sensors

Article
An Obstacle-Tolerant Path Planning Algorithm for
Mobile-Anchor-Node-Assisted Localization
Rong-Guei Tsai and Pei-Hsuan Tsai * ID

Institute of Manufacturing Information and Systems, National Cheng Kung University, Tainan City 70101,
Taiwan; P98031056@mail.ncku.edu.tw
* Correspondence: Phtsai@mail.ncku.edu.tw; Tel.: +886-6-275-7575

Received: 14 January 2018; Accepted: 15 March 2018; Published: 16 March 2018

Abstract: The location information obtained using a sensor is a critical requirement in

wireless sensor networks. Numerous localization schemes have been proposed, among which
mobile-anchor-node-assisted localization (MANAL) can reduce costs and overcome environmental
constraints. A mobile anchor node (MAN) provides its own location information to assist the
localization of sensor nodes. Numerous path planning schemes have been proposed for MANAL,
but most scenarios assume the absence of obstacles in the environment. However, in a realistic
environment, sensor nodes cannot be located because the obstacles block the path traversed by the
MAN, thereby rendering the sensor incapable of receiving sufficient three location information from
the MAN. This study proposes the obstacle-tolerant path planning (OTPP) approach to solve the
sensor location problem owing to obstacle blockage. OTPP can approximate the optimum beacon
point number and path planning, thereby ensuring that all the unknown nodes can receive the
three location information from the MAN and reduce the number of MAN broadcast packet times.
Experimental results demonstrate that OTPP performs better than Z-curves because it reduces the
total number of beacon points utilized and is thus more suitable in an obstacle-present environment.
Compared to the Z-curve, OTPP can reduce localization error and improve localization coverage.

Keywords: localization; mobile anchor node; path planning; wireless sensor networks

1. Introduction
Wireless sensor networks (WSNs) consist of a number of sensing devices with external physical
information and network communication capabilities. WSNs are deployed in an area to be detected,
and they collect and report environmental information for users. The scope of application of WSNs
includes the military, intelligent h, health care, and ecological detection [1]. In the application of
WSNs, localization is a critical technique that enables unknown nodes (or ordinary nodes) to obtain
their own location information to further optimize applications such as routing protocols and object
tracking [2,3].
The Global Positioning System (GPS) is a tool that obtains location information. However,
each unknown node equipped with a GPS transceiver will result in a substantial increase in costs.
A viable method is utilizing a few anchor nodes. Anchor nodes are nodes that know their locations,
being equipped with GPS transceivers. Anchor nodes broadcast messages including their location
information to unknown nodes. Unknown nodes then use distance measurement technology to obtain
the relative distance or angle relationship with the anchor nodes. Thereafter, the positions of these
unknown nodes are calculated via localization algorithms [2–6]. The minimum number of required
anchor nodes must ensure that each unknown node receives at least three messages from different
anchor nodes, which is still costly. An alternative is using a dynamic anchor node, also known as
a mobile anchor node (MAN), with a pre-planned path to continuously broadcast location information

Sensors 2018, 18, 889; doi:10.3390/s18030889 www.mdpi.com/journal/sensors

ensure that all the unknown nodes can receive three (non-colinear) packets, thereby enabling
unknown nodes to calculate their own positions. Most environments have obstacles that involve
communication-signal shadowing. For example, electromagnetic waves collide with obstacles on the
transmission path, thereby resulting in reflection, diffraction, and scattering as well as in the
unknown nodes being rendered incapable of receiving packets from a MAN.
Sensors 2018, 18, 889 2 of 19
To solve the problem of obstacle blocking, reference [8] proposed an intuitive method, in which
a MAN changes its movement direction when it encounters obstacles, detours around the obstacles,
and broadcasts
to replace multiplepackets to anchor
static unknown nodes.
nodes. TheHowever,
locations this method
where cannot
the MAN guaranteemessages
broadcasts that all the
are
unknown nodes
called beacon will receive
points. three packets
This localization for locating
architecture their mobile-anchor-node-assisted
is called position. In addition, a MAN localization
is required
to additionally
(MANAL) [2]. broadcast
MANALsubstantial
compriseslocation information
a GPS-equipped MANwhenand
changing the direction
numerous unknown of movement,
nodes [7,8],
thereby resulting
as shown in Figure 1. in excessive wastage of power for the MAN and unknown nodes.

Figure 1. Network architecture of MANAL.

Figure 1. Network architecture of MANAL.

