations. Thus the High-Level Security Mechanism is used, composed of a cryp-
the sub-system samples solve the problem of zero failure. The tographic function that offers security. The advantage of ACSO
analysis of reliability-based approaches using classical methods, is its adaptiveness, which mainly reduces the time consumption
optimization approaches, and machine learning techniques are for selecting the best CH. Minimizing nodes’ energy usage while
given in Table 9. detecting an attack and for minimizing the false alarm rate, a
Multi-Kernel Extreme Learning Machine (MK-ELM) is used [321,
5.4. Security in clustering 322].

5.4.1. Classical methods 5.4.3. Machine learning techniques

Cross-layer design and an energy-harvesting mechanism are Support Vector Machines (SVM) classification method defines
used for secure cluster-based routing protocol [316]. It uses a a decision function (or separation function) that is executed at
distributed scheme in which parameters are exchanged between CH to detect anomalous sensor, and for data classification [225].
various layers that ensure energy efficiency usage. The energy The data classification in real-time can be established through the
harvesting mechanism extracts energy and stores it. This stored learning phase, as shown in Fig. 13. A learning phase establishes
energy further takes decisions for the routing issues and the the decision function that classifies a new data set based on
node’s state. The different modules of a secure routing protocol statistical learning techniques. The learning database consists of
are shown in Fig. 12. The cluster formation module partitioned a collection of standard and faulty data.
the network into clusters, where the nodes form multi-hop clus- A Fault Detection Scheme (FDS) identifies the faulty sensor
ters for coordination. The energy model reduces energy con- nodes in which the verification process is performed in two-
sumption when nodes communicate with the cluster head. The phases [232]. The first phase verification is performed locally
cross-layer design is used because it provides energy parameters inside the sensor nodes using the Bayesian model, while the
using a cross-layer interface at the network layer. A security second phase verification is performed in the cluster head. This
mechanism is used to prevent a variety of active and passive second phase deals with the decision results and Probability Joint
attacks. (PJ), provided by the first phase to reach a global decision and
A Secure and Efficient Intrusion-Fault tolerant protocol (SEIF) broadcast to its corresponding CHs. The CHs are responsible for
is an efficient and lightweight security scheme based on the choosing a final decision based on the similarity between PJs of
multi-path protocol [317]. SEIF offers in-network implementation all nodes that belong to the same cluster.
that never refers to the base station for security checks and route Moreover, the k-means clustering algorithm over the training
creation. Besides, SEIF guarantees round initialization, sub-branch data is used for intrusion detection [296,323]. There are two

27
 J. Amutha, S. Sharma and S.K. Sharma
Table 9
Analysis of Reliability-based approaches in clustering using classical methods, optimization approaches and machine learning techniques.
Ref. Protocol CH selection Cluster size Cluster Sensor node Network Communica- Feature/Objective Deployment Load Approach Simulation Nature & No. Area of Type of sink Simulation Future direction/Remarks
density type lifetime tion balancing of nodes interest &location environment
performance
parameters

[180] Hybrid Based on Equal Not specified Not specified Yes Not specified Discover an optimal route and Random Yes Optimization Simulation Static, Square, Static, Distance, time, High rate of reliability &
Method LEACH enhance the route transmission 100 nodes 400 × 400 Center reliability decreases energy consumption
based on protocol across nodes m2
clustering

[309] DSN Virtual Equal Variable Homogeneous Yes Multi-hop Formation of reliable cluster of Random No Classical C program Static, 2000 × 1000 Static, Delay and Networks with different TR
backbone sensor nodes 40 nodes m2 Corner Reliability and routing methods can be
investigated

[310] DCRRP Remaining Equal Not specified Not specified Yes Multi-hop To enhance network reliability Random Yes Opnet Static, Square, Static Delay and The efficiency of the entire
28

energy & Classical Modeler 11.5 50 nodes 100 × 100 throughput network is optimized due to
distance m2 dynamic sink

The fuzzy Fuzzy cluster Equal Not specified Not specified Yes Multiple base The reliability of each fuzzy Data sets No ML Matlab Eight data sets NMI and DB Optimization method can be
[313] clustering level based clustering cluster is measured based on used to obtain final clustering
ensemble on weights the estimation of its
approach unreliability

[314] RE-TOPSIS Based on Equal Variable Homogeneous Yes Multi-hop Reliable, reduced energy Random Yes ML NS-2 Static, Square, Static, Average energy Six attributes are needed to
index and consumption and 100 nodes 100 × 100 Center consumption, compute the best rank index
threshold judicious selection of CH m2 (50,50) m throughput, PDR value for the selection of
values cluster head

[315] Real time Not specified Equal Variable, Not specified Yes Not specified Reliability is calculated by the Not specified Yes ML Real-time approach Absolute error The appropriate value of the k
reliability scale & shape computation of each k-means in K-means clustering can be
estimation parameter sub-system determined

RE-TOPSIS - Reliability-based Enhanced Technique for the Ordering of Preference by Similarity Ideal Solution, PDR - Packet Delivery Ratio, DSN - Distributed Sensor Network NMI - Normalized Mutual Information between two clusterings, DB - Davies–Bouldin index of clustering, DCRRP - Distributed Clustering
Reliable Routing Protocol, TR- Transmission Range.

Computer Science Review 40 (2021) 100376

J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

phases used for this purpose: offline and online stages. The pre- after the deployment of nodes [331]. It comprises three phases:
processing and removal of outliers are performed offline. The sink the cluster head phase, members joining phase, and data com-
node of WSN executes these functions. During the online stage, munication phase. Intra-cluster communications are directly per-
the cluster head performs the function of anomaly detection in formed, but the inter-cluster communication depends on the
every cluster. The anomaly detector’s output is loaded into the distance of the CHs from the BS. Both energy and network models
security system, where the alarm will trigger if any intrusion are used in the communication phase to determine the energy
is detected. A hybrid model and a feature elimination mecha- and check the sensor node and the network’s various conditions.
nism for intrusion detection using a decision tree classifier are Moreover, the URBD algorithm is used to optimize the lifetime
employed using uni-variate features for feature selection [324, of the network and used for scalability. The limitation is that it
325]. A K-medoid clustering technique is another anomaly detec- does not meet other objectives of clustering like connectivity and
tion scheme used for misdirection and blackhole attacks [326]. fault tolerance. Also, the sensor nodes are distributed randomly
Threshold values and network parameters are used to detect the without considering the obstacles.
anomalies with high accuracy. Similarly, to detect the black hole Another distributed algorithm known as Location-Based Un-
attack, k-means clustering is used with the LEACH protocol [327]. equal Clustering Algorithm (LUCA) reduces the overall utilization
An Intelligent and Secured Fuzzy Clustering Algorithm using of energy by all sensor nodes in WSNs [332]. Every sensor eval-
Balanced Load Sub-cluster formation (ISFC-BLS) determines the uates the size of the cluster based on location information and
nodes that join the cluster in a secure manner [328]. ISFC al- generates CHs uniformly. Furthermore, it forms more massive
gorithm uses a Fuzzy S-means scheme to identify the CH. The clusters far from the sink node and smaller clusters near the
CH is responsible for trust-based path routing and energy con- sink node. The impact of clustering deteriorates as the distance
sumption. This scheme also uses an ACO technique to select the from the BS increases owing to the trade-off between the inter
optimal path by using three metrics: node coverage area, path, and intra-cluster data transfer. LUCA determines location-aware
and node trust value. A Hierarchical Trust Management (HTM) nodes through GPS, which is unreliable for many practical appli-
utilizes fuzzy logic that focuses on direct trust computation, in- cations and failed to address the algorithm in the probabilistic
direct trust, and credit-based calculations [329]. The analysis clustering approach.
of security-based algorithms in clustering using classical meth- Constructing Optimal Clustering Architecture (COCA) creates
ods, optimization approaches, and machine learning techniques the optimal clustering architecture, routing, and energy-efficient
are given in Table 10. Moreover, features, simulation metrics, CH [333]. The lifetime of the network is partitioned into several
and future directions of security-based algorithms are given in communication rounds, in which one of them is further parti-
Table 11. tioned into topology formation and transmission process. After
the topology formation, every node broadcasts multiple reports
6. Unequal clustering in wireless sensor networks to the BS by using the inter-cluster configuration. All nodes’
overall energy consumption is minimized within the transmis-
An unequal clustering approach is used to prevent the energy- sion process, and a new communication round begins. At the
hole or hot-spot problem (Fig. 14) in wireless sensor networks beginning of each round, clusters and CHs are reconstructed and
by load balancing between the cluster heads. Unequal clustering reselected, respectively. Even-energy-depletion is accomplished
minimizes the cluster size nearer to BS and maximizes the cluster by the energy consumption of nodes within a unit and in dif-
size far from BS. Smaller size clusters close to the BS have fewer ferent units. The limitation is that the small amount of energy
Cluster Members (CMs) and generates less traffic, thus rely more consumption caused by topology formation is not considered, and
on inter-cluster traffic. Likewise, larger-sized clusters far from the the theoretical result generated is not accurate in real situations
BS have more CMs and devote more time to intra-cluster com- because of the radio propagation model.
munication. Thus, unequal clustering permits all cluster heads to Unequal Cluster-based Routing (UCR) avoids hot-spot issue
have an equal amount of energy, eliminating the hot-spot prob- by dividing into unequal clusters via clustering, and multi-hop
lem [44]. Various researchers present different unequal clustering routing [334]. UCR comprises Energy-Efficient Unequal Clustering
strategies for an energy-efficient WSN. However, the absence of a (EEUC) and energy-aware routing protocol for topology man-
survey of unequal clustering algorithms based on classical, opti- agement and inter-cluster data transfer. The CHs are elected by
mization and machine learning approaches motivated to perform considering the remaining energy of the node using the EEUC
this task. Unequal clustering strategies are explored in detail in algorithm. As the sensor node’s competition radius decreases,
this section, and an analysis is done based on the different cluster the distance to the sink node also decreases, resulting in the
properties. formation of smaller cluster sizes closer to the sink node. Thus
CHs consume low energy during intra-cluster communication and
6.1. Classical methods high energy during inter-cluster communication. If the trans-
mitting power is identified, a sensor node can determine the
An Improved Energy-Aware Distributed Unequal Clustering shortest path to another node based on the received frequency
(EADUC) removes the hot-spot problem, eliminates the re-cluste- range. Moreover, error occurs because of the noise in real network
ring overhead by considering energy as a relay metric, and opti- applications.
mizes scalability, and prolongs the lifetime of the network [330]. An Energy Degree Distance Unequal Clustering Algorithm (ED-
In improved EADUC protocol, after the node deployment, the DUCA) eliminates hot-spot issue by ‘Sierpenski’ triangle, thereby
distance from the base station is calculated by each node that equalizes energy utilization and increases the network lifetime
operates in rounds. Intra-cluster communication in all clusters [335]. EDDUCA consists of three phases: during cluster formation,
is performed simultaneously, and the CH broadcasts the data ‘Sierpenski’ triangle partitions into unequal clusters, as shown in
directly to the sink node. However, in this case, the limitation is Fig. 15. During CH selection phase, the degree of nodes, remaining
that the relay nodes’ selection is focused on the energy estimation energy, and distance metrics are considered. In the data transmis-
of every feasible node. sion phase, every node broadcasts its data to CH and transmits
Like the improved EADUC protocol, the Unequal cluster-Radius it to BS. However, EDDUCA failed to assess the deployment of
approach Based on node Density (URBD) is a distributed algo- heterogeneous nodes that have a high energy capacity and have
rithm that operates in rounds, and the first round begins only considered only three factors for CH selection.
29
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Fig. 13. Learning phase.

Fig. 14. Unequal clustering in wireless sensor networks.

Fig. 15. Sierpenski triangle.

An Energy Balancing Unequal Clustering Protocol (EB-UCP) when all the aggregated information by the CHs are received
eliminates the hot-spot issue by a probabilistic approach and at the sink node. UHEED algorithm utilizes competition radius
performs uniform energy distribution for all nodes [336]. Various as a metric to form unequal clusters; thus, nodes’ early death
probabilities are assigned according to the distance from the is avoided. The limitation is that the UHEED scheme failed to
BS. Low probability indicates a small number of CHs, and CHs analyze the relationship between various lifetime parameters and
are selected according to the higher probability nodes. Every the network density.
CH initially gathers the information from its neighbors and then An Unequal Clustering Theory and Connected Graph Theory
delivers it to the BS through multi-hop data transfer. The network for Routing Algorithm (UCCGRA) utilizes a voting scheme to avoid
model of EB-UCP depicts a layered network in which each layer the hot-spot issue, thus improves energy efficiency by routing and
incorporates defined clusters. The limitation of EB-UCP is that, cluster head election [338]. UCCGRA is a distributed competitive
when they have low energy data transfer to the BS and are scheme that supports special features like scalability, stability,
committed to heavy relay traffic, all first layer nodes become CHs. large scale network, distributed control, and vulnerability to fail-
The Unequal Hierarchical Energy-Efficient Distributed (UHEED) ure. However, CHs are principally elected only based on the
scheme determines the optimal cluster size by using the distance weight of every sensor node.
metrics and avoids the hot-spot problem [337]. UHEED comprises
several rounds, in which each round initiates by activating the 6.2. Optimization approaches
clustering process, and after the cluster formation, the network
enters the data communication phase. This phase comprises An Energy Balanced Unequal Clustering (EBUC) is a centralized
intra-cluster and inter-cluster communication. The round ends protocol that divides the network into unequal clusters, and CHs

30
 J. Amutha, S. Sharma and S.K. Sharma
Table 10
Analysis of security-based algorithms in clustering using classical methods, optimization approaches and machine learning techniques.
Ref. Secure Approach Cluster size Routing protocol issues addressed in clustering of WSN Security objectives Secure mechanism that detect attacks
algorithm

LEC DA Dep FT Sc LB Re Aut C I A Sp Si Wo Sy He Bl Gr

[316] CLDEH Classical Equal Y Cluster Random N Y N Y N Y Y Y N N Y N N Y Y

[317] SEIF Classical Equal Y Tree Random Y Y Y Y Y Y N Y Y N N N N N N
31

[318] INSENS Classical Equal Y Tree Random Y Y N Y N N N N Y N Y N Y N N

[319] Sec-TEEN Classical Equal Y Cluster Random Y Y Y N Y Y N N N N N N Y N N
[320] FFOA Optimization Equal Y Not specified Random N Y N N N N N N Y N N N N N N
[326] K-medoid Machine learning Equal Y Cluster Random N Y N Y N N N N N N N N N Y N
[328] ISFC-BLS Fuzzy and ACO Equal Y Cluster Uniform N Y Y Y Y Y Y N N N Y Y Y Y N

[329] HTMS Machine learning Equal Y Cluster Random Y Y N Y N Y N N Garnished attack, Bad mouthing attack and Self-promoting attack

LEC - Less Energy Consumption, DA - Data Aggregation, Dep - Deployment, FT - Fault Tolerance, Sc- Scalability, Bl- Blackhole attack, LB- Load Balancing, Re - Reliability, Aut - Authentication, C-Confidentiality, I- Integrity, Gr-Greyhole attack, A - Availability, Sp - Spoofing attack, Si - Sinkhole attack, Wo -
Wormhole attack, He - Hello flood attack, Sy - Sybil attack, CLDEH - CrossLayer Design and Energy Harvesting, SEIF -Secure and Efficient Intrusion-Faultntolerant protocol, OHC - One way Hash Chain, INSENS - INtrusion-tolerant routing protocol for wireless SEnsor NetworkS, Sec-TEEN - Secure Threshold
sensitive Energy Efficient sensor Network, Fruit Fly Optimization Algorithm, S/N - Spoofing or Node Capture Attack, ISFC-BLS -Intelligent and Secured Fuzzy Clustering Algorithm using Balanced Load Sub-cluster formation, HTMS - Hierarchical Trust Management Scheme, Y - Yes, N - No.

