This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 1

Location of Things (LoT): A Review and Taxonomy

of Sensors Localization in IoT Infrastructure
Rathin Chandra Shit, Suraj Sharma, Deepak Puthal, and Albert Y. Zomaya, Fellow, IEEE

Abstract—Internet of Things (IoT) is a novel design paradigm, a two step process of distance measurement between sensor
intended as a network of billions to trillions of tiny sensors nodes followed by calculation based on measured distances.
communicating with each other to offer innovative solutions to Various type of classification of localization algorithms has
real time problems. These sensors form a network named as
Wireless Sensor Networks (WSN) to monitor physical environ- been proposed such as Range Based algorithms calculate
ment and disseminate collected data back to the base station location information from range based measurement tech-
through multiple hops. WSN has the capability to collect and niques like Received Signal Strength (RSS) [5], Time of
report data for a specific application. The location information Arrival (ToA) [6], Time Difference of Arrival (TDoA) [7] and
plays an important role for various wireless sensor network Angle of Arrival (AoA) [8]. Range free algorithms calculate
applications. A majority of the applications are related to location
based services (LBS). The development of sensor technology, the location information from the connectivity information.
processing techniques and communication systems give rise to a Some physical measurement based Localization schemes are
development of the smart sensor for the adaptive and innovative classified as Coarse-Grained and Fine-Grained as illustrated in
application. So a single localization technique is not adequate Fig. 1. Another way of classification is based on cooperative
for all application. In this paper, a recent extensive analysis
of localization techniques and hierarchical taxonomy and their
applications in the different context is presented. This taxonomy
of the localization technique is classified based on presence
of offline training in localization namely self determining and
training dependent approaches. Finally, various open research
issue related to localization schemes for IoT are compared and
proposed various directions for future research.
Fig. 1. Physical measurements used in localization schemes
Index Terms—Wireless Sensor Network (WSN), Internet of
Things (IoT), Localization, LMFF, RBFM, SDP, MDS, Finger-
and non cooperative. In cooperative method the neighbouring
printing
sensor node collectively process information to find location
whereas in non cooperative method they process information
I. I NTRODUCTION individually. Based on processing point, the algorithms can be
network centric positionings like Centralized scheme or self-

I NTERNET of Things is a network of sensor and actuator

nodes called ’things’. ’Things’ refers to device or sensors
that sense physical world signals and records it in digital forms
positioning like Distributed scheme.
There are plenty of literature showing various methods
and algorithms for localization. Most literature do not have
[1]. Location-based services (LBS) in variety of recent applica- proper classification of the problems in localization and recent
tions including military [2], e-Health, environment monitoring, advancements. Few literature concentrate on mobility [27] [28]
IoT, cyber-physical systems, home office automation [3] [4], [29] and other concentrate on processing aspects of algorithms
weather forecasting, early warning and rescue operation etc. [10] [37] [35].
Global Positioning System (GPS) can be the solution to all of The objective of this paper is to compare the research
these services. But GPS uses high hardware cost, high power works related to localization and classify them in a framework
consumption and poor performance in indoor environment. applicable to IoT infrastructure and provide the taxonomy.
So a large number or researches are conducted to optimise Acronym and definitions are provided in the Table. I.
the location based services. Localization is the process of Many literature survey have been conducted on localization
ascertaining the location of unknown sensor nodes in network methods. The comparative analysis of the published survey
with the help of some reference nodes called anchors or beacon work is illustrated in Table. II. Costa et. al. [9] have presented
nodes. The reference nodes obtain location information with a set of Multi-Dimensional Scaling (MDS) algorithms and
inbuilt GPS module or manual deployment. Localization is specified centralized and distributed methods for localization.
A generalized survey of localization on stationary wireless
R. C. Shit and S. Sharma are with the International Institute sensor network is conducted by Mao et. al. [10] [15]. Gwo
of Information Technology, Bhubaneswar, Orissa, 751003 India e-mail:
rathin088@[Link], suraj@[Link] Yu et. al. [11] classified localization algorithms based on
D. Puthal is with the University of Technology Sydney, Australia e-mail: Multi-Dimensional Scaling (MDS) in the hierarchical network.
[Link]@[Link] The work only concentrates on accuracy and comparisons
A. Y. Zomaya is with The University of Sydney, Australia e-mail: al-
[Link]@[Link] of various MDS schemes. Santos et. al. [12] explained the
Manuscript received Month XX, XXXX; revised Month XX, XXXX. process and computational steps involved in the process of

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 2

TABLE I
L IST OF ACRONYMS A ND C ORRESPONDING D EFINATIONS

Acronym Defination Acronym Defination

3D-MALS) Three Dimensional Multi-power Area Localization Scheme LMAT Localization Algorithm with a Mobile Anchor node based on Trilateration
3DUL Three-Dimensional Underwater Localization LMCS Localization with a Mobile Beacon based on Compressive Sensing
3DVFDPP Three Dimensional Virtual Force Dynamic Path Planning LMFF Levenberg-Marquardt Feed- forward
AAL AUV-Aided Localization LSHL Large-Scale Hierarchical Localization
ABC Assumption-Based Coordinates LSLS Large-Scale Localization Scheme
ADMM Alternating Direction Method of Multipliers LWUPLM Localization With Unknown Path-Loss Model
ADO Arrival and Departure Overlap MACL Mobile Anchor Node Centroid localization
AFL Anchor Free Localization MALS Mobile Assisted Localization by Stitching
ALS Area-based Localization Scheme MARB Moving Aerial Robotic Beacon
ANN Artificial Neural Network MASL Motion Aware Self Localization
APT Accurate Pedestrian Tracking MBAL Mobile Beacon Assisted Localization
BR Bayesian Regularization MBL Mobile Beacon Localization
BRF-BTG BReadth-First- Back-Tracking Greedy MBL (NDC) Mobile Beacon Assisted localization algorithm based on Network Density Clustering
BRL Bayesian Ranging Based Method MCL Monte Carlo Localization
BSR Beacon Signal Ring MDP Markov Decision Process
C-ML Cognitive Maximum Likelihood MLP Multi Layer Perceptron
CAPS Cell-ID Aided Positioning System MMDP Multiagent Markov Decision process
CL Collaborative Localization MSL Multiple Source Localization
DECPOMDP Decentralized Partially Observable Markov Decision Process MSL Multi-Stage Localization
DESR Dual Embedding Spectral Regression NAL Nine Anchor Localization
DETL Detachable Elevator Transceiver Localization PA Peer Assisted
DNRL Dive and Rise Localization PD Probability Distribution
DOL Distributed Online Localization PI Perpendicular Intersection
DREAMS Deterministic Beacon Mobility Scheduling PL Passive Localization
DS theory Dempster-Shafer theory POMDP Partially Observable Markov Decesion Process
DTN Dynamic Triangular Network PWF Peak based Wi-Fi Fingerprinting
DV-ARND Distance Vector-Adaptive Regulated Neighborhood Distance RAA Range-free Aerial Anchor
DV-Hop Distance Vector Hop RAPS Rate Adaptive Positioning System
DWKNN Distance Weighted k-Nearest Neighbors RBFN Radial Basis Function Network
ECHO Extended Computation scHeme of cOordinate RD Random Direction
ECM Expectation Conditional Maximization RNST Reference Node Selection based Triangulation
EEL Energy Efficient Localization RODL Robust Distributed Network Localization
EnLoc Energy-efficient localization RP (LM) Resilient Back-propagation (Levenberg-Marquardt)
FAL Five Anchor Localization RW Random Walk
FF Feed-forward RWP Random Way Point
fGn Fractional Gaussian Noise SCG Scaled Conjugate Gradient
GAC Gps Accelerometer Compass SG Stochastic Game
GD Gradient Descent SLMP Scalable Localization with Mobility Prediction
GM Group Mobility SLNN Source Localization with Nuclear Norm
GMAN Group of Mobile Anchor Nodes SOCP Second Order Cone Programming
GR Generalized Regression SPNP Six Possible Next Positions
HiRLoc High-Resolution robust Localization UPS Underwater Positioning Scheme
HL Hexahedral Localization USP Underwater Sensor Positioning
HL Hyperbola-based Localization VBT Visibility Binary Tree
HSL Half Symmetric Lens VC Virtual Compass
LCB Localizable Collaborative Body WCL Weighted Centroid Localization
LDB Localization with Directional Beacons WPS Wide Coverage Positioning
LLSE Linear Least Square Estimation ZCL ZigBee-based collaborative localization

localization. They consider only three localization algorithms localization as an energy-based method and investigated node
namely directionality based algorithms, terrain and hop-terrain self-localization methods. A discussion on challenges in non-
and compared their benefits, the source of error and accuracy. line-of-sight, node selection criteria, scheduling the sensor
Mobile WSNs are analyzed and classified by Amundson et. node and evaluation of localization criteria are also conducted.
al. [13]. This paper discussed the benefit of mobility, the Stone et. al. [20] provided a three-tier grading method in
difference within static and mobile wireless sensor network, that first they sort algorithm based on distributed, distributed-
the architecture of mobile WSN, localization steps and effect. centralized and centralized and then classified using protocol
Faheem et. al. [14] gave an overview of data-dissemination in techniques. Alrajeh et. al. [21] classified and analyzed local-
mobile sink environment. They described data dissemination ization methods as range-based and range-free. Dong et. al.
methods based on mobility model, mobile sink routing and [22] conducted a comparative survey of mobility issue and
application type. Pal et. al. [16] proposed a classification of medium access control (MAC) protocols of WSN. Han et. al.
localization algorithms and shown factors that affect the pro- [23] [36] analyzed approaches based on static nodes, mobile
cess. Kulaib et. al. [17] presented a distance-based localization nodes, range-based and range-free. Mesmoudi et. al. [24] have
algorithm review. They took ten algorithms that have distinct evaluated localization strategies as range-based, range-free and
characteristic and classified the algorithms as centralized, dis- hybrid methods. Patel et. al. [25] have studied localization
tributed and distributed-centralized. It shows that centralized approaches based on different range measurement techniques.
localization algorithms produce better location estimation than Kuriakose et. al. [26] ranked localization algorithms based
distributed and distributed-centralized algorithms. Li et. al. on the properties anchor node in the network. Tunca et. al.
[18] classified a set of localization and tracking algorithms [27] surveyed based on sink mobility in WSN. Yu Gu et.
based on centroid algorithm, trilateration algorithm and maxi- al. [28] conducted a comprehensive review on sink mobility
mum likelihood estimation. Cheng et. al. [19] explained target management. Mistry et. al. [29] explained localization based

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 3

TABLE II
C OMPARISION OF S URVEY W ORKS FOR L OCALIZATION T ECHNOLOGY
Survey Works Focus Static Coverage Mobility Fingerprint Probability Machine Remarks
Coverage Based Estimation Learning
J.A. Costa 2006 [9] Localization Extensive No No No No Fine-grained based on anchor properties and processing in-
formation covered.
[Link] 2007 [10] Localization Extensive No No No No Comprehensive organizations of static localization designs
[Link] 2007 [11] Localization Extensive No No No No Range based and range-free (Hybrid and Full) methods re-
ported
F. Santos 2008 [12] Localization Limited Extensive No No No Categorized different schemes based on coordination, mea-
surement, and estimation
I. Amudson 2009 [13] Localization No Specific No No No Features on MWSNs but poor algorithmic specifications
Y. Faheem 2009 [14] Data Dissemina- Yes Specific No No No Organizations on data distribution procedures of mobile sink
tion
G. Mao 2009 [15] Localization Extensive No No No No Static localization methods only explained
A. Pal 2010 [16] Localization Extensive No No No No Classification of Static Localization
A.R Kulaib 2011 [17] Localization Static Coverage No No No No Distributed ,distributed-centralized or centralized distance
based approach
[Link] 2011 [18] Localization Extensive No No No No Classification based on single and multi-hop algorithms.
L. Cheng 2012 [19] Localization Extensive No No No No Classified algorithms based on range,number of anchors and
processing types.
K. Stone 2012 [20] Localization MDS Based No No No No Based on different types of Multi-dimensional Scaling ap-
proaches.
N.A. Alrajeh 2013 [21] Localization Extensive range Extensive No No No Mobility of anchor and sensor node is discussed with exten-
based sive mobility model
Q. Dong 2013 [22] MAC Protocol Extensive Extensive No No No Survey on Medium Access Protocols.
G. Han 2013 [23] Localization Extensive Extensive No No No Mobility of Nodes is considered for localization.
A. Mesmoudi 2013 [24] Localization Extensive Limited No No No Distinctive Classification with details of the attributes.
R. Patel 2014 [25] Localization Limited No No No No Range free localization , no classification
J. Kuriakose 2014 [26] Localization Limited No No No No RSSI based localization Schemes.
C. Tunca 2014 [27] Sink Routing Extensive Extensive No No No Comprehensive comparison of routing protocols for mobile
sink nodes.
Y. Gu 2014 [28] Sink Mobility Extensive Extensive No No No Classification based on sink mobility management.
R.P. Mistry 2015 [29] Localization Extensive Extensive No No No Classification based on Range Measurement.
A. Alsheikh [30] Localization Limited Limited No Yes No Markov decesion Model for Localization in WSN
A.K.M. Hossain 2015 [31] Localization Limited Limited Yes No No Calibration free indoor Positioning
S. He 2016 [32] Localization Extensive Limited Yes No No Wi-Fi Fingerprint based Indoor Localization
Q.D. Vo 2016 [33] Localization Extensive Limited Yes No No Fingerprint based outdoor Localization
S.K. Gharghan 2016 [34] Localization Extensive No No No Yes Comparative study of Machine Learning based Estimatiors
for Localization.
T. Chowdhury 2016 [35] Localization Extensive Extensive No no No Discusses algorithms from static and mobility concept
G. Han 2016 [36] Localization Limited Extensive no no No Discusses Mobile anchor node assisted Localization
Our work Localization Extensive Extensive Extensive Extensive Extensive Categorize and Discuss algorithms for IoT Infrastructure

on RSSI measurement and classified different algorithms to tions of this survey paper is to provide an insight and review of
reduce localization error and improve accuracy. localization technology targeting to IoT infrastructure. The rest
Alsheikh et. al. [30] reviewed the localization methods to of the paper is organised as follows: The localization problem
address data transfer, topology formulation, power optimiza- in IoT infrastructure is presented in Sec II. The classification
tion, resource optimization, sensing coverage, object discov- of localization algorithms are presented in Section III to X, In
ery and security challenges. Hossain et. al. [31] reviewed Section XI practical application and further research directions
emerging fingerprint based localization in indoor positioning is discussed. Finally, Section XII presents the concluding
system. Suining He et. al. [32] conducted survey on recent remarks followed by references.
advancement in fingerprint based localizations and presented
the effectiveness of temporal or spatial signal patterns. Q. D.
II. L OCALIZATION P ROBLEM IN I OT I NFRASTRUCTURE
Vo et. al. [33] conducted a classification of existing fingerprint-
based localization strategies which intelligently sense and Internet of things conceptualised a big connectivity model
match different parameters from the environment to identify where all devices and services from all the places connected
the location. Gharghan et. al. [34] examined different machine to the internet all time. The ’things’ of IoT are heteroge-
learning based approach for localization and the influence of nous devices consists of sensors and actuators. The sensor
anchor node density on localization accuracy in the indoor senses signals from environment and the actuator act on
environment. Chowdhury et. al. [35] classified localization environment. The IoT opens an important dimension called
techniques based on the algorithms type, comparative analysis, ’Location of Things’ (LoT) where things get geographical
and application. position information. There are billions of ’things’ in IoT
As discussed, the majority of survey is only based on either and continuously adding. To organise this big data from huge
range-based and range-free [11] [19] [25] [29] or Centralized amount of heterogenous devices the location information plays
and Distributed [9] [17] [35]. Majority of the localization an important role. IoT pulls data from huge devices, filter out
issues together are not covered by the existing papers each and organise them by means of location information. Hence
paper presents some aspect. [34] [33] [32] [31]. In this survey location acts as a search engine for organising big data and
paper the parameter definition, classification of algorithms, devices.
comparative analysis of available algorithms are reviewed and The localization principle developed 20 years back with the
summarised along with future research direction. The proposed development of GPS for military application further it is com-
survey paper is broadly compared to the other review papers mercialised. Currently all device connected to internet uses
presented in the literature and shown in Table II. The contribu- location information e.g. Google Map, Uber, Waze, Foursquare

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 4

etc. Hence the location based services (LBS) comes into play • [Link] Layer: User layer consists of user which use
as a paradigm shift from traditional localization i.e GPS. GPS services through application. The characteristics of users
is just a subset of LBS since it cannot localize ’things’ (sensor are user ID, location, application ID etc. A single user
and actuators) always. There are huge number of scenarios like can use multiple applications.
indoor network, underwater network, unfavourable weather In the given LoT model of IoT infrastructure location infor-
condition when GPS do not work. So alternative techniques mation of ’things’ are very important for providing services
required. Hence development of huge number of techniques and application. The cloud should know the location informa-
and protocol is done for these scenarios and localization plays tion of user and sensors to provide services. To control and
an important role for services in IoT infrastructure. A major coordinate the ’things’ (sensor and actuators) of IoT the virtual
sink or cloud require location information a priori. Hence
there is requirement of localization and prediction of ’things’
(sensor and actuators) in IoT infrastructure. To schedule sensor
for respective user location of user is important to the cloud
or virtual sink. The advantages of this paradigm shift over
traditional WSN are
• Energy efficiency of services due to optimization of
sensors and services.
• This infrastructure provides on demand services based on
quality of information.
Hence localization plays a key role in Sensor-cloud based
IoT infrastructure.

III. M ETHODOLOGIES
In the last few years research is going on for finding the
efficient algorithm for localization. The localization of sensor
node is a two step process of first physical measurement with
some reference and second estimate the location from the
Fig. 2. Location based interactive model of IoT (showing IoT-Cloud-WSN physical measurement data. Sometimes nodes with limited
integration) hardware do not conduct physical measurement rather they
try to calculate the position from neighbour node information.
part of IoT infrastructure consists of heterogenous sensors so The following approach for localization are existing in the
it can be called as sensing service on demand based on user. literature. These are based on range measurement, available
IoT infrastructure is a paradigm shift from traditional sensor of anchor nodes, fine-grained, coarse-grained, incremental and
network as shown in Fig. 2. It consists of a cloud or virtual sink concurrent localization method. The basis of classification of
which schedule and control physical sensor network based on the above scheme are based on either anchor node or range
location of users. Traditional WSN have grown into Sensor- measurement.
Cloud architecture in IoT infrastructure. Here efficient sensor From the existing literature and IoT application it is found
network are used for Sensing as a Service (SeaaS) model that there is a need of review of existing technology to
of the IoT infrastructure. The architecture of location based address the localization problem to the IoT infrastructure. The
interactive model of IoT infrastructure [38], [39] is presented selection of proper localization method in the IoT scenario
in Fig 2. The architecture consists of four layers. aims to address the restriction and avoid the pitfalls. The
• [Link] WSN Layer: This layer consists of physical proposed taxonomy classifies the localization approach based
sensor network. The characteristics of these sensors have on offline training. Since IoT infrastructure analyses huge data
sensor ID, state, location and ownership. and devices. There is huge scope of application offline training
• [Link] or Virtual Sink: The sensor layer forward their for localization. So training step is the key for taxonomical
sensed data to cloud or virtual sink. The characteristics hierarchy of localization method in IoT infrastructure. The
of cloud is cloud ID, QoS, resource and price. Cloud classification is presented in Fig 3. The localization approach
provides sensing service from different sensor network are reviewed into two category considering IoT scenario as (1)
from different WSN with different owner. The cloud Self Determining Method (2) Training Dependent Method.
provide on demand sensing service Sensing as a Service
(SeaaS). A. Self Determining Method
• [Link] Layer: Application layer provide different Self determining method can be sub divided into four
applications to the user and access services from cloud. categories namely (1) Geometric Method. (2) Mobility model
The characteristics of application layer is application ID, (3) Path planning (4) Statistical Approximation. The geometric
region of interest, sensor data of interest and Quality of method consider various geometric parameters to estimate
Services (QoS) requirements like sensing intervals, delays location. This method check triangular information like late-
etc. ration, angulation or connectivity among sensor nodes. This

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 5

Fig. 3. Taxonomy of Localization Methods

method is categorised based on connectivity, centroid calcula- IV. G EOMETRIC A PPROACH OF L OCALIZATION
tion, landmark design, region overlap, clustering etc. Mobility
This approach of localization exploit geometric info to find
model take advantages the mobility pattern of anchor nodes of
location. These methods are categorised as follows.
the network . The mobility pattern can be random or specified.
This is further classified on the basis of mobility patterns
like random walk, random direction, self-organisation etc. as A. Connectivity
shown in Fig. 3. Path planning is advanced version of anchor The range-free algorithm determines the pairwise distance
mobility model in which anchor node move in a specified path of two nodes from the connectivity information. There are
in the network to localize the whole network. Few scenario different types of classification such as connectivity [40] [41],
of this categories like static path model and dynamic path centroid [49] [50] [51], energy attenuation [52] [53] [54] and
model have huge application in robotic sensor and actuator Region overlap [55] [56] [57] and others [58] [59] [60] [61]
network. Statistical approximation techniques uses standard as shown in the Table III. Connectivity method consolidates
approximation methods like Least Square, Maximum Likeli- Graph theory with node’s localization. Graphs with all the
hood, Multi-Dimensional Scaling, Semidefinite Programming points joined are called connected graph. Network topology
etc. to estimate location. plays a significant role in connectivity based algorithms. There
are two algorithms DV-Hop [40] and Localizable Collabora-
B. Training Dependent Method tive Body (LCB) [41] which uses connectivity information.
These method uses an offline training phase before online Initially the anchor node broadcasts its coordinate and hop
localization step. These are classified into (1) Fingerprinting count in vector [Link] node who receives the vector packet
(2) Stochastic Model (3) Machine Learning approach. Fin- again broadcast it with increased hop count. In this way all
gerprinting method is further classified into signal, visual, nodes have a list of anchor node coordinate and the distance
motion and hybrid types of fingerprints. Stochastic Models in hop count. Unknown node calculate the distance with the
can be of POMDP (Partially Observable Markov decision anchor nodes and find its position as shown in Fig. 4 (a). For
process), MDP (Markov decision process), SG (Stochas- efficient calculation, there is the requirement of at least three
tic Game), HMDP (Hierarchical Markov decision process) anchors in the range of unknown nodes. The density of anchor
and CMDP (Constrained Markov decision process). Machine decreases the localization error.
learning based approach is classified into LMFF (Levenberg- LCB also uses graph theory to localize unknown nodes
Marquardt Feed-Forward), RBFN (Radial Basis Function Net- with multihop anchors. LCB overcome the restriction of three
work), BRGD (Bayesian Regularization Gradient Descent), anchors for calculating the position of unknown nodes with
RPML (Resilient Back-propagation Multilayer) and MLP multihop anchors. In LCB all the anchors first announce their
(Multi-Layer Perceptron). location information. The remote node acquires the location

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 6

TABLE III
G EOMETRIC A PPROACHES OF L OCALIZATION
Localization Method Algorithm Sensor Density Anchor Density Accuracy Power Consumption Remarks
DV-Hop [40] High 3 or more High High High Energy Consumption but computationally sim-
Connectivity
ple
LCB [41] High 3 or more High Medium Medium Energy Consumption computationally com-
plex
DV-ARND [42] High 3 or more High Medium Cooperative method uses connectivity range mea-
surement
Trilateration
ECHO [43] Medium 3 or more High Medium Cooperative method
DS Theory [44] Medium 4 or more High Medium Non cooperative method uses joint RSS and AoA
range measurement
RODL [45] Medium 3 or more Medium High Cooperative method uses TDoA range measurement
M-Mobility [46] Medium 3 or more Medium High Cooprative method relies connectivy information
Multilateration
AFL [47] Medium 3 or more Medium Medium Acoustic sensors uses connectivity information co-
operatively
Triangulation ABC [48] High 3 or more High Medium Uses RSSI information cooperatively for estimation
Centroid [49] Low 3 or more Low High High Energy Consumption but computationally sim-
Centroid ple
ABC [50] Low 3 or more Medium Medium Medium Energy Consumption with complex compu-
tation
Threedimensional Centroid [51] Low 4 or more High High High Energy Consumption and simple algorithm
Source Energy Attenuation [52] Average 3 or more Medium Medium Medium Energy Consumption computationally sim-
ple
Energy Attenuation
BSR [53] Average 3or more High High High Energy Consumption and computationally sim-
ple
Energy Interval [54] Average 3or more High High High Energy Consumption and computationally sim-
ple
HiRLoc [55] Small 3or more Medium Medium Medium Energy Consumption with moderate com-
plex
APIS [56] Low 3or more High Low Low Energy Consumption and complex
Region Overlap
Voronoi [57] Low 3or more Medium Medium Medium Energy Consumption and moderate compu-
tation
Bionics Honeybee Orientation [58] Medium 3or more Medium High High Energy Consumption and computationally sim-
ple
Verification Weighted Centroid Algorithm [59] Medium 3or more Medium Medium Medium Energy Consumption and computationally
simple
RNST [60] Low 3or more High Low Low Energy Consumption and computationally sim-
Landmark Design
ple
Landmark Sparse [61] Low 3or more Medium High High Energy Consumption and computationally sim-
ple
Cluster Target Tracking Algorithm [62] No effect 3or more High High Low Energy Consumption and complex
Distributed mobile Localization [37] No effect 2 High Medium Medium Energy Consumption and moderate compu-
Historical Information
tation
DTN [63] No effect 3 or more High High Medium Energy Consumption and moderate compu-
tation

information and modify the topology of the network and build

a BN-tree. Unknown node estimate their position by the co-
ordinates of the anchor nodes and modify the BN-tree as
shown in Fig. 4 (b). LCB method reduces computing and
communication cost, but it causes commutative localization
error due to cooperative localization between unknown nodes.