Computer Science Review 40 (2021) 100376

J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

preserve more inter-cluster energy to avoid the hot-spot prob- 7.1. Research findings
lem [339]. EBUC chooses high-energy nodes as CHs, determined
by the PSO algorithm that reduces the cost function. For inter- High-Quality Clustering Algorithm (HQCA -WSN) method ex-
cluster data transfer, a CH selects a relay node by using the hibits better scalability, improves cluster quality, and reduces
greedy algorithm. The EBUC protocol performs its functions in the error rate during clustering [10]. Genetic Algorithm-Adaptive
two phases; the set-up phase and the steady-state phase. After Neuro-Fuzzy Inference System (GA-ANFIS) is an effective method
the set-up process, the BS sends a message containing the cluster for CH selection, data routing, and detecting the malicious node
and multi-hop routing information. By using this message, each in a WSN [141]. Node Ranked (NR) LEACH avoids the random
node ensures its task, and every CH recognizes its relay node. collection of cluster heads, and their unequal distribution [243].
To avoid the imbalanced distribution of energy and fault Distributed Cluster Head Scheduling (DCHS) can be effectively
tolerance issues caused by EBUC, another protocol known as adopted for energy sensitive applications in WSN [249]. The Dy-
PSO based Unequal and Fault-tolerant Clustering (PSO-UFC) is namic CH Selection Method (DCHSM) solves the problem of cov-
used [340]. This protocol solves the hot-spot problem, the cost erage overlapping in the cluster communication [250].
function balances the inter-cluster and intra-cluster energy con- Random selection of cluster heads leads to an unbalanced
sumption, multi-hop routing tree construction ensures the net- distribution of network load [255]. HSA requires many iterations
work connectivity, and selection of surrogate cluster head ad- for the selection of optimal cluster head [295]. Artificial Bee
dresses the fault tolerance issue. Moreover, in the PSO-UFC pro- Colony optimization (ABCO) based CHs selection approaches are
tocol, the network’s connectivity is preserved by selecting an ideal in exploitation or exploration, but not both. Rule-based ap-
additional cluster head known as Surrogate Cluster Head (SCH) proaches in machine learning techniques are mainly used to save
when there is a sudden failure in the Master Cluster Heads the energy of sensor nodes, while transmission of information,
(MCHs). PSO-UFC selects more MCHs nearer to the sink node to thereby prolong the network lifetime. A hybrid meta-heuristic
solve the hot-spot problem. The limitation of PSO-UFC is that it approach using the Improved Genetic Algorithm (IGA) and Binary
failed to address the scenario in variable traffic load. Ant Colony Algorithm (BACA) ensures optimal coverage, mini-
A Sink Mobility based Energy Balancing Unequal Clustering mizes the sensing of redundant information, and optimizes the
(SMEBUC) protocol forms unequal clusters and balances the con- multi-objective function by determining the minimum number
sumption of energy by using the Shuffled Frog Leaping Algorithm of sensors [202]. Square grid deployment is simple to implement,
(SFLA) [341]. It identifies a relay node for inter-cluster routing achieves better coverage performance, and supports grid merging
and uses the sink’s location to eliminate the hot spot problem. for clustering than hexagonal grid deployment [288].
The communication between the CH and the BS comprises two The K-medoid algorithm discourages malicious attacks and
stages; formation of cluster and data transmission. To balance vulnerabilities, leads to network reliability improvement [326].
each sensor node’s energy, SMEBUC incorporates the CH compe- In the case of a trust-based key management system, a rekeying
tition scheme in the clustering process, and the BS switches to scheme can be used, in which the key(s) to the sensor node
its default location. However, the performance is degraded with can be assigned and renewed based on trust value update [329].
packet delivery ratio and end-to-end delay due to node and path Location-aware nodes increase the energy overhead and are un-
loss routing redundancy. suitable for many real-time applications [332]. The competi-
tion radius scheme minimizes the intra-cluster energy consump-
6.3. Machine learning techniques tion [343]. Due to the random distribution of nodes, a similar
clustering mechanism leads to unbalanced energy consumption,
A distributed competitive algorithm, known as Fuzzy Energy- and to solve this problem; Unequal clustering algorithms are
Aware Unequal Clustering (EAUCF) deals with the hot-spot issue, devised [347].
and the variations in the CH radius estimation is handled by the
fuzzy logic approach [274]. EAUCF is designed with stationary 7.2. Challenges and open issues
sensor nodes. Moreover, parameters such as local distance, den-
sity, degree of nodes, distance to the sink, and remaining energy • The Selection of C parameter in the C-means algorithm is a
can be used further to improve this algorithm’s performance. crucial issue due to the optimal CH selection based on node
Distributed Unequal Clustering using Fuzzy logic (DUCF) balances density.
the distribution of nodes uniformly within the clusters to avoid • For wireless sensor networks, mobility has often been re-
hot-spot issue [277]. The input metrics to Fuzzy Inference System garded as an important challenge. The nature of mobility can
(FIS) are the remaining energy, distance to sink node, and degree induce numerous issues such as accessibility, reliability, sta-
of nodes. The output metrics are the size of the clusters and the bility, etc. Node mobility facilitates the process of clustering
choice to become CH. Mamdani and centroid techniques are used very challenging, as node membership changes dynamically,
for fuzzification and defuzzification, respectively, as shown in triggering the growth of clusters over time.
Fig. 16. Similarly, various fuzzy-based algorithms provide uniform • Heterogeneous sensors cause technical issues related to rout-
energy among the CHs, avoids the hot-spot issue by consid- ing because the specific application needs a different inte-
ering more input parameters and fuzzy rule, and extend the gration of sensors for monitoring humidity, pressure, tem-
network lifetime [302,342–346]. The various unequal clustering perature, detecting motion, and tracking moving objects.
approaches based on classical methods, optimization approaches, • Load balancing is a fundamental challenge that can arise
and machine learning techniques are shown in Table 12. because of the unequal distribution of nodes in the sensor.
The ‘‘Hot spot problem’’ can be solved by load balancing, but
7. Research findings, challenges, open issues, and future direc- the distribution of packets to the BS is a challenge.
tions • The error rate during cluster formation is an important issue
concerned with the formation of high-quality clusters.
This section deals with the research findings, challenges, and • An energy-efficient and secure inter and intra-cluster commu-
future directions, which helps the researchers explore further nication is another critical challenge in clustering algorithms
aspects of clustering in WSNs based on cluster formation, cluster due to the deployment of tiny sensor nodes which are
head selection, routing, reliability, and security. unattended in most cases.
32
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Table 11
Features, simulation parameters and future directions of security-based algorithms.
Ref. Feature Experiment (or) Simulation Nature and no. of nodes Area of interest (AOI) Simulation environment Future direction/Remarks
performance parameters
2
[316] Selects secure NS-2 Static, 200 nodes Square, 100 × 100 m Packet loss, routing overhead, Hash function, cryptographic
energy-efficient path to delay technique can be used to
cluster head improve the scheme

[317] Provides authentication Avrora tool Static, 100 nodes Square Meantime to failure, resiliency Performance analysis can be
through symmetric compared with metrics related
cryptography mechanism to security attacks
known OHC

[318] INSENS avoids malicious NS-2 Berkeley MICA2 motes Effects of a node failure In a vast network,
attacks & limits the communication between nodes
damage an intruder can and BS takes place in
cause hierarchies

[319] Avoids collisions during MATLAB Static, 100 nodes Square Data overload, energy Symmetric key is used to
communication between consumption, alive nodes protect the aggregate result
CH and BS between CH and BS

[320] Discover optimal nodes on Case-study and simulation Static, 200 nodes Square, 100 × 100 m2 Compromised traffic, energy The security of FFOA
lower attacking rounds cost, attacking rounds and techniques can be further
execution time enhanced in the future

[326] Anomaly detection method NS-2 Static, 18 nodes Square, 500 × 500 m2 End to end delay, send and To detect more anomalies in
receive traffic the network

[328] Enhances energy efficiency Simulation Static, 51 nodes 1300 × 1000 m2 Packet drop, PDR, energy ISFC-BLS can be enhanced by
and makes the network consumption, and end to end transferring high priority
secure Delay packets in a secured manner

[329] Reduces communication MATLAB Static, 55 nodes Square, 1.6 × 1.6 km2 Memory overhead, trust value Rekeying phenomenon can be
overhead, computational analysis, communication incorporated for trust-based
time & memory utilization overhead system

Fig. 16. DUCF fuzzy inference system.

• Selection of cluster head, suitable cluster size, and cluster man- • Most studies consider the sensing and transmission range of
agement are some of the challenges to be tackled when the nodes as a binary sensing model. However, this assertion
developing clustering algorithms. does not suit perfectly since sensing ranges depend on a
• As the network scale increasing on a large-scale, network highly irregular location, and the range of communication
scalability becomes a fundamental issue. may vary irregularly. Therefore, rather than compromise
• Each topology has specific advantages and limitations, how with this binary sensing model, it is a challenging task to
to combine their advantages into a single solution to the follow some other realistic model.
issue of clustering remains a challenge.
• Detecting and maintaining routes is a critical issue since the 7.3. Future directions
failure of node and energy constraints give rise to topologi-
cal alterations. • Transformation of current simulation scenarios into cross-
• Though WSNs are known to be highly scalable networks, layer and real-world applications are critical problems that
one of the critical challenge affecting these networks is the can be solved through optimization techniques.
network management. • Future directions can focus on computing the quality of clus-
• To determine the appropriate nodes as optimal CH have ters since it is more critical in evaluating the performance of
to undergo different evaluation criteria that make the task the clustering process.
more complicated.
• Determining the optimal frequency of cluster head rota-
• Although the sensor nodes are configured with a global posi-
tion maximizes the network lifetime during the clustering
tioning system to determine their locations, the challenge is
process.
to monitor the location of the sensors and in turn, the design of
• Mobile and multiple sinks can be adapted to increase en-
routing protocols at the preliminary phase of route discovery.
ergy efficiency, reduce energy-hole problems, and extend
• Providing the desired coverage and connectivity with a lim-
the network lifetime. However, this results in the flooding
ited number of nodes is indeed very challenging in WSNs.
of information. Future research can focus on reducing this
• Emerging demand in underwater sensor networks, collec-
control overhead.
tion of oceanographic data, pollution monitoring, etc. at-
tracts researchers to the three-dimensional (3D) networks. • Loop generation while routing data packets impact directly
The 3D network configuration is far more complicated than on the network performance because it increases the delay
the 2D network because, as in real-life situations, only the in the packet, bandwidth and energy resources get wasted.
assumption does not fulfill all the 3D constraints. Hence, loop generation can be avoided or reduced during
• While schemes such as k-coverage and m-connectivity might packet transmission.
continue functioning even if specific nodes could be affected • The integration of mathematical modeling, along with simu-
with extended network lifetime and overcome the network lation studies, can be focused so that fair comparison results
fault, the challenge addressed is the increased cost of the of the algorithms can be generated.
WSNs.
33
 J. Amutha, S. Sharma and S.K. Sharma
Table 12
Unequal clustering approaches based on classical, optimization and machine learning techniques.
Ref. Proto- Type Cluster CH selection CH selection Energy Cluster com- Feature/Objective Deployment Location- Role Energy model Simulation Simulation Nature & Node Area of Type of Sink Future
col/Algorithm size type efficiency munication of nodes aware Relay/aggregation environment type & No. of Interest (AOI) & location direction/Remarks
performance nodes
parameters

[131] NCRO- Dis. Unequal Average distance, Optimization Yes Inter and Enhance the lifetime Random No Relay, Radio model Matlab Residual energy, Static, Square, Static, Fault-tolerance is
UCRA mean distance & Intra of the network & aggregation network lifetime, Homogeneous, 500 × 500 (250, 250) not considered in
and energy ratio load balancing number of alive 200 to 800 nodes m2 m, designing
nodes, packets (500, 250) m NCRO-UCRA
received,
convergence rate

[274] EAUCF Dis. Unequal Based on Fuzzy (ML) Yes Intra-cluster Load balancing, Random Yes Relay, Data First-order Simulation Alive sensor Static, Square, Static EAUCF may be
remaining energy maximize lifetime aggregation radio model nodes, number of Homogeneous, 200 × 200 (100, 100) adapted to
and sink node clusters, total 100 nodes m2 m, manage the
distance remaining energy (100, 250) m mobile nodes

[277] DUCF Dis. Unequal Remaining energy, Fuzzy Yes Inter and Load balancing, Random No Relay, First-order Matlab Energy Static, Square, Static, DUCF can be
sink node (ML) Intra improves network aggregation radio model consumption per Homogeneous, 200 × 200 Middle, verified by more
distance and lifetime round, number of 100 nodes m2 corner and parameters such
degree of node message outside AOI as node centrality
transmissions, and coverage
network lifetime redundancy

[302] DFCR Dis. Unequal Sensor nodes near Fuzzy (ML) Yes Intra cluster Energy-efficient and Random Yes Relay, First-order Matlab Fuzzy Network lifetime, Static, Square, Static, Future work
base station and energy balancing aggregation radio model Toolbox energy efficiency Homogeneous, 100 × 100 (50, 0) m, focuses on
higher energy 100 nodes m2 (50, 50) m handling DFCR
using mobile
WSNs.

[330] Impro- Dis. Unequal Highest residual Deterministic Yes Inter and Load balancing, Uniform and No Relay, First-order Matlab Number of cluster Static, Square, Pre-defined The selection of a
ved energy (Classical) Intra improve network non-uniform aggregation radio model heads, alive Heterogeneous 200 × 200 and Static, relay node for
EADUC lifetime nodes, average 100 nodes m2 (250, 100) m forwarding data
energy is based in terms
consumption per of energy.
round, network
remaining energy,
network lifetime

[331] URBD Dis. Unequal Minimum distance Hybrid Yes Inter and Optimize the Random No Aggregation First-order Matlab Network lifetime, Static, Square, Static, Future work can
and remaining (Classical) Intra scalability and lifetime radio model number of control Homogeneous, 200 × 200 (100, 100) m focus on
energy of the network messages 100 nodes m2 multi-objective
clustering
34

[332] LUCA Dis. Unequal Residual energy Random Yes Inter and Reduces energy Random Yes Aggregation Not specified Matlab Total energy Static, Square Static, Chance of CH
(Classical) Intra consumption consumption for Homogeneous Inside AOI selection depends
different network on the position
sizes information

[333] COCA Dis. Unequal Sensor nodes Hybrid Yes Inter and Maximize sensor Random No Relay, Radio energy NS-2 Optimal number Static, Square, 104 Static Challenge is to
having the same (Classical) Intra network lifetime aggregation model of transmission Homogeneous, m2 find the optimal
probability rounds, lifetime variable nodes value
analysis for
different network
sizes, remaining
energy

[334] UCR Dis. Unequal Competition Hybrid Yes Inter and Reduce the hot spot Uniform No Relay, Not specified Simulation Number of Static, (0,0) Static, Rotation of CH
diameter (Classical) Intra problem aggregation clusters, network Homogeneous, –(400,200) (500,100) m and energy are
lifetime, energy 600 nodes m2 not appropriate
efficiency, number for balancing
of alive sensor power
nodes, consumption

[335] EDDU-CA Dis. Unequal Energy, degree Compound Yes Inter and Balance energy Uniform No Aggregation Not specified Matlab Number of Static, Square, Static Future work focus
and distance (Classical) Intra consumption and clusters, average Homogeneous, 200 × 200 on heterogeneous

Computer Science Review 40 (2021) 100376

extends network energy 300 nodes m2 nodes
lifetime consumption,
number of nodes
alive

[337] UHEED Dis. Unequal Higher Hybrid Yes Inter and Preventing the early Random No Relay, Radio model Simulation Network lifetime, Static, 2000 × 2000 Outside the To check the
residual energy (Classical) Intra death of nodes Aggregation residual energy Homogeneous, grid grid simulation results,
1000 nodes dimension a mathematical
model can be
designed

[338] UCCG-RA Dis. Unequal Based on the Compound Yes Inter and Increases the link Random No Relay, Radio model OM-NET++ Network lifetime Static, Square, Static, Vote-based and
weight of each (classical) Intra quality & extends aggregation Homogeneous, 500 × 500 (250, 500) m graph-based
node network lifetime 200 nodes m2 protocol manage
CH energy
consumption

[339] EBUC Cen. Unequal High-energy Heuristic - Yes Inter and Balance the energy Uniform No Relay, First-order Matlab Number of alive Static, Square, Static, Parameters of PSO
nodes PSO Intra and extends the aggregation radio model sensor nodes, Homogeneous, 500 × 500 Inside, can be considered
(Optimization) network distribution of 200 nodes m2 outside AOI for performance
lifetime dead nodes evaluation

(continued on next page)

 J. Amutha, S. Sharma and S.K. Sharma
Table 12 (continued).
Ref. Proto- Type Cluster CH selection CH selection Energy Cluster com- Feature/Objective Deployment Location- Role Energy model Simulation Simulation Nature & Node Area of Type of Sink Future
col/Algorithm size type efficiency munication of nodes aware Relay/aggregation environment type & No. of Interest (AOI) & location direction/Remarks
performance nodes
parameters

[340] PSO-UFC Dis. Unequal Residual energy & PSO - (Opti- Yes Inter and Addressing the issues Random No Relay, First-order Matlab Network lifetime, Static, Square, Static, Future work may
better location mization) Intra of hot-spot and fault aggregation radio model residual energy, Homogeneous, 200 × 200 Center & focus on variable
tolerance total energy 200 nodes m2 right corner traffic load
consumption

[342] EUDFC Dis. Unequal Based on the Fuzzy (ML) Yes Not specified Maximize the lifetime Random No Data First-order Matlab Scenario with Static, Square, Pre-defined Heterogeneous
turn-based Aggregation radio model different locations Homogeneous, 100 × 100 and Static nodes can be
scheduling of base station 100 nodes m2 considered

[343] FUCA Dis. Unequal Distance, energy Fuzzy (ML) Yes Intra cluster Energy conservation is Random No Aggregation Free space Matlab Residual energy, Static, Square, Static, Not affected by
& density of attained model alive nodes, dead Homogeneous, 100 × 100 Center and the density &
nodes nodes 100, 200 nodes m2 corner location of sink
node

[344] LAUCF Dis. Unequal FCM value, Fuzzy C- Yes Inter and CH rotation strategy Random No Aggregation Radio energy Matlab Number of alive Static, Square, Static, Provides better
Chebyshev Means Intra balance the energy dissipation nodes, network Heterogeneous 200 × 200 (100,100) m, distribution of the
distance (ML) consumption model lifetime, residual 100 nodes m2 (100,250) m number of
and threshold energy, optimum clusters
values energy dissipation
of network, data
transmission rate,
throughput

[345] EBCAG Dis. Unequal Residual energy Gradient Yes Inter and Avoids hot-spot issue Uniform No Relay, First-order Simulation Average energy Static, Circle Static, In real-time
35

of each node Descent Intra aggregation Radio Model consumption, Homogeneous, (0,0) m applications,
(ML) variance of 600 nodes uniform
amount of energy, deployment is not
number of CHs, feasible
alive nodes, dead
nodes, network
lifetime

[346] IFUC Dis. Unequal Distance, energy, Fuzzy Yes Inter and Provides energy Random No Relay, Multi-path Matlab Number of dead Static, Square, Static, Future work can
and density (ML) Intra efficiency and ensures aggregation channel nodes, variance of Homogeneous, 200 × 200 (100,250) m focus on
balanced network fading model energy 400 nodes m2 analyzing more
metrics during
the process of
clustering

PEBECS Dis. Unequal Based on the Deterministic Yes Inter and Improve the scalability Random No Relay, First-order Theoretical Ratio of node Static, Square Static, CHs are selected
[348] node weight, (Classical) Intra and prolongs the aggregation radio model analysis and alive via four Homogeneous Outside the by a heuristic
residual energy network lifetime simulation different sensor field algorithm with
algorithms node weight

[347] EAKDE Dis. Unequal Distance to sink Fuzzy (ML) Yes Intra cluster Balance the loss of Random Yes Relay, First-order Matlab R2017 Network lifetime, Static, Square, Static, EAKDE can be
node and energy between CHs aggregation radio model energy efficiency Homogeneous, 100 × 100 (0, 50) m, further
residual energy 200 nodes m2 (50, 50) m investigated based

Computer Science Review 40 (2021) 100376

on other metrics

EAUCF - Energy-Aware Unequal Clustering Algorithm, LUCA - Location-based Unequal Clustering Algorithm, EUDFC - Enhanced Unequal Distributed Type-2 Fuzzy Clustering Algorithm, LAUCF - Low energy Adaptive Unequal Clustering protocol using Fuzzy c-means, FCM - Fuzzy c-means, URBD - Unequal
cluster-Radius approach Based on node Density, Improved EADUC - Improved Energy-Aware Distributed Unequal Clustering, EBUC - Energy Balanced Unequal Clustering, PSO-UFC - Particle Swarm Optimization based Unequal and Fault-tolerant Clustering protocol, DFCR - Distributed Fuzzy logic-based unequal
Clustering approach and Routing algorithm, EAKDE - Energy-Aware adaptive Kernel Density Estimation, FUCA - Fuzzy based Unequal Clustering Algorithm, COCA -Constructing Optimal Clustering Architecture, UCR - Unequal Cluster-based Routing, EDDUCA - Energy Degree Distance Unequal Clustering Algorithm,
UHEED - Unequal Hierarchical Energy-Efficient Distributed Clustering, PEBECS - Partition Energy Balanced and Efficient Clustering Scheme, EBCAG - Energy-Balancing unequal Clustering Approach for Gradient-based routing, IFUC - Improved Fuzzy Unequal Clustering algorithm, DUCF - Distributed Unequal
Clustering using Fuzzy logic, NCRO- UCRA - Novel Chemical Reaction Optimization-based Unequal Clustering and Routing Algorithm, UCCGRA - Unequal Clustering Theory and Connected Graph Theory for Routing Algorithm Dis. - Distributed, Cen.-Centralized.
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

• The integration of optimization approaches and machine learn- each round. Typically, network lifetime, intra-cluster, and inter-
ing techniques in WSNs can be used to optimize and extend cluster distances are the basic features of the optimization-based
the network performance in the clustering process. algorithms’ decision variables. Machine learning offers a wide
• The implementation of multi-objective firefly algorithms range of techniques to improve the ability of WSN to respond to
evaluate the optimal CH in WSNs. A Hybrid Artificial Bee the complex behavior of its environment. To address the issues
Colony and Bacterial Foraging Algorithm can be devised for from various research areas, Table 3 analyzes reviews that have
energy balanced CH selection to evaluate its performance adopted machine learning methods. It has become apparent from
with the contributed topic of research. To solve the CH the discussion that many design issues in WSNs have been re-
selection problem and to design more suitable algorithms solved using several machine learning methods. However, various
for initial clustering, various PSO variants can be used to other challenges like developing distributed lightweight mecha-
minimize the number of computation and extend network nisms, hierarchical patterns in clustering, and adopting machine
lifetime. learning in resource management problems of WSNs are still
• Considering several cluster configuration metrics into ac- open and require additional research initiatives.
count, such as node degree, mobility, battery power, clus- The current works are reported in respective tables to high-
ter size and optimizing their parametric values for various light various research gaps. There are numerous research chal-
application scenarios. lenges for WSN, while energy efficiency has prevailed as the
• Meta-heuristic algorithms based on Cuckoo and Harmony primary requirement in various WSN applications. Based on the
Search can be improved to solve the various issues like delay specific applications, the sensor node contains power-hungry el-
sensitivity and network with obstacles. ements, like GPS - for identifying their locations and camera - for
• Security issues can be addressed both at node and agent level capturing images [349]. Several energy conservation techniques
to enhance the routing algorithms further. such as cooperative communication, data fusion, overhead re-
• Ring Routing can be improved so that it works with several duction, clustering have been studied for decades. Among these,
mobile sinks more effectively. Moreover, different mobility the clustering scheme has gained the maximum attention be-
patterns such as nomadic or controlled sink mobility rather cause it minimizes the number of transmissions and the dis-
than the constant speed sink mobility can be explored. tance of nodes to the sink node. Moreover, clustering achieves
• Genetic algorithms can be enhanced by combining sleep and energy-efficiency, scalability, and energy efficiency, especially in
awake principles since all nodes do not remain involved large-scale multi-hop environments.
throughout the system. To address the combined challenge Hence, this review is initiated by analyzing the existing survey
of clustering and routing, GA can be improved in a dis- papers that focus on WSNs clustering and presented in Table 1.
The various schemes, algorithms, protocols, challenges, limita-
tributed manner.
tions are addressed for each review paper, and the observations
• The impact of noisy channels, physical and MAC layer chal-
drawn from the extensive literature work are summarized at the
lenges on the performance of clustering approaches can be
end of the related work section. The taxonomy of cluster analysis
examined by optimization algorithms.
based on various strategies such as environment, characteristics,
• There is still more research work that can focus on the
methods, process, objectives, and routing are examined by var-
selection of reinforcement-based routing parameter values,
ious attributes. The strategies used for various aspects of the
such as Q-value learning rates, NQ value update rates, par-
clustering process are described along with the methods used, ap-
ent reset threshold, can indeed be optimized for specific
plications, advantages, and limitations. The comparison between
applications.
different optimization approaches used in WSNs, in terms of
mimic behavior, operator, feature, criteria for CH selection, and
7.4. Summary limitations are summarized in Table 2.
Moreover, several works related to various aspects are com-
Based on the comparative methods provided in Tables 1–12, prehensively discussed. The precise comparisons are carried out
there have been several efforts taken by researchers to develop by appropriate criteria for all aspects and tabulated for easy refer-
effective clustering-based protocols, methods, and algorithms. In ence. Thus, comparison of survey papers for more than a decade;
this review, each of these schemes is addressed based on different most detailed and fine-grained taxonomy of various methods to
strategies like classical methods, optimization approaches, and WSN clustering; strategies used along with applications, limita-
machine learning techniques for clustering process in WSNs. In tions, and advantages; various aspects based on each strategy;
classical based strategies, CH’s selection is mainly performed in comparative study of each aspect; analysis of security-based al-
a distributed or centralized manner that employs a standard gorithms; evaluation of reliability-based approaches; investiga-
formula by considering various metrics of the sensor nodes. These tion of unequal clustering schemes; research findings, challenges,
strategies do not address the relevant criteria and their impor- open issues, and future directions; in which all these considers
tance of CH selection. Besides, in these methods, scalability is the proposed review innovative.
not considered efficiently. However, in recent studies, this issue
is solved by combining multiple variables and introducing multi- 8. Conclusion
level routing. Most of these strategies are not application-specific,
and their performance is already defined for all applications. Considering the energy efficiency of nodes in WSNs and the
Classical routing protocols are unreliable due to changes in the importance of clustering as an efficient solution for minimiz-
power supply of WSNs that are restricted by energy. There are ing energy consumption, this review focuses on clustering and
different 2N -1 solutions for N sensor nodes. Clustering is, there- cluster-based algorithms. These algorithms provide insights into
fore, an NP-hard problem in which optimization techniques can the various aspects like cluster formation, cluster head selec-
be efficient. tion, routing, reliability, security, and unequal clustering. Various
Optimization strategies are effective than the classical based strategies study these aspects in terms of classical, optimization,
strategies in choosing relevant CHs, yet, they have high com- and machine learning techniques. For a more in-depth insight
putational complexity and impose a high delay. The reason is into each viewpoint, the reliable parameters with the correspond-
that in optimization approaches, CH selection is performed in ing views are considered and presented in their respective tables.
36
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