B. Lateration/Trilateration/Multilateration
Wang et. al. [42] improved the DV-hop algorithm which
uses regulated neighbourhood distance (RND) based on disk
communication model of localization. RND method corrects
(a) the ambiguity in hop distance with the help of neighbouring
node’s proximity information. It is an adaptive method called
DV-ARND. An iterative method for localization is proposed
[43] in which a node’s position is represented in barycentric
coordinate system. It is an extension of Distributed Iterative
LOCalization (DILOC) method [64]. This method solves the
problem considering the pseudo linear system. Dempster-
Shafer (DS) theory [44] is a novel method based on data fusion
and Bayesian probability theory. A new method called Standby
[65] is adopted with RSS and AoA which is used in DS
Theory called as basic probability assignment (BPA) [65]. BPA
(b)
considers best and worst case scenario with the lower bound,
Fig. 4. Connectivity approach of localization (a)DV hop Localization algo- upper bound and confidence. This method first measures RSS,
rithm. (b)LCB algorithm evolves from BN tree with dark dots are landmark
and hollow dots are unknown nodes. AoA, and standby distance. Then the filtering of the data is
done and minimum and maximum values are taken for lower
and upper bound of BPA. The BPAs are then aggregated,

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 7

(a) (b) (c)

Fig. 5. Geometric Centroid Method (a)Tetrahedron method (b)ABC Method (c)Three dimensional Centroid algorithm

and most possible distance is predicted with lower and upper WSN. The method considers a geometric relationship with
bound. Moore et. al. [45] proposed a Robust Distributed communication constraints to find the position. First three-
Network Localization (RODL). It is an anchor-free localization dimensional graph is constructed taking into the anchor profile
algorithm and considers noisy range measurement and flip then it is converted to plane graph, and finally, a centroid of the
ambiguity. First it forms cluster in the form of quadrilateral. plane graph gives the estimated position. In this method the
After each cluster is localized, overlapping quadrilateral are localization accuracy is 99% with minimum 6 anchor nodes.
found using numerical optimization. Finally, local clusters are
combined using common nodes.
E. Energy Attenuation
Energy attenuation method uses signal energy pattern of the
C. Angulation/Triangulation anchors to find the position of unknown nodes as shown in
Assumption Based Coordinate (ABC) [48] method uses Fig.6 (a). The distance of the unknown node from the anchor
range measurement to find a local position of each node. In this is extracted from the attenuated signal energy. Sound energy
method, a node collects information about range measurement is used to study this method. In source energy attenuation
from all the neighboring nodes and localize itself. Each local method [52], first the objective function is constructed with
position information is shared with the neighbors and forms a the help of maximum likelihood method then the position
global map. of the unknown node is estimated with the Gauss-Newton
Method. By increasing the SNR (Signal to Noise Ratio),
the positioning accuracy can be increased immensely. The
D. Centroid
biggest problem in this method is the impact of noise in
Centroid approach uses connectivity knowledge for location signal propagation which involves error in the localization
estimation of the nodes. The anchor nodes periodically broad- process. Beacon Signal Ring (BSR) [53] approach is a novel
cast their location. For the calculation of location information method in which the anchors broadcast signals of multiple
unknown node find the centroid of the received locations of power level to remote nodes. Each remote node receives and
the anchor nodes. In [49] the author explained the centroid monitor power report and estimates the signal range based
method in which the unknown node estimates the coordinate on the acquired information. This method has higher accuracy
of the centroid of the tetrahedron as shown in Fig. 5 (a). There than the previous method but to transmit in different power the
is two process of calculating the centroid one is a traditional requirement of energy is more. Energy Interval [54] method
method and other is to find the centroid of the tetrahedron. estimate the coordinate of the unknown nodes based on Log-
The localization error of tetrahedron method is 0.54R, whereas normal distribution representation of energy attenuation. Here
traditional methods give error 0.7R. So the tetrahedron cen- a region mapping is done between the received signal power
troid method improves 29% over traditional method. Since and the transmission range. From the received energy the
the calculation runs in many rounds for unknown nodes, so it unknown node estimates its transmission range. Then the
causes high energy consumption. In [50] the algorithm incor- region information is transformed to distance information of
porate DV-hop with Assumption-Based Coordinates (ABC) for the overlapping region. The centroid of this overlapping region
localization. This is the augmented form of the DV-hop method provides the position of the unknown node. The calculation is
in which first unknown nodes calculate distance information done with many rounds may rise the energy consumption.
from the connectivity information using the DV-hop method
and then apply ABC algorithm to find position information
as shown in Fig. 5 (b). This is the simple computation with F. Region Overlapping
high error. Another novel method is three-dimensional centroid Region Overlap method estimates the position of the un-
algorithm [51] which considers the unknown nodes in 3-D known node by calculating the centroid of the overlapping

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 8

(a) (b) (c)

(d) (e) (f)

(g) (h)
Fig. 6. Geometric Approaches (a)Energy grade overlap for energy attenuation method (b)Region Overlapping (c)Spherical shells overlap (d)Voronoi Graph
(e)Left initial position and U2 updated landmark diagram. (f)Distributed mobile localization algorithm. (g)DTN Localization algorithm. (h)Target tracking
algorithm.

region as shown in Fig. 6 (b). This method does not calculate position. Number of the data packet in HirLoc is less compared
the distance information. Hence, it reduces the communication to APIS [56] method. Voronoi graph [57] method is another
cost and saves the energy consumption. There are three promising method in which the unknown node sort the RSSI
algorithms in this approach HiRLoc [55], APIS [56], Voronoi (Received Signal Strength Indicator) in descending order of
[57]. The HiRLoc [55] algorithm uses variable transmission the anchor and calculate the Voronoi region of the anchor
power and variable angle of the directional antenna to reduce using Unit Disk Graph followed by finding the centroid of
the overlapping region of the unknown nodes. The effect of the Voronoi overlap region. This centroid is the position of
anchor node become very less in this method. However, the the unknown node. Here the Localization error increases as
communication cost is more as compared to other methods the communication space and anchor density decreases.
due to transmission and the directional antenna. To reduce the
communication cost a similar method called Spherical shell
G. Bionics
overlap method APIS is proposed in [56]. In this method
anchor node broadcasts location information in the spherical Bionics algorithms consider the biological motion and try
form and the unknown node receives and calculates the to mimic it to localize sensor nodes. The algorithm establishes
a relationship between the position of the unknown node and

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 9

the anchor based on the laws of biological motion. The habits

of the bee in the physical world for searching nectar is used
for the localization of unknown node [58]. This algorithm first
computes the distance from nearby anchor according to signal
strength then using the cosine law estimate the relative angle
and obtain the relative position based on the bees model. Here
accuracy is high and hardware cost is also high.

H. Verification
Verification localization method estimates pairwise distance
from the RSSI information. First, the algorithm estimates the
pairwise distance of the nodes from RSSI information then
verify the weight value using distance and RSSI and further
estimates the position. In this method, localization error is less
as compared to weighted centroid algorithm [59].
Fig. 7. Comparison of different geometric localization approach with respect
I. Landmark Design to accuracy and power consumption
Landmark placement has a significant effect on localization
error. These placement methods use the different geometry
L. Summary and Insight
of positioning of the anchors which reduces the localization
error of unknown nodes. Reference Node Selection based The existing geometrical method exploit the geometric
Triangulation (RNST) [60] shows less error in localization architecture of the network to optimise accurately and power
if the anchors construct an equilateral triangle. Landmark consumption. The reviewed article compared on the basis of
Upgrade method is another method to increase the accuracy of accuracy and power consumption and a comparative chart is
localization. In WSN the number of the anchor is limited so plotted and analysed. Landmark design and region overlap
all the unknown nodes do not get three neighboring anchors. method gives promising result than other method in terms of
This problem is solved with updating accurate estimated non- accuracy and power consumption as shown in Fig. 7.
achor unknown nodes to anchor nodes. This method causes
more communication cost but increases accuracy [61]. V. M OBILITY M ODEL
The different mobility pattern of anchor nodes are exploited
J. Cluster in this method to find the geographical location. These are
Cluster-Based Algorithm have lower computational com- explained as follows.
plexity. These algorithm first divides the region into the
clusters. Each cluster is associated with one anchor node. A. Random Walk Model
Each anchor localize unknown nodes of the cluster and finally,
all the localized information is merged, and final estimation
is done. The distributed target tracking algorithm [62] is a
cluster based algorithm which first divides the region into the
cluster with one anchor node. The anchor node takes care
of establishing the connectivity assigning the task and target
tracking. When the unknown node moves into the different
group, the respective anchor node of the cluster cooperatively
estimate the position.

K. Historical Information
The concept behind the historical information method is to (a) (b)
take the historical information to estimate the current position. Fig. 8. (a) Random Walk Mobility Mode. (b) Possible area in DOL algorithm
Here first maintain a queue of at least three location of the
unknown node and form the linear motion equation followed Random Walk Mobility Model [66] [67] [68] is used by
by the estimation of the location [37]. Dynamic Triangular the anchor node to move in the network from one point to
Network (DTN) [63] Algorithm predict the location with the another point with a random velocity and random direction.
help of RSSI information of the unknown node. Here the Here current speed and direction are independent of the past
algorithm first considers the possible location with respect to speed and direction.
the anchor nodes and then estimates the distance from two Moving Aerial Robotic Beacon (MARB) [66] localization
anchor nodes. The error is calculated from estimated distance is based upon the RSSI measurement technique. There is a
and actual distance and compared. The lowest distance from relation between distance and RSSI value of the unknown
the anchor is considered as the position of the unknown node. nodes to the anchor nodes. A Bayesian framework is used

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 10

TABLE IV
M OBILITY M ODEL APPROACH OF L OCALIZATION
Localization Method Algorithm No. of Anchors Anchor Utilization Accuracy Power Consumption Remarks
MARB [66] Single Low High Medium Not Complete Localizable due to Random Walk
RW MBL-MDS [67] Single Low High Medium Sufficient Reference Required
DOL [68] Single Low Medium Low Simple technology used
Ssu [69] Single Low Medium Low No Extra Hardware since Range-free
RAA [70] Single Low Medium Low No Extra Hardware since Range-free
RWP
Yu [71] Single Low Medium Medium Mobile Trajectory is straight
LMCS [72] Single Low High High More Energy Consumption but Better Performance
in Obstacle Environment
Finegrained [73] Single Low High Low Better than Ssu [67]and Yu [69] but more cumulative
RD
localization error
ADO [74] Single Low Depends RSS model Medium Choice of RSS model is important
FAL [75] Five High Medium Low Simple technology used
GM
NAL [76] Nine High High Low Less energy consumption due to no distance estima-
tion
Self Organising Time-based [77] More High Average High High travelling speed with High energy consumption
PD MCL [78] More Low High High Medium travelling speed with High energy consump-
tion

to estimate the location of the unknown nodes where the more.

conditional probability is the measure of the position. Here the
position is estimated and updated recursively. Mobile Beacon B. Random Waypoint Model
K. F. Ssu et. al. [69] demonstrated a range free method
of localization of mobile anchor nodes using perpendicular
bisector of the chord in which the endpoint of the chord
represents the two anchor nodes. These two sets of the
perpendicular bisector of the chord intersect at a point, which
represents the localization of the unknown node.
Range-free Aerial Anchor node (RAA) [70] localization
method is proposed for the network with the mobile anchor
node. In RAA method a circular cross section is formed by the
different beacon position of the anchor node.A perpendicular
line is drawn from the centre point of the circular cross section
to the unknown node plane. This point is estimated as the
(a) (b) position of the unknown node.
Yu et. al. [71] modified the previous algorithms [70] [69]
and proposed an algorithm with a flying mobile anchor nodes.
Many circular cross section can be drawn by different anchor
nodes. The perpendicular lines are drawn from the unknown
node to the centre point of the circular cross section of
different mobile anchors. The intersection point of different
perpendicular lines is estimated as the position of unknown
node. Localization with a Mobile Beacon based on Compres-
sive Sensing (LMCS) [72] is a mobile anchor based range free
scheme. LMCS uses comprehensive sensing (CS) to prepare
the related order of the unlocalized nodes. From the related
(c) (d) order, LMCS selects the weight of each point and computes
the location of the unknown node by weighted centroid.
Fig. 9. Random Way Point (a)Perpendicular bisector of Chord (b)Localization
With Mobile (c)Localization with aerial (d)Localization with flying anchors
C. Random Direction Model
Localization (MBL) [67] uses the classical Multidimensional Random Direction (RD) movement model [73] [74] assumes
Scaling (MDS) with connectivity information in mobile anchor that anchor move in a random path till the boundary of the
nodes. Here the mobile anchor broadcasts the beacon packets deployed network region as shown in Fig.10 (a). After a pause
in each point and the unknown node receives and estimates anchor node again take a new direction randomly from zero
the position using MDS. It is applicable in 3D Network. to 180 degree to move.
In Distributed Online Localization (DOL) [68] first, the The fine-grained method [73] is also a range free method.
mobile anchor nodes broadcast the beacon packets. In the It is a updated method described by K. F. Ssu et. al. [69] and
reception of beacon packets, the unknown node approximates C. Zhao et. al. [71]. In this method possible positions can be
its position using rectangular bounding box. Further, it finds calculated using the analytical method. The position of the
the centroid of the estimated location of the several beacon unknown node can be finalized from the possible positions
packets which reduces the complexity, but the error is quite using the mobile anchor node.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 11

(a) (a) (b)

Fig. 11. Group Mobility Model:(a)Possible Area in FAL(b)Possible area in
NAL

concept of this algorithm is to determine the position of the

remote nodes for a short period. A self-organization method
[77] consider unknown nodes static for a certain interval and
send beacon packet in this interval to localize. When unknown
node get enough beacon to calculate the position, it uses the
trilateration method to calculate its position.

F. Probability Distribution Model

(b)
Probability distribution method uses prior probability to
Fig. 10. Random direction Mobility Model (a)Model showing the path
(b)Localization using arrival and leaving information of the beacon estimate the location of the unknown nodes. Dynamic Monte
Carlo Localization (MCL) [78] method has two stages of
localization; prediction and filtration. In prediction stage, the
Arrival and Departure Overlap (ADO) [74] is a distributed node forecasts its position based upon preceding probability
localization method. ADO uses different mobility models distribution of the network reserved. In filtration stage, the
sparse straight line, dense straight line and random movement node discards the incompatible position information from the
pattern. The location of the unknown node is estimated as the network. The main disadvantage is that it require a reserved
intersection point of received beacon signals from the mobile information. The complexity of this method is very high.
anchor nodes as shown in Fig. 10 (b).

D. Group Mobility Model

Group mobility model consist mobility of more than one
anchors at a time as shown in Fig. 11. This method com-
pares the movement of the transmitted RSS and geometrical
information. Here unknown nodes can estimate their position
without much interaction. The primary two methods are Five
Anchor Localization (FAL) [75] method in which four anchors
located in a square with one unknown node at centre based on
RSSI. Another method is the Nine Anchor Localization (NAL)
[76] which uses eight anchors in a circle with one unknown
node at the centre using the RSSI. Multiple anchor localization
uses a particular geometric pattern and mobility model of the
anchor node in the network.

E. Self Organizing Model Fig. 12. Comparison of different Mobility-Based localization system with
Time-based algorithms rely on the continuous movement respect to accuracy and energy efficiency
of the anchor nodes to localize unknown nodes. The main

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 12

TABLE V
PATH P LANNING A PPROACH OF L OCALIZATION
Localization Method Algorithm Area Coverage Number of Mobile Anchors Localization Accuracy Power Consumption Anchor Utilization Remarks
Lines [79] No Sngle 90% Low Low Simple path planning but needs extra hardware for
long communication range.
SCANS, DOUBLE-SCAN, HILBERT [80] No Single 90% Medium Low Many collinear anchors wasted.
CIRCLES [81] No Single 70% Medium Low Non-collinear anchors with Shorter path lenght than
[82]
MACL [83] No Single 50% Medium Low Non-collinear anchors with shorter path length.
Static Path GMAN [84] Yes Multiple 90% Medium Low Synchronisation in movement of anchor.
K-Coverage [85] Yes Single 70% High Low Ensure k-coverage but long trajectory length.
PI [86] Yes Single 90% Medium Low Energy consumption for turns not considered
LMAT [87] Yes Single 70% Medium Low Energy consumption for turns not considered
S-Type [88] Yes Single 90% High Low Much energy consumed for turn but solved collinear-
ity problem.
SCAN Based [89] Yes Single 95% High Loow Longer path length but fully localisable network.
WCL [90] No Multiple 90% High Low Beacon collinear problem present
Layered Scan [91] No Single 90% High Low Applicable for 3D localisation more path length more
energy consumption
HL [92] Yes Single 80% High Low Applicable for 3D localisation more path length more
energy consumption
MBAL [93] No Single 50% Medium High Works well in ununiform WSNs but not consider
obstacle environment
BRF and BTG [94] No Single 95% Medium High High anchor utilisation
MBL (nde) [95] No Single 90% High High Dynamic algorithm but Highest complexity
Dynamic Path Virtual Force [85] No Single 90% Medium High Works well in ununiform WSNs but needs direc-
tional antennae
MALS [96] No Single 90% Medium High Applicable for Large scale WSNs but high localisa-
tion delay.
SPNP [97] No Single 50% Low High Less computational load but high localisation delay.
DREAMS [98] No Single 90% Medium High Node failure is a major problem.
Anchor Guiding [99] No Single 90% High High Can balance the localisation error but computational
complexity is high.
3DVFDPP [100] No Single 90% High High Avoid movement in no sensor field but calculate
virtual force in all directions leads complexity.
Virtual Ruler [101] No Multiple 80% High High applicable to obstructed WSNs with ultrasound trans-
Dynamic Path+Obstacle mitter
Snake Like [102] No Single 90% Medium High Use minimum beacons but applicable only for known
obstacles.
Visibilty-Binary Tree [103] No Single 90% Medium High Shortest travelling path but long localisation delay.