Moreover, to provide reliable information and enrich the readers, [17] B.P. Deosarkar, N.S. Yadav, R. Yadav, Clusterhead selection in clustering
this survey aims at research papers from various reputed journals. algorithms for wireless sensor networks: A survey, in: 2008 International
Conference on Computing, Communication and Networking, IEEE, 2008,
In effort, research findings, challenges, open issues, and future
pp. 1–8.
directions provide a unique perspective and a primary source for [18] Z. Jin, Y. Jian-Ping, Z. Si-Wang, L. Ya-Ping, L. Guang, A survey on position-
researchers to explore more in this area and help discover new based routing algorithms in wireless sensor networks, Algorithms 2 (1)
topics in wireless sensor networks. (2009) 158–182.
[19] M. Dorigo, M. Birattari, T. Stutzle, Ant colony optimization, IEEE Comput.
CRediT authorship contribution statement Intell. Mag. 1 (4) (2006) 28–39.
[20] S.A. Adubi, S. Misra, A comparative study on the ant colony optimiza-
tion algorithms, in: 2014 11th International Conference on Electronics,
J. Amutha: Resources, Writing - original draft preparation, Computer and Computation (ICECCO), IEEE, 2014, pp. 1–4.
Visualization, Investigation, Validation, Writing - reviewing and [21] W.-H. Liao, Y. Kao, C.-M. Fan, Data aggregation in wireless sensor
editing. Sandeep Sharma: Conceptualization, Methodology, Su- networks using ant colony algorithm, J. Netw. Comput. Appl. 31 (4) (2008)
pervision, Resources, Writing - original draft preparation, Vi- 387–401.
[22] T.A. Runkler, Ant colony optimization of clustering models, Int. J. Intell.
sualization, Investigation, Validation, Writing - reviewing and
Syst. 20 (12) (2005) 1233–1251.
editing. Sanjay Kumar Sharma: Conceptualization, Methodology, [23] T. İnkaya, S. Kayalıgil, N.E. Özdemirel, Ant colony optimization based
Supervision, Writing - reviewing and editing. clustering methodology, Appl. Soft Comput. 28 (2015) 301–311.
[24] B.C. Mohan, R. Baskaran, A survey: Ant colony optimization based recent
Declaration of competing interest research and implementation on several engineering domain, Expert Syst.
Appl. 39 (4) (2012) 4618–4627.
[25] S.K. Singh, M. Singh, D. Singh, A survey of energy-efficient hierarchical
The authors declare that they have no known competing finan- cluster-based routing in wireless sensor networks, Int. J. Adv. Netw. Appl.
cial interests or personal relationships that could have appeared (IJANA) 2 (02) (2010) 570–580.
to influence the work reported in this paper. [26] R.V. Kulkarni, A. Förster, G.K. Venayagamoorthy, Computational intelli-
gence in wireless sensor networks: A survey, IEEE Commun. Surv. Tutor.
13 (1) (2011) 68–96.
Acknowledgments
[27] K. Vivek, C. Narottam, S. Surender, A survey on clustering algorithms for
heterogeneous wireless sensor networks, Int. J. Adv. Netw. Appl. 2 (4)
We want to acknowledge Gautam Buddha University, Greater (2011) 745–754.
Noida, India, for providing us with the institutional support. We [28] M. Saleem, G.A. Di Caro, M. Farooq, Swarm intelligence based routing
extend our sincere thank to the editor and all the anonymous protocol for wireless sensor networks: Survey and future directions,
reviewers for providing helpful comments and suggestions. Inform. Sci. 181 (20) (2011) 4597–4624.
[29] S. Selvakennedy, S. Sinnappan, Y. Shang, A biologically-inspired clustering
protocol for wireless sensor networks, Comput. Commun. 30 (14–15)
References (2007) 2786–2801.
[30] S. Naeimi, H. Ghafghazi, C.-O. Chow, H. Ishii, A survey on the taxonomy
[1] I.F. Akyildiz, W. Su, Y. Sankarasubramaniam, E. Cayirci, Wireless sensor of cluster-based routing protocols for homogeneous wireless sensor
networks: a survey, Comput. Netw. 38 (4) (2002) 393–422. networks, Sensors 12 (6) (2012) 7350–7409.
[2] J. Yick, B. Mukherjee, D. Ghosal, Wireless sensor network survey, Comput. [31] X. Liu, A survey on clustering routing protocols in wireless sensor
Netw. 52 (12) (2008) 2292–2330. networks, Sensors 12 (8) (2012) 11113–11153.
[3] W. Dargie, C. Poellabauer, Fundamentals of Wireless Sensor Networks: [32] A.M. Zungeru, L.-M. Ang, K.P. Seng, Classical and swarm intelligence based
Theory and Practice, John Wiley & Sons, 2010. routing protocols for wireless sensor networks: A survey and comparison,
[4] T. Rault, A. Bouabdallah, Y. Challal, Energy efficiency in wireless sensor J. Netw. Comput. Appl. 35 (5) (2012) 1508–1536.
networks: A top-down survey, Comput. Netw. 67 (2014) 104–122. [33] S. Tyagi, N. Kumar, A systematic review on clustering and routing
[5] J. Amutha, S. Sharma, J. Nagar, Wsn strategies based on sensors, techniques based upon leach protocol for wireless sensor networks, J.
deployment, sensing models, coverage and energy efficiency: Review, Netw. Comput. Appl. 36 (2) (2013) 623–645.
approaches and open issues, Wirel. Pers. Commun. (2019) 1–27. [34] N.K. Mishra, V. Jain, S. Sahu, Survey on recent clustering algorithms in
[6] D.K. Sah, T. Amgoth, Parametric survey on cross-layer designs for wireless wireless sensor networks, Int. J. Sci. Res. Publ. 3 (4) (2013) 1–4.
sensor networks, Comp. Sci. Rev. 27 (2018) 112–134.
[35] N.A. Pantazis, S.A. Nikolidakis, D.D. Vergados, Energy-efficient routing
[7] J. Zhu, C.-H. Lung, V. Srivastava, A hybrid clustering technique using
protocols in wireless sensor networks: A survey, IEEE Commun. Surv.
quantitative and qualitative data for wireless sensor networks, Ad Hoc
Tutor. 15 (2) (2012) 551–591.
Netw. 25 (2015) 38–53.
[36] M. Afsar, M.-H. Tayarani-N, Clustering in sensor networks: A literature
[8] N. Sabor, M. Abo-Zahhad, S. Sasaki, S.M. Ahmed, An unequal multi-hop
survey, J. Netw. Comput. Appl. 46 (2014) 198–226.
balanced immune clustering protocol for wireless sensor networks, Appl.
[37] W. Guo, W. Zhang, A survey on intelligent routing protocols in wireless
Soft Comput. 43 (2016) 372–389.
sensor networks, J. Netw. Comput. Appl. 38 (2014) 185–201.
[9] M. Youssef, A. Youssef, M. Younis, Overlapping multihop clustering for
[38] M.R. Senouci, A. Mellouk, K. Asnoune, F.Y. Bouhidel, Movement-assisted
wireless sensor networks, IEEE Trans. Parallel Distrib. Syst. 20 (12) (2009)
sensor deployment algorithms: A survey and taxonomy, IEEE Commun.
1844–1856.
[10] A.A. Baradaran, K. Navi, Hqca-wsn: high-quality clustering algorithm Surv. Tutor. 17 (4) (2015) 2493–2510.
and optimal cluster head selection using fuzzy logic in wireless sensor [39] V. Sucasas, A. Radwan, H. Marques, J. Rodriguez, S. Vahid, R. Tafazolli,
networks, Fuzzy Sets and Systems 389 (2020) 114–144. A survey on clustering techniques for cooperative wireless networks, Ad
[11] P. Neamatollahi, M. Naghibzadeh, S. Abrishami, Fuzzy-based clustering- Hoc Netw. 47 (2016) 53–81.
task scheduling for lifetime enhancement in wireless sensor networks, [40] Z. Fei, B. Li, S. Yang, C. Xing, H. Chen, L. Hanzo, A survey of multi-objective
IEEE Sens. J. 17 (20) (2017) 6837–6844. optimization in wireless sensor networks: Metrics, algorithms, and open
[12] S. Ebrahimi, S. Tabatabaei, Using clustering via soccer league competition problems, IEEE Commun. Surv. Tutor. 19 (1) (2017) 550–586.
algorithm for optimizing power consumption in wsns (wireless sensor [41] E. Zitzler, L. Thiele, Multiobjective evolutionary algorithms: a comparative
networks), Wirel. Pers. Commun. (2020) 1–16. case study and the strength Pareto approach, IEEE Trans. Evol. Comput.
[13] O. Younis, M. Krunz, S. Ramasubramanian, Node clustering in wireless 3 (4) (1999) 257–271.
sensor networks: recent developments and deployment challenges, IEEE [42] Z. Zhang, K. Long, J. Wang, F. Dressler, On swarm intelligence inspired
Netw. 20 (3) (2006) 20–25. self-organized networking: its bionic mechanisms, designing principles
[14] D. Dechene, A. El Jardali, M. Luccini, A. Sauer, A Survey of Clustering and optimization approaches, IEEE Commun. Surv. Tutor. 16 (1) (2013)
Algorithms for Wireless Sensor Networks, Project Report, The University 513–537.
Of Western Ontario London, Ontario, Canada, 2006. [43] A.S. Rostami, M. Badkoobe, F. Mohanna, A.A.R. Hosseinabadi, A.K. San-
[15] A.A. Abbasi, M. Younis, A survey on clustering algorithms for wireless gaiah, et al., Survey on clustering in heterogeneous and homogeneous
sensor networks, Comput. Commun. 30 (14–15) (2007) 2826–2841. wireless sensor networks, J. Supercomput. 74 (1) (2018) 277–323.
[16] C.-M. Liu, C.-H. Lee, L.-C. Wang, Distributed clustering algorithms for data- [44] S. Arjunan, S. Pothula, A survey on unequal clustering protocols in
gathering in wireless mobile sensor networks, J. Parallel Distrib. Comput. wireless sensor networks, J. King Saud Univ. Comput. Inf. Sci. 31 (3)
67 (11) (2007) 1187–1200. (2019) 304–317.

37
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

[45] A. Shahraki, A. Taherkordi, Ø. Haugen, F. Eliassen, Clustering objectives [74] M. Younis, K. Akkaya, Strategies and techniques for node placement in
in wireless sensor networks: A survey and research direction analysis, wireless sensor networks: A survey, Ad Hoc Netw. 6 (4) (2008) 621–655.
Comput. Netw. (2020) 107376. [75] M. Abo-Zahhad, O. Amin, M. Farrag, A. Ali, Survey on energy consumption
[46] A. Singh, V. Kotiyal, S. Sharma, J. Nagar, C.-C. Lee, A machine learning models in wireless sensor networks, Open Trans. Wirel. Sens. Netw. 1 (1)
approach to predict the average localization error with applications to (2014).
wireless sensor networks, IEEE Access 8 (2020) 208253–208263. [76] M. Adnan, M.A. Razzaque, I. Ahmed, I.F. Isnin, et al., Bio-mimic opti-
[47] A. Singh, J. Nagar, S. Sharma, V. Kotiyal, A Gaussian process regression mization strategies in wireless sensor networks: A survey, Sensors 14
approach to predict the k-barrier coverage probability for intrusion (1) (2014) 299–345.
detection in wireless sensor networks, Expert Syst. Appl. (2021) 114603. [77] A. Abd Aziz, Y.A. Sekercioglu, P. Fitzpatrick, M. Ivanovich, A survey on
[48] A. Singh, S. Sharma, J. Singh, Nature-inspired algorithms for wireless distributed topology control techniques for extending the lifetime of
sensor networks: A comprehensive survey, Comp. Sci. Rev. 39 (2021) battery powered wireless sensor networks, IEEE Commun. Surv. Tutor.
100342. 15 (1) (2012) 121–144.
[49] S. Sharma, J. Nagar, Intrusion detection in mobile sensor networks: A case [78] C. Sergiou, P. Antoniou, V. Vassiliou, A comprehensive survey of conges-
study for different intrusion paths, Wirel. Pers. Commun. 115 (3) (2020) tion control protocols in wireless sensor networks, IEEE Commun. Surv.
2569–2589. Tutor. 16 (4) (2014) 1839–1859.
[50] S. Sharma, R. Mishra, A cross layer approach for intrusion detection in [79] M.A. Kafi, D. Djenouri, J. Ben-Othman, N. Badache, Congestion control
manets, Int. J. Comput. Appl. 93 (9) (2014). protocols in wireless sensor networks: A survey, IEEE Commun. Surv.
[51] S. Sharma, J. Singh, R. Kumar, A. Singh, Throughput-save ratio optimiza- Tutor. 16 (3) (2014) 1369–1390.
tion in wireless powered communication systems, in: 2017 International
[80] J. Xu, G. Yang, Z. Chen, Q. Wang, A survey on the privacy-preserving data
Conference on Information, Communication, Instrumentation and Control
aggregation in wireless sensor networks, China Commun. 12 (5) (2015)
(ICICIC), IEEE, 2017, pp. 1–6.
162–180.
[52] W.-S. Jung, K.-W. Lim, Y.-B. Ko, S.-J. Park, Efficient clustering-based data
[81] M. Conti, J. Willemsen, B. Crispo, Providing source location privacy in
aggregation techniques for wireless sensor networks, Wirel. Netw. 17 (5)
wireless sensor networks: A survey, IEEE Commun. Surv. Tutor. 15 (3)
(2011) 1387–1400.
(2013) 1238–1280.
[53] R.E. Mohamed, A.I. Saleh, M. Abdelrazzak, A.S. Samra, Survey on wireless
[82] R. Bista, J.-W. Chang, Privacy-preserving data aggregation protocols for
sensor network applications and energy efficient routing protocols, Wirel.
wireless sensor networks: a survey, Sensors 10 (5) (2010) 4577–4601.
Pers. Commun. 101 (2) (2018) 1019–1055.
[54] F. Wang, J. Liu, Networked wireless sensor data collection: issues, [83] J.M. de Fuentes, L. González-Manzano, O. Mirzaei, Privacy models in
challenges, and approaches, IEEE Commun. Surv. Tutor. 13 (4) (2010) wireless sensor networks: A survey, J. Sensors 2016 (2016).
673–687. [84] J. Jose, J. Jose, M. Princy, A survey on privacy preserving data aggregation
[55] C.A. Coello, An updated survey of ga-based multiobjective optimization protocols for wireless sensor networks, J. Comput. Inf. Technol. 22 (1)
techniques, ACM Comput. Surv. 32 (2) (2000) 109–143. (2014) 1–20.
[56] S. Ehsan, B. Hamdaoui, A survey on energy-efficient routing techniques [85] H. Xie, Z. Yan, Z. Yao, M. Atiquzzaman, Data collection for security
with qos assurances for wireless multimedia sensor networks, IEEE measurement in wireless sensor networks: a survey, IEEE Internet Things
Commun. Surv. Tutor. 14 (2) (2011) 265–278. J. 6 (2) (2018) 2205–2224.
[57] N. Li, N. Zhang, S.K. Das, B. Thuraisingham, Privacy preservation in [86] C. Zhu, C. Zheng, L. Shu, G. Han, A survey on coverage and connectivity
wireless sensor networks: A state-of-the-art survey, Ad Hoc Netw. 7 (8) issues in wireless sensor networks, J. Netw. Comput. Appl. 35 (2) (2012)
(2009) 1501–1514. 619–632.
[58] Y. Zeng, J. Cao, J. Hong, S. Zhang, L. Xie, Secure localization and location [87] P. Schaffer, K. Farkas, Á. Horváth, T. Holczer, L. Buttyán, Secure and
verification in wireless sensor networks: a survey, J. Supercomput. 64 (3) reliable clustering in wireless sensor networks: A critical survey, Comput.
(2013) 685–701. Netw. 56 (11) (2012) 2726–2741.
[59] Y. Gu, F. Ren, Y. Ji, J. Li, The evolution of sink mobility management in [88] E. Fasolo, M. Rossi, J. Widmer, M. Zorzi, In-network aggregation tech-
wireless sensor networks: A survey, IEEE Commun. Surv. Tutor. 18 (1) niques for wireless sensor networks: a survey, IEEE Wirel. Commun. 14
(2015) 507–524. (2) (2007) 70–87.
[60] R.V. Kulkarni, G.K. Venayagamoorthy, Particle swarm optimization in [89] W.B. Heinzelman, A.P. Chandrakasan, H. Balakrishnan, An application-
wireless-sensor networks: A brief survey, IEEE Trans. Syst. Man Cybern. specific protocol architecture for wireless microsensor networks, IEEE
C 41 (2) (2010) 262–267. Trans. Wirel. Commun. 1 (4) (2002) 660–670.
[61] S. Jabbar, R. Iram, A.A. Minhas, I. Shafi, S. Khalid, M. Ahmad, Intelligent [90] T.M. Behera, U.C. Samal, S.K. Mohapatra, Energy-efficient modified leach
optimization of wireless sensor networks through bio-inspired comput- protocol for IoT application, IET Wirel. Sensor Syst. 8 (5) (2018) 223–228.
ing: survey and future directions, Int. J. Distrib. Sens. Netw. 9 (2) (2013) [91] S.H. Kang, T. Nguyen, Distance based thresholds for cluster head se-
421084. lection in wireless sensor networks, IEEE Commun. Lett. 16 (9) (2012)
[62] I. Butun, S.D. Morgera, R. Sankar, A survey of intrusion detection systems 1396–1399.
in wireless sensor networks, IEEE Commun. Surv. Tutor. 16 (1) (2013) [92] A. Wang, D. Yang, D. Sun, A clustering algorithm based on energy
266–282. information and cluster heads expectation for wireless sensor networks,
[63] S.M. Mohamed, H.S. Hamza, I.A. Saroit, Coverage in mobile wireless sensor Comput. Electr. Eng. 38 (3) (2012) 662–671.
networks (m-wsn): A survey, Comput. Commun. 110 (2017) 133–150. [93] O. Younis, S. Fahmy, Heed: a hybrid, energy-efficient, distributed cluster-
[64] X. Liu, Routing protocols based on ant colony optimization in wireless ing approach for ad hoc sensor networks, IEEE Trans. Mob. Comput. 3 (4)
sensor networks: A survey, IEEE Access 5 (2017) 26303–26317. (2004) 366–379.
[65] J.W. Jung, M.A. Weitnauer, On using cooperative routing for lifetime opti-
[94] Y. Tao, Y. Zhang, Y. Ji, Flow-balanced routing for multi-hop clustered
mization of multi-hop wireless sensor networks: Analysis and guidelines,
wireless sensor networks, Ad Hoc Netw. 11 (1) (2013) 541–554.
IEEE Trans. Commun. 61 (8) (2013) 3413–3423.
[95] A. Shahraki, M.K. Rafsanjani, A.B. Saeid, Hierarchical distributed manage-
[66] G. Khanna, S.K. Chaturvedi, A comprehensive survey on multi-hop wire-
ment clustering protocol for wireless sensor networks, Telecommun. Syst.
less networks: milestones, changing trends and concomitant challenges,
65 (1) (2017) 193–214.
Wirel. Pers. Commun. 101 (2) (2018) 677–722.
[96] K. Cengiz, T. Dag, Improving energy-efficiency of wsns through lefca, Int.
[67] Z. Huang, H. Okada, K. Kobayashi, M. Katayama, A study on cluster
J. Distrib. Sens. Netw. 12 (8) (2016) 8139401.
lifetime in multi-hop wireless sensor networks with cooperative miso
scheme, J. Commun. Netw. 14 (4) (2012) 443–450. [97] K.C. Dag, Tamer, Energy aware multi-hop routing protocol for wsns, IEEE
[68] H. Yetgin, K.T.K. Cheung, M. El-Hajjar, L.H. Hanzo, A survey of network Access 6 (2017) 2622–2633.
lifetime maximization techniques in wireless sensor networks, IEEE [98] M. Sabet, H. Naji, An energy efficient multi-level route-aware cluster-
Commun. Surv. Tutor. 19 (2) (2017) 828–854. ing algorithm for wireless sensor networks: A self-organized approach,
[69] M. Radi, B. Dezfouli, K.A. Bakar, M. Lee, Multipath routing in wireless Comput. Electr. Eng. 56 (2016) 399–417.
sensor networks: survey and research challenges, Sensors 12 (1) (2012) [99] S. Verma, N. Sood, A.K. Sharma, Genetic algorithm-based optimized
650–685. cluster head selection for single and multiple data sinks in heterogeneous
[70] I. Snigdh, N. Gupta, Quality of service metrics in wireless sensor networks: wireless sensor network, Appl. Soft Comput. 85 (2019) 105788.
A survey, J. Inst. Eng. (India) B 97 (1) (2016) 91–96. [100] M. Al-Shalabi, M. Anbar, T.-C. Wan, Z. Alqattan, Energy efficient multi-hop
[71] R. Sumathi, M. Srinivas, A survey of qos based routing protocols for path in wireless sensor networks using an enhanced genetic algorithm,
wireless sensor networks, J. Inf. Process. Syst. 8 (4) (2012) 589–602. Inform. Sci. 500 (2019) 259–273.
[72] M.A. Yigitel, O.D. Incel, C. Ersoy, Qos-aware mac protocols for wireless [101] R. Khanna, H. Liu, H.-H. Chen, Self-organisation of sensor networks using
sensor networks: A survey, Comput. Netw. 55 (8) (2011) 1982–2004. genetic algorithms, Int. J. Sensor Netw. 1 (3–4) (2006) 241–252.
[73] R.A. Uthra, S.K. Raja, Qos routing in wireless sensor networks—A survey, [102] L. Bhasker, Genetically derived secure cluster-based data aggregation in
ACM Comput. Surv. 45 (1) (2012) 1–12. wireless sensor networks, IET Inf. Secur. 8 (1) (2014) 1–7.