G. Summary and Insight Circle and S-curve are two other methods [81]to reduce
Mobility model concentrates the movement pattern of the localization collinearity. This types of scan leave the corner of
anchor nodes. The more the coverage of the mobile anchors the region.
more the sensor can be localized in a network. The six Mobile Anchor Centroid Localization (MACL) [83] is pro-
categories of methods like random walk, random direction, posed, in which the anchor node traverse in a spiral path and
self organizationn, Random waypoint, Group Mobility and periodically broadcast the beacon packet which contains the
Probability Distribution are reviewed. It is found that some current location.
method requires sufficient reference when other uses simple Group Mobile Anchor Node (GMAN) [84] contains three
hardware with RSSI. A comparative analysis chart is drawn anchor nodes which construct an equilateral triangle with all
of those method and shown in Fig. 12. RW and RD methods anchor at the vertex. In two-tier network architecture, the
found to be outperform other methods in terms of accuracy anchor group move in the x-axis or in a random pattern.
and energy consumption. K-coverage [85] is employed to diminish beacon density and
trajectory range. This method contains two steps. In the first
VI. PATH P LANNING A PPROACH step, an optimal 3-coverage is used for deployment of virtual
Path Planning tries to improvise the localization accuracy beacons. In the second step, the ant colony optimization is used
with a best possible trajectory of the mobile anchor node. to create a movement pattern for a mobile anchor to pass the
Problem of path planning is to design a trajectory which virtual beacons. Perpendicular Intersection (PI) [86] method
• Passes maximum nodes in the network. uses perpendicular intersection to determine node position.
• Provides each node with sufficient number of anchors. The mobile anchor node starts at one position and move zigzag
• Reduces the energy consumption using the shortest tra- manner with an angle 0 < θ < π/3
jectory length. Localization with Mobile Anchor using Trilateration
The path planning method can be of two types; first is static in (LMAT) [87] presents the progress of mobile anchor nodes
which the path is predefined and second is dynamic in which in triangular trajectory to improve the localization accuracy.
path can be changed. The major advantage of this method is that it can estimate the
Static Path Planning can be of two types such as two position even if the beacon points are collinear.
dimensional and three dimensional. In two-dimensional path S-type is another method [88] in which the mobile anchor
planning anchor traverse in the two-dimensional plane [79]- node moves in an S-type movement in the network area.
[88]. The√network area is divided into the small square of size
In Lines [79] the anchor node moves in the X-axis of the R/ 2. This method can maintain the shortest path for the
region and broadcasts gradient signals to localize the unknown trajectory and collinearity problem is solved. SCAN Based
node. It is independent of unknown node density. method [89] is based on scan algorithm in which nodes can
Based on the scan it can be SCAN, DOUBLE SCAN, and identify three or more anchors to form two nonparallel chord
HILBERT [80]. It maximize the network coverage. The simple of a circle. The length of each chord is allowed to take a
SCAN moves the anchor and scan in the x-axis. The DOUBLE specified value to reduce the error. The anchor node will move
SCAN moves the anchor in both direction but HILBERT taking a scan and when finding an obstacle it will bypass
SCAN moves the anchor node in the Hilbert pattern. nearby obstacle crossing in the right-hand direction. After

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 13

(a) (b) (c) (d)

(e) (f) (g) (h)

(i) (j) (k) (l)

Fig. 13. Static Path Planning of anchor node for localization:a)SCAN (b) Double Scan (c) Hilbert (d) Circles (e) S-Curves (f) Trajectory of MACL (g)
Trajectory of GMAN (h) Trajectory of PI(i) Trajectory of LMAT (j) Trajectory of S-type (k) 3D Scan (l) HL trajectory

bypassing, the anchor node returns to its previous path. to localize in the dynamic and real-time sensor network. So
A Weighted Centroid Localization (WCL) [90] method has there is the requirement of active route plan. Since motionless
been proposed. In WCL four mobile anchor nodes forms the route preparation considered the sensor nodes to be disposed
tetrahedron and moves as the layered scan trajectory. Hence it uniformly. But in real scenario sensors are distributed non-
can cover the three dimensional area of the network. uniformly so it is very hard to localize. Static path plan takes
The Layered Scan [91] is the modification of the 2D scan in the long route, time delay and under utilization of beacon
3D. Here the 3D region is divided into layers and each layer messages. So to improve these issues, there is a proposal
is scan using 2D scan methods and S-curve method to reduce of dynamic path planning. It is of two types one is without
collinearity and co-planarity. obstacle and second is with obstacle. So obstacle detection
and avoidance are the requirements in the current research.
Hexahedral Localization (HL) [92] is a 3D path planning
based localization in which space is divided into hexahedrons. Mobile Beacon-Assisted Localization (MBAL) [93] is con-
Unknown nodes can be localized using the perpendicular sists of three steps. First anchor movement step, Second sensor
characteristic of the route. Mobile beacons move in the path localization step and finally the path decision step. The mobile
and continuously broadcast their position information and the anchors move in a regular triangle and broadcast three beacon
unknown nodes take the position information and calculate packets. When the remote nodes demand more beacon packet,
their position. then the anchor node determines the trajectory of the request,
The use of single anchor node is more economical than and move towards the demand to reduce the path length.
multiple anchor nodes, but it gives issues like collinear and A method has been proposed [94] for Dynamic path plan-
co-planner. Multiple anchors can solve these issues with better ning. In this the network is considered as an undirected graph
localization in the 3D sensor network but still they are unable where vertices of the graph are unknown nodes. Breadth First

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 14

(BRF) and Back-Tracking Greedy (BTG) algorithm is used to

form the path for the anchor node by traversing the undirected
graph. The movement of the anchor node varies dynamically
according to the path.
Mobile beacon-assisted localization (MBL) [95] considers
the node distribution information of the network. It divides the
network based on clusters. Mobile anchor first traverses all
the group head using the genetic algorithm. Then the anchor
moves in a regular hexagon in particular cluster including
cluster head in the middle of the hexagon. Optimal movement
is obtained from the local and global path. It improves (a) (b)
localization accuracy and reduces the power consumption.
Dynamic anchor movement trajectory based on the potential
field is a new idea [85]. An interaction between mobile
anchor nodes and unknown nodes is created in this method. In
non-uniformly placed network, sensor nodes are placed with
omnidirectional antenna and anchor nodes with the directional
antenna. Anchor node sends information in a fixed direction
and receives feedback to calculate total virtual force. This total
virtual force helps the anchor to decide which direction it has (c) (d)
to move.
Fig. 14. Dynamic Path Planning of anchor node for localization:(a)MBAL
Accomodating non-uniform and irregular sensor network (b) SPNP (c) Virtual Ruler (d) Snake Like Obstacle Avoidance
with the mobile anchor is proposed in Mobile Assisted local-
ization by stitching (MALS) [96]. It takes the rigidity theory
to partition the network with small localizable units. The final Visibility binary tree (VBT) [103] is an obstacle avoidance
path for the anchor node is obtained by joining all the units. method. In VBT all of the tangent paths between obstacle is
Six Possible Next Position (SPNP) [97] method is proposed drawn. Then the redundant paths are reduced. Finally, it apply
by Li et. al. This method utilize anchor routes to localize an ad-hoc searching algorithms to form an optimal binary tree
unknown nodes. First, three anchor nodes are placed randomly to calculate the shortest path.
as they form an equilateral triangle. Then the next three anchor
node are placed in the position where they can cover more A. Summary and Insight
number of unknown nodes. This method is computationally
Another similar method to mobility pattern based approach
efficient and consume low power.
is path planning approach gains huge interest in current
Deterministic Beacon Mobility Scheduling (DREAMS) [98] localization techniques. Reviewing existing literature of path
method presents a new idea in which each anchor (beacon) planning and considering static, dynamic and obstacle avoid-
traverse a path in the network and produce a traversal tree. ing path planning of anchor nodes. It is found that the anchor
The traversal tree is used to find the path of the mobile anchor node is very limited in these method. Few algorithms use extra
(beacon) by distance-based heuristic method. hardware to maintain the anchor node movement. Majority
An anchor guiding method [99] is proposed based on of the algorithm used collinearity principle of anchor nodes.
the size of the estimated region. The unknown node and Anchor utilisation has a huge impact on accuracy and power
anchor nodes have some estimated rectangle [Link] the consumption in these approaches.
unknown node receive the beacon from the anchor a new
region called broadcasting region is formed. Anchor node VII. E STIMATION /S TATISTICAL A PPROXIMATION
try to move in the broadcasted region and consider this A PPROACH
as the estimated rectangle region. Three-dimensional virtual
force dynamic path planning (3D VFDPP) [100] assumes that The methods in this category uses various statistical ap-
there sensor nodes make a team. Sensing zone of the mobile proach to estimate the location information. This method is
anchor node is partitioned into eight zones. The anchor grow further classified into following.
according to the virtual force from a certain direction. It will
have the tendency to move in that direction. A. Multidimensional Scaling (MDS)
Since there is not always the cases in sensor network Multidimensional Scaling (MDS) [104] is a range based
without the obstacle, so few algorithms are proposed for distributed algorithm in which pairwise distance between the
obstacle based path planning for anchor nodes. nodes is calculated to form a local map. It is further stitched
A Snake Like [102] anchor movement algorithm is proposed to form the whole map to get the spatial coordinates. There
in which anchor node moves as snake. When it hit some are two types of MDS: Classical MDS and Iterative MDS.
obstacle, it detours the path and saves the obstacle point for MDS-MAP [105] is a good solution for anchor free range-
future movement. free localization problem. It consists of three steps. First,

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 15

TABLE VI
E STIMATION /S TATISTICAL A PPROXIMATION A PPROACH
Method Algorithm Centralized/Distributed Range Measurement Architecture Accuracy Remark
Distributed MDS [104] Distributed RSSI Cooperative Medium Immobile nodes with scalable
MDS-MAP [105] Centralized Connectivity Cooperative Medium Immobile nodes not scalable
MDS MDS-MAP (D) [106] Distributed Connectivity Cooperative Medium Medium node density
Map-Stitching [107] Centralized Connectivity Cooperative High Work on dense network
Patch and Stitch [108] Centralized Connectivity Cooperative Medium Work on dense network and scalable
LWUPLM (Localization with Unknown Distributed RSSI Non Cooperative Medium Immobile scalable nodes
Path loss Model) [109]
EEL (Energy Efficient Localization) [110] Distributed RSSI Non Cooperative Medium Immobile scalable nodes
LS
LLSE [111] Distributed RSSI/ToA Non Cooperative Medium Immobile scalable nodes
SOCP+SDP/SOCP [112] Centralized RSSI Both High Immobile scalable nodes
MSL [113] Centralized ToA Non Cooperative Medium Immobile scalable nodes
Distributed LS [114] Distributed Non Specified Range Measurement Cooperative Medium Immobile scalable nodes
C-ML [115] Distributed RSSI Non Cooperative High Work on static netowrk
fGn [116] Centralized Acoustic Non Cooperative High Mobile scalable nodes
MLE Distributed ECM [117] Distributed ToA Cooperative Medium Mobile scalable nodes
Mobility Aided SDP [118] Distributed ToA+RSSI Non Cooperative Medium Mobile scalable nodes
ADMM [119] Distributed Connectivity Cooperative Medium Mobile scalable nodes
SDP SLNN [120] Centralized ToA Non Cooperative High Not scalable
ILS BRL (Bayesian Ranging Method) [121] Centralized RSSI Non Cooperative Medium Immobile scalable nodes
Spectral Regression DESR [122] Centralized Connectivity Cooperative High Immobile scalable nodes
IML PL [123] Both ToA Non Cooperative Medium Immobile scalable nodes
RA HSL [124] Distributed RSSI+Connectivity Non Cooperative High Immobile scalable nodes
FP DWKNN [125] Distributed RSSI Non Cooperative High Mobile and scalable nodes
MCL Hit Ball [126] Distributed RSSI Non Cooperative Medium Mobile and scalable nodes
MASL MASL [127] Centralized ToA (One way ranging) 3D Mobile High No Anchors and Active Messaging
HL HL [128] [129] Centralized TDoA 2D Stationary High Stationary Anchor with Active Messaging
ALS ALS [130] [131] Centralized Range Free 2D Stationary High Anchors with variable power levers and
active messaging
3D-MALS 3D-MALS [132] Centralized ToA(One-way Ranging) 3D Mobile High Mobile anchors (Electro-mechanical mo-
tion) with active messaging
CL CL [133] Centralized ToA (One way ranging) 3D Mobile High No Anchors and Active Messaging
AAL AAL [134] Distributed ToA (two-way Ranging) 3D Hybrid Medium Propelled mobile anchor (AUV) with silent
messaging
LDB LDB [135] [136] Distributed Range-free 3D Hybrid Medium Propelled mobile anchor (AUV) with silent
messaging
DNRL DNRL [137] Distributed ToA (One-way ranging) 3D Mobile Medium Non-propelled mobile anchors with silent
messaging
MSL MSL [138] Distributed ToA (One-way ranging) 3D Mobile High Non-propelled mobile anchors and refer-
ence nodes with active messaging
LSHL LSHL [139] Distributed ToA (one-way Ranging) 3D Stationary High Surface buoys, underwater anchors and ref-
erence nodes with active messaging
DETL DETL [140] Distributed ToA (One-way ranging) 3D Mobile High Surface buoys with DETs, underwater an-
chors and refreence nodes with active mes-
saging
3DUL 3DUL [141] Distributed ToA (two-way ranging) 3D Hybrid High Three initial anchors and reference nodes
with active messaging
AFL AFL [142] [143] Distributed Not specified 3D Stationary High Anchor-free (One initial seed) with active
messaging
UPS UPS [144] [145] Distributed TDoA 3D stationary Low Four stationary anchors with silent messag-
ing
WPS WPS [146] Distributed TDoA 3D Stationary Low Four or five stationary anchor with silent
messaging
LSLS LSLS [147] Distributed TDoA 3D Stationary High Stationary Anchors with active messaging
USP USP [148] [149] [150] Distributed Not specified 3D Stationary High Stationary Anchors with active messaging
SLMP SLMP [151] Distributed ToA (One-way ranging) 3D Mobile High Surface buoys,Underwater anchor and re-
freence node with active messaging

the smallest route is calculated by connecting a couple of to prevent flip error in localization. The translation, rotation
nodes and a pairwise distance matrix is formed. Then using and reflection are done in the stitching transformation. The
this pairwise distance matrix 2D or 3D map is constructed minimum stitching errors are achieved by reflection and re-
which shows the position of all the nodes. Finally, the absolute flectionless transformation is called flip ambiguity. If both the
position of all the nodes are estimated using the known operation is observed it is called flip conflict. This method
nodes. This algorithm performs better than other algorithm, but reduces the flip error and conflicts. MDS and multilateration
performance goes down when more number of anchor nodes are used in path construction for localization. Finally global
presented in the network. coordinate is calculated using the reference coordinate.
MDS-MAP is further improved into distributed form called
MDS-MAP (D) [106] using positioning local maps. First, this B. Least Square (LS) Estimation
algorithm calculates local map using MDS-MAP then these
local maps are connected based on adjacent node common Path Loss Model [109] is used to localize the unknown
among these local maps. The local maps are aligned further, node. In this model ratio of transmitting power and receiving
and least square method is used to minimize the distance power is considered to calculate the distance. The model of
between the neighbor nodes. the path loss is given as follows.
A map stitching based localization method [107] is devel-
Pr (di ) = Pr (d0 ) − 10.η.log(di /d0 ) + Xσ (1)
oped. In this method a local map is formed for each component
of the network using multi-lateration and MDS. Then map Where Pr (di ) is the received signal power and Pr (d0 ) is the
stitching is done using a core node from a map either using received signal power at a reference point d0 , and η is the path
some critera or randomly. Here the core node is a speacial loss exponent.
node choosen for map-stitching. The path loss exponent is the only unknown for an envi-
Kwon et. al. [108] proposed a new method of map stitching ronment. A path loss exponent is found when the unknown

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 16

node’s position can be calculated using the linear least square function. Wavelet-based methods [155] calculates Hurst ex-
estimator. ponent of noisy signals. E. Dranka et. al. [116] proposed a
A localization approach [110] based on energy efficiency is source lcalization method. In this method each signal sample
proposed. In this method, average energy reception of anchor is represented by fractional Gaussian noise (fGn) which is
is considered. Different power allocation shows a substantial capable of modeling any degree of correlation utilizing Hurst
effect on localization accuracy. The author claim that an exponent and energy determination. The estimation process
optimal power allocation can reduce the localization error. A is called Hurst Maximum Likelihood Energy (HMLE) [116]
tetrahedron based semi-3D range-based localization method localization. This estimator uses a gain matrix, attenuation ma-
[111] computes the pairwise distance between nodes as well trix, acoustic energy source vector and error vector. From this
as anchors which form a tetrahedron. The volume of the estimator, the joint probability distribution is calculated, and
tetrahedron is calculated with Heron’s formula [152]. Then a log-likelihood function is formulated. This method performed
transformation of the coordinate is done with transformation better in noisy environment. Yin et. al. modified the ML
matrix in 3D points followed by linear least square estimation estimator using expectation-conditional maximization (ECM)
to find the unknown nodes. A comparison of different maxi- [117] criterion to estimate locations of the sensor nodes. This
mum likelihood based localization with non-convex estimation method models the measurement error as Gaussian mixture
is proposed [112]. The non-convex objective function has parameter. It is shown that the mobility information of the node
multiple local minima of the ML estimator which is critical has an impact on accuracy in localization schemes [118]. This
in problem formulation of the localization algorithm. The algorithms first use RSS or ToA range measurement model
convex relaxation is used to optimize the proposed estimator. and then uses ML estimator with different SDP (Semi-definite
Second-order cone programming is used for non-cooperative Programming) relaxation approach for non-convex objective
localization and Semidefinite programming is used for coop- [Link], SDP is used in noisy measurements. Si-
erative localization. Both the methods consider path-loss and monetto et. al. [119] proposed a maximum likelihood based
unknown transmitted power. They have shown that the method convex relaxation with certain novel characteristics.
have a better result than the existing algorithms. Multiple
source localization is proposed by Shen et. al. [113]. This E. Bayesian Ranging Model (BRL)
method uses time of arrival (ToA) as measurement technique Bayesian Ranging Model (BRL) [121] uses ranging mea-
to localize multiple nodes. ToA measurement is modelled surement using the Bayesian model. This method is a modi-
with mixed optimization problem. Further it is optimized fication of Empirical Bayes model [156]. BRL method uses a
with convex optimization and represented with a permutation minimum mean square error (MMSE) estimator for final es-
matrix. Then the permutation matrix is modified, which further timation of conditional mean which includes shrinkage factor
divide this problem to subproblems taking initial value from for correction of measurement. Iterative least square is used
the convex optimization result, by which improve the local- for position estimation and inclusion of shrinking factor makes
ization accuracy. A cooperative method is presented taking a robust method.
Fisher information matrix (FIM) [114] using the NLOS bias
model. After comparing the least square (LS), square range LS F. Monte Carlo Localization (MCL)
(SRLS) and square range weighted LS (SR-WLS) algorithms The node position is estimated using the current beacon
position error bound (PEB) method, it is found that the SRLS and historical beacon [126]. It consists of three phases; In first
and SR-WLS perform better. phase the intersection region of one-hop anchor node with his-
torical anchor node is estimated. The area covered by the target
C. Semi-definite Programming (SDP) is called constrained region. In second phase invalid samples
are filtered out by RSS based constrained region. There may be
A range based maximum-likelihood (ML) algorithm is
three types of region Current-Current RSS constrained region
proposed by Oguz-Ekim et. al. [120]. The location estima-
(CC-region), Current-Historical-RSS constrained region (CH-
tion is done with maximum likelihood method followed by
region) and historical-historical RSS constrained region (HH-
optimization using SDP (Semidefinite Programming).
region). Finally, the location estimation is carried out by taking
the centroid of all the valid samples.
D. Maximum Likelihood Estimation (MLE) Priyantha et. al. [47] introduced a distributed anchor-free
Maximum Likelihood method using cognitive sense esti- scheme considering fold freedom of nodes. After every global
mation (C-ML) with the received signal strength indicator translation, there is a proper orientation of the nodes called
(RSSI) is proposed [115]. This method finds the environment fold free graph. Anchor free localization (AFL) solve the false
whenever it is homogeneous or heterogeneous followed by a minima by construction and fold free contour of the nodes.
hypothesis testing of generalized likelihood ratio test (GLRT) Then a mass-spring optimization technique is applied. It is
[153]. more accurate and converging to local minima.
The efficiency of the maximum likelihood estimator is en-
hanced by acoustic energy based localization [116]. Acoustic G. Motion-Aware Self Localization (MASL) Technique
noise corrupt the source signal which is described by the The distance between the mobile nodes in the mobile
correlation degree of degraded signal or Hurst exponent [154]. network changes over time. A dynamic estimation for mea-
It is the level of the decay rate of auto-correlation coefficient surement is proposed in MASL scheme [127] by Mirza et. al.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 17

Estimating location of sensor nodes with propagation delay is K. Collaborative Localization (CL)
complicated but MASL provides an accurate estimation. In this
Mirza et. al. [133] proposed a prediction-based Collabo-
method the measured data are utilized in the post processing
rative Localization (CL) method for mobile wireless sensor
stage in a central station. The distance information is fed to an
network. This method is applicable for terrestrial and un-
iterative algorithm. Each step the algorithm improves position
derwater wireless sensor network. Underwater WSNs collect
information by dividing the area of operation with smaller
information from the depth sensor nodes and passes it to
grids with a higher probability of node presence. MASL is
the surface sensor nodes. Profilers and Followers are two
anchor-free and computationally simple. Online monitoring
types of nodes present in this protocol. Profilers are deeper
and synchronization are its principal drawbacks.
sensor nodes in underwater WSN which are localized with
ToA technique. The location of the Profiler sensor nodes
H. Hyperbola-based Localization (HL)
gives a prediction of the future location of the Follower
The traditional source localization problem is adapted for sensor nodes. The Follower sensor nodes are drifting with the
two-dimensional underwater acoustic WSNs [128] [129]. The water current similar to Profiler sensor nodes. Here the author
computation is based on hyperbola-based estimator [128] assumes the underwater wireless sensor network descend with
[129]. In HL systems a long range signal is sent to central node constant velocity. The problem of this type of network is
by the unknown nodes. Then the central node estimates the synchronization.
location of the unknown nodes. It consumes excessive energy
due to use of long range signals. The anchor nodes are placed
in the corners of the network so it cannot be extended into L. AUV-Aided Localization (AAL)
three-dimensional sensor network.
Erol [Link]. proposed AUV-Aided Localization (AAL) [134]
I. Area-based Localization Scheme (ALS) for three-dimensional hybrid underwater sensor network. AUV
is a vehicle which can find its location by dead-reckoning. This
Yao et. al. [130] proposed the area based localization is expensive navigation tools and need frequent calibration
schemes which is further modified for stable two-dimensional with the help of GPS. The AUV broadcasts its location infor-
underwater acoustic WSNs by Chandrasekhar et. al. [131]. mation while moving under water. The unknown sensor nodes
ALS is a coarse-grained localization method which yields an receive the AUV location. From three consecutive broadcast
estimated area where the node present in contrast to the exact messages, unknown sensor node estimates its location using
coordinates of the sensor node. Here anchor node partitioned lateration. It is a two-way ranging method so donot need
the area with non-overlapping areas by sending different power synchronization, but the energy consumption is more.
level. Anchor nodes listen to each others signal keeping the
list of anchor nodes with power level and send the data to sink
node. The sink associates the position of the anchor nodes so it M. Localization with Directional Beacons (LDB)
can estimate the position of the unknown nodes. It is applicable
when anchors can modify their power levels. It is range free, The LDB [135] [136] is a three-dimensional hybrid method
computationally simple and non-synchronized. This method proposed by Luo [Link]. similar to AAL. It has application in
does not apply for online localization estimation. In this an underwater sensor network. In this method the AUV is
method the unknown nodes send the location information to first localize itself with GPS then move to underwater for
sink node. Hence, it has high communication cost with high dead-reckoning. This approach is different from AAL in the
energy loss. sense AUV uses directional acoustic transceiver to broadcast
its coordinates and angle. From the angle information, the
J. Three Dimensional Multi-power Area Localization Scheme unknown node calculate location. It is a range free method
(3D-MALS) and more energy-efficient than AAL method.
Zhou et. al. [132] extended the ALS method further in three-
dimensional sensor network called 3D-MALS. 3D-MALS
N. Dive and Rise Localization (DNRL) Protocol
method is a combination of vertical mobility method [137] and
variable transmission level method [131] of the anchor nodes. DNRL is a mobile anchor based localization where anchor
In underwater wireless sensor network there is an elevetor like nodes are called DiveNRise (DNR) beacons. The anchors can
special transceiver called Detachable Elevator Transceivers move up and down in underwater sensor network with the
(DETs) [132]. The DETs broadcast their location information hydraulic principles and help the unknown nodes to localize
with varying power level and continuously descend in water. themselves [137]. Anchor nodes are enabled with GPS and find
The unknown node receives anchor position and their respec- their location when floating to the surface and then they move
tive lowest power. Further the unknown node send the position to a pre-calibrated depth broadcasting their location informa-
and power level of the anchor nodes to the sink node. Sink tion. The unknown nodes listen to the broadcast messages from
node calculates the area of the unknown nodes by the power the anchor and utilise ToA to calculate their positions using lat-
level and location of anchor nodes. Since all unknown node eration as shown in Fig. 15 (a). The advantage of this method
communicate to sink node the communication cost is very high is high energy efficiency and low communication overhead.
in this method. The drawback is use of ToA which need synchronization.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 18

(a) (b) (c)

Fig. 15. Application Scenarios (a)DNRL method (b)LSLS Method (c) TDoA measurement used in UPS Method