38
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

[103] P. Kuila, S.K. Gupta, P.K. Jana, A novel evolutionary approach for load [130] P.S. Rao, P.K. Jana, H. Banka, A particle swarm optimization based energy
balanced clustering problem for wireless sensor networks, Swarm Evol. efficient cluster head selection algorithm for wireless sensor networks,
Comput. 12 (2013) 48–56. Wirel. Netw. 23 (7) (2017) 2005–2020.
[104] X. Yuan, M. Elhoseny, H.K. El-Minir, A.M. Riad, A genetic algorithm-based, [131] P.S. Rao, H. Banka, Novel chemical reaction optimization based unequal
dynamic clustering method towards improved wsn longevity, J. Netw. clustering and routing algorithms for wireless sensor networks, Wirel.
Syst. Manage. 25 (1) (2017) 21–46. Netw. 23 (3) (2017) 759–778.
[105] S.A.F. Aghda, M. Mirfakhraei, Improved routing in dynamic environments [132] M. Razzaq, S. Shin, Fuzzy-logic dijkstra-based energy-efficient algorithm
with moving obstacles using a hybrid fuzzy-genetic algorithm, Future for data transmission in wsns, Sensors 19 (5) (2019) 1040.
Gener. Comput. Syst. (2020). [133] P.A. Forero, A. Cano, G.B. Giannakis, Distributed clustering using wireless
[106] A. Alibeiki, H. Motameni, H. Mohamadi, A new genetic-based approach sensor networks, IEEE J. Sel. Top. Sign. Proces. 5 (4) (2011) 707–724.
for maximizing network lifetime in directional sensor networks with [134] A.Y. Ng, M.I. Jordan, Y. Weiss, On spectral clustering: Analysis and an
adjustable sensing ranges, Pervasive Mob. Comput. 52 (2019) 1–12. algorithm, in: Advances in Neural Information Processing Systems, 2002,
[107] N.A. Morsy, E.H. AbdelHay, S.S. Kishk, Proposed energy efficient algorithm pp. 849–856.
for clustering and routing in wsn, Wirel. Pers. Commun. 103 (3) (2018) [135] X. Liu, Atypical hierarchical routing protocols for wireless sensor
2575–2598. networks: A review, IEEE Sens. J. 15 (10) (2015) 5372–5383.
[108] P. Lalwani, S. Das, H. Banka, C. Kumar, Crhs: clustering and routing [136] S. Alam, G. Dobbie, Y.S. Koh, P. Riddle, S.U. Rehman, Research on particle
in wireless sensor networks using harmony search algorithm, Neural swarm optimization based clustering: a systematic review of literature
Comput. Appl. 30 (2) (2018) 639–659. and techniques, Swarm Evol. Comput. 17 (2014) 1–13.
[109] P. Lalwani, H. Banka, C. Kumar, Crwo: Clustering and routing in wireless [137] S. Shamshirband, A. Amini, N.B. Anuar, M.L.M. Kiah, Y.W. Teh, S. Furnell,
sensor networks using optics inspired optimization, Peer-to-Peer Netw. D-ficca: A density-based fuzzy imperialist competitive clustering algo-
Appl. 10 (3) (2017) 453–471. rithm for intrusion detection in wireless sensor networks, Measurement
[110] M.T. Mekonnen, K.N. Rao, Cluster optimization based on metaheuristic 55 (2014) 212–226.
algorithms in wireless sensor networks, Wirel. Pers. Commun. 97 (2) [138] J. Amutha, J. Nagar, S. Sharma, A distributed border surveillance (dbs)
(2017) 2633–2647. system for rectangular and circular region of interest with wireless sensor
[111] M. Ezhilarasi, V. Krishnaveni, An evolutionary multipath energy-efficient networks in shadowed environments, Wirel. Pers. Commun. (2020) 1–21.
routing protocol (emeer) for network lifetime enhancement in wireless [139] J. Yang, M. Xu, W. Zhao, B. Xu, A multipath routing protocol based
sensor networks, Soft Comput. 23 (18) (2019) 8367–8377. on clustering and ant colony optimization for wireless sensor networks,
[112] F. Gao, W. Luo, X. Ma, Energy constrained clustering routing method Sensors 10 (5) (2010) 4521–4540.
based on particle swarm optimization, Cluster Comput. 22 (3) (2019) [140] R. Ranganathan, B. Somanathan, K. Kannan, Fuzzy-based cluster head
7629–7635. amendment (fcha) approach to prolong the lifetime of sensor networks,
Wirel. Pers. Commun. 110 (3) (2020) 1533–1549.
[113] W.-X. Xie, Q.-Y. Zhang, Z.-M. Sun, F. Zhang, A clustering routing protocol
for wsn based on type-2 fuzzy logic and ant colony optimization, Wirel. [141] S. Al Hayali, J. Rahebi, O.N. Ucan, O. Bayat, Increasing energy efficiency
Pers. Commun. 84 (2) (2015) 1165–1196. in wireless sensor networks using ga-anfis to choose a cluster head and
assess routing and weighted trusts to demodulate attacker nodes, Found.
[114] S. Arjunan, P. Sujatha, Lifetime maximization of wireless sensor network
Sci. (2019) 1–20.
using fuzzy based unequal clustering and aco based routing hybrid
[142] R. Logambigai, A. Kannan, Fuzzy logic based unequal clustering for
protocol, Appl. Intell. 48 (8) (2018) 2229–2246.
wireless sensor networks, Wirel. Netw. 22 (3) (2016) 945–957.
[115] P. Maheshwari, A.K. Sharma, K. Verma, Energy efficient cluster based
[143] H. Çam, S. Özdemir, P. Nair, D. Muthuavinashiappan, H.O. Sanli, Energy-
routing protocol for wsn using butterfly optimization algorithm and ant
efficient secure pattern based data aggregation for wireless sensor
colony optimization, Ad Hoc Netw. 110 (2020) 102317.
networks, Comput. Commun. 29 (4) (2006) 446–455.
[116] S. Soni, M. Shrivastava, Novel learning algorithms for efficient mobile sink
[144] N. Mittal, U. Singh, R. Salgotra, Tree-based threshold-sensitive energy-
data collection using reinforcement learning in wireless sensor network,
efficient routing approach for wireless sensor networks, Wirel. Pers.
Wirel. Commun. Mob. Comput. 2018 (2018).
Commun. 108 (1) (2019) 473–492.
[117] A. Alaybeyoglu, A distributed fuzzy logic-based root selection algorithm
[145] K. Guleria, A.K. Verma, Comprehensive review for energy efficient hier-
for wireless sensor networks, Comput. Electr. Eng. 41 (2015) 216–225.
archical routing protocols on wireless sensor networks, Wirel. Netw. 25
[118] F. Fanian, M.K. Rafsanjani, A.B. Saeid, Fuzzy multi-hop clustering protocol:
(3) (2019) 1159–1183.
Selection fuzzy input parameters and rule tuning for wsns, Appl. Soft
[146] K. Yadav, V. Pal, G. Singh, R. Yadav, An efficient load balancing clustering
Comput. (2020) 106923.
scheme for data centric wireless sensor networks, Int. J. Commun. Netw.
[119] F. Fanian, M.K. Rafsanjani, Memetic fuzzy clustering protocol for wireless Secur. (IJCNS) 1 (3) (2012) 24–28.
sensor networks: Shuffled frog leaping algorithm, Appl. Soft Comput. 71 [147] L. Chitnis, A. Dobra, S. Ranka, Fault tolerant aggregation in heterogeneous
(2018) 568–590. sensor networks, J. Parallel Distrib. Comput. 69 (2) (2009) 210–219.
[120] M. Shokouhifar, A. Jalali, Optimized sugeno fuzzy clustering algorithm for [148] M. Younis, M. Youssef, K. Arisha, Energy-aware management for
wireless sensor networks, Eng. Appl. Artif. Intell. 60 (2017) 16–25. cluster-based sensor networks, Comput. Netw. 43 (5) (2003) 649–668.
[121] S. Radhika, P. Rangarajan, On improving the lifespan of wireless sensor [149] Y.T. Hou, Y. Shi, H.D. Sherali, S.F. Midkiff, On energy provisioning and
networks with fuzzy based clustering and machine learning based data relay node placement for wireless sensor networks, IEEE Trans. Wireless
reduction, Appl. Soft Comput. 83 (2019) 105610. Commun. 4 (5) (2005) 2579–2590.
[122] M. Premkumar, T. Sundararajan, Dldm: Deep learning-based defense [150] T. Shankar, S. Shanmugavel, A. Rajesh, Hybrid hsa and pso algorithm
mechanism for denial of service attacks in wireless sensor networks, for energy efficient cluster head selection in wireless sensor networks,
Microprocess. Microsyst. 79 (2020) 103278. Swarm Evol. Comput. 30 (2016) 1–10.
[123] A. Förster, A.L. Murphy, Froms: A failure tolerant and mobility enabled [151] S.S. Katre, S.K. Gosavi, Challenges and issues in wireless sensor network–a
multicast routing paradigm with reinforcement learning for wsns, Ad Hoc review, Int. Res. J. Eng. Technol. (IRJET) 5 (4) (2018).
Netw. 9 (5) (2011) 940–965. [152] M.I. Chidean, E. Morgado, E. del Arco, J. Ramiro-Bargueno, A.J. Caamaño,
[124] K. Thangaramya, K. Kulothungan, R. Logambigai, M. Selvi, S. Ganapathy, A. Scalable data-coupled clustering for large scale wsn, IEEE Trans. Wireless
Kannan, Energy aware cluster and neuro-fuzzy based routing algorithm Commun. 14 (9) (2015) 4681–4694.
for wireless sensor networks in IoT, Comput. Netw. 151 (2019) 211–223. [153] Y.-G. Yue, P. He, A comprehensive survey on the reliability of mobile
[125] N. Mazumdar, A. Nag, S. Nandi, Hdds: Hierarchical data dissemination wireless sensor networks: Taxonomy, challenges, and future directions,
strategy for energy optimization in dynamic wireless sensor network Inf. Fusion 44 (2018) 188–204.
under harsh environments, Ad Hoc Netw. (2020) 102348. [154] J.E. Mbowe, G.S. Oreku, Quality of service in wireless sensor networks,
[126] B.M. Sahoo, H.M. Pandey, T. Amgoth, Gapso-h: A hybrid approach towards Wirel. Sensor Netw. 6 (2) (2014) 19–26.
optimizing the cluster based routing in wireless sensor network, Swarm [155] R. Raghav, K. Thirugnansambandam, D.K. Anguraj, Beeware routing
Evol. Comput. 60 (2020) 100772. scheme for detecting network layer attacks in wireless sensor networks,
[127] F. Fanian, M.K. Rafsanjani, Cluster-based routing protocols in wireless Wirel. Pers. Commun. (2020) 1–21.
sensor networks: A survey based on methodology, J. Netw. Comput. Appl. [156] K. Akkaya, M. Younis, A survey on routing protocols for wireless sensor
142 (2019) 111–142. networks, Ad Hoc Netw. 3 (3) (2005) 325–349.
[128] S. Ghalib, R. Mishra, A.S. Baghel, S. Sharma, Routing protocol development [157] X. Liu, An optimal-distance-based transmission strategy for lifetime
for quality of service optimization of video-on-demand system over maximization of wireless sensor networks, IEEE Sens. J. 15 (6) (2014)
mobile ad hoc networks, Int. J. Commun. Syst. 31 (2) (2018) e3452. 3484–3491.
[129] B. Abidi, A. Jilbab, M. El Haziti, Routing protocols for wireless sensor [158] A. Sabri, K. Al-Shqeerat, Hierarchical cluster-based routing protocols for
networks: A survey, in: Advances in Ubiquitous Computing, Elsevier, wireless sensor networks-a survey, Int. J. Comput. Sci. Issues (IJCSI) 11
2020, pp. 3–15. (1) (2014) 93.