O. Multi-Stage Localization (MSL) be true until a further update from the anchor. Communication
Erol et. al. proposed a Multi-Stage Localization [138] cost and energy consumption depends upon the pattern of the
method. It is the extension of DNRL [137] method by pro- mobility.
moting the localized nodes as anchor nodes. First unlocalized
node estimates its location with distance measurement from Q. Detachable Elevator Transceiver Localization
three non-co planner anchor nodes. Then it is promoted to (DETL)Protocol
anchor nodes and help other unknown nodes to localize. The DETL [140] method proposed by Chen [Link] uses the same
drawback of this method is high communication overhead due architecture as LSHL [139] method with an extension of
to an iterative process. Hence this approach is less power eliminating anchor localization. Surface Buoys estimate their
efficient than DNRL. Propagation of error is added in this location from GPS. DET units moves up and down with
method. broadcasting coordinates at several depths alike to DNRL.
DET coordinates and distance helps localize anchor nodes.
Normal sensor nodes are localized like LSHL. DETL is a
P. Large-Scale Hierarchical Localization (LSHL) Protocol
solution for anchor localization for deep and narrow nodes
Large-Scale Hierarchical Localization (LSHL) Protocol is with the help of broadcasting short range acoustic links. Major
proposed by Zhou et. al. [139] for underwater wireless sensor drawback is the requirement of number of DETs.
network. This network contains three types of nodes: Surface
Buoys, Anchor Nodes and Sensor Nodes. Surface Buoys R. Three-Dimensional Underwater Localization (3DUL)
have GPS receiver and they localize themselves with line-of-
For hybrid underwater sensor network an iterative local-
sight satellite. Anchor nodes find localization through surface
ization method is presented by Isik et. al. [141]. The lo-
buoys and move underwater in the network. The common
calization process starts with the surface anchor nodes that
sensor nodes determine their location from the mobile anchor
receives location from GPS transceiver. The surface anchor
nodes with the help of ToA method. When the anchor is
nodes broadcasts its location information to the next sensor
two hops away the non-localized node find its position from
node in one hop distance. The sensor nodes receive position
the extended Euclidean distance estimation algorithm [139].
information of anchor nodes, estimate their position with two-
Key issues faced by this network is high energy consumption
way ToA estimator and further promoted to anchor nodes. This
and high communication overhead compared to DNRL and
method repeats until reach the last layer of a network similar to
MSL. The method presented in LSHL [139] with Surface
MSL and LSHL. The drawback of this approach is localization
Buoys, Anchor nodes and Sensor Nodes are further improved
delay of the lower sensor nodes and error propagation of
with SLMP [151] method proposed by Zhou et. al. Mobility
the estimated sensor node from top anchor nodes to bottom
patterns and sequential position information help anchor nodes
unknown nodes.
to estimate their current location. Anchor periodically check
the patterns since it can obsolete in due course of time. When
the pattern is not valid anchor updates the pattern. Surface S. Anchor-Free Localization (AFL)
Buoys localize themselves with GPS and transfer location Anchor-Free Localization [157] is proposed by Capkun
information to anchor nodes. After prediction, the anchor et. al. for WSNs. This method is extended to underwater
node use Surface Buoys coordinate and distance information sensor network by Othman et. al. [142] [143]. Here the
to estimate its location. If the discrepancy within predicted author proposes the concept of the seed node (S). Localization
and estimated location is smaller than the threshold then the starts with node discovery process initiated by seed node
mobility model of the anchor is assumed to be valid. When (S1). S1 broadcast its position and receives the distance
it is not valid, it will run mobility prediction algorithm to information from the neighbor nodes. Then second seed node,
determine new mobility pattern. The common nodes predict S2 is selected from the neighbor nodes. The same process
their location using the mobility pattern and it is expected to is repeated. The distance of the new seed is broadcasted by

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 19

other seeds when a new seed is selected. The intersection of it is situated. It maps two dimensions from three dimensions.
three seed forms an area, and the sensor present in the area There is a chance of overlapping location of anchor nodes. If
find its position with trilateration. This method forms a local overlapping anchors presents, then the unknown node picks
coordinate system in the process of localization. This approach another collection of anchor nodes. In each repetition of USP
faces high energy consumption, communication overhead and the nodes which are already localized show their location
propagation delay. messages and improve their location information accepting
neighbor location messages. If only two anchors available
T. Underwater Positioning Scheme (UPS) in neighbor it uses bilateration which may not find unique
location. In this case the sensor wait for another localized
Cheng et. al. [144] [145] extends the terrestrial positioning
anchor message. The drawback of this method is all nodes
scheme for localization to underwater scenarios. UPS is a
should be synchronized.
localization method based on TDoA scheme. This approach
uses four anchors which sequentially broadcast beacon signal.
One is picked as a master anchor and broadcast beacon and X. Other Estimators
other anchor receive the beacon with timing information. Then Gepshtein et. al. proposed a novel method of localization
a sensor node receives the information of the beacon packet considering dual embedding spectral regression (DESR) [122].
of the anchor and estimate their position using TDoA and Lasla et. al. [124] proposed a half symmetric lens (HLS)
trilateration method as shown in Fig 15 (c). This method is less method using RSSI information exchange. In this method a
energy consuming with low communication overhead. This pair of anchor forms a symmetric lens shape. Further it is di-
approach cannot localize if sensor node is outside the area vided into two halves and ensure if unknown node is inside any
enclosed by four anchor nodes [144]. of them using RSSI. Then the unknown node is estimated with
the help of grid scan. A distributed fingerprint based method
U. Wide Coverage Positioning (WPS) [125] follows accelerometer data for better accuracy. In this
The UPS method presented before [144] is unable to local- method area is partitioned and RSSI fingerprint is used to
ize sensor nodes found outside the area enclosed by the four form local map followed by estimation of nodes using nearest
sensor nodes. This problem is solved by Tan et. al. [146] with neighbour (NN) [158]. After this the local map is converted
using one more anchor node. WPS localize with four anchors into a global map. The estimation accuracy is improved by
called UPS (4) when localizable and when not localizable it considering mobility information of some selected nodes.
localize with five anchor nodes called UPS (5). UPS (4) and
UPS (5) combined work to localize the network and reduce Y. Summary and Insight
communication overhead for nodes those are localizable with
It is not always possible theoretically calculate the posi-
four nodes. The energy consumption in this method is same
tion geometrically. There are constraints in different network
as in original UPS [144] method. The drawback of this system
and infrastructure. Hence statistical approximation of location
is localization delay.
based on existing data are required. Huge volume of methods
in the literature are reviewed and classified. The accuracy of
V. Large-Scale Localization Scheme (LSLS) the method depends upon the range measurement. The better
The coverage of UPS [144] is increased by augmenting is the range measurement the estimation output is accurate.
iterative localization phase and complimentary phase is called So from reviewing the literature it can be concluded that the
LSLS [147] proposed by Wang [Link] and explained in Fig 15. accuracy depends upon range measurement techniques. An-
(b). Initially, the underwater nodes localize using UPS. The other aspect of this estimator is that the distributed algorithms
next phase is iterative localization phase. The localized nodes propagate localization error. Hence centralized algorithms are
in the first phase are promoted to anchor nodes and assist suitable for IoT infrastructure.
in localizing other nodes. The last stage is the complimentary
phase. In this step unallocated nodes start a localization request VIII. F INGERPRINTING
selecting a set of reference nodes and repeat UPS method.
The localization based on fingerprint approach takes a
The advantages of UPS are inherited by LSLS. Coverage is
signature and match it with a set of geo-tagged signatures to
increased by iterative phase and unique localization is raised
find the position of the device. In the device, the signature can
in complimentary phase [146]. LSLS method has greater
be recorded using various sensors. The user’s activity patterns
communication cost and energy loss than UPS since two more
can be detected using intelligent sensors. For visual signature,
stages are attached.
the camera is used to capture an image of the landmark
and resemble it against the geo-tagged image to identify the
W. Underwater Sensor Positioning (USP) location. Similarly, the audio signature can be detected by a
For three-dimensional underwater network USP [148] [149] microphone. The WiFi can be used to get the signature and
[150] is proposed by Teymorian et. al. This method assumes find the movement pattern of the devices. The main aim of
that all sensor node know their depth information by pressure fingerprint-based localization is to match the signature and
sensor. Considering the depth information an unknown node efficiently localize a device. This can be further classified as
find the available anchor nodes in the horizontal plan in which follows.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 20

TABLE VII
F INGERPRINTING A PPROACH OF L OCALIZATION
Fingerprint Model Algorithm Sensor Measurement Accuracy Power Consumption Remarks
PWF [159] RSSI peak in a temporal sequence Medium Medium Peak detection and location estimation is difficult if
the user movement is fast
Walkie-Markie [160] RSSI sequence High Low For accurate data collection the user must move in
one direction.
UnLoc [161] Wi-Fi Landmark Medium Medium This approach is applicable best with dense landmark
Signal Pattern
since the size of wifi landmarks cannot be too large
HALLWAY [162] Order of Wi-Fi RSSI value High Medium The granularity of using received signal order may
not be very high; the RSSI order remains similar at
times.
Wi-Fi signal Coverage Intersection and Division [163] [164] Simillar signal values from signal sector Low High If Wi-Fi AP installation are co-located, the over-
within sector intersection lapped region could be too large and may not provide
tight constraints.
Place Lab [165] WiFi Antenna GSM Antenna 15 to 20 m Medium
Place Engine [166] WiFi Antenna 5 to 100 m Medium Very less response time less than 1 sec.
[Link] [167] WiFi Antenna less than 5 m Medium
CAPS [168] GSM Antenna Not Available Low
CellSense [169] GSM Antenna 27 to 42 m Medium Highly energy saving method 6 times than Signal
Fingerprint Approach
VC [164] Wi-Fi direct Bluetooth Medium Low For absolute positioning distance measurement may
not be very accurate.
Collaborative Pattern PA [170] Sound Estimation High High Require accurate pairwise distance measure-
[Link] network graph may suffer from
measurement error and require synchronisation
Centaur [171] Sound Estimation High High Peer synchronization required design for static de-
vice.
ZCL [172] ZigBee Low Medium The user must be near to each other, cannot provide
randomly moving user.
Social-Loc [173] Wi-Fi direct Bluetooth Low Low Thresholds of detection may experience from the
noise and information using RSSI may not be ac-
curate.
[Link] [174] Camera Less than 20 m Medium Close to Realtime
Visual Analysis
[Link] [175] Camera Within 50 m Medium Percentage of accuracy is more than 96 percent.
Zee [176] Auto Correlation based Step counts Heading Medium High Employ map information to separate incorrect parti-
direction cle crowdsourced signal data may carry noise.
XINS [177] Peak detection based Step counts heading High High Utilize particle filter to fuse different signals. Work
direction the best when mixed signals are available for location
Motion Analysis fixing.
Graph Fusion [178] Peak detection based Step counts heading Medium Medium Analyze the indoor map model.
direction, online stride length estimation.
HMM Fusion [179] Step counts. heading direction Low Low Extensive training data set required for HMM train-
ing which is expensive.
Moloc [180] Auto correlation based Step counts crowd- High Low Need to accumulate user motion profile for localiza-
sourced motion [Link] direction tion.
MapCraft [181] Step counts,Heading direction. Medium Low Rely on extensive training sets; Complex training
GAC [182] Compass. Accelerometer Within intracity Low Less use of GPS
CompAcc [183] Compass Accelerometer Less than 11 m Low For fallback mechanism AGPS used.
APT [184] Gyroscope Accelerometer Less than 5m Low GPS is infrequently used
M. Anisetti [185] GSM Antenna, Camera 25 to 38m Low
WhellLoe [186] Accelerometer, Magnetometer, GSM An- Less than 40m Low Localization estimation delay is very less, less than
tenna 40 ms .
Dejavu [187] WiFi Antenna GSM Antenna, Accelerom- 8.4 to 16m Low Both energy efficient and highly accurate
Hybrid Model
ete, Magnetometer
EnLoc [188] WiFi Antenna, GSM Antenna Less than 12m Low
RAPS [189] Accelerometer, GSM Antenna Not Available Low Manually activated GPS
Location Study [190] Accelerometer Not Available Low
SmartLoc [191] Gyroscope, Accelerometer Less than 20m Low Special driving pattern used when GPS is weak.
A. Hallquist [192] Camera, Accelerometer, Compass Less than 10m Low An extension of [Link] [175]

map wrongly into a different position. Higher accuracy can

be obtained by considering both temporal and spatial pattern.
Temporal pattern is signal sequence pattern during the move-
ment in indoor area. Signal pattern of the route is used to
find the location as compared to single signal vector. Spatial
pattern is geographical distribution of signal or simply RSSI.
Peak based Wi-Fi Fingerprinting (PWF) [159] takes the
highest signal state for estimation. The highest in the signal
sequence informs the Wi-Fi access point is near to the sensor
node. So powerful RSS gives tremendous determination than
weak RSS. To find the highest a user have to collect a sequence
of RSS values then it finds the corresponding signal map and
highest RSS. The PWF method find accurate location when
access points are on the ceiling of indoor path.
Taking the entire distribution of data for analysis is better
than a particular peak value in case of a noisy environment.
Walkie-Markie [160] take the whole signal sequence for
classify the location. Wi-Fi RSSI patterns are first recorded in
Fig. 16. Fingerprint Based Localization Walkie-Markie. For a given area the distribution of RSSI forms
a pattern. By resembling users RSSI distribution throughout
walking, Walkie-Markie finds the position and outline the
A. Signal Pattern
knowledge of the target. In the indoor localization the Wi-Fi
Signal pattern based fingerprint localization method depends access points are placed at specific points. Excellent signal is
on the RSSI vectors. Measurement inaccuracy leads the target received at some points, which is called landmarks. With some

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 21

Fig. 17. Signal Pattern for Fingerprinting

Fig. 18. Localization approach with collaborative pairwise distance measure-
ment from WiFi signal RSSI.

simple investigation, these landmarks are saved in a database.

UnLoc [161] and MapCraft [181] use this measurement of pairwise distance the topology is constructed for different
access points to correctly localize. These two methods contain users. Then all the involved user’s position is estimated in
stationary sensor and motion information of signal patterns. the distance-based method. Virtual Compass [164] proposes
Simultaneous analysis of all the Access Points are considered a new algorithm to locate neighbor users. It is used on Wi-
to reduce measurement errors. HALLWAY [162] method clas- Fi Direct and Bluetooth. This method is just an extension
sify signature pattern from different access point based on the of Vivaldi algorithm [193] which find the coordinate so that
signal strength. Device dependent small signal fluctuation is all nodes provide pairwise distance. All the pairwise distance
reduced by RSS distribution which indicates the location of the was given, and Vivaldi algorithm just forms a network and
region. In indoor area when signal pattern is same this method construct the network topology. A new approach called sound-
is unable to localize. So the fingerprint is to be thoroughly based scheme [194] [195] to distance measurement. Peer-
preprocessed to obtain robust variation of location pattern. assisted (PA) localization [170] and Centaur [171] are the
The intersection of signal pattern from various access points two basic works using the sound assisted localization are
are analyzed for estimating the target [163]. The estimation proposed. The distance measurement is done with Beep [194]
error is reduced using the probabilistic approach of signal technique. Mutual distance and Wifi RSSIs are collected by
pattern [164]. The target is first outlined to the whole area of PA localization method. Initial location estimation is done with
a fingerprint, and then the system determines the connecting the traditional fingerprint of Wi-Fi signal of the user. The
point with maximum comparable signal patterns as position network topology is formed with this pairwise distance. The
estimation. Temporal and spatial patterns can be used in indoor location of all users are estimated by PA localization method.
localization with good AP coverage [161] [159] [164]. Collaborative Localization using the Bayesian network is
proposed by Centaur [171]. The rigid graph is not relied on
Centaur’s method as it is vulnerable to significant error in
B. Collaborative Pattern
measurement. Instead it used the maximum likelihood of all
Recently many works are carried out on collaborative local- associated devices. This method is only applicable for the
ization. It is evolved from the context of social interactions like static devices. Pairwise distance is not giving accurate result
people may gather in social scenarios [172]. Social interaction in the noisy environment as well as in the case of mobility, So
shows a location pattern. Currently various mobile devices and proximity [172] based method is proposed. Proximity is not
sensor can easily discover their neighbors with the help of a fundamental concept, so the probabilistic inference is used
different protocols like bluetooth, Wi-Fi Aware, Wi-Fi direct for better estimation.
[164], Near Field Communication (NFC) [173] and sound 2) Proximity-based scheme: There are many shortcoming
wave [170]. Depending upon the mutual distance measurement in the joint distance measurement of actual measurements. For
the collaborative localization is classified into two sections: dynamic measurement, proximity is often utilized. Proximity
distance-based and proximity-based algorithms. information is obtained from Wi-Fi AP list or MAC address
1) Distance Based Scheme: Advanced sensor technology or Bluetooth [196]. ZigBee-based collaborative localization
measures the pairwise distance between the user. From this (ZCL) [172] is a fine-grained system for absolute location

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 22

estimation. ZCL first uses its neighbour-detection sensor with and an electronic compass. A dead-reckoning technique is
its radio and then it computes a confidence score for the applied with the measured data and estimate the pattern of
target within the neighbor. Through the algorithm the system path signature which is further united to form the whole path.
corrects the neighboring estimations meant on the contrast This method is simple but consuming time in calibration of
among confidence counts. The candidates with low confi- sensors. Hence it is not applicable for large-scale applications.
dence filtered out. The drawback of this system are it can Zhu et. al. [184] proposed Accurate Pedestrian Tracking
undergo multipath impact, and the neighborhood knowledge (APT) method with high accuracy compared to GPS. APT
can be inaccurate. This problem can be rectified using multiple uses map matching with a strong DR algorithm with error-
samples and averaging filtering in ZCL. ZCL performs best tolerant method. This DR algorithm finds the acceleration
when the users are in a small region or moving uniformly. pattern of the mobile device. Experimentally it is shown that
Social-Loc [173] method propose a localization in random APT is better than GPS. Motion assisted localization is an
movement environment. It observe the signal patterns during indoor localization which uses hybrid techniques. Recently the
the encounters like the two user crossing each other. Then it development of motion sensors in the mobile devices are very
uses these encounters to correct the localization errors. First advanced, it further improved the indoor localization based on
traditional fingerprint based localization is initialised with each motion assistance. The recent advancement on motion assisted
user estimate multiple reference points with different prior localization is explained as follows. The essential part of accu-
probability. This collaborative localization method shows the rate motion-assisted localization is to monitor the movement
highest accuracy. pattern of the object. The major challenges are imperfect
calibration and noisy measurement of the sensors. Hence
C. Visual Analysis the motion detection is the main aim of the motion-assisted
localization. Magnetometer, gyroscopes and accelerometers
The main underlying concept behind visual fingerprint
are used for movement detections. These detection techniques
analysis based localization is to match the user generated
are further improved by signal filtering techniques. For real
query image with the image from the geo-tagged database.
deployment, the motion sensor requires specific calibration
The main drawback of this method is that when the size
which can be offline [178], [161] or online [197].
of fingerprint database increases the comparable algorithm
The fusion of the detected motion are essential for the
performance decreases in terms of accuracy, high latency,
localization accuracy. The correlation like temporal or special
and energy efficiency. Many solutions are proposed for this
between measured signal should be capture for fusion. The
problem. Extensive range image retrieval for user localization
model should be computationally efficient, which is important
is proposes by Zhang et. al. [175]. This approach divides the
for the motion-assisted localization [180] [181]. Finding the
large database of geo-tagged images to overlapping cells and
target location is the main focus for traditional fusion method
uses coarse position estimation to reduce the search space. The
[178]. The incorrect position is filtered out with fusion filters
query image only matched with images located in the area of
which filter signal fluctuation or fingerprint ambiguity. Some
interest which improve the search performance. Transmission
recent works on fusion methods are given below.
of data between client and server is the major challenge
1) Kalman filter: Kalman filter (KF) [198] is used in
for visual-fingerprinting based localization. The statistics of
discrete time system. Under linear Gaussian environment
the database is exploited and important characteristics which
Kalman filter obtain a better result than least square estimation
contribute maximum information on localization is identified.
[199]. Kalman filter solves fusion problem for linear, and for
It is used by Schroth et. al. [174] to avoid redundant image
nonlinear cases. It is extended to Extended Kalman Filter
and network delay.
(EKF) [200]. Moreover, Unscented KF and Fingerprint KF
(FKF) [201] utilizes the linear unbiased estimator by linking
D. Motion Analysis all the current and past signal ranges.
The motiion sensor in smartphone and embedded devices 2) Particle filter: Particle filter is more generalized version
can detect movement of a mobile user. This method is of tracking methods based on nonlinear motion representation.
highly noisy, and estimation of distance and position is often Initially all the measured point are collected then inconsistent
inaccurate. To overcome this difficulty several resolutions localtion points are filtered out comparing with walking dis-
have been recommended by supplementing GPS module and tance. The computational cost of particle filter is very high
synchronization by M. Youssef et. al. [182] of data from compared to Kalman filter. Graph-Fusion [178] introduces a
motion sensors. The main idea of reducing error is to turn on system which clarifies the predictions of particle filter. As
GPS module in some interval and synchronize GPS data with applying several particles significantly raises the complexity
the motion sensor data. The trade-off is in between energy and [179]. The main focus on utilizing extra efficient fusion
accuracy. The more accuracy can be obtained by synchronizing patterns to substitute particle filter. Zee [176] and XINS [177]
the GPS module. This process require more power consump- are two standard works applying particle filter. Zee employs
tion. Experiment shows that the proposed method [182] is able the map constraints to filter the particles and restrict the
to save energy of the system exponentially. An infrastructure- exploration region of target localization.
independent localization system [183] is built by CompAcc 3) Advanced and expert fusion models: Recently some
using DR algorithm and Assisted-GPS (AGPS). Speed and expert models are proposed [180], [178], [181] to locate the
orientation of the device are measured by an accelerometer target by: 1) Simplifying indoor map formations [178] with