39
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

[159] L. Blazevic, J.-Y. Le Boudec, S. Giordano, A location-based routing method [188] A. Jain, A. Pathak, Ant colony optimization and excess energy calculations
for mobile ad hoc networks, IEEE Trans. Mob. Comput. 4 (2) (2005) based fast converging energy efficient routing algorithm for wsns, Wirel.
97–110. Pers. Commun. 109 (4) (2019) 2305–2328.
[160] N. Sabor, S. Sasaki, M. Abo-Zahhad, S.M. Ahmed, A comprehensive [189] V. Gupta, S.K. Sharma, Cluster head selection using modified aco, in:
survey on hierarchical-based routing protocols for mobile wireless sensor Proceedings of Fourth International Conference on Soft Computing for
networks: Review, taxonomy, and future directions, Wirel. Commun. Mob. Problem Solving, Springer, 2015, pp. 11–20.
Comput. 2017 (2017) 1–23. [190] S. Kumar, Optimization of ant based cluster head election algorithm in
[161] S.A. Mingoti, J.O. Lima, Comparing som neural network with fuzzy wireless sensor networks, Int. J. Comput. Appl. 144 (2) (2016) 5–9.
c-means, K-means and traditional hierarchical clustering algorithms, [191] S. Okdem, D. Karaboga, Routing in wireless sensor networks using an ant
European J. Oper. Res. 174 (3) (2006) 1742–1759. colony optimization (aco) router chip, Sensors 9 (2) (2009) 909–921.
[162] D. Kumar, T.C. Aseri, R. Patel, Eehc: Energy efficient heterogeneous [192] A.M. Bongale, C. Nirmala, A.M. Bongale, Hybrid cluster head election
clustered scheme for wireless sensor networks, Comput. Commun. 32 (4) for wsn based on firefly and harmony search algorithms, Wirel. Pers.
(2009) 662–667. Commun. 106 (2) (2019) 275–306.
[163] D. Yi, H. Yang, Heer–a delay-aware and energy-efficient routing protocol [193] G. Yogarajan, T. Revathi, Nature inspired discrete firefly algorithm for
for wireless sensor networks, Comput. Netw. 104 (2016) 155–173. optimal mobile data gathering in wireless sensor networks, Wirel. Netw.
[164] H. Luo, F. Ye, J. Cheng, S. Lu, L. Zhang, Ttdd: Two-tier data dissemination 24 (8) (2018) 2993–3007.
in large-scale wireless sensor networks, Wirel. Netw. 11 (1–2) (2005) [194] A.A.A. Ari, B.O. Yenke, N. Labraoui, I. Damakoa, A. Gueroui, A power
161–175. efficient cluster-based routing algorithm for wireless sensor networks:
Honeybees swarm intelligence based approach, J. Netw. Comput. Appl.
[165] M. Ye, C. Li, G. Chen, J. Wu, An energy efficient clustering scheme in
69 (2016) 77–97.
wireless sensor networks, Ad Hoc Sensor Wirel. Netw. 17 (2006) 33.
[195] Z. Al Aghbari, A.M. Khedr, W. Osamy, I. Arif, D.P. Agrawal, Routing in
[166] L. Buttyán, P. Schaffer, Position-based aggregator node election in wireless
wireless sensor networks using optimization techniques: A survey, Wirel.
sensor networks, Int. J. Distrib. Sens. Netw. 6 (1) (2010) 679205.
Pers. Commun. (2019) 1–28.
[167] J.H. Holland, et al., Adaptation in Natural and Artificial Systems: an
[196] A. Kaushik, M. Goswami, M. Manuja, S. Indu, D. Gupta, A binary pso
Introductory Analysis with Applications to Biology, Control, and Artificial
approach for improving the performance of wireless sensor networks,
Intelligence, MIT press, 1992.
Wirel. Pers. Commun. (2020) 1–35.
[168] M. Dorigo, L.M. Gambardella, Ant colonies for the travelling salesman [197] H. Izakian, B.T. Ladani, A. Abraham, V. Snasel, et al., A discrete particle
problem, Biosystems 43 (2) (1997) 73–81. swarm optimization approach for grid job scheduling, Int. J. Innovative
[169] X.-S. Yang, Nature-Inspired Metaheuristic Algorithms, Luniver press, 2010. Comput. Inf. Control 6 (9) (2010) 1–15.
[170] K.M. Passino, Biomimicry of bacterial foraging for distributed optimization [198] P. Kuila, P.K. Jana, Energy efficient clustering and routing algorithms for
and control, IEEE Control Syst. Mag. 22 (3) (2002) 52–67. wireless sensor networks: Particle swarm optimization approach, Eng.
[171] R. Eberhart, J. Kennedy, A new optimizer using particle swarm theory, Appl. Artif. Intell. 33 (2014) 127–140.
in: MHS’95. Proceedings of the Sixth International Symposium on Micro [199] R.S. Elhabyan, M.C. Yagoub, Two-tier particle swarm optimization pro-
Machine and Human Science, Ieee, 1995, pp. 39–43. tocol for clustering and routing in wireless sensor network, J. Netw.
[172] J. Kennedy, R. Eberhart, Particle swarm optimization, in: Proceedings of Comput. Appl. 52 (2015) 116–128.
ICNN’95-International Conference on Neural Networks, Vol. 4, IEEE, 1995, [200] D.R. Edla, M.C. Kongara, R. Cheruku, A pso based routing with novel
pp. 1942–1948. fitness function for improving lifetime of wsns, Wirel. Pers. Commun.
[173] D. Karaboga, An Idea Based on Honey Bee Swarm for Numerical Op- 104 (1) (2019) 73–89.
timization, Technical Report, Technical Report-Tr06, Erciyes university, [201] J. RejinaParvin, C. Vasanthanayaki, Particle swarm optimization-based
2005. clustering by preventing residual nodes in wireless sensor networks, IEEE
[174] S. Saremi, S. Mirjalili, A. Lewis, Grasshopper optimisation algorithm: Sensors J. 15 (8) (2015) 4264–4274.
theory and application, Adv. Eng. Softw. 105 (2017) 30–47. [202] A. Singh, S. Sharma, J. Singh, R. Kumar, Mathematical modelling for reduc-
[175] S. Potthuri, T. Shankar, A. Rajesh, Lifetime improvement in wireless sensor ing the sensing of redundant information in wsns based on biologically
networks using hybrid differential evolution and simulated annealing inspired techniques, J. Intell. Fuzzy Systems 37 (5) (2019) 6829–6839.
(desa), Ain Shams Eng. J. 9 (4) (2018) 655–663. [203] S. Famila, A. Jawahar, Improved artificial bee colony optimization-based
[176] S.K. Gupta, P.K. Jana, Energy efficient clustering and routing algorithms clustering technique for wsns, Wirel. Pers. Commun. 110 (4) (2020)
for wireless sensor networks: Ga based approach, Wirel. Pers. Commun. 2195–2212.
83 (3) (2015) 2403–2423. [204] P.S. Mann, S. Singh, Optimal node clustering and scheduling in wireless
[177] G.-S. Yao, Z.-X. Dong, W.-M. Wen, Q. Ren, A routing optimization strategy sensor networks, Wirel. Pers. Commun. 100 (3) (2018) 683–708.
for wireless sensor networks based on improved genetic algorithm, J. [205] S.D. Chavan, A.V. Kulkarni, Improved bio inspired energy efficient clus-
Appl. Sci. Eng. Technol. 19 (2) (2016) 221–228. tering algorithm to enhance qos of wsns, Wirel. Pers. Commun. 109 (3)
[178] A. Chakraborty, S.K. Mitra, M.K. Naskar, A genetic algorithm inspired (2019) 1897–1910.
routing protocol for wireless sensor networks, Int. J. Comput. Intell. [206] A. Tanwar, A.K. Sharma, R.V.S. Pandey, Fractional-grasshopper optimiza-
Theory Pract. (6) (2011) 1–8. tion algorithm for the sensor activation control in wireless sensor
networks, Wirel. Pers. Commun. (2020) 1–24.
[179] K. Khoshraftar, B. Heidari, A hybrid method based on clustering to
[207] P. Kuila, P.K. Jana, Approximation schemes for load balanced clustering
improve the reliability of the wireless sensor networks, Wirel. Pers.
in wireless sensor networks, J. Supercomput. 68 (1) (2014) 87–105.
Commun. (2020) 1–21.
[208] C.P. Low, C. Fang, J.M. Ng, Y.H. Ang, Efficient load-balanced clustering
[180] R. Hamidouche, Z. Aliouat, A.M. Gueroui, Genetic algorithm for improving
algorithms for wireless sensor networks, Comput. Commun. 31 (4) (2008)
the lifetime and qos of wireless sensor networks, Wirel. Pers. Commun.
750–759.
101 (4) (2018) 2313–2348.
[209] M. Bhanderi, H. Shah, Machine learning for wireless sensor network: A
[181] A. Bari, S. Wazed, A. Jaekel, S. Bandyopadhyay, A genetic algorithm based
review, challenges and applications, Adv. Electron. Electr. Eng. 4 (2014)
approach for energy efficient routing in two-tiered sensor networks, Ad
475–486.
Hoc Netw. 7 (4) (2009) 665–676.
[210] X. Chang, J. Huang, S. Liu, G. Xing, H. Zhang, J. Wang, L. Huang, Y. Zhuang,
[182] S.K. Gupta, P. Kuila, P.K. Jana, Gar: An energy efficient ga-based routing Accuracy-aware interference modeling and measurement in wireless
for wireless sensor networks, in: International Conference on Distributed sensor networks, IEEE Trans. Mob. Comput. 15 (2) (2015) 278–291.
Computing and Internet Technology, Springer, 2013, pp. 267–277. [211] W.M. Tan, P. Sullivan, H. Watson, J. Slota-Newson, S.A. Jarvis, An indoor
[183] A. Mohajerani, D. Gharavian, An ant colony optimization based routing test methodology for solar-powered wireless sensor networks, ACM
algorithm for extending network lifetime in wireless sensor networks, Trans. Embedded Comput. Syst. (TECS) 16 (3) (2017) 1–25.
Wirel. Netw. 22 (8) (2016) 2637–2647. [212] X. Song, C. Wang, J. Gao, X. Hu, Dlrdg: distributed linear regression-based
[184] K. Guleria, A.K. Verma, Meta-heuristic ant colony optimization based hierarchical data gathering framework in wireless sensor network, Neural
unequal clustering for wireless sensor network, Wirel. Pers. Commun. Comput. Appl. 23 (7–8) (2013) 1999–2013.
105 (3) (2019) 891–911. [213] W. Zhao, S. Su, F. Shao, Improved DV-hop algorithm using locally
[185] E. Amiri, H. Keshavarz, M. Alizadeh, M. Zamani, T. Khodadadi, Energy weighted linear regression in anisotropic wireless sensor networks, Wirel.
efficient routing in wireless sensor networks based on fuzzy ant colony Pers. Commun. 98 (4) (2018) 3335–3353.
optimization, Int. J. Distrib. Sens. Netw. 10 (7) (2014) 768936. [214] J. Shu, S. Liu, L. Liu, L. Zhan, G. Hu, Research on link quality estima-
[186] M. Jafari, H. Khotanlou, A routing algorithm based an ant colony, local tion mechanism for wireless sensor networks based on support vector
search and fuzzy inference to improve energy consumption in wireless machine, Chin. J. Electron. 26 (2) (2017) 377–384.
sensor networks, Int. J. Electr. Comput. Eng. 3 (5) (2013) 640. [215] A. Garofalo, C. Di Sarno, V. Formicola, Enhancing intrusion detection in
[187] X. Liu, S. Li, M. Wang, An ant colony based routing algorithm for wireless wireless sensor networks through decision trees, in: European Workshop
sensor network, Int. J. Future Gener. Commun. Netw. 9 (6) (2016) 75–86. on Dependable Computing, Springer, 2013, pp. 1–15.