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 23

minimum measured data 2) Analyze the localization estimate location where GPS is not available then it avoids GPS to turn
while performing adequate calculations using dynamic or clear on the sensors. Experiments showed that battery lifetime in
probabilistic models. These model uses Hidden Markov Model RAPS is 3.8 times longer than GPS. LocationStudy is another
(HMM) [179] or Conditional Random Field (CRF) [181]. The system which improves energy efficiency by accelerometer
HMM Fusion [179] aims to apply Hidden Markov Model based method [190]. This method uses mobility discovery
(HMM) to mix the sensor and interpret the fusion method. algorithm which detects mobility pattern and switch on or
HMM requires extensive training data set and the offline of the location sensors. Inertial sensors like accelerometer
training method is yet computationally complex [180]. Prior to or compass are used to measure speed and direction of the
HMM model, MoLoc [180] forms the probabilistic transition mobile user and determine the location of the user using
connecting diverse locations in the site based on the user’s dead-reckoning algorithm [184] [183]. However, those sensors
stepping length and direction. MapCraft [181], [202] proposed are extremely noisy and give errors in estimation. To reduce
a system meant on Conditional Random Field (CRF) more these noise and accumulated error, the SmartLoc [191] uses
complex than HMM. predictive regression model for estimation of trajectory. This
method utilizes a self-learning paradigm which calibrates the
localization event. It further improves the accuracy even with
E. Hybrid Model
the weak GPS signal.
The accuracy and energy efficiency of fingerprint-based
localization is improved by hybrid schemes. This method
always requires fingerprint generation process before localiza-
tion. In fingerprint generation process sensed data are collected
from multiple sensors to combine and construct the hybrid
fingerprints. Different types of hybrid systems for localization
are described as follows.
1) Use of Multiple Fingerprint Modes: The area where poor
signal and low geolocation are present that can be enhanced
by mixing location information acquired by RSSI with the
landmark matching [185]. This method is designed on time-
forwarding algorithm using a database association technique.
Further it is improved by the landmark recognition technique
proposed by J. Zhang et. al [175]. A sensor fusion approach
using the GPS, accelerometer and compass is proposed by Fig. 19. Comparative plot of different Fingerprinting approach.
Hallquist et. al. [192]. This method first finds the image that
resembles with the query image taken by the device from the
database [175]. Then the suitable matched image and its global F. Summary and Insight
coordinates are used to improve the query image. Further, the
Accuracy and complexity of algorithms are two basis of
position is computed with homography transformation matrix
selecting better algorithms. In IoT infrastructure it is complex
among the query and matching image. WeelLoc [186] is a
to achieve both attributes. The limited power capability and
continuous location service. It is an indirect method which
computing power of IoT device with heterogenous nature of
uses mobility trace through interpolation or extrapolation and
the device various challenges arises. A comparative study and
matches with the map via HMM and Viterbi decoding method.
analysis of fingerprinting algorithms on the basis of accuracy
Dejavu et. al. [187] provides both energy efficiency and accu-
and energy efficiency is plotted. It is found that motion
racy of localization of outdoor environment. The main idea of
information based fingerprinting approach is better in terms
this method is based on the fact that different places have a
of energy and accuracy and outperforms the others methods
different unique signature which is used to distinguish between
shown in Fig. 19.
them. This approach used dead-reckoning based techniques
and proved that it is better in localization and energy efficiency
than GPS. IX. S TOCHASTIC M ODEL /M ARKOV P ROCESS BASED
2) Augmenting Standard Positioning Techniques With Fin- L OCALIZATION
gerprints: Energy efficiency is a major concern for wireless A novel optimization method for decesion making in sce-
sensor network localization system. Many solutions are pro- narios where uncertainty play important role is modelled with
posed for energy efficient fingerprinting method. The Rate- Markov Decesion Process. The interaction of the sensor nodes
Adaptive Positioning System (RAPS) is a novel techniques with environment can be modelled with Markov Decesion
for energy and accuracy tradeoff [189]. RAPS maintain a Process. The sensor node localization is carried out with
localization history and velocity of the mobile nodes. When different extensions of Markov Models [217]: POMDP (Par-
the accuracy is below certain threshold it turn on the GPS mod- tially Observable Markov Decesion Process), MMDP (Multi-
ule. It uses duty-cycle based accelerometer and estimate the agent Markov Decesion process), DEC-POMDPS (Decentral-
movement. This also uses Bluetooth connection to diminish ized Partially Observable Markov Decesion Process) and SG
position ambiguity among neighboring nodes. If RAPS detects (Stochastic Game). This is shown in Fig. 20.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 24

TABLE VIII
S TOCHASTIC M ODEL BASED L OCALIZATION
Stochastic Models Algorithm Decision States Actions Rewards/Costs Application Context
[203] Centralized Node Activation (sleep,active) Select Active Nodes Energy,Detection Probability Cooperative Object Tracking
[204] Centralized Estimated Object’s Location, Sleep times Select Active Nodes Energy Consumption,Detection Probability Cooperative Object Tracking
[205] Centralized Targets locations, node activations Select active nodes Energy consumption, detection probability Multiple Target Tracking
POMDP [206] Centralized Estimated Object’s Location Select Active Nodes Energy Consumption,Detection Probability Cooperative Object Tracking
[207] Centralized Targets locations and velocity Select active nodes Nodes interception risk,detection accuracy Multiple Target Tracking
[208] Centralized Human body activities Select active nodes Energy Consumption,detection Probability Health and body Networks
[209] Centralized Human body activities Select active nodes Energy Consumption and detection proba- Health and body Networks
bility
[210] Centralized Estimated Adversary’s Region Select Active Nodes Energy Consumption,Detection Probability Cooperative Object Tracking
[211] Distributed Sensor’s State (sleep,fully or partially ac- Select Active Nodes Energy ,Detection Probability Clustered Tracking Systems
MDP
tive)
[212] Distributed Targets Location and velocities Send or discard a message Detection probability Prioritized data delivery
[213] Centralized Asset’s location Move (north,east,west,south) Transportation Delay Health and body Networks
HMDP [214] Distributed CH’s state (Sensing,listening or tracking) Select Active Nodes Sensing Rate ,Detection Probability Clustered Tracking Systems
(Lower tier) buffer occupancy,congestion Select active noes and detection threshold Network congestion ,detection probability Clustered Tracking Systems
CMDP [215] Centralized
matrix
(Upper tier)priority matrix,competing user Assign a spectrum Priority Clustered Tracking Systems
SG [216] Centralized Quantized Spectrum Bandwidth Select Active Nodes Energy ,Sucessfull transmission Cooperative Object Tracking

Atia et. al. [206] proposed the method considering the object
tracking in overlapped and non-overlapped region. When many
sensors detect the target, it is called overlapped and when a
single sensor detects the target it is called non-overlapped.
In both the cases energy and detection accuracy has a trade-
off. Fuemmeler et. al. [204] proposed a method in which they
are considering sensor location outside the coverage area. A
central controller with POMDP manage the sleep and active
time of the sensor node for the better target prediction.
The limited and shared channel spectrum and its effect on
object detection are analyzed by Huang et. al. [216]. In this
protocol spectrum bandwidth is managed by a coordinator,
which consider only active nodes. Spectrum management is
used for transmission decision and object detection. Further
the above methods [210] [206] [204] are extended with a
centralized offline learning method called Q-learning method.
Fig. 20. Markov Models: MDP-Markov Decision Process,POMDP-Partially
Observable Markov Decision Process, MMDP-Multiagent Markov Decision
Biometric sensor detects body activities. Biometric WSN
process, DECPOMDP-Decentralized Partially Observable Markov Decision collects biological data from different body activities with
Process, SG-Stochastic Game the help of biometric devices like electrocardiogram (ECG),
electroencephalograph (EEG), pulse oximeters etc. The real-
When there is hardware limitation for sensor node’s local- time tracking of the physical condition of the human body for
ization POMDP is applied. To apply POMDP the sensor nodes disease monitoring is developed by Au et. al. [208]. Network
in the network has to maintain the history of the observed lifetime is an issue for these systems since sensor selection
location. MMDP is applied when all the sensor nodes in the and action requires continuous scheduling. POMDP algorithms
network cooperatively optimize to find the location. DEC- are used for scheduling of real-time body area sensor of health
POMDP is similar to MMDP but sensor nodes localize a part monitoring system. Detection of the body activities and Energy
of the network in individual step. It is a decentralized version consumption of the sensor network are two challenges. The
of MMDP. SG is a non-cooperative method of interaction of model proposed by the authors considers human activities as
sensor nodes for localization. The sensor node’s objective in a state space and various commands for sensor actions like
this method is to maximize localization accuracy individually. active or sleep state constituted action space. Zois et. al. [209]
developed activity detection in wireless body area network.
The author considered a heterogeneous network and applied
A. Partially Observable Markov Decision Process (POMDP)
optimum node selection with sensor node sleep scheduling.
For the densely deployed WSNs the duty cycle management Different human activities like standing up, running, walking
policy is presented by Fuemmeler et. al. [203]. There are two etc. are considered and POMDP is formulated. Further it
modes: sleep and active modes. The detection is done by only is solved with dynamic programming to find the optimum
a few nodes so the other sensor nodes can be switched to selection. Switching between different activities is modeled
sleep mode. The inactive sensor node cannot be active by any with their probability as a transition matrix.
external signal. It will be activated after interior sleep timer.
A minimum number of active nodes are mere in the network
at any instant of time. This developed system is based on B. Markov Decision Process (MDP)
POMDP model [217]. Energy saving and location detection In monitoring applications, the localization of malicious
accuracy combined to form the cost function for optimize objects is presented by Zhan et. al. [210]. The cooperating
detection performance. node finds the adversary location by MDP. MDP’s states

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 25

represents the possible detection area. The intersection of these congestion. The sensor node detects delay and passes it to
detection areas represents the possible location of nodes. This the cluster head. The upper tier manages spectrum access by
method determines a set of nodes to be activated for the considering event arrival rate, link quality, services priority
detection of the malicious object. and collision probability.
In surveillance systems, the target tracking is analyzed by
Misra et. al. [211]. They proposed three modes of sensor node
namely sleep mode, partially active mode (not processing), E. Summary and Insight
fully active mode (processing). The difference in partially Existing literature shows that both finite time and infinite
active and fully active mode is fully active mode process time markov decision process can be solved in polynomial
the information. The wakeup request sent by cluster head to time. Extending markov decision process in different scenarios
switch nodes to fully active mode. The secondary users are leads different complexity. The complexity of markov decision
allowed to access the spectrum when it is not used by the process is P-Complete as well as of MMDP and POMDP.
primary users. Cluster head selects the minimum number of But DECMDP and DECPOMDP are further more complex.
active nodes using MDP to optimize the energy. Each cluster Hence in stochastic modelling the worst case complexity plays
head detects the objects based on the message received by a major impact in application scenarios. Majority of this
neighboring clusters. methods are applied in object tracking systems. Among all
Pietrabissa et. al. [213] modeled a hospital system and POMDP method have huge application in the literature like
formed a tracking system of various medical assets with cooperative object tracking, multiple object tracking, health
radio-frequency identification (RFID). Since it is an indoor and body monitoring due to its low complexity and centralized
application, so the RFID signals have affected by walls and applicability. Hence it can be suitable for IoT infrastructure.
other scenarios. WSN collects information from a deployed
region and sends it to the sink. The network life can be
X. M ACHINE L EARNING A PPROACH
improved by sending data those are important and dropping
the data which are unimportant for localization. So this can Localization using learning algorithms are not highlighted
be formulated as maximizing the probability of sensitive data in many studies. Learning based localization uses a distance
delivery and dropping of useless data. Pino-Povedano et. al. measurement matrix which is mostly RSSI and a learning
[212] proposed a method for dropping the data which are algorithm. A comparative analysis is presented in Table IX.
unimportant for target tracking. The author proposed a method The noise like multipath, non-line of sight and fading impairs
to select the data to be declined by the importance of the RSSI measurement. The noise is modeled with log-normal
message power consumption in the node and transmission link shadowing model (LNSM). The measurement with noise
cost. On successful transmission of the data from source to model is applied to learning-based optimization to predict the
the sink one unit of reward is given and feedback is sent from noise-free measurement of the range. Training and testing of
sink to source. Feedback, on the other hand, increase the data these learning methods reduce the channel impairments.
load of the network. Hence, this work is further extended for The traditional approach uses LNSM-based distance es-
suboptimal feedback based on two hop. timators. These methods are simple but prone to channel
impairments. The advantages of this approach is that it uses
C. Hierarchical Markov decision process (HMDP) very less anchor nodes for static anchor. In case of mobile
anchor nodes a single anchor node is sufficient to localize
Resource availability is the main consideration for object
the network. Various learning methods like PSO-ANN uses
detection and tracking in cluster-based systems. Both special
multiple anchors to improve the distance estimation. Normally
and temporal characteristic of an object is detected by a
sensor nodes are deployed in a region where the supply of
target tracking algorithm presented by Yeow et. al. [214].
energy is restricted. The node simply use a battery and when
The agents are divided into Lower level agent (LLA) and
the power is drained out the sensor lose its sensing and
Higher Level Agents (HLA). LLAs predict the target location
communication ability. So various strategies are developed
followed by activation of HLAs. HLAs are cluster heads with
for energy saving. One of the best methods is transmission
sleep, sensing, tracking and listening states. When HLAs are
power control (TPC). This method is based on the distance
active they track objects. If one HLA gets information about
between sensor node and anchor node. The transmission power
neighbouring HLA tracking object, it switches to listening
is reduced with respect to distance between anchor and sensor
state.
to save the energy.
The degree of interest in various scientific studies related to
D. Constrained Markov Decision Processes (CMDP) location-based services gained importance recently. A method
Detection in Cognitive Radio WSNs is proposed by Jamal using RSSI and linear least squares (LLS) was conducted by
et. al. [215] using two tier CMDP model. The system offered Yaming et. al. [109]. The LNSM path loss model generates
to be better in detection accuracy, managed the network the RSSI values. In this method the estimated distance error
congestion and spectrum allocation. It has two tier system was found to be 2.72 m. Another method based on LNSM and
architecture. The cluster head consists of upper tier which Polynomial Modelling with a mobile anchor node to generate
passes the messages to the base station. Secondary user consist RSSI values is conducted by Pratap et. al. [232]. In this
of the lower tier. This model balances accuracy and network approach, the positional error is found to be 2.2m.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 26

TABLE IX
M ACHINE L EARNING M ODEL BASED L OCALIZATION A PPROACH
Machine Learning Model Algorithm Architecture/Framework Estimation Metric Estimated Parameter Type of Study Environment Tested Area Localization Error
BR [218] ZigBee (MICAz) RSSI Location Simulation and Experimental Outdoor 10m x 10m 0.6938m
RP [219] IEEE 802.14.4 (Crossbow RSSI Location Experimental and Simulation Indoor Room Contains 24x10 position 0.3m
IRIS XM2110CA)
GR [220] Access Point RSSI Location Simulation and Experimental Indoor 20m x 20m 1.298m
MLP [221] ZigBee (CC2431) RSSI Distance Experimental and ANN Simulation Indoor 5m x 5m 2m
RBFN [222] ZigBee (FT-6250/FT- LQI Location Experimental Indoor 7.26m x 16.5 m 2.8m
6251)
[223] Zigbee (IRIS) RSSI Location Experimental Indoor 24m x 10 m Cumulative Error Calculation
FF ANN [224] N/A RSSI Location Simulation N/A 100m x100m 1.1862m
[225] Simulation in MATLAB RSSI and Hop Count Distance Simulation N/A 50m x 50m with Transmission 6-7m
Range (25m)
[226] MATLAB Simulation RSSI Location Simulation N/A 100mx100m and 300m x 300m 0.25-0.75m
LMFF
[34] ZigBee (XBee S2) RSSI Distance Experimental and offline ANN training Outdoor and Indoor 65m (Outdoor) 46m (Indoor) 0.022m (outdoor) 0.208m (indoor)
BR+ GD [227] ZigBee (Telosb) RSSI and LQI Location Experimental and offline training Indoor 12.19m x20.12 m and 6.28m 1.65 m
x11.28 m
BR+LM [228] N/A RSSI Location Simulation N/A 300m x 300m 0.49m
MLP+GR [229] N/A Wavelet-based features (WBF) Distance and Location Simulation Indoor 120 untrained data and 350 trained 2m (distance)
data
BR+RP+SCG [230] ZigBee (XBee S2) RSSI Location Experimental Indoor 8m x 6.4m 0.87m
LM+BR+RP+SCG+GD [231] ZigBee (XBee S2) RSSI Location Experimental and Offline ANN training Indoor 5m x 4m 0.3m

Estimation of the position of the sensor nodes in wireless to train the neural network. This method is found to be better
sensor network is calculated mostly with the use of lateration than weighted k-nearest neighbor (WkNN) method. When five
or angulation. The position of the sensor node is determined by anchor nodes are used WkNN method shows a better result
three anchor nodes in two-dimensional sensor network using than ANN. In case of three anchor nodes ANN shows a
trilateration [233]. In the case of mobile sensor and anchor better result than WkNN. Hence higher anchor shows a better
nodes the distance between sensor and anchor changes over result in WkNN. Further this gives high cost of the network
time so this method is not efficient. The hop count (HC) deployment.
algorithms can be used for estimation of the distance between Chuang [Link] [225] propose a new Neural Network based
unknown mobile nodes. However, the limitation is that the node localization method. In this approach, the RSSI values
network should be uniformly distributed. The accuracy of this are collected by LNSM method. The distance of HCs are the
uniform network is dependent upon the hop density of the shortest routes which are calculated by Dijkstra algorithm.
network. Irregularly deployed network localization with HC This method compared with PSO and ANN localization meth-
method causes inaccuracy with even high system cost. So ods and found out to be better than them.
for mobile sensor and anchor node based network, statistical
localization methods like Kalman filter is more accurate in A. Summary and Insight
prediction of position. Kalman filter applies a recursive strat-
In most of the learning based localization research, the
egy for range measurement and variance error reduction until
distance accuracy or localization accuracy is unsatisfactory.
final convergence. The application of this method is limited to
It opens a new research scope for energy saving localization
linear systems [225].
technology design which can reduce power consumption and
Numerous artificial intelligence techniques promised better increase network lifetime.
estimation of position recently like PSO and ANN. Sensor
node localization with two PSO algorithms is proposed by XI. A PPLICATIONS , C HALLENGES AND F UTURE
Kulkarni et. al. [234]. Mobile ad-hoc network (MANET) R ESEARCH IN I OT I NFRASTRUCTURE
utilize first PSO to localize the objective and second PSO to With incorporation of modern localization techniques in
converge the nodes around the target. This type of network- IoT infrastructure several important and popular application
centric collaborative localization shows the efficiency of PSO evolved to form a smart world. Increasing the computing
Algorithms. Artificial Neural Network (ANN) based estima- capabilities and incorporating artificial intelligence techniques
tors gaining importance in recent years. ANN shows faster to the device made them smart devices like smart phone, smart
convergence speed. The cost of calculation is very low com- TV, smart wearable devices etc. Movement of mechanical
pared to other estimators [224]. Irfan et. al. [227] propose devices with incorporation of artificial intelligence and modern
two neural network based algorithms, Gradient Descent and robotic techniques makes them smart with augmentation of
Bayesian Regularization for indoor application for mobile IoT. Incorporation of IoT makes the health services to be
network. ZigBee utilizes RSSI and link quality indicator (LQI) smart with better outcomes. The buildings becomes smart with
for neural network training and testing. The results show that addition of RFID tags and Bluetooth technologies to provide
1.65 m position accuracy can be obtained by this method. location aware smart spaces [250], [251]. The management
Link quality indicator (LQI) [222] has importance in indoor and monitoring of traffic is possible with addition of LoT
robotic sensor network localization and uses three ZigBee sen- techniques and computational intelligence. The handling of
sor node utilizing feed-forward sensor network. The average electrical usages in application as well as self adaptation of
positioning of sensor error on this robotic experiment is found electric grid made them smart grid is possible only through
to be 2.8 m. Positioning of mobile nodes in case of the indoor LoT. The evolution of the world of making it smart world
network by using Levenberg-Marquardt (LM) training based by incorporating intelligent computational capabilities to solve
feed-forward neural network algorithms estimator is proposed problems demands location information. In this section we will
by Gogolak et. al. [223]. In this method mobile node records analyze the technical aspects of different IoT applications and
RSSI value of five anchor nodes. Then these values are used their need for location information.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 27

TABLE X
A PPLICATIONS , C HALLENGES AND F UTURE R ESEARCH OF L OCALIZATION IN I OT I NFRASTRUCTURE

Applications in IoT Sensor Used For Localization Challenges Future Research

Smart Devices(Phone,TV,Wearable,Tablet) GPS, Wi-Fi signal, Cell Phone signals, Diversified accuracy and precision. Smart World [235]
Bluetooth, magnetometer, barometer, ac-
celerometer and cameras
Moving Smart Devices Camera, Speedometer, RF communication Simultaneous Localization and Mapping(SLAM) Autonomous Driving [236], [237], [238]
devices, RFID readers and optical measur-
ing meters.
Smart Health 3D X-ray [239], 3D ultra-sound imaging High accuracy and Precesion Health Monitoring and Microbots [240]
[240], 3D Magnetic Resonance Imaging [241]
(MRI) [241], and ComputerAided Tomog-
raphy(CAT) [242] , hybrid RF[243]
Smart Space Active and Passive RFID chips power efficiency Smart RFID Tagging , RF Energy Transport
[244]
Smart Transportation Networking of Transporting System Autonomous Vehicle [245]
Smart Infrastructure Electrical grid sensor, water quality mea- Smart Grid And Smart Infrastructure [246]
surement sensor etc.
Smart Monitoring Temperature sensors, Humidity sensor, fire- Signal Coverage and Target Positioning Cognitive Radio Sensor Network(CRSN)
alarm sensor, toxical gas and chemical sen- [247], [248], [249]
sors

A. Smart Device Localization and Future Research extract patterns of biological information for better prediction
Formation of smart world or environment is possible with in IoT healthcare system [255], [256], [257]. High localization
the smart sensor technology with location information. IoT accuracy is the requirement for health device and monitoring
devices like smart phone, tablets, smart watch, smart glasses, systems inside human body. The sensor used for localization
smart TV either directly or indirectly uses location information and mapping of things inside human body are Ultrasound
for better services. Diversified Accuracy and Precision are the imaging [240], Computer Aided Tomography(CAT) [242], 3D
major challenges as different IoT device use different accuracy Magnetic Resonance Imaging(MRI) [241], 3D X-ray [239] and
measure. For example in indoor localization the requirement hybrid RF[243].
is few meters whereas in outdoor it is tens of meter. There are Human behaviour recognition is gaining importance in IoT
two types of sensor like RF location sensor and Mechanical infrastructure [258]. The movement of human body parts
location sensor used. RF sensor are GPS receiver, WiFi sensor, changes in signal reflections which leads to variation of
Cellular Signal Sensor and Bluetooth sensor. The mechani- Channel State Information (CSI). Analyzing CSI data and
cal sensor used are magnetometer, barometer, accelerometer, comparing existing model human behaviour can be analyzed.
cameras and microphone. Heterogeneous environment and This is possible with feature extraction from CSI data. These
multiple sensors in devices demands location for different data can only be organized with geographical localization
applications [235]. based fingerprinting and help in model development.