40
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

[216] H. He, Z. Zhu, E. Mäkinen, Task-oriented distributed data fusion in [244] J.-S. Lee, W.-L. Cheng, Fuzzy-logic-based clustering approach for wireless
autonomous wireless sensor networks, Soft Comput. 19 (8) (2015) sensor networks using energy predication, IEEE Sens. J. 12 (9) (2012)
2305–2319. 2891–2897.
[217] A.A. Rezaee, F. Pasandideh, A fuzzy congestion control protocol based [245] A. Zahedi, M. Arghavani, F. Parandin, A. Arghavani, Energy efficient
on active queue management in wireless sensor networks with medical reservation-based cluster head selection in wsns, Wirel. Pers. Commun.
applications, Wirel. Pers. Commun. 98 (1) (2018) 815–842. 100 (3) (2018) 667–679.
[218] A. Mehmood, Z. Lv, J. Lloret, M.M. Umar, Eldc: An artificial neural [246] A. Saidi, K. Benahmed Pr, Secure cluster head election algorithm and
network based energy-efficient and robust routing scheme for pollution misbehavior detection approach based on trust management technique
monitoring in wsns, IEEE Trans. Emerg. Top. Comput. 8 (1) (2020) for clustered wireless sensor networks, Ad Hoc Netw. (2020) 102215.
106–114. [247] A. Ismail, R. Amin, Malicious cluster head detection mechanism
[219] S.K. Gharghan, R. Nordin, M. Ismail, J. Abd Ali, Accurate wireless sensor in wireless sensor networks, Wirel. Pers. Commun. 108 (4) (2019)
localization technique based on hybrid pso-ann algorithm for indoor and 2117–2135.
outdoor track cycling, IEEE Sens. J. 16 (2) (2015) 529–541. [248] S. Mahajan, J. Malhotra, S. Sharma, An energy balanced qos based cluster
[220] P. Chanak, I. Banerjee, Fuzzy rule-based faulty node classification and head selection strategy for wsn, Egypt. Inform. J. 15 (3) (2014) 189–199.
management scheme for large scale wireless sensor networks, Expert [249] G. Kannan, T.S.R. Raja, Energy efficient distributed cluster head scheduling
Syst. Appl. 45 (2016) 307–321. scheme for two tiered wireless sensor network, Egypt. Inform. J. 16 (2)
[221] T. Ma, F. Wang, J. Cheng, Y. Yu, X. Chen, A hybrid spectral clustering (2015) 167–174.
and deep neural network ensemble algorithm for intrusion detection in
[250] D. Jia, H. Zhu, S. Zou, P. Hu, Dynamic cluster head selection method for
sensor networks, Sensors 16 (10) (2016) 1701.
wireless sensor network, IEEE Sens. J. 16 (8) (2016) 2746–2754.
[222] Y. Lee, Classification of node degree based on deep learning and routing
[251] N. Mittal, U. Singh, B.S. Sohi, An energy-aware cluster-based stable
method applied for virtual route assignment, Ad Hoc Netw. 58 (2017)
protocol for wireless sensor networks, Neural Comput. Appl. 31 (11)
70–85.
(2019) 7269–7286.
[223] M. Gholipour, A.T. Haghighat, M.R. Meybodi, Hop-by-hop congestion
[252] N. Mittal, U. Singh, B.S. Sohi, A stable energy efficient clustering protocol
avoidance in wireless sensor networks based on genetic support vector
for wireless sensor networks, Wirel. Netw. 23 (6) (2017) 1809–1821.
machine, Neurocomputing 223 (2017) 63–76.
[224] Z. Wang, H. Zhang, T. Lu, Y. Sun, X. Liu, A new range-free localisation in [253] M.O. Oladimeji, M. Turkey, S. Dudley, Hach: Heuristic algorithm for
wireless sensor networks using support vector machine, Int. J. Electron. clustering hierarchy protocol in wireless sensor networks, Appl. Soft
105 (2) (2018) 244–261. Comput. 55 (2017) 452–461.
[225] S. Zidi, T. Moulahi, B. Alaya, Fault detection in wireless sensor networks [254] P. Sundareswaran, K. Vardharajulu, R. Rajesh, Dech: Equally distributed
through svm classifier, IEEE Sens. J. 18 (1) (2017) 340–347. cluster heads technique for clustering protocols in wsns, Wirel. Pers.
[226] R. Feng, X. Han, Q. Liu, N. Yu, A credible Bayesian-based trust manage- Commun. 84 (1) (2015) 137–151.
ment scheme for wireless sensor networks, Int. J. Distrib. Sens. Netw. 11 [255] A. Kardi, R. Zagrouba, Rach: A new radial cluster head selection algorithm
(11) (2015) 678926. for wireless sensor networks, Wirel. Pers. Commun. (2020) 1–14.
[227] V. Jafarizadeh, A. Keshavarzi, T. Derikvand, Efficient cluster head selection [256] B. Rambabu, A.V. Reddy, S. Janakiraman, Hybrid artificial bee colony and
using Naïve Bayes classifier for wireless sensor networks, Wirel. Netw. 23 monarchy butterfly optimization algorithm (habc-mboa)-based cluster
(3) (2017) 779–785. head selection for wsns, J. King Saud Univ. Comput. Inf. Sci. (2019).
[228] J.J. Perez-Solano, S. Felici-Castell, Improving time synchronization in [257] Q. Ni, Q. Pan, H. Du, C. Cao, Y. Zhai, A novel cluster head selection
wireless sensor networks using Bayesian inference, J. Netw. Comput. Appl. algorithm based on fuzzy clustering and particle swarm optimization,
82 (2017) 47–55. IEEE/ACM Trans. Comput. Biol. Bioinform. 14 (1) (2015) 76–84.
[229] B. Zhou, Q. Chen, T.J. Li, P. Xiao, Online variational Bayesian filtering- [258] G.P. Gupta, S. Jha, Integrated clustering and routing protocol for wireless
based mobile target tracking in wireless sensor networks, Sensors 14 (11) sensor networks using cuckoo and harmony search based metaheuristic
(2014) 21281–21315. techniques, Eng. Appl. Artif. Intell. 68 (2018) 101–109.
[230] S. Hwang, R. Ran, J. Yang, D.K. Kim, Multivariated Bayesian compres- [259] P.S. Rao, H. Banka, Energy efficient clustering algorithms for wireless
sive sensing in wireless sensor networks, IEEE Sens. J. 16 (7) (2015) sensor networks: novel chemical reaction optimization approach, Wirel.
2196–2206. Netw. 23 (2) (2017) 433–452.
[231] T.L.T. Nguyen, F. Septier, H. Rajaona, G.W. Peters, I. Nevat, Y. Delignon, [260] S. Nikoletseas, On the energy balance problem in distributed sensor
A Bayesian perspective on multiple source localization in wireless sensor networks, Comp. Sci. Rev. 4 (2) (2010) 65–79.
networks, IEEE Trans. Signal Process. 64 (7) (2015) 1684–1699. [261] S. Kaur, R. Mahajan, Hybrid meta-heuristic optimization based energy
[232] C. Titouna, M. Aliouat, M. Gueroui, Fds: fault detection scheme for efficient protocol for wireless sensor networks, Egypt. Inform. J. 19 (3)
wireless sensor networks, Wirel. Pers. Commun. 86 (2) (2016) 549–562. (2018) 145–150.
[233] W. Li, P. Yi, Y. Wu, L. Pan, J. Li, A new intrusion detection system based on [262] P.S. Mann, S. Singh, Artificial bee colony metaheuristic for energy-efficient
knn classification algorithm in wireless sensor network, J. Electr. Comput.r clustering and routing in wireless sensor networks, Soft Comput. 21 (22)
Eng. 2014 (2014) 1–8. (2017) 6699–6712.
[234] Y. Li, L.E. Parker, Nearest neighbor imputation using spatial–temporal [263] D. Karaboga, S. Okdem, C. Ozturk, Cluster based wireless sensor network
correlations in wireless sensor networks, Inf. Fusion 15 (2014) 64–79. routing using artificial bee colony algorithm, Wirel. Netw. 18 (7) (2012)
[235] P. Neamatollahi, S. Abrishami, M. Naghibzadeh, M.H.Y. Moghaddam, O. 847–860.
Younis, Hierarchical clustering-task scheduling policy in cluster-based
[264] D.C. Hoang, P. Yadav, R. Kumar, S.K. Panda, Real-time implementation of a
wireless sensor networks, IEEE Trans. Ind. Inf. 14 (5) (2017) 1876–1886.
harmony search algorithm-based clustering protocol for energy-efficient
[236] X. Xu, R. Ansari, A. Khokhar, A.V. Vasilakos, Hierarchical data aggregation
wireless sensor networks, IEEE Trans. Ind. Inform. 10 (1) (2013) 774–783.
using compressive sensing (hdacs) in wsns, ACM Trans. Sensor Netw. 11
[265] K. Vijayalakshmi, P. Anandan, A multi objective tabu particle swarm
(3) (2015) 1–25.
optimization for effective cluster head selection in wsn, Cluster Comput.
[237] J. Qin, W. Fu, H. Gao, W.X. Zheng, Distributed k-means algorithm and
22 (5) (2019) 12275–12282.
fuzzy c-means algorithm for sensor networks based on multiagent
[266] P. Sengottuvelan, N. Prasath, Bafsa: Breeding artificial fish swarm algo-
consensus theory, IEEE Trans. Cybern. 47 (3) (2016) 772–783.
rithm for optimal cluster head selection in wireless sensor networks,
[238] F. Zhu, J. Wei, Localization algorithm for large-scale wireless sensor
Wirel. Pers. Commun. 94 (4) (2017) 1979–1991.
networks based on fcmtsr-support vector machine, Int. J. Distrib. Sens.
Netw. 12 (10) (2016) 1550147716674010. [267] M. Baskaran, C. Sadagopan, Synchronous firefly algorithm for cluster head
[239] A. Morell, A. Correa, M. Barceló, J.L. Vicario, Data aggregation and principal selection in wsn, Sci. World J. 2015 (2015).
component analysis in wsns, IEEE Trans. Wireless Commun. 15 (6) (2016) [268] T.S. Murugan, A. Sarkar, Optimal cluster head selection by hybridisation
3908–3919. of firefly and grey wolf optimisation, Int. J. Wirel. Mob. Comput. 14 (3)
[240] D.P. Kumar, T. Amgoth, C.S.R. Annavarapu, Machine learning algorithms (2018) 296–305.
for wireless sensor networks: A survey, Inf. Fusion 49 (2019) 1–25. [269] T. Bhatia, S. Kansal, S. Goel, A. Verma, A genetic algorithm based distance-
[241] M.A. Alsheikh, S. Lin, D. Niyato, H.-P. Tan, Machine learning in wireless aware routing protocol for wireless sensor networks, Comput. Electr. Eng.
sensor networks: Algorithms, strategies, and applications, IEEE Commun. 56 (2016) 441–455.
Surv. Tutor. 16 (4) (2014) 1996–2018. [270] S. Hussain, A.W. Matin, O. Islam, Genetic algorithm for hierarchical
[242] A. Braman, G. Umapathi, A comparative study on advances in leach wireless sensor networks., J. Netw. 2 (5) (2007) 87–97.
routing protocol for wireless sensor networks: A survey., Int. J. Adv. Res. [271] P. Kuila, P.K. Jana, A novel differential evolution based clustering al-
Comput. Commun. Eng. 3 (2) (2014) 5683–5690. gorithm for wireless sensor networks, Appl. Soft Comput. 25 (2014)
[243] A. Al-Baz, A. El-Sayed, A new algorithm for cluster head selection in leach 414–425.
protocol for wireless sensor networks, Int. J. Commun. Syst. 31 (1) (2018) [272] E.A. Khalil, A.A. Bara’a, Stable-aware evolutionary routing protocol for
e3407. wireless sensor networks, Wirel. Pers. Commun. 69 (4) (2013) 1799–1817.

41
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

[273] A.Y. Barnawi, I.M. Keshta, Energy management in wireless sensor net- [300] A.A. Bara’a, E.A. Khalil, A new evolutionary based routing protocol for
works based on naive bayes, mlp, and svm classifications: A comparative clustered heterogeneous wireless sensor networks, Appl. Soft Comput. 12
study, J. Sensors 2016 (2016). (7) (2012) 1950–1957.
[274] H. Bagci, A. Yazici, An energy aware fuzzy approach to unequal clustering [301] J.-L. Liu, C.V. Ravishankar, Leach-ga: Genetic algorithm-based energy-
in wireless sensor networks, Appl. Soft Comput. 13 (4) (2013) 1741–1749. efficient adaptive clustering protocol for wireless sensor networks, Int.
[275] B.M. Khan, R. Bilal, R. Young, Fuzzy-topsis based cluster head selection in J. Mach. Learn. Comput. 1 (1) (2011) 79.
mobile wireless sensor networks, J. Electr. Syst. Inf. Technol. 5 (3) (2018) [302] N. Mazumdar, H. Om, Distributed fuzzy approach to unequal clustering
928–943. and routing algorithm for wireless sensor networks, Int. J. Commun. Syst.
[276] S. Su, S. Zhao, An optimal clustering mechanism based on fuzzy-c means 31 (12) (2018) e3709.
for wireless sensor networks, Sustain. Comput. Inform. Syst. 18 (2018) [303] S. Gajjar, M. Sarkar, K. Dasgupta, Famacrow: Fuzzy and ant colony
127–134. optimization based combined mac, routing, and unequal clustering cross-
[277] B. Baranidharan, B. Santhi, Ducf: Distributed load balancing unequal layer protocol for wireless sensor networks, Appl. Soft Comput. 43 (2016)
clustering in wireless sensor networks using fuzzy approach, Appl. Soft 235–247.
Comput. 40 (2016) 495–506. [304] J.A. Boyan, M.L. Littman, Packet routing in dynamically changing net-
[278] B. Jain, G. Brar, J. Malhotra, Ekmt-k-means clustering algorithmic solution works: A reinforcement learning approach, in: Advances in Neural
for low energy consumption for wireless sensor networks based on min- Information Processing Systems, 1994, pp. 671–678.
imum mean distance from base station, in: Networking Communication [305] M. Littman, A. Moore, Reinforcement Learning: A Survey, Journal of
and Data Knowledge Engineering, Springer, 2018, pp. 113–123. Artificial Intelligence Research 4, syf, 1996, pp. 237–285.
[306] Y. Zhang, Q. Huang, A learning-based adaptive routing tree for wireless
[279] S. El Khediri, W. Fakhet, T. Moulahi, R. Khan, A. Thaljaoui, A. Kachouri,
sensor networks, J. Commun. 1 (2) (2006) 12–21.
Improved node localization using K-means clustering for wireless sensor
[307] X. Zhu, Y. Lu, J. Han, L. Shi, Transmission reliability evaluation for wireless
networks, Comp. Sci. Rev. 37 (2020) 100284.
sensor networks, Int. J. Distrib. Sens. Netw. 12 (2) (2016) 1346079.
[280] T.-K. Dao, T.-T. Nguyen, J.-S. Pan, Y. Qiao, Q.-A. Lai, Identification fail-
[308] F. Wang, J. Liu, On reliable broadcast in low duty-cycle wireless sensor
ure data for cluster heads aggregation in wsn based on improving
networks, IEEE Trans. Mob. Comput. 11 (5) (2011) 767–779.
classification of svm, IEEE Access 8 (2020) 61070–61084.
[309] H.M. AboElFotoh, S.S. Iyengar, K. Chakrabarty, Computing reliability and
[281] D. Sinha, R. Kumari, S. Tripathi, Semisupervised classification based message delay for cooperative wireless distributed sensor networks
clustering approach in wsn for forest fire detection, Wirel. Pers. Commun. subject to random failures, IEEE Trans. Reliab. 54 (1) (2005) 145–155.
109 (4) (2019) 2561–2605. [310] S. Tabatabaei, A.M. Rigi, Reliable routing algorithm based on clustering
[282] S.M. Bozorgi, A.S. Rostami, A.A.R. Hosseinabadi, V.E. Balas, A new clus- and mobile sink in wireless sensor networks, Wirel. Pers. Commun. 108
tering protocol for energy harvesting-wireless sensor networks, Comput. (4) (2019) 2541–2558.
Electr. Eng. 64 (2017) 233–247. [311] C. Chowdhury, N. Aslam, G. Ahmed, S. Chattapadhyay, S. Neogy, L. Zhang,
[283] A. Agrawal, V. Singh, S. Jain, R.K. Gupta, Gcrp: Grid-cycle routing protocol Novel algorithms for reliability evaluation of remotely deployed wireless
for wireless sensor network with mobile sink, AEU-Int. J. Electron. sensor networks, Wirel. Pers. Commun. 98 (1) (2018) 1331–1360.
Commun. 94 (2018) 1–11. [312] H. Mostafaei, Energy-efficient algorithm for reliable routing of wireless
[284] D. Xie, X. Wu, D. Li, J. Sun, Multiple mobile sinks data dissemination sensor networks, IEEE Trans. Ind. Electron. 66 (7) (2018) 5567–5575.
mechanism for large scale wireless sensor network, China Commun. 11 [313] A. Bagherinia, B. Minaei-Bidgoli, M. Hossinzadeh, H. Parvin, Reliability-
(13) (2014) 1–8. based fuzzy clustering ensemble, Fuzzy Sets and Systems (2020).
[285] A.W. Khan, A.H. Abdullah, M.A. Razzaque, J.I. Bangash, Vgdra: a virtual [314] S. Murugaanandam, V. Ganapathy, Reliability-based cluster head selection
grid-based dynamic routes adjustment scheme for mobile sink-based methodology using fuzzy logic for performance improvement in wsns,
wireless sensor networks, IEEE Sensors J. 15 (1) (2014) 526–534. IEEE Access 7 (2019) 87357–87368.
[286] O. Banimelhem, S. Khasawneh, Gmcar: Grid-based multipath with con- [315] W. Cai, J. Zhao, M. Zhu, A real time methodology of cluster-system theory-
gestion avoidance routing protocol in wireless sensor networks, Ad Hoc based reliability estimation using k-means clustering, Reliab. Eng. Syst.
Netw. 10 (7) (2012) 1346–1361. Saf. (2020) 107045.
[287] N. Nasser, Y. Chen, Seem: Secure and energy-efficient multipath routing [316] N.A. Alrajeh, S. Khan, J. Lloret, J. Loo, Secure routing protocol using cross-
protocol for wireless sensor networks, Comput. Commun. 30 (11–12) layer design and energy harvesting in wireless sensor networks, Int. J.
(2007) 2401–2412. Distrib. Sens. Netw. 9 (1) (2013) 374796.
[288] X. Meng, X. Shi, Z. Wang, S. Wu, C. Li, A grid-based reliable routing [317] Y. Challal, A. Ouadjaout, N. Lasla, M. Bagaa, A. Hadjidj, Secure and efficient
protocol for wireless sensor networks with randomly distributed clusters, disjoint multipath construction for fault tolerant routing in wireless
Ad Hoc Netw. 51 (2016) 47–61. sensor networks, J. Netw. Comput. Appl. 34 (4) (2011) 1380–1397.
[289] S. Sasirekha, S. Swamynathan, Cluster-chain mobile agent routing al- [318] J. Deng, R. Han, S. Mishra, Insens: Intrusion-tolerant routing for wireless
gorithm for efficient data aggregation in wireless sensor network, J. sensor networks, Comput. Commun. 29 (2) (2006) 216–230.
Commun. Netw. 19 (4) (2017) 392–401. [319] A.A. Ibrahim, A.K. Tamer, A. Abdelshakour, Sec-teen: a secure routing
[290] J. Ben-Othman, B. Yahya, Energy efficient and qos based routing protocol protocol for enhanced efficiency in wireless sensor networks, in: Inno-
for wireless sensor networks, J. Parallel Distrib. Comput. 70 (8) (2010) vations and Advances in Computer, Information, Systems Sciences, and
849–857. Engineering, Springer, 2013, pp. 621–631.
[320] R. Bhatt, P. Maheshwary, P. Shukla, P. Shukla, M. Shrivastava, S. Changlani,
[291] K. Sohrabi, J. Gao, V. Ailawadhi, G.J. Pottie, Protocols for self-organization
Implementation of fruit fly optimization algorithm (FFOA) to escalate the
of a wireless sensor network, IEEE Pers. Commun. 7 (5) (2000) 16–27.
attacking efficiency of node capture attack in wireless sensor networks
[292] X. Huang, Y. Fang, Multiconstrained qos multipath routing in wireless
(wsn), Comput. Commun. 149 (2020) 134–145.
sensor networks, Wirel. Netw. 14 (4) (2008) 465–478.
[321] W. Zhang, D. Han, K.-C. Li, F.I. Massetto, Wireless sensor network
[293] X. Cai, Y. Duan, Y. He, J. Yang, C. Li, Bee-sensor-c: an energy-efficient and
intrusion detection system based on mk-elm, Soft Comput. (2020) 1–14.
scalable multipath routing protocol for wireless sensor networks, Int. J.
[322] G. Pachauri, S. Sharma, Anomaly detection in medical wireless sensor
Distrib. Sens. Netw. 11 (3) (2015) 976127.
networks using machine learning algorithms, Procedia Comput. Sci. 70
[294] A. Sarkar, T.S. Murugan, Cluster head selection for energy efficient and
(2015) 325–333.
delay-less routing in wireless sensor network, Wirel. Netw. 25 (1) (2019) [323] J. Nagar, S. Sharma, K-barrier coverage-based intrusion detection for
303–320. wireless sensor networks, in: Cyber Security, Springer, 2018, pp. 373–385.
[295] A. Shankar, N. Jaisankar, M.S. Khan, R. Patan, B. Balamurugan, Hybrid [324] S. Aljawarneh, M. Aldwairi, M.B. Yassein, Anomaly-based intrusion de-
model for security-aware cluster head selection in wireless sensor tection system through feature selection analysis and building hybrid
networks, IET Wirel. Sensor Syst. 9 (2) (2018) 68–76. efficient model, J. Comput. Sci. 25 (2018) 152–160.
[296] M. Wazid, A.K. Das, An efficient hybrid anomaly detection scheme using [325] H. Nkiama, S.Z.M. Said, M. Saidu, A subset feature elimination mechanism
K-means clustering for wireless sensor networks, Wirel. Pers. Commun. for intrusion detection system, Int. J. Adv. Comput. Sci. Appl. 7 (4) (2016)
90 (4) (2016) 1971–2000. 148–157.
[297] G.M. Borkar, L.H. Patil, D. Dalgade, A. Hutke, A novel clustering approach [326] B. Ahmad, W. Jian, Z.A. Ali, S. Tanvir, M.S.A. Khan, Hybrid anomaly
and adaptive svm classifier for intrusion detection in wsn: A data mining detection by using clustering for wireless sensor network, Wirel. Pers.
concept, Sustain. Comput. Inform. Syst. 23 (2019) 120–135. Commun. 106 (4) (2019) 1841–1853.
[298] S.M. Amini, A. Karimi, S. Shehnepoor, Improving lifetime of wireless [327] K. Kalaiselvan, G. Singh, Detection and isolation of black hole attack in
sensor network based on sinks mobility and clustering routing, Wirel. wireless sensor networks, Int. J. Innov. Res. Sci. Eng. Technol. 4 (5) (2015)
Pers. Commun. 109 (3) (2019) 2011–2024. 23–31.
[299] C. Tunca, S. Isik, M.Y. Donmez, C. Ersoy, Ring routing: An energy-efficient [328] B. Bhushan, G. Sahoo, Isfc-bls (intelligent and secured fuzzy clustering al-
routing protocol for wireless sensor networks with a mobile sink, IEEE gorithm using balanced load sub-cluster formation) in wsn environment,
Trans. Mob. Comput. 14 (9) (2014) 1947–1960. Wirel. Pers. Commun. 111 (3) (2020) 1667–1694.

42
J. Amutha, S. Sharma and S.K. Sharma Computer Science Review 40 (2021) 100376

[329] R. Das, D. Dash, M.K. Sarkar, Htms: Fuzzy based hierarchical trust J. Amutha is currently a Ph.D. research scholar in
management scheme in wsn, Wirel. Pers. Commun. (2020) 1–34. Computer Science and Engineering, University School
[330] V. Gupta, R. Pandey, An improved energy aware distributed unequal of ICT, Gautam Buddha University, Greater Noida, India.
clustering protocol for heterogeneous wireless sensor networks, Eng. Sci. She has completed her M.Tech degree in Computer
Technol. Int. J. 19 (2) (2016) 1050–1058. Science and Engineering from Gautam Buddha Univer-
[331] J. Hamidzadeh, M.H. Ghomanjani, An unequal cluster-radius approach sity, Greater Noida, India in 2019. Her research interest
based on node density in clustering for wireless sensor networks, Wirel. includes wireless sensor networks, cluster-based and
Pers. Commun. 101 (3) (2018) 1619–1637. routing algorithms in wireless sensor networks.
[332] S. Lee, H. Choe, B. Park, Y. Song, C.-K. Kim, Luca: An energy-efficient
unequal clustering algorithm using location information for wireless
sensor networks, Wirel. Pers. Commun. 56 (4) (2011) 715–731.
[333] H. Li, Y. Liu, W. Chen, W. Jia, B. Li, J. Xiong, Coca: Constructing optimal
clustering architecture to maximize sensor network lifetime, Comput.
Commun. 36 (3) (2013) 256–268. Sandeep Sharma received a B.Tech. degree in electron-
[334] G. Chen, C. Li, M. Ye, J. Wu, An unequal cluster-based routing protocol in ics engineering from RGPV, Bhopal, India, in 1997 and
wireless sensor networks, Wirel. Netw. 15 (2) (2009) 193–207. an M.Tech. degree in digital communication from Devi
[335] A.B.F. Guiloufi, N. Nasri, A. Kachouri, An energy-efficient unequal cluster- Ahilya University, Indore, India, in 2005. He is a Ph.D. in
ing algorithm using ‘sierpinski triangle’for wsns, Wirel. Pers. Commun. 88 electronics and communication engineering from Gau-
(3) (2016) 449–465. tam Buddha University, Greater Noida, India, in 2016.
[336] J. Yang, D. Zhang, An energy-balancing unequal clustering protocol for Since 2010, he is working as a Faculty in electronics
wireless sensor networks, Inf. Technol. J. 8 (1) (2009) 57–63. and communication engineering department at Univer-
[337] E. Ever., R. Luchmun., L. Mostarda., A. Navarra., P. Shah., UHEED - An sity School of ICT, Gautam Buddha University, Greater
unequal clustering algorithm for wireless sensor networks, in: Proceed- Noida, India. His research interest includes wireless
ings of the 1st International Conference on Sensor Networks - Volume 1: sensor networks, wireless network security, physical
SENSORNETS, INSTICC, SciTePress, 2012, pp. 185–193, http://dx.doi.org/ layer authentication, intrusion detection in wireless networks, cross-layer design
10.5220/0003804001850193. and machine learning applications in WSNs. He has published 25 research papers
[338] H. Xia, R.-h. Zhang, J. Yu, Z.-k. Pan, Energy-efficient routing algorithm in reputed international journals and 41 papers published in the international
based on unequal clustering and connected graph in wireless sensor conferences.
networks, Int. J. Wirel. Inf. Netw. 23 (2) (2016) 141–150. Dr. Sharma is a recipient of the Best Conference Paper in the international
[339] C.-J. Jiang, W.-R. Shi, X.-l. Tang, et al., Energy-balanced unequal clustering conference ICCCS, 2016 and the Young Scientist Award in 2019 for his research
protocol for wireless sensor networks, J. China Univ. Posts Telecommun. work. He is an active reviewer of IET Communications, IEEE Wireless Comm.
17 (4) (2010) 94–99. Letters, IEEE Access, Journal of Information Technology (Springer), Personal and
[340] T. Kaur, D. Kumar, Particle swarm optimization-based unequal and fault Ubiquitous Computing (Springer), Multimedia Tools and Applications (Springer),
tolerant clustering protocol for wireless sensor networks, IEEE Sens. J. 18 Complex and Intelligent Systems(Springer), International Journal of Computer
(11) (2018) 4614–4622. Applications in Technology (Inderscience), International Journal of Communica-
[341] F. Xunli, D. Feiefi, Shuffled frog leaping algorithm based unequal clus- tion Systems (Wiley), Journal of The Institution of Engineers (India): Series B,
tering strategy for wireless sensor networks, Appl. Math. Inf. Sci. 9 (3) Journal of Intelligent and Fuzzy Systems.
(2015) 1415–1426.
[342] A.J. Yuste-Delgado, J.C. Cuevas-Martinez, A. Triviño-Cabrera, Eudfc-
Sanjay Kumar Sharma is currently working with Gau-
enhanced unequal distributed type-2 fuzzy clustering algorithm, IEEE
tam Buddha University (GBU) as a Professor and Dean,
Sens. J. 19 (12) (2019) 4705–4716.
University School of ICT at GBU. He is a member
[343] D. Agrawal, S. Pandey, Fuca: Fuzzy-based unequal clustering algorithm to
of Executive Council of Nuclear Track Society of In-
prolong the lifetime of wireless sensor networks, Int. J. Commun. Syst.
dia, Life member Society for Scientific Values (SSV),
31 (2) (2018) e3448.
Life Member Indian Association for Physics Teachers
[344] R. Dutta, S. Gupta, M.K. Das, Low-energy adaptive unequal clustering
(IAPT) Secretary, Vigyan Bharti, Haryana, Life member,
protocol using fuzzy c-means in wireless sensor networks, Wirel. Pers.
Swadeshi Science Movement, Delhi (SSMD). He holds
Commun. 79 (2) (2014) 1187–1209.
an Academic honor of being Chairman, Institution of
[345] T. Liu, Q. Li, P. Liang, An energy-balancing clustering approach for
Electronics and Telecommunication Engineers (IETE),
gradient-based routing in wireless sensor networks, Comput. Commun.
Ambala a Sub-Center (Haryana). He has successfully
35 (17) (2012) 2150–2161.
completed a large number of in-house research and development projects. His
[346] S. Mao, C. Zhao, Z. Zhou, Y. Ye, An improved fuzzy unequal clustering
research interest includes: Basic and Applied Aspects of SSSNTs, Heavy ions, IOT,
algorithm for wireless sensor network, Mob. Netw. Appl. 18 (2) (2013)
Smart Material and Sensors, Artificial Intelligence, besides Nanotechnology. He
206–214.
has actively contributed to many research projects like Search of super heavy
[347] F. Liu, Y. Chang, An energy aware adaptive kernel density estimation
elements in nature using meteoritic crystal funded by CSIR, New Delhi (India)
approach to unequal clustering in wireless sensor networks, IEEE Access
at Kurukshetra University, Kurukshetra and A.M.U., Aligarh; Single pore sensor
7 (2019) 40569–40580.
for water pollution funded by Ministry of Environment and Forest, Govt. of
[348] Y. Wang, T. Yang, D. Zhang, An energy efficient and balance hierarchical
India at G.N.D. University, Amritsar; Calibration and validation of data products
unequal clustering algorithm for large scale sensor network, Inform.
obtained from remote sensing satellites (with Image Processing Division of Space
Technol. J. 8 (1) (2009) 28–38.
Application Centre, Ahmedabad) funded by Indian Space Research Organization
[349] G. Abdul-Salaam, A.H. Abdullah, M.H. Anisi, A. Gani, A. Alelaiwi, A com-
(ISRO), Bangalore (India) and Web based engineering education system: a
parative analysis of energy conservation approaches in hybrid wireless
consortium for technical institutes funded by MHRD, New Delhi (India) at
sensor networks data collection protocols, Telecommun. Syst. 61 (1)
National Institute of Technology, Kurukshetra.
(2016) 159–179.

43