D. Smart RFID and RF Energy Transport

B. Mobile Device(Robot) Localization and Autonomous Driv-
ing The smart world is possible with the smart RFID technology
with passive and active tags. RF energy transport have huge
The smart world is possible with development of Robotic application in IoT infrastructure. Trillions of RFID tags in
platforms of different size of ground robots or flying robots. indoor and outdoor scenarios have been implemented in IoT
The applications include smart manufacturing, security, aerial scenarios. RFID tags are expected to be deployed in almost all
photography, smart health technology, military mission etc. devices ranging from small indoor objects to smart building
Since every robot is moving so real time location information [259], [260] embedded with RFID readers. There are active
is required. There is a requirement of simultaneous localization and passive devices and which is further extended with RF
and mapping(SLAM) in either 2D or 3D. The robots benefits energy transport technology[244]. These RFID based local-
from SLAM methods for smart navigation in 2D [252], [253] ization system can augment the existing geolocation scenarios
and 3D [236], [237], [238] scenarios. Camera, Speedometer, with better accuracy.
RFID sensors, and Optical sensors are basically used but for
medical bots inside human body the camera sensor is preferred
E. Transportation System Localization and Autonomous Vehi-
[254]. The number of sensor and type of sensor used depend
cle
upon the application.
The key enabling technology of smart city and smart world
is the smart transportation system [245]. The fuel consumption
C. Health Monitoring and Microbots and cost of transport is hugely improved by using location
The localization in medical term is to find out the location information of ground and air transports and utilized it with a
of tumors, pain, bleeding, lesion etc inside human body. Other better network model. Finding the relative position and speed
localization applications like availability of doctors, location of information have huge impact on smart vehicular navigation
patients and their activity detection. Medical equipments and system. These navigation information is helpful for assisting
their interconnection over IoT to form a smart health system drivers by social event analysis [261] land use monitoring
Locating peoples and objects in geolocation scenarios to [262] smart public security [245] urban traffic planning [263]

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 28

etc. For all these application localization plays an important [2] M. P. DJurivsc, Z. Tafa, G. Dimic, and V. Milutinovic, “A survey
role. of military applications of wireless sensor networks,” in Embedded
Computing (MECO), 2012 Mediterranean Conference on, pp. 196–199,
IEEE, 2012.
[3] H. Lee, C. Wu, and H. Aghajan, “Vision-based user-centric light control
F. Smart Grid and Smart Infrastructure for smart environments,” Pervasive and Mobile Computing, vol. 7,
The IoT infrastructure implemented in national level re- no. 2, pp. 223–240, 2011.
[4] J. Bangali and A. Shaligram, “Energy efficient smart home based
cently and to form it there are different services like electric on wireless sensor network using labview,” American Journal of
smart grid, smart water system etc [264]. For smooth operation Engineering Research (AJER), vol. 2, no. 12, pp. 409–413, 2013.
and managing resources localization play important role like [5] L. Girod, V. Bychkovskiy, J. Elson, and D. Estrin, “Locating tiny
sensors in time and space: A case study,” in Computer Design:
detecting faults in smart grid [265], [246]. Sensor used for VLSI in Computers and Processors, 2002. Proceedings. 2002 IEEE
water quality monitoring or the load in grid the quantity of International Conference on, pp. 214–219, IEEE, 2002.
information is not usable without the sensor location informa- [6] A. Harter, A. Hopper, P. Steggles, A. Ward, and P. Webster, “The
tion. anatomy of a context-aware application,” Wireless Networks, vol. 8,
no. 2/3, pp. 187–197, 2002.
[7] L. Girod and D. Estrin, “Robust range estimation using acoustic
and multimodal sensing,” in Intelligent Robots and Systems, 2001.
G. Cognitive Radio Network and Smart Monitoring Proceedings. 2001 IEEE/RSJ International Conference on, vol. 3,
The smart monitoring of accumulate real time data such pp. 1312–1320, IEEE, 2001.
[8] D. Niculescu and B. Nath, “Ad hoc positioning system (aps) using
as temperature, humidity, fire alarm and toxic gas which aoa,” in INFOCOM 2003. Twenty-Second Annual Joint Conference
require event driven communication generally yields bursty of the IEEE Computer and Communications. IEEE Societies, vol. 3,
traffic [247]. Hence Cognitive Radio Sensor Network(CRSN) pp. 1734–1743, Ieee, 2003.
[9] J. A. Costa, N. Patwari, and A. O. Hero III, “Distributed weighted-
is a promising method for heterogeneous network like IoT. multidimensional scaling for node localization in sensor networks,”
CRSN have two types of sensor nodes: primary radio and ACM Transactions on Sensor Networks (TOSN), vol. 2, no. 1, pp. 39–
cognitive radio nodes. The primary radio have the right to 64, 2006.
[10] G. Mao, B. Fidan, and B. D. Anderson, “Wireless sensor network
use the frequency band. When primary radio is not using localization techniques,” Computer networks, vol. 51, no. 10, pp. 2529–
the band the cognitive radio can only use the frequency band 2553, 2007.
with their cognitive ability [248], [249]. The CRSN improve [11] G.-J. Yu and S.-C. Wang, “A hierarchical mds-based localization
algorithm for wireless sensor networks,” in Mobile and Wireless
the spectrum aware reliability, opportunistic energy efficient Communications Summit, 2007. 16th IST, pp. 1–5, IEEE, 2007.
transport, real time cognitive reliable transport. Hence smart [12] F. Santos and I. Tecnico, “Localization in wireless sensor networks,”
monitoring of real time data is possible with cognitive radio ACM Journal Name, vol. 5, pp. 1–19, 2008.
sensor network in future IoT applications. [13] I. Amundson and X. D. Koutsoukos, “A survey on localization for
mobile wireless sensor networks,” in Mobile Entity Localization and
There are various interactive discipline which combine Tracking in GPS-less Environnments, pp. 235–254, Springer, 2009.
cloud computing, fog computing, mobility and interaction [14] Y. Faheem, S. Boudjit, and K. Chen, “Data dissemination strategies
among them to build smart IoT system. The role of localization in mobile sink wireless sensor networks: A survey,” in Wireless Days
(WD), 2009 2nd IFIP, pp. 1–6, IEEE, 2009.
is important in building this smart IoT system. The application [15] G. Mao, Localization Algorithms and Strategies for Wireless Sensor
of localization in emerging smart city, health monitoring, RFID Networks: Monitoring and Surveillance Techniques for Target Track-
tags, smart transportation system and cognitive sensor network ing: Monitoring and Surveillance Techniques for Target Tracking. IGI
Global, 2009.
system is analyzed in context of challenges and future research [16] A. Pal, “Localization algorithms in wireless sensor networks: Current
direction. approaches and future challenges,” Network Protocols and Algorithms,
vol. 2, no. 1, pp. 45–73, 2010.
[17] A. Kulaib, R. Shubair, M. Al-Qutayri, and J. W. Ng, “An overview of
XII. C ONCLUSION localization techniques for wireless sensor networks,” in Innovations
in Information Technology (IIT), 2011 International Conference on,
In this review of localization algorithms, a comparative pp. 167–172, IEEE, 2011.
strategies are presented for broad application of localization [18] X. Li, Y. Zhang, K. Xu, G. Fan, and H. Wu, “Research of localization
technologies in different forms of network. The future of and tracking algorithms based on wireless sensor network,” Journal
of Information & Computational Science, vol. 8, no. 4, pp. 708–715,
localization and its implementation for IoT is manifested. Re- 2011.
cent localization technologies are analysed and a taxonomical [19] L. Cheng, C. Wu, Y. Zhang, H. Wu, M. Li, and C. Maple, “A survey
model is presented based on IoT Infrastructure. The emphasis of localization in wireless sensor network,” International Journal of
Distributed Sensor Networks, vol. 8, no. 12, p. 962523, 2012.
of classification is done based on the presence of offline [20] K. Stone and T. Camp, “A survey of distance-based wireless sensor
training state in the algorithms. Various Comparison tables network localization techniques,” International Journal of Pervasive
are given for different approaches. How everything can be Computing and Communications, vol. 8, no. 2, pp. 158–183, 2012.
localized and its futuristic application of IoT-based localization [21] N. A. Alrajeh, M. Bashir, and B. Shams, “Localization techniques in
wireless sensor networks,” International Journal of Distributed Sensor
is justified. Different noise distribution and their effect on Networks, 2013.
localization are exposed. Thus the use of localization has [22] Q. Dong, W. Dargie, et al., “A survey on mobility and mobility-aware
different approach for different scenarios. mac protocols in wireless sensor networks,” IEEE Communications
Surveys and Tutorials, vol. 15, no. 1, pp. 88–100, 2013.
[23] G. Han, H. Xu, T. Q. Duong, J. Jiang, and T. Hara, “Localization
R EFERENCES algorithms of wireless sensor networks: a survey,” Telecommunication
Systems, pp. 1–18, 2013.
[1] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, “Wire- [24] A. Mesmoudi, M. Feham, and N. Labraoui, “Wireless sensor net-
less sensor networks: a survey,” Computer networks, vol. 38, no. 4, works localization algorithms: a comprehensive survey,” arXiv preprint
pp. 393–422, 2002. arXiv:1312.4082, 2013.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 29

[25] R. Patel, R. Joshi, P. Gosai, and J. Patel, “A survey on localization for 1st international conference on Embedded networked sensor systems,
wireless sensor network,” Int. J. Comput. Sci. Trends Technol.(IJCST), pp. 340–341, ACM, 2003.
vol. 2, no. 1, pp. 79–83, 2014. [48] C. Savarese, J. M. Rabaey, and J. Beutel, “Location in distributed
[26] J. Kuriakose, S. Joshi, R. V. Raju, and A. Kilaru, “A review on local- ad-hoc wireless sensor networks,” in Acoustics, Speech, and Signal
ization in wireless sensor networks,” in Advances in signal processing Processing, 2001. Proceedings.(ICASSP’01). 2001 IEEE International
and intelligent recognition systems, pp. 599–610, Springer, 2014. Conference on, vol. 4, pp. 2037–2040, IEEE, 2001.
[27] C. Tunca, S. Isik, M. Y. Donmez, and C. Ersoy, “Distributed mobile [49] H. Chen, P. Huang, M. Martins, H. C. So, and K. Sezaki, “Novel
sink routing for wireless sensor networks: A survey,” IEEE communi- centroid localization algorithm for three-dimensional wireless sensor
cations surveys & tutorials, vol. 16, no. 2, pp. 877–897, 2014. networks,” in Wireless Communications, Networking and Mobile Com-
[28] Y. Gu, F. Ren, Y. Ji, and J. Li, “The evolution of sink mobility manage- puting, 2008. WiCOM’08. 4th International Conference on, pp. 1–4,
ment in wireless sensor networks: A survey,” IEEE Communications IEEE, 2008.
Surveys & Tutorials, vol. 18, no. 1, pp. 507–524, 2016. [50] J. Shu, L. Liu, Y. Chen, and H. Hu, “A novel three-dimensional
[29] H. P. Mistry and N. H. Mistry, “Rssi based localization scheme in localization algorithm in wireless sensor networks,” in Wireless Com-
wireless sensor networks: a survey,” in Advanced Computing & Com- munications, Networking and Mobile Computing, 2008. WiCOM’08.
munication Technologies (ACCT), 2015 Fifth International Conference 4th International Conference on, pp. 1–3, IEEE, 2008.
on, pp. 647–652, IEEE, 2015. [51] X. Lai, J. Wang, G. Zeng, M. Wu, J. She, and S. Yang, “Distributed
[30] M. Abu Alsheikh, D. T. Hoang, D. Niyato, H.-P. Tan, and S. Lin, positioning algorithm based on centroid of three-dimension graph for
“Markov Decision Processes With Applications in Wireless Sensor wireless sensor networks,” Journal of System Simulation, vol. 20,
Networks: A Survey,” IEEE Communications Surveys & Tutorials, no. 15, pp. 4104–4111, 2008.
vol. 17, pp. 1239–1267, jan 2015. [52] H. Yu, X. Chen, and J. Fan, “Gauss-newton method based on energy
[31] A. M. Hossain and W.-S. Soh, “A survey of calibration-free indoor target localization,” Computer Engineering and Applications, vol. 43,
positioning systems,” Computer Communications, vol. 66, pp. 1–13, no. 27, pp. 124–126, 2007.
jul 2015. [53] E. R. Marques, J. Pinto, S. Kragelund, P. S. Dias, L. Madureira,
[32] S. He and S.-H. G. Chan, “Wi-Fi Fingerprint-Based Indoor Positioning: A. Sousa, M. Correia, H. Ferreira, R. Gonçalves, R. Martins, et al.,
Recent Advances and Comparisons,” IEEE Communications Surveys & “Auv control and communication using underwater acoustic networks,”
Tutorials, vol. 18, pp. 466–490, jan 2016. in OCEANS 2007-Europe, pp. 1–6, IEEE, 2007.
[33] Q. D. Vo and P. De, “A Survey of Fingerprint-Based Outdoor Localiza- [54] W. Qin, Y. Feng, and X. Zhang, “Localization algorithm for wireless
tion,” IEEE Communications Surveys & Tutorials, vol. 18, pp. 491–506, sensor network based on characteristics of energy attenuation,” Journal
jan 2016. of Chinese Computer Systems, vol. 30, no. 6, pp. 1082–1088, 2009.
[34] S. K. Gharghan, R. Nordin, M. Ismail, and J. A. Ali, “Accurate Wireless [55] L. Lazos and R. Poovendran, “Hirloc: high-resolution robust localiza-
Sensor Localization Technique Based on Hybrid PSO-ANN Algorithm tion for wireless sensor networks,” IEEE Journal on selected areas in
for Indoor and Outdoor Track Cycling,” IEEE Sensors Journal, vol. 16, communications, vol. 24, no. 2, pp. 233–246, 2006.
pp. 529–541, jan 2016. [56] L. Lv, Y. Cao, X. Gao, and H. Luo, “Three dimensional localization
schemes based on sphere intersections in wireless sensor network (pp.
[35] T. J. Chowdhury, C. Elkin, V. Devabhaktuni, D. B. Rawat, and
48–51),” Beijing: Beijing Posts and Telecommunications University,
J. Oluoch, “Advances on localization techniques for wireless sensor
2006.
networks: A survey,” Computer Networks, vol. 110, pp. 284–305, dec
[57] W. Jichun, H. Liusheng, X. Hongli, X. Ben, and L. Shanliang, “A novel
2016.
range free localization scheme based on voronoi diagrams in wireless
[36] G. Han, J. Jiang, C. Zhang, T. Q. Duong, M. Guizani, and G. K.
sensor networks [j],” Journal of Computer Research and Development,
Karagiannidis, “A Survey on Mobile Anchor Node Assisted Localiza-
vol. 1, p. 014, 2008.
tion in Wireless Sensor Networks,” IEEE Communications Surveys and
[58] S.-j. Zheng, L. Kai, and Z. Zheng, “Three dimensional localization
Tutorials, vol. 18, pp. 2220–2243, jan 2016.
algorithm based on nectar source localization model in wireless sensor
[37] K. Langendoen and N. Reijers, “Distributed localization in wireless
network,” Application Research of Computers, vol. 25, no. 8, pp. 2512–
sensor networks: a quantitative comparison,” Computer Networks,
2513, 2008.
vol. 43, no. 4, pp. 499–518, 2003.
[59] Y. Wang, L.-s. Huang, and M.-j. Xiao, “Localization algorithm for
[38] T. Dinh, Y. Kim, and H. Lee, “A location-based interactive model for wireless sensor network based on rssi verification,” Journal of Chinese
internet of things and cloud (iot-cloud),” in Ubiquitous and Future Computer Systems, vol. 30, no. 1, pp. 59–62, 2009.
Networks (ICUFN), 2016 Eighth International Conference on, pp. 444– [60] G. Han, D. Choi, and W. Lim, “Reference node placement and selection
447, IEEE, 2016. algorithm based on trilateration for indoor sensor networks,” Wireless
[39] S. Misra, S. Chatterjee, and M. S. Obaidat, “On theoretical modeling Communications and Mobile Computing, vol. 9, no. 8, pp. 1017–1027,
of sensor cloud: A paradigm shift from wireless sensor network,” IEEE 2009.
Systems Journal, vol. 11, no. 2, pp. 1084–1093, 2017. [61] M. Liu, T.-t. Wang, and Z.-b. Zhou, “Self-localization algorithm for
[40] D. Niculescu and B. Nath, “Dv based positioning in ad hoc networks,” sensor networks of sparse anchors,” Computer Engineering, vol. 35,
Telecommunication Systems, vol. 22, no. 1, pp. 267–280, 2003. no. 22, pp. 119–121, 2009.
[41] K. Liu, S. Wang, F. Zhang, F. Hu, and C. Xu, “Efficient localized [62] D. Li and Y. H. Hu, “Energy-based collaborative source localization
localization algorithm for wireless sensor networks,” in Computer using acoustic microsensor array,” EURASIP Journal on Advances in
and Information Technology, 2005. CIT 2005. The Fifth International Signal Processing, vol. 2003, no. 4, p. 985029, 2003.
Conference on, pp. 517–523, IEEE, 2005. [63] R. C. Luo, O. Chen, and S. H. Pan, “Mobile user localization in wireless
[42] B. Wang, G. Wu, S. Wang, and L. T. Yang, “Localization based on sensor network using grey prediction method,” in Industrial Electronics
adaptive regulated neighborhood distance for wireless sensor networks Society, 2005. IECON 2005. 31st Annual Conference of IEEE, pp. 6–
with a general radio propagation model,” IEEE Sensors Journal, pp, IEEE, 2005.
vol. 14, no. 11, pp. 3754–3762, 2014. [64] U. A. Khan, S. Kar, and J. M. Moura, “Distributed sensor localization
[43] Y. Diao, Z. Lin, and M. Fu, “A barycentric coordinate based distributed in random environments using minimal number of anchor nodes,” IEEE
localization algorithm for sensor networks,” IEEE Transactions on Transactions on Signal Processing, vol. 57, no. 5, pp. 2000–2016, 2009.
Signal Processing, vol. 62, no. 18, pp. 4760–4771, 2014. [65] T. Augustin, “Generalized basic probability assignments,” International
[44] C. Elkin, R. Kumarasiri, D. B. Rawat, and V. Devabhaktuni, “Lo- Journal of General Systems, vol. 34, no. 4, pp. 451–463, 2005.
calization in wireless sensor networks: A dempster-shafer evidence [66] F. Caballero, L. Merino, I. Maza, and A. Ollero, “A particle filtering
theoretical approach,” Ad Hoc Networks, vol. 54, pp. 30–41, 2017. method for wireless sensor network localization with an aerial robot
[45] D. Moore, J. Leonard, D. Rus, and S. Teller, “Robust distributed beacon,” in Robotics and Automation, 2008. ICRA 2008. IEEE Inter-
network localization with noisy range measurements,” in Proceedings national Conference on, pp. 596–601, IEEE, 2008.
of the 2nd international conference on Embedded networked sensor [67] E. Kim, S. Lee, C. Kim, and K. Kim, “Mobile beacon-based 3d-
systems, pp. 50–61, ACM, 2004. localization with multidimensional scaling in large sensor networks,”
[46] M. İ. Akbaş, M. Erol-Kantarcı, and D. Turgut, “Localization for IEEE Communications Letters, vol. 14, no. 7, pp. 647–649, 2010.
wireless sensor and actor networks with meandering mobility,” IEEE [68] A. Galstyan, B. Krishnamachari, K. Lerman, and S. Pattem, “Dis-
Transactions on Computers, vol. 64, no. 4, pp. 1015–1028, 2015. tributed online localization in sensor networks using a moving target,”
[47] N. B. Priyantha, H. Balakrishnan, E. Demaine, and S. Teller, “Anchor- in Information Processing in Sensor Networks, 2004. IPSN 2004. Third
free distributed localization in sensor networks,” in Proceedings of the International Symposium on, pp. 61–70, IEEE, 2004.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 30

[69] C.-H. Ou, K.-F. Ssu, and H. C. Jiau, “Range-free localization with [91] H. Cui, Y. Wang, and J. Lv, “Path planning of mobile anchor in
aerial anchors in wireless sensor networks,” International Journal of three-dimensional wireless sensor networks for localization,” Journal of
Distributed Sensor Networks, vol. 2, no. 1, pp. 1–21, 2006. Information and Computational Science, vol. 9, no. 8, pp. 2203–2210,
[70] C.-H. Ou and K.-F. Ssu, “Sensor position determination with flying 2012.
anchors in three-dimensional wireless sensor networks,” IEEE Trans- [92] L. Liu, H. Zhang, X. Geng, and X. Shu, “Hexahedral localization
actions on Mobile Computing, vol. 7, no. 9, pp. 1084–1097, 2008. (hl): A three-dimensional hexahedron localization based on mobile
[71] C. Zhao, Y. Xu, H. Huang, and B. Cui, “Localization with a mobile beacons,” The Scientific World Journal, vol. 2013, 2013.
beacon based on compressive sensing in wireless sensor networks,” [93] K. Kim and W. Lee, “Mbal: A mobile beacon-assisted localization
International Journal of Distributed Sensor Networks, 2013. scheme for wireless sensor networks,” in Computer Communications
[72] E. Guerrero, H. Xiong, Q. Gao, G. Cova, R. Ricardo, and J. Estévez, and Networks, 2007. ICCCN 2007. Proceedings of 16th International
“Adal: a distributed range-free localization algorithm based on a mobile Conference on, pp. 57–62, IEEE, 2007.
beacon for wireless sensor networks,” in Ultra Modern Telecommuni- [94] H. Li, J. Wang, X. Li, and H. Ma, “Real-time path planning of mobile
cations & Workshops, 2009. ICUMT’09. International Conference on, anchor node in localization for wireless sensor networks,” in Informa-
pp. 1–7, IEEE, 2009. tion and Automation, 2008. ICIA 2008. International Conference on,
[73] K. Liu and J. Xiong, “A fine-grained localization scheme using a mobile pp. 384–389, IEEE, 2008.
beacon node for wireless sensor networks,” Journal of Information [95] F. Zhao, H.-y. Luo, and L. Quan, “A mobile beacon-assisted localization
Processing Systems, vol. 6, no. 2, pp. 147–162, 2010. algorithm based on network-density clustering for wireless sensor
[74] B. Xiao, H. Chen, and S. Zhou, “Distributed localization using a networks,” in Mobile Ad-hoc and Sensor Networks, 2009. MSN’09.
moving beacon in wireless sensor networks,” IEEE Transactions on 5th International Conference on, pp. 304–310, IEEE, 2009.
Parallel and Distributed Systems, vol. 19, no. 5, pp. 587–600, 2008.
[96] H. Wang, W. Qi, K. Wang, P. Liu, L. Wei, and Y. Zhu, “Mobile-assisted
[75] L. Xiao, W. Jin-kuan, and W. Yun, “A novel localization algorithm
localization by stitching in wireless sensor networks,” in Communica-
based on received signal strength for mobile wireless sensor networks,”
tions (ICC), 2011 IEEE International Conference on, pp. 1–5, IEEE,
in Microwave and Millimeter Wave Technology, 2008. ICMMT 2008.
2011.
International Conference on, vol. 1, pp. 92–95, IEEE, 2008.
[76] Z. Zhang, Z. Sun, G. Wang, R. Yu, and S. MEI, “Localization in [97] S. Li, D. Lowe, X. Kong, and R. Braun, “Wireless sensor network
wireless sensor networks with mobile anchor nodes [j],” Journal of localization algorithm using dynamic path of mobile beacon,” in Com-
Tsinghua University (Science and Technology), vol. 4, p. 021, 2007. munications (APCC), 2011 17th Asia-Pacific Conference on, pp. 344–
[77] B. Neuwinger, U. Witkowski, and U. Rückert, “Ad-hoc communication 349, IEEE, 2011.
and localization system for mobile robots,” in Advances in Robotics, [98] X. Li, N. Mitton, I. Simplot-Ryl, and D. Simplot-Ryl, “Dynamic beacon
pp. 220–229, Springer Nature, 2009. mobility scheduling for sensor localization,” IEEE Transactions on
[78] A. Baggio and K. Langendoen, “Monte carlo localization for mobile Parallel and Distributed Systems, vol. 23, no. 8, pp. 1439–1452, 2012.
wireless sensor networks,” Ad hoc networks, vol. 6, no. 5, pp. 718–733, [99] C.-T. Chang, C.-Y. Chang, and C.-Y. Lin, “Anchor-guiding mechanism
2008. for beacon-assisted localization in wireless sensor networks,” IEEE
[79] R. Zhang, L. Zhang, and Y. Feng, “Very low energy consumption Sensors Journal, vol. 12, no. 5, pp. 1098–1111, 2012.
wireless sensor localization for danger environments with single mobile [100] J. Lv, Y. Wang, N. Wei, and H. Cui, “Dynamic path planning method
anchor node,” Wireless personal communications, vol. 47, no. 4, for anchor node in three-dimensional wireless sensor networks,” in
pp. 497–521, 2008. Instrumentation and Measurement, Sensor Network and Automation
[80] D. Koutsonikolas, S. M. Das, and Y. C. Hu, “Path planning of mobile (IMSNA), 2013 2nd International Symposium on, pp. 900–904, IEEE,
landmarks for localization in wireless sensor networks,” Computer 2013.
Communications, vol. 30, no. 13, pp. 2577–2592, 2007. [101] Y. Ding, C. Wang, and L. Xiao, “Using mobile beacons to locate sen-
[81] R. Huang and G. V. Zaruba, “Static path planning for mobile beacons sors in obstructed environments,” Journal of Parallel and Distributed
to localize sensor networks,” in Pervasive Computing and Communi- Computing, vol. 70, no. 6, pp. 644–656, 2010.
cations Workshops, 2007. PerCom Workshops’ 07. Fifth Annual IEEE [102] G. Han, J. Jiang, C. Zhang, T. Q. Duong, M. Guizani, and G. K.
International Conference on, pp. 323–330, IEEE, 2007. Karagiannidis, “A survey on mobile anchor node assisted localization in
[82] A. Tuncer and M. Yildirim, “Dynamic path planning of mobile robots wireless sensor networks,” IEEE Communications Surveys & Tutorials,
with improved genetic algorithm,” Computers & Electrical Engineer- vol. 18, no. 3, pp. 2220–2243, 2016.
ing, vol. 38, no. 6, pp. 1564–1572, 2012. [103] A. T. Rashid, A. A. Ali, M. Frasca, and L. Fortuna, “Path planning with
[83] Z. Hu, D. Gu, Z. Song, and H. Li, “Localization in wireless sensor obstacle avoidance based on visibility binary tree algorithm,” Robotics
networks using a mobile anchor node,” in Advanced Intelligent Mecha- and Autonomous Systems, vol. 61, no. 12, pp. 1440–1449, 2013.
tronics, 2008. AIM 2008. IEEE/ASME International Conference on, [104] X. Ji and H. Zha, “Sensor positioning in wireless ad-hoc sensor
pp. 602–607, IEEE, 2008. networks using multidimensional scaling,” in INFOCOM 2004. Twenty-
[84] B. Zhang, Z. Zhang, et al., “Collaborative localization algorithm for third AnnualJoint Conference of the IEEE Computer and Communica-
wireless sensor networks using mobile anchors,” in Computational tions Societies, vol. 4, pp. 2652–2661, IEEE, 2004.
Intelligence and Industrial Applications, 2009. PACIIA 2009. Asia- [105] Y. Shang, W. Ruml, Y. Zhang, and M. P. Fromherz, “Localization
Pacific Conference on, vol. 1, pp. 309–312, IEEE, 2009. from mere connectivity,” in Proceedings of the 4th ACM international
[85] Q. Fu, W. Chen, K. Liu, W. Chen, et al., “Study on mobile beacon symposium on Mobile ad hoc networking & computing, pp. 201–212,
trajectory for node localization in wireless sensor networks,” in Infor- ACM, 2003.
mation and Automation (ICIA), 2010 IEEE International Conference
[106] Y. Shang, W. Ruml, and M. P. Fromherz, “Positioning using local
on, pp. 1577–1581, IEEE, 2010.
maps,” Ad Hoc Networks, vol. 4, no. 2, pp. 240–253, 2006.
[86] Z. Guo, Y. Guo, F. Hong, Z. Jin, Y. He, Y. Feng, and Y. Liu,
“Perpendicular intersection: Locating wireless sensors with mobile [107] O.-H. Kwon and H.-J. Song, “Localization through map stitching
beacon,” IEEE Transactions on Vehicular Technology, vol. 59, no. 7, in wireless sensor networks,” IEEE Transactions on Parallel and
pp. 3501–3509, 2010. distributed systems, vol. 19, no. 1, pp. 93–105, 2008.
[87] G. Han, H. Xu, J. Jiang, L. Shu, T. Hara, and S. Nishio, “Path planning [108] O.-H. Kwon, H.-J. Song, and S. Park, “Anchor-free localization through
using a mobile anchor node based on trilateration in wireless sensor flip-error-resistant map stitching in wireless sensor network,” IEEE
networks,” Wireless Communications and Mobile Computing, vol. 13, Transactions on Parallel and distributed systems, vol. 21, no. 11,
no. 14, pp. 1324–1336, 2013. pp. 1644–1657, 2010.
[88] H. Chen, B. Liu, P. Huang, J. Liang, and Y. Gu, “Mobility-assisted node [109] Y. Xu, J. Zhou, and P. Zhang, “Rss-based source localization when
localization based on toa measurements without time synchronization path-loss model parameters are unknown,” IEEE Communications
in wireless sensor networks,” Mobile Networks and Applications, Letters, vol. 18, no. 6, pp. 1055–1058, 2014.
vol. 17, no. 1, pp. 90–99, 2012. [110] F. Yaghoubi, A.-A. Abbasfar, and B. Maham, “Energy-efficient rssi-
[89] C.-H. Ou and W.-L. He, “Path planning algorithm for mobile anchor- based localization for wireless sensor networks,” IEEE Communica-
based localization in wireless sensor networks,” IEEE Sensors Journal, tions Letters, vol. 18, no. 6, pp. 973–976, 2014.
vol. 13, no. 2, pp. 466–475, 2013. [111] J.-K. Lee, Y. Kim, J.-H. Lee, and S.-C. Kim, “An efficient three-
[90] H. Cui and Y. Wang, “Four-mobile-beacon assisted localization in dimensional localization scheme using trilateration in wireless sensor
three-dimensional wireless sensor networks,” Computers & Electrical networks,” IEEE Communications Letters, vol. 18, no. 9, pp. 1591–
Engineering, vol. 38, no. 3, pp. 652–661, 2012. 1594, 2014.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 31

[112] S. Tomic, M. Beko, and R. Dinis, “Rss-based localization in wireless Applications, 2007. WASA 2007. International Conference on, pp. 44–
sensor networks using convex relaxation: Noncooperative and cooper- 54, IEEE, 2007.
ative schemes,” IEEE Transactions on Vehicular Technology, vol. 64, [135] H. Luo, Z. Guo, W. Dong, F. Hong, and Y. Zhao, “Ldb: Localization
no. 5, pp. 2037–2050, 2015. with directional beacons for sparse 3d underwater acoustic sensor
[113] H. Shen, Z. Ding, S. Dasgupta, and C. Zhao, “Multiple source networks.,” JNW, vol. 5, no. 1, pp. 28–38, 2010.
localization in wireless sensor networks based on time of arrival mea- [136] H. Luo, Y. Zhao, Z. Guo, S. Liu, P. Chen, and L. M. Ni, “Udb: Using
surement.,” IEEE Trans. Signal Processing, vol. 62, no. 8, pp. 1938– directional beacons for localization in underwater sensor networks,”
1949, 2014. in Parallel and Distributed Systems, 2008. ICPADS’08. 14th IEEE
[114] T. Van Nguyen, Y. Jeong, H. Shin, and M. Z. Win, “Least square International Conference on, pp. 551–558, IEEE, 2008.
cooperative localization,” IEEE Transactions on Vehicular Technology, [137] M. Erol, L. F. Vieira, and M. Gerla, “Localization with dive’n’rise
vol. 64, no. 4, pp. 1318–1330, 2015. (dnr) beacons for underwater acoustic sensor networks,” in Proceedings
[115] F. Bandiera, A. Coluccia, and G. Ricci, “A cognitive algorithm for of the second workshop on Underwater networks, pp. 97–100, ACM,
received signal strength based localization.,” IEEE Trans. Signal Pro- 2007.
cessing, vol. 63, no. 7, pp. 1726–1736, 2015. [138] M. Erol, L. F. Vieira, A. Caruso, F. Paparella, M. Gerla, and S. Oktug,
[116] E. Dranka and R. Coelho, “Robust maximum likelihood acoustic “Multi stage underwater sensor localization using mobile beacons,”
energy based source localization in correlated noisy sensing environ- in Sensor Technologies and Applications, 2008. SENSORCOMM’08.
ments,” IEEE Journal of Selected Topics in Signal Processing, vol. 9, Second International Conference on, pp. 710–714, IEEE, 2008.
no. 2, pp. 259–267, 2015.
[139] Z. Zhou, J.-H. Cui, and S. Zhou, “Localization for large-scale under-
[117] F. Gustafsson and A. M. Zoubir, “Cooperative localization in wsns
water sensor networks,” in International Conference on Research in
using gaussian mixture modeling: Distributed ecm algorithms,” IEEE
Networking, pp. 108–119, Springer, 2007.
Transactions on Signal Processing, vol. 63, no. 6, pp. 1448–1463, 2014.
[140] K. Chen, Y. Zhou, and J. He, “A localization scheme for underwater
[118] S. Salari, S. Shahbazpanahi, and K. Ozdemir, “Mobility-aided wire-
wireless sensor networks,” International Journal of Advanced Science
less sensor network localization via semidefinite programming,” IEEE
and Technology, vol. 4, 2009.
Transactions on Wireless Communications, vol. 12, no. 12, pp. 5966–
5978, 2013. [141] M. Isik and O. Akan, “A three dimensional localization algorithm for
[119] A. Simonetto and G. Leus, “Distributed maximum likelihood sensor underwater acoustic sensor networks,” IEEE Transactions on Wireless
network localization.,” IEEE Trans. Signal Processing, vol. 62, no. 6, Communications, vol. 8, pp. 4457–4463, sep 2009.
pp. 1424–1437, 2014. [142] A.-K. Othman, A. Adams, and C. C. Tsimenidis, “Node discovery
[120] P. Oguz-Ekim, J. P. Gomes, J. Xavier, M. Stosic, and P. Oliveira, “An protocol and localization for distributed underwater acoustic networks,”
angular approach for range-based approximate maximum likelihood in Telecommunications, 2006. AICT-ICIW’06. International Conference
source localization through convex relaxation,” IEEE Transactions on on Internet and Web Applications and Services/Advanced International
Wireless Communications, vol. 13, no. 7, pp. 3951–3964, 2014. Conference on, pp. 93–93, IEEE, 2006.
[121] A. Coluccia and F. Ricciato, “Rss-based localization via bayesian [143] A.-K. Othman, “Gps-less localization protocol for underwater acoustic
ranging and iterative least squares positioning,” IEEE Communications networks,” in Wireless and Optical Communications Networks, 2008.
Letters, vol. 18, no. 5, pp. 873–876, 2014. WOCN’08. 5th IFIP International Conference on, pp. 1–6, IEEE, 2008.
[122] S. Gepshtein and Y. Keller, “Sensor network localization by augmented [144] X. Cheng, H. Shu, Q. Liang, and D. H.-C. Du, “Silent positioning in
dual embedding,” IEEE Transactions on Signal Processing, vol. 63, underwater acoustic sensor networks,” IEEE Transactions on vehicular
no. 9, pp. 2420–2431, 2015. technology, vol. 57, no. 3, pp. 1756–1766, 2008.
[123] O. Jean and A. J. Weiss, “Passive localization and synchronization [145] X. Cheng, H. S. H. Shu, and Q. Liang, “A range-difference based
using arbitrary signals,” IEEE Transactions on Signal Processing, self-positioning scheme for underwater acoustic sensor networks,” in
vol. 62, no. 8, pp. 2143–2150, 2014. Wireless Algorithms, Systems and Applications, 2007. WASA 2007.
[124] N. Lasla, M. F. Younis, A. Ouadjaout, and N. Badache, “An effective International Conference on, pp. 38–43, IEEE, 2007.
area-based localization algorithm for wireless networks,” IEEE Trans- [146] H.-P. Tan, A. F. Gabor, Z. A. Eu, and W. K. G. Seah, “A wide coverage
actions on Computers, vol. 64, no. 8, pp. 2103–2118, 2015. positioning system (wps) for underwater localization,” in Communica-
[125] X. Lv, F. Mourad-Chehade, and H. Snoussi, “Decentralized localization tions (ICC), 2010 IEEE International Conference on, pp. 1–5, IEEE,
using radio-fingerprints and accelerometer in wsns,” IEEE Transactions 2010.
on Aerospace and Electronic Systems, vol. 51, no. 1, pp. 242–257, [147] W. Wang, D. Peng, H. Wang, H. Sharif, and H.-H. Chen, “Energy-
2015. constrained distortion reduction optimization for wavelet-based coded
[126] J.-F. Huang, G.-Y. Chang, and G.-H. Chen, “A historical-beacon-aided image transmission in wireless sensor networks,” IEEE Transactions
localization algorithm for mobile sensor networks,” IEEE Transactions on Multimedia, vol. 10, no. 6, pp. 1169–1180, 2008.
on Mobile Computing, vol. 14, no. 6, pp. 1109–1122, 2015. [148] A. Y. Teymorian, W. Cheng, L. Ma, X. Cheng, X. Lu, and Z. Lu, “3d
[127] D. Mirza and C. Schurgers, “Motion-aware self-localization for un- underwater sensor network localization,” IEEE Transactions on Mobile
derwater networks,” in Proceedings of the third ACM international Computing, vol. 8, no. 12, 2009.
workshop on Underwater Networks, pp. 51–58, ACM, 2008. [149] W. Cheng, A. Y. Teymorian, L. Ma, X. Cheng, X. Lu, and Z. Lu,
[128] T. Bian, R. Venkatesan, and C. Li, “An improved localization method “Underwater localization in sparse 3d acoustic sensor networks,” in
using error probability distribution for underwater sensor networks,” INFOCOM 2008. The 27th Conference on Computer Communications.
in Communications (ICC), 2010 IEEE International Conference on, IEEE, pp. 236–240, IEEE, 2008.
pp. 1–6, IEEE, 2010.
[150] A. Y. Teymorian, W. Cheng, L. Ma, and X. Cheng, “An underwater
[129] T. Bian, R. Venkatesan, and C. Li, “Design and evaluation of a
positioning scheme for 3d acoustic sensor networks,” in Proc. of Second
new localization scheme for underwater acoustic sensor networks,”
Workshop on Underwater Networks, Montreal, Quebec, Canada, 2007.
in Global Telecommunications Conference, 2009. GLOBECOM 2009.
[151] Z. Zhou, Z. Peng, J.-H. Cui, Z. Shi, and A. Bagtzoglou, “Scalable
IEEE, pp. 1–5, IEEE, 2009.
localization with mobility prediction for underwater sensor networks,”
[130] Q. Yao, S.-K. Tan, Y. Ge, B.-S. Yeo, and Q. Yin, “An area localization
IEEE Transactions on Mobile Computing, vol. 10, no. 3, pp. 335–348,
scheme for large wireless sensor networks,” in Vehicular Technology
2011.
Conference, 2005. VTC 2005-Spring. 2005 IEEE 61st, vol. 5, pp. 2835–
2839, IEEE, 2005. [152] K. Kendig, “Is a 2000-year-old formula still keeping some secrets?,”
[131] V. Chandrasekhar and W. Seah, “An area localization scheme for The American Mathematical Monthly, vol. 107, p. 402, may 2000.
underwater sensor networks,” in OCEANS 2006-Asia Pacific, pp. 1– [153] A. Willsky and H. Jones, “A generalized likelihood ratio approach
8, IEEE, 2007. to the detection and estimation of jumps in linear systems,” IEEE
[132] Y. Zhou, B.-j. Gu, K. Chen, J.-b. Chen, and H.-b. Guan, “An range- Transactions on Automatic Control, vol. 21, pp. 108–112, feb 1976.
free localization scheme for large scale underwater wireless sensor [154] H. E. Hurst, “Long-term storage capacity of reservoirs,” Trans. Amer.
networks,” Journal of Shanghai Jiaotong University (Science), vol. 14, Soc. Civil Eng., vol. 116, pp. 770–808, 1951.
no. 5, p. 562, 2009. [155] P. Abry and D. Veitch, “Wavelet analysis of long-range-dependent
[133] D. Mirza and C. Schurgers, “Collaborative localization for fleets of traffic,” IEEE transactions on information theory, vol. 44, no. 1, pp. 2–
underwater drifters,” in OCEANS 2007, pp. 1–6, IEEE, 2007. 15, 1998.
[134] M. Erol, L. F. M. Vieira, and M. Gerla, “Auv-aided localization for [156] E. L. Lehmann and G. Casella, Theory of point estimation. Springer
underwater sensor networks,” in Wireless Algorithms, Systems and Science & Business Media, 2006.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 32

[157] S. Čapkun, M. Hamdi, and J.-P. Hubaux, “Gps-free positioning in [178] S. Hilsenbeck, D. Bobkov, G. Schroth, R. Huitl, and E. Steinbach,
mobile ad hoc networks,” Cluster Computing, vol. 5, no. 2, pp. 157– “Graph-based data fusion of pedometer and wifi measurements for
167, 2002. mobile indoor positioning,” in Proceedings of the 2014 ACM Inter-
[158] F. Aldhubaib and N. V. Shuley, “Radar target recognition based national Joint Conference on Pervasive and Ubiquitous Computing,
on modified characteristic polarization states,” IEEE Transactions on pp. 147–158, ACM, 2014.
Aerospace and Electronic Systems, vol. 46, no. 4, pp. 1921–1933, 2010. [179] J. Seitz, J. Jahn, J. G. Boronat, T. Vaupel, S. Meyer, and J. Thielecke,
[159] Y. Kim, H. Shin, and H. Cha, “Smartphone-based wi-fi pedestrian- “A hidden markov model for urban navigation based on fingerprinting
tracking system tolerating the rss variance problem,” in Pervasive and pedestrian dead reckoning,” in Information Fusion (FUSION), 2010
Computing and Communications (PerCom), 2012 IEEE International 13th Conference on, pp. 1–8, IEEE, 2010.
Conference on, pp. 11–19, IEEE, 2012. [180] W. Sun, J. Liu, C. Wu, Z. Yang, X. Zhang, and Y. Liu, “Moloc: On
[160] G. Shen, Z. Chen, P. Zhang, T. Moscibroda, and Y. Zhang, “Walkie- distinguishing fingerprint twins,” in Distributed Computing Systems
markie: indoor pathway mapping made easy,” in Proceedings of the (ICDCS), 2013 IEEE 33rd International Conference on, pp. 226–235,
10th USENIX conference on Networked Systems Design and Imple- IEEE, 2013.
mentation, pp. 85–98, USENIX Association, 2013. [181] Z. Xiao, H. Wen, A. Markham, and N. Trigoni, “Lightweight map
[161] H. Wang, S. Sen, A. Elgohary, M. Farid, M. Youssef, and R. R. matching for indoor localisation using conditional random fields,” in
Choudhury, “No need to war-drive: unsupervised indoor localization,” Information Processing in Sensor Networks, IPSN-14 Proceedings of
in Proceedings of the 10th international conference on Mobile systems, the 13th International Symposium on, pp. 131–142, IEEE, 2014.
applications, and services, pp. 197–210, ACM, 2012. [182] M. Youssef, M. A. Yosef, and M. El-Derini, “Gac: energy-efficient hy-
[162] Y. Jiang, Y. Xiang, X. Pan, K. Li, Q. Lv, R. P. Dick, L. Shang, and brid gps-accelerometer-compass gsm localization,” in Global Telecom-
M. Hannigan, “Hallway based automatic indoor floorplan construction munications Conference (GLOBECOM 2010), 2010 IEEE, pp. 1–5,
using room fingerprints,” in Proceedings of the 2013 ACM international IEEE, 2010.
joint conference on Pervasive and ubiquitous computing, pp. 315–324, [183] I. Constandache, R. R. Choudhury, and I. Rhee, “Towards mobile phone
ACM, 2013. localization without war-driving,” in Infocom, 2010 proceedings ieee,
[163] S. He and S.-H. G. Chan, “Sectjunction: Wi-fi indoor localization based pp. 1–9, IEEE, 2010.
on junction of signal sectors,” in Communications (ICC), 2014 IEEE [184] X. Zhu, Q. Li, and G. Chen, “Apt: Accurate outdoor pedestrian tracking
International Conference on, pp. 2605–2610, IEEE, 2014. with smartphones,” in INFOCOM, 2013 Proceedings IEEE, pp. 2508–
[164] K. Kaji and N. Kawaguchi, “Design and implementation of wifi indoor 2516, IEEE, 2013.
localization based on gaussian mixture model and particle filter,” in [185] M. Anisetti, C. A. Ardagna, V. Bellandi, E. Damiani, M. Döller,
Indoor Positioning and Indoor Navigation (IPIN), 2012 International F. Stegmaier, T. Rabl, H. Kosch, and L. Brunie, “Landmark-assisted
Conference on, pp. 1–9, IEEE, 2012. location and tracking in outdoor mobile network,” Multimedia Tools
[165] A. LaMarca, Y. Chawathe, S. Consolvo, J. Hightower, I. Smith, and Applications, vol. 59, no. 1, pp. 89–111, 2012.
J. Scott, T. Sohn, J. Howard, J. Hughes, F. Potter, et al., “Place lab: [186] H. Wang, Z. Wang, G. Shen, F. Li, S. Han, and F. Zhao, “Wheelloc:
Device positioning using radio beacons in the wild,” in International Enabling continuous location service on mobile phone for outdoor
Conference on Pervasive Computing, pp. 116–133, Springer, 2005. scenarios,” in INFOCOM, 2013 Proceedings IEEE, pp. 2733–2741,
[166] “Placeengine.” [Link] IEEE, 2013.
[167] P. Cherntanomwong, J.-i. Takada, and H. Tsuji, “Signal subspace
[187] H. Aly and M. Youssef, “Dejavu: an accurate energy-efficient outdoor
interpolation from discrete measurement samples in constructing a
localization system,” in Proceedings of the 21st ACM SIGSPATIAL
database for location fingerprint technique,” IEICE transactions on
International Conference on Advances in Geographic Information
communications, vol. 92, no. 9, pp. 2922–2930, 2009.
Systems, pp. 154–163, ACM, 2013.
[168] J. Paek, K.-H. Kim, J. P. Singh, and R. Govindan, “Energy-efficient
[188] I. Constandache, S. Gaonkar, M. Sayler, R. R. Choudhury, and L. Cox,
positioning for smartphones using cell-id sequence matching,” in
“Enloc: Energy-efficient localization for mobile phones,” in INFOCOM
Proceedings of the 9th international conference on Mobile systems,
2009, IEEE, pp. 2716–2720, IEEE, 2009.
applications, and services, pp. 293–306, ACM, 2011.
[169] M. Ibrahim and M. Youssef, “Cellsense: An accurate energy-efficient [189] J. Paek, J. Kim, and R. Govindan, “Energy-efficient rate-adaptive
gsm positioning system,” IEEE Transactions on Vehicular Technology, gps-based positioning for smartphones,” in Proceedings of the 8th
vol. 61, no. 1, pp. 286–296, 2012. international conference on Mobile systems, applications, and services,
pp. 299–314, ACM, 2010.
[170] H. Liu, Y. Gan, J. Yang, S. Sidhom, Y. Wang, Y. Chen, and F. Ye, “Push
the limit of wifi based localization for smartphones,” in Proceedings [190] T. O. Oshin, S. Poslad, and A. Ma, “Improving the energy-efficiency of
of the 18th annual international conference on Mobile computing and gps based location sensing smartphone applications,” in Trust, Security
networking, pp. 305–316, ACM, 2012. and Privacy in Computing and Communications (TrustCom), 2012
[171] R. Nandakumar, K. K. Chintalapudi, and V. N. Padmanabhan, “Cen- IEEE 11th International Conference on, pp. 1698–1705, IEEE, 2012.
taur: locating devices in an office environment,” in Proceedings of [191] C. Bo, X.-Y. Li, T. Jung, X. Mao, Y. Tao, and L. Yao, “Smartloc:
the 18th annual international conference on Mobile computing and Push the limit of the inertial sensor based metropolitan localization
networking, pp. 281–292, ACM, 2012. using smartphone,” in Proceedings of the 19th annual international
[172] L.-w. Chan, J.-r. Chiang, Y.-c. Chen, C.-n. Ke, J. Hsu, and H.-h. Chu, conference on Mobile computing & networking, pp. 195–198, ACM,
“Collaborative localization: Enhancing wifi-based position estimation 2013.
with neighborhood links in clusters,” in International Conference on [192] A. Hallquist and A. Zakhor, “Single view pose estimation of mobile
Pervasive Computing, pp. 50–66, Springer, 2006. devices in urban environments,” in Applications of Computer Vision
[173] J. Jun, Y. Gu, L. Cheng, B. Lu, J. Sun, T. Zhu, and J. Niu, “Social-loc: (WACV), 2013 IEEE Workshop on, pp. 347–354, IEEE, 2013.
Improving indoor localization with social sensing,” in Proceedings of [193] F. Dabek, R. Cox, F. Kaashoek, and R. Morris, “Vivaldi: A decen-
the 11th ACM Conference on Embedded Networked Sensor Systems, tralized network coordinate system,” in ACM SIGCOMM Computer
p. 14, ACM, 2013. Communication Review, vol. 34, pp. 15–26, ACM, 2004.
[174] G. Schroth, R. Huitl, D. Chen, M. Abu-Alqumsan, A. Al-Nuaimi, [194] C. Peng, G. Shen, Y. Zhang, Y. Li, and K. Tan, “Beepbeep: a high
and E. Steinbach, “Mobile visual location recognition,” IEEE Signal accuracy acoustic ranging system using cots mobile devices,” in Pro-
Processing Magazine, vol. 28, no. 4, pp. 77–89, 2011. ceedings of the 5th international conference on Embedded networked
[175] J. Zhang, A. Hallquist, E. Liang, and A. Zakhor, “Location-based image sensor systems, pp. 1–14, ACM, 2007.
retrieval for urban environments,” in Image Processing (ICIP), 2011 [195] M. Uddin and T. Nadeem, “Rf-beep: A light ranging scheme for smart
18th IEEE International Conference on, pp. 3677–3680, IEEE, 2011. devices,” in Pervasive Computing and Communications (PerCom),
[176] A. Rai, K. K. Chintalapudi, V. N. Padmanabhan, and R. Sen, “Zee: 2013 IEEE International Conference on, pp. 114–122, IEEE, 2013.
Zero-effort crowdsourcing for indoor localization,” in Proceedings of [196] S. Liu, Y. Jiang, and A. Striegel, “Face-to-face proximity estima-
the 18th annual international conference on Mobile computing and tionusing bluetooth on smartphones,” IEEE Transactions on Mobile
networking, pp. 293–304, ACM, 2012. Computing, vol. 13, no. 4, pp. 811–823, 2014.
[177] Y. Gao, Q. Yang, G. Li, E. Y. Chang, D. Wang, C. Wang, H. Qu, [197] F. Li, C. Zhao, G. Ding, J. Gong, C. Liu, and F. Zhao, “A reliable and
P. Dong, and F. Zhang, “Xins: The anatomy of an indoor positioning accurate indoor localization method using phone inertial sensors,” in
and navigation architecture,” in Proceedings of the 1st international Proceedings of the 2012 ACM Conference on Ubiquitous Computing,
workshop on Mobile location-based service, pp. 41–50, ACM, 2011. pp. 421–430, ACM, 2012.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 33

[198] Z. Chen, H. Zou, H. Jiang, Q. Zhu, Y. C. Soh, and L. Xie, “Fusion neural network.,” Arabian Journal for Science & Engineering (Springer
of wifi, smartphone sensors and landmarks using the kalman filter for Science & Business Media BV), vol. 37, no. 4, 2012.
indoor localization,” Sensors, vol. 15, no. 1, pp. 715–732, 2015. [221] A. Azenha, L. Peneda, and A. Carvalho, “A neural network approach
[199] Z. Yang, X. Feng, and Q. Zhang, “Adometer: Push the limit of for radio frequency based indoors localization,” in IECON 2012-38th
pedestrian indoor localization through cooperation,” IEEE Transactions Annual Conference on IEEE Industrial Electronics Society, pp. 5990–
on Mobile Computing, vol. 13, no. 11, pp. 2473–2483, 2014. 5995, IEEE, 2012.
[200] J. Yim, S. Jeong, K. Gwon, and J. Joo, “Improvement of kalman filters [222] L. Luoh, “Zigbee-based intelligent indoor positioning system soft
for wlan based indoor tracking,” Expert Systems with Applications, computing,” Soft Computing, vol. 18, no. 3, pp. 443–456, 2014.
vol. 37, no. 1, pp. 426–433, 2010. [223] L. Gogolak, S. Pletl, and D. Kukolj, “Neural network-based indoor lo-
[201] S. Ali-Loytty, T. Perala, V. Honkavirta, and R. Piché, “Fingerprint calization in wsn environments,” Acta Polytechnica Hungarica, vol. 10,
kalman filter in indoor positioning applications,” in Control Applica- no. 6, pp. 221–235, 2013.
tions,(CCA) & Intelligent Control,(ISIC), 2009 IEEE, pp. 1678–1683, [224] A. Payal, C. Rai, and B. Reddy, “Artificial neural networks for
IEEE, 2009. developing localization framework in wireless sensor networks,” in
[202] Z. Xiao, H. Wen, A. Markham, and N. Trigoni, “Indoor tracking Data Mining and Intelligent Computing (ICDMIC), 2014 International
using undirected graphical models,” IEEE Transactions on Mobile Conference on, pp. 1–6, IEEE, 2014.
Computing, vol. 14, no. 11, pp. 2286–2301, 2015. [225] P.-J. Chuang and Y.-J. Jiang, “Effective neural network-based node
[203] J. A. Fuemmeler and V. V. Veeravalli, “Smart sleeping policies for localisation scheme for wireless sensor networks,” IET Wireless Sensor
energy-efficient tracking in sensor networks,” in Networked Sensing Systems, vol. 4, no. 2, pp. 97–103, 2014.
Information and Control, pp. 267–287, Springer, 2008. [226] A. Payal, C. S. Rai, and B. R. Reddy, “Analysis of some feedforward
[204] J. A. Fuemmeler, G. K. Atia, and V. V. Veeravalli, “Sleep control for artificial neural network training algorithms for developing localization
tracking in sensor networks,” IEEE Transactions on Signal Processing, framework in wireless sensor networks,” Wireless Personal Communi-
vol. 59, no. 9, pp. 4354–4366, 2011. cations, vol. 82, no. 4, pp. 2519–2536, 2015.
[205] Y. Li, L. W. Krakow, E. K. Chong, and K. N. Groom, “Approximate [227] N. Irfan, M. Bolic, M. C. Yagoub, and V. Narasimhan, “Neural-
stochastic dynamic programming for sensor scheduling to track mul- based approach for localization of sensors in indoor environment,”
tiple targets,” Digital Signal Processing, vol. 19, no. 6, pp. 978–989, Telecommunication Systems, vol. 44, no. 1, pp. 149–158, 2010.
2009.
[228] A. Payal, C. Rai, and B. Reddy, “Comparative analysis of bayesian reg-
[206] G. K. Atia, V. V. Veeravalli, and J. A. Fuemmeler, “Sensor scheduling ularization and levenberg-marquardt training algorithm for localization
for energy-efficient target tracking in sensor networks,” IEEE Transac- in wireless sensor network,” in Advanced Communication Technology
tions on Signal Processing, vol. 59, no. 10, pp. 4923–4937, 2011. (ICACT), 2013 15th International Conference on, pp. 191–194, IEEE,
[207] Z.-n. Zhang and G.-l. Shan, “Uts-based foresight optimization of sensor 2013.
scheduling for low interception risk tracking,” International Journal of
[229] C. Nerguizian and V. Nerguizian, “Indoor fingerprinting geolocation
Adaptive Control and Signal Processing, vol. 28, no. 10, pp. 921–931,
using wavelet-based features extracted from the channel impulse re-
2014.
sponse in conjunction with an artificial neural network,” in Industrial
[208] L. K. Au, A. A. Bui, M. A. Batalin, X. Xu, and W. J. Kaiser,
Electronics, 2007. ISIE 2007. IEEE International Symposium on,
“Carer: Efficient dynamic sensing for continuous activity monitoring,”
pp. 2028–2032, IEEE, 2007.
in Engineering in Medicine and Biology Society, EMBC, 2011 Annual
[230] S. Kumar, S. M. Jeon, and S. R. Lee, “Localization estimation using
International Conference of the IEEE, pp. 2228–2232, IEEE, 2011.
artificial intelligence technique in wireless sensor networks,” , vol. 39,
[209] D.-S. Zois and U. Mitra, “A unified framework for energy efficient
no. 9, pp. 820–827, 2014.
physical activity tracking,” in Signals, Systems and Computers, 2013
Asilomar Conference on, pp. 69–73, IEEE, 2013. [231] S. Kumar and S. R. Lee, “Localization with rssi values for wireless
sensor networks: An artificial neural network approach,” in Interna-
[210] S. Zhan and J. Li, “Active cross-layer location identification of attackers
tional Electronic Conference on Sensors and Applications, vol. 1,
in wireless sensor networks,” in Computer Engineering and Technology
Multidisciplinary Digital Publishing Institute, 2014.
(ICCET), 2010 2nd International Conference on, vol. 3, pp. V3–240,
IEEE, 2010. [232] P. K. Sahu, E. H.-K. Wu, and J. Sahoo, “Durt: Dual rssi trend based
[211] S. Misra and S. Singh, “Localized policy-based target tracking using localization for wireless sensor networks,” IEEE Sensors Journal,
wireless sensor networks,” ACM Transactions on Sensor Networks vol. 13, no. 8, pp. 3115–3123, 2013.
(TOSN), vol. 8, no. 3, p. 27, 2012. [233] K. Thongpul, N. Jindapetch, and W. Teerapakajorndet, “A neural
[212] S. Pino-Povedano, R. Arroyo-Valles, and J. Cid-Sueiro, “Selective network based optimization for wireless sensor node position estima-
forwarding for energy-efficient target tracking in sensor networks,” tion in industrial environments,” in Electrical Engineering/Electronics
Signal Processing, vol. 94, pp. 557–569, 2014. Computer Telecommunications and Information Technology (ECTI-
[213] A. Pietrabissa, C. Poli, D. G. Ferriero, and M. Grigioni, “Optimal plan- CON), 2010 International Conference on, pp. 249–253, IEEE, 2010.
ning of sensor networks for asset tracking in hospital environments,” [234] R. V. Kulkarni, G. K. Venayagamoorthy, A. Miller, and C. H. Dagli,
Decision Support Systems, vol. 55, no. 1, pp. 304–313, 2013. “Network-centric localization in manets based on particle swarm op-
[214] W.-L. Yeow, C.-K. Tham, and W.-C. Wong, “Energy efficient multiple timization,” in Swarm Intelligence Symposium, 2008. SIS 2008. IEEE,
target tracking in wireless sensor networks,” IEEE Transactions on pp. 1–6, IEEE, 2008.
Vehicular Technology, vol. 56, no. 2, pp. 918–928, 2007. [235] K. Pahlavan, F. Akgul, Y. Ye, T. Morgan, F. Alizadeh-Shabdiz, M. Hei-
[215] A. Jamal, C.-K. Tham, and W. C. Wong, “Event detection and channel dari, and C. Steger, “Taking positioning indoors wi-fi localization and
allocation in cognitive radio sensor networks,” in Communication gnss,” Inside GNSS, vol. 5, no. 3, pp. 40–47, 2010.
Systems (ICCS), 2012 IEEE International Conference on, pp. 157–161, [236] J. Fuentes-Pacheco, J. Ruiz-Ascencio, and J. M. Rendón-Mancha,
IEEE, 2012. “Visual simultaneous localization and mapping: a survey,” Artificial
[216] J. W. Huang, Q. Zhu, V. Krishnamurthy, and T. Basar, “Distributed Intelligence Review, vol. 43, no. 1, pp. 55–81, 2015.
correlated q-learning for dynamic transmission control of sensor net- [237] P. Henry, M. Krainin, E. Herbst, X. Ren, and D. Fox, “Rgb-d mapping:
works,” in Acoustics Speech and Signal Processing (ICASSP), 2010 Using kinect-style depth cameras for dense 3d modeling of indoor
IEEE International Conference on, pp. 1982–1985, IEEE, 2010. environments,” The International Journal of Robotics Research, vol. 31,
[217] R. D. Smallwood and E. J. Sondik, “The optimal control of partially no. 5, pp. 647–663, 2012.
observable markov processes over a finite horizon,” Operations Re- [238] K. Pahlavan, Y. Ye, U. Khan, and R. Fu, “Rf localization inside human
search, vol. 21, pp. 1071–1088, oct 1973. body: Enabling micro-robotic navigation for medical applications,” in
[218] S. Yun, J. Lee, W. Chung, E. Kim, and S. Kim, “A soft computing Localization and GNSS (ICL-GNSS), 2011 International Conference
approach to localization in wireless sensor networks,” Expert Systems on, pp. 133–139, IEEE, 2011.
with Applications, vol. 36, no. 4, pp. 7552–7561, 2009. [239] N. Marya, A. Karellas, A. Foley, A. Roychowdhury, and D. Cave,
[219] L. Gogolak, S. Pletl, and D. Kukolj, “Indoor fingerprint localization “Computerized 3-dimensional localization of a video capsule in the
in wsn environment based on neural network,” in Intelligent Systems abdominal cavity: validation by digital radiography,” Gastrointestinal
and Informatics (SISY), 2011 IEEE 9th International Symposium on, endoscopy, vol. 79, no. 4, pp. 669–674, 2014.
pp. 293–296, IEEE, 2011. [240] S. Yim and M. Sitti, “3-d localization method for a magnetically actu-
[220] M. Rahman, Y. Park, and K.-D. Kim, “Rss-based indoor localization ated soft capsule endoscope and its applications,” IEEE Transactions
algorithm for wireless sensor network using generalized regression on Robotics, vol. 29, no. 5, pp. 1139–1151, 2013.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 34

[241] T. D. Than, G. Alici, H. Zhou, and W. Li, “A review of localiza- [263] P. Castro, D. Zhang, and S. Li, “Urban traffic modelling and prediction
tion systems for robotic endoscopic capsules,” IEEE Transactions on using large scale taxi gps traces,” Pervasive Computing, pp. 57–72,
Biomedical Engineering, vol. 59, no. 9, pp. 2387–2399, 2012. 2012.
[242] R. Kuth, J. Reinschke, and R. Rockelein, “Method for determining the [264] J. J. Forest, Homeland security: Critical infrastructure, vol. 3. Green-
position and orientation of an endoscopy capsule guided through an wood Publishing Group, 2006.
examination object by using a navigating magnetic field generated by [265] X. Fang, S. Misra, G. Xue, and D. Yang, “Smart grid—the new and
means of a navigation device,” July 7 2006. improved power grid: A survey,” IEEE communications surveys &amp;
[243] K. Pahlavan, G. Bao, Y. Ye, S. Makarov, U. Khan, P. Swar, D. Cave, tutorials, vol. 14, no. 4, pp. 944–980, 2012.
A. Karellas, P. Krishnamurthy, and K. Sayrafian, “Rf localization for
wireless video capsule endoscopy,” International Journal of Wireless
Information Networks, vol. 19, no. 4, pp. 326–340, 2012.
[244] H. Li, K. Ota, M. Dong, and H.-H. Chen, “Efficient energy transport
in 60 ghz for wireless industrial sensor networks,” IEEE Wireless
Communications, vol. 24, no. 5, pp. 143–149, 2017.
[245] G. Pan, G. Qi, W. Zhang, S. Li, Z. Wu, and L. T. Yang, “Trace analysis
and mining for smart cities: issues, methods, and applications,” IEEE Rathin Chandra Shit received the [Link]. degrees
Communications Magazine, vol. 51, no. 6, pp. 120–126, 2013. in Electronics and Telecommunication engineering
[246] M. He and J. Zhang, “A dependency graph approach for fault detection from the Biju Patnaik University of Technology,
and localization towards secure smart grid,” IEEE Transactions on Rourkela, India, in 2010 and [Link]. degree in
Smart Grid, vol. 2, no. 2, pp. 342–351, 2011. Information and Communication Technology from
[247] O. B. Akan, O. B. Karli, and O. Ergul, “Cognitive radio sensor Veer Surendra Sai University of Technology, Burla,
networks,” IEEE network, vol. 23, no. 4, 2009. India, in 2016, and is currently working toward the
[248] S. H. R. Bukhari, M. H. Rehmani, and S. Siraj, “A survey of channel Ph.D. degree from International Institute of Informa-
bonding for wireless networks and guidelines of channel bonding tion Technology, Bhubaneswar, India.
for futuristic cognitive radio sensor networks,” IEEE Communications Also, he had been a Senior Research Fellow with
Surveys &amp; Tutorials, vol. 18, no. 2, pp. 924–948, 2016. the Radar Systems Group, Integrated Test Range,
[249] A. Ahmad, S. Ahmad, M. H. Rehmani, and N. U. Hassan, “A survey Defence Research and Development Organisation, India under the Ministries
on radio resource allocation in cognitive radio sensor networks,” IEEE of Defence, Government of India. During this period, he was involved in
Communications Surveys &amp; Tutorials, vol. 17, no. 2, pp. 888–917, many projects including: Tracking Radar Systems, Receiver Front-end Design
2015. of Radar System, Radar Data Processing Software Development. Currently,
his research interests in Wireless Sensor Network, focused on Localization
[250] S. Helal, W. Mann, H. El-Zabadani, J. King, Y. Kaddoura, and
and Tracking and Location Based Services in IoT Infrastructure.
E. Jansen, “The gator tech smart house: A programmable pervasive
space,” Computer, vol. 38, no. 3, pp. 50–60, 2005.
[251] S.-H. Baeg, J.-H. Park, J. Koh, K.-W. Park, and M.-H. Baeg, “Building
a smart home environment for service robots based on rfid and sensor
networks,” in Control, Automation and Systems, 2007. ICCAS’07.
International Conference on, pp. 1078–1082, IEEE, 2007.
[252] L. Bruno and P. Robertson, “Observability of path loss parameters in
wlan-based simultaneous localization and mapping,” in Indoor Posi- Suraj Sharma received the Ph.D. degree from the
tioning and Indoor Navigation (IPIN), 2013 International Conference National Institute of Technology Rourkela, India.
on, pp. 1–10, IEEE, 2013. He is currently an Assistant Professor with the
[253] J. Huang, D. Millman, M. Quigley, D. Stavens, S. Thrun, and A. Ag- Department of Computer Science and Engineering,
garwal, “Efficient, generalized indoor wifi graphslam,” in Robotics and International Institute of Information Technology at
Automation (ICRA), 2011 IEEE International Conference on, pp. 1038– Bhubaneswar, India. He has authored several journal
1043, IEEE, 2011. and international conference papers. His research
[254] G. Iddan, G. Meron, A. Glukhovsky, and P. Swain, “Wireless capsule interest includes Information security, Internet of
endoscopy,” Nature, vol. 405, no. 6785, p. 417, 2000. Things and Wireless Sensor Networks.
[255] Y. Geng, J. Chen, R. Fu, G. Bao, and K. Pahlavan, “Enlighten wearable
physiological monitoring systems: On-body rf characteristics based
human motion classification using a support vector machine,” IEEE
transactions on mobile computing, vol. 15, no. 3, pp. 656–671, 2016.
[256] G. Hackmann, W. Guo, G. Yan, Z. Sun, C. Lu, and S. Dyke,
“Cyber-physical codesign of distributed structural health monitoring
with wireless sensor networks,” IEEE Transactions on Parallel and
Distributed Systems, vol. 25, no. 1, pp. 63–72, 2014.
[257] K. Pahlavan, X. Li, and J.-P. Makela, “Indoor geolocation science and Deepak Puthal received the Ph.D. degree in com-
technology,” IEEE Communications Magazine, vol. 40, no. 2, pp. 112– puter and information systems from University of
118, 2002. Technology Sydney (UTS), Australia. He is cur-
[258] S. Yousefi, H. Narui, S. Dayal, S. Ermon, and S. Valaee, “A survey rently a Lecturer (Assistant Professor) with the
on behavior recognition using wifi channel state information,” IEEE Faculty of Engineering and IT, UTS, Australia. He
Communications Magazine, vol. 55, no. 10, pp. 98–104, 2017. has authored in several international conferences
[259] Y. Ma, K. Pahlavan, and Y. Geng, “Comparison of poa and toa based and journals, including IEEE and ACM transac-
ranging behavior for rfid application,” in Personal, Indoor, and Mobile tions. His research interests include cyber security,
Radio Communication (PIMRC), 2014 IEEE 25th Annual International Internet of Things, distributed computing, and big
Symposium on, pp. 1722–1726, IEEE, 2014. data analytics. He received the IEEE Distinguished
[260] S. C. Spinella, A. Iera, and A. Molinaro, “On potentials and limitations Doctoral Dissertation Award for the year 2017. He
of a hybrid wlan-rfid indoor positioning technique,” International is an Associate Editor of the IEEE Consumer Electronics Magazine, the
Journal of Navigation and Observation, vol. 2010, 2010. Internet Technology Letters (Wiley) and the KSII Transactions on Internet
[261] M. Karaliopoulos and C. Rohner, “Trace-based performance analysis and Information Systems. He also served as a Co-Guest Editor of several
of opportunistic forwarding under imperfect node cooperation,” in reputed journals, including the Concurrency and Computation: Practice and
INFOCOM, 2012 Proceedings IEEE, pp. 2651–2655, IEEE, 2012. Experience, the Wireless Communications and Mobile Computing, and the
[262] J. Hagenauer and M. Helbich, “Mining urban land-use patterns from IEEE Consumer Electronics Magazine.
volunteered geographic information by means of genetic algorithms
and artificial neural networks,” International Journal of Geographical
Information Science, vol. 26, no. 6, pp. 963–982, 2012.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2798591, IEEE
Communications Surveys & Tutorials
IEEE JOURNAL OF XXXXX , VOL. XX, NO. X, MONTH XXXX 35

Albert Y. Zomaya (M’90-SM’97-F’04) is currently

the Chair Professor of High Performance Comput-
ing & Networking in the School of Information
Technologies, University of Sydney. He is also the
Director of the Centre for Distributed and High
Performance Computing which was established in
late 2009. Professor Zomaya was an Australian
Research Council Professorial Fellow during 2010-
2014. He published more than 500 scientific papers
and articles and is author, co-author or editor of more
than 20 books.
He is the Founding Editor in Chief of the IEEE Transactions on Sustainable
Computing and previously he served as Editor in Chief for the IEEE Trans-
actions on Computers (2011-2014). Currently, Professor Zomaya serves as an
associate editor for 22 leading journals, such as, the ACM Computing Surveys,
IEEE Transactions on Computational Social Systems, IEEE Transactions on
Cloud Computing, and Journal of Parallel and Distributed Computing. He
delivered more than 180 keynote addresses, invited seminars, and media
briefings and has been actively involved, in a variety of capacities, in the
organization of more than 700 national and international conferences.
Professor Zomaya is the recipient of the IEEE Technical Committee on
Parallel Processing Outstanding Service Award (2011), the IEEE Technical
Committee on Scalable Computing Medal for Excellence in Scalable Com-
puting (2011), and the IEEE Computer Society Technical Achievement Award
(2014). He is a Chartered Engineer, a Fellow of AAAS, IEEE, IET (UK),
and an IEEE Computer Society’s Golden Core member. Professor Zomaya’s
research interests lie in parallel and distributed computing, networking, and
complex systems.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See [Link] for more information.