To reduce costs and guarantee that all the unknown nodes in the sensing area receive at least
three Path planning
messages from is aMAN,
criticalthere
issuearein MANAL. The fundamental
optimal placements requirement
for beacon points.for a MAN
The is topath
optimal ensure
of
that all the unknown nodes can receive three (non-colinear) packets,
MAN is based on the optimal placement of beacon points. Obviously, the MAN’s path problem thereby enabling unknown nodesis
to calculate their own positions. Most environments have obstacles that
one of the Hamiltonian path problems [9] that visits each vertex of the graph exactly once. The involve communication-signal
shadowing. path
Hamiltonian For example,
problem is electromagnetic waves collideproblem
a well-known NP-complete with obstacles
[10,11] and on the transmission
therefore MAN’s path,
path
thereby resulting in reflection, diffraction, and scattering as well
is also an NP-complete problem. Therefore, this study presents an obstacle-tolerant path as in the unknown nodes being
planning
rendered
(OTPP) incapablewhich
approach, of receiving
uses thepackets
divide from a MAN. strategy to divide the entire area into multiple
and conquer
unit squares, thereby optimizing the position of reference
To solve the problem of obstacle blocking, the beacon [8]point
proposed
and the anrequired
intuitivenumber
method,ofinbeacon
which
a MAN changes its movement direction when it encounters obstacles,
points for each unit square. Lastly, all the beacon points are visited by MAN. To summarize, the detours around the obstacles,
and broadcasts
contributions packets
of this study toare
unknown nodes. However, this method cannot guarantee that all the
as follows:
unknown nodes will receive three packets for locating their position. In addition, a MAN is required
1. In the obstacle-present
to additionally environment,
broadcast substantial OTPP
location guarantees
information whenthatchanging
all the unknown nodes
the direction ofcan receive
movement,
three location information (packets) from a MAN and
thereby resulting in excessive wastage of power for the MAN and unknown nodes. the number of beacon points of MAN in
OTTP is less
To reduce thanand
costs in Z-curve.
guarantee that all the unknown nodes in the sensing area receive at least
2. OTPP uses a rectangle
three messages from MAN, there as are
the optimal
boundary of each for
placements of beacon
the obstacles;
points. The anyoptimal
convexpathpolygon
of MAN is
transformed into a rectangle, and two path trajectories for MAN
is based on the optimal placement of beacon points. Obviously, the MAN’s path problem is one ofare proposed.
3. In the obstacle-present
the Hamiltonian path problems environment, this
[9] that visits eachstudy is the
vertex of thefirst to propose
graph exactly aonce.
static path
The planning
Hamiltonian
path problem is a well-known NP-complete problem [10,11] and therefore MAN’s path points,
method to optimize the number of beacon points. OTPP optimizes the number of beacon is also
thereby reducing
an NP-complete problem. the Therefore,
number ofthis broadcast times an
study presents of obstacle-tolerant
MAN and receiving packet times
path planning (OTPP)of
unknown
approach, whichnodesusestothe facilitate
dividethe andreduction
conquerin power consumption.
strategy to divide the entire area into multiple unit
squares, thereby optimizing the position of the beacon point and the required number of beacon points
for each unit square. Lastly, all the beacon points are visited by MAN. To summarize, the contributions
of this study are as follows:
1. In the obstacle-present environment, OTPP guarantees that all the unknown nodes can receive
three location information (packets) from a MAN and the number of beacon points of MAN in
OTTP is less than in Z-curve.
2. OTPP uses a rectangle as the boundary of each of the obstacles; any convex polygon is transformed
into a rectangle, and two path trajectories for MAN are proposed.
3. In the obstacle-present environment, this study is the first to propose a static path planning
method to optimize the number of beacon points. OTPP optimizes the number of beacon points,
thereby reducing the number of broadcast times of MAN and receiving packet times of unknown
nodes to facilitate the reduction in power consumption.
Sensors 2018, 18, 889 3 of 19

4. OTPP performs better than Z-curve because it uses less total number of beacon points and is thus
more suitable in an obstacle-present environment. Compared to the Z-curve, OTPP can reduce
localization error and improve localization coverage.
5. Online path planning methods cannot guarantee that all the unknown sensors in the sensing
area can be positioned. Compared to them, our approach, OTPP, does not only guarantee full
coverage of sensor localization but also has scalability and flexibility to adapt to environmental
changes by adopting the divide-and-conquer method.

The rest of this paper is organized as follows: Section 2 reviews the related works and summarizes
the existing path planning method for MANAL. Section 3 explains the problems for existing path
planning in the obstacle-present environment. Section 4 presents the OTPP approach. Section 5
provides an analysis of the performance of the Z-curve, and OTPP through simulations, and Section 6
contains our conclusions and future works.

2. Related Works
Path planning algorithms can be categorized into two, namely offline [7,8,12–18] and online
schemes [19–27]. The offline path planning schemes are based on conditional restrictions to determine
the globally optimal path because a MAN visits the path and the location of the broadcast has been
pre-planned. Thus, the computational capability of a MAN is low, and the number of broadcast times
of a MAN can be optimized, thereby facilitating the reduction of energy consumption of the MAN
and unknown nodes. The online path planning schemes are based on observable environments or
sensor deployment situations. A MAN dynamically adjusts the feasible path to satisfy the system
requirements, although online path planning has considerable flexibility. A MAN must have high
computing capability because it must make decisions (e.g., move path trajectory or broadcast timing)
during movements. In addition, the online path planning method can only provide the optimum
solution locally, thereby failing to guarantee that all the unknown nodes can receive three location
information of the MAN to estimate their locations [2]. Table 1 summarizes the comparison between
the offline and online path planning algorithms for MANAL.

Table 1. Comparison of path planning algorithms.

Localization
Path Planning Algorithm Type NM CC AC Accuracy
Method
SCAN [7], DOUBLE-SCAN [7],
Offline Single Range-based Low No Low
Hilbert [7]
Z-curve [8] Offline Single Range-based Low Yes High
CIRCLES and S-CURVES [12] Offline Single Range-based Low No Low
MACL [13] Offline Single Range-based Low No Low
GMAN [14] Offline Multiple Range-based Low Yes High
LMAT [15] Offline Single Range-based Low Yes High
PI [16] Offline Single Range-free Low Yes High
SCAN-based [17] Offline Single Range-free Low Yes High
MMN [19] Online Multiple Range-based High No Low
Li [20] Online Single Range-based High No Low
MBL (ndc) [19] Online Single Range-based High No Low
MBAL [22] Online Single Range-based High No Low
DREAMS [23] Online Single Range-based High No Low
Visibility Binary Tree (VBT) [24] Online Single Range-based High No Low
BRF and BTG [25] Online Single Range-based High No Low
Snake-like [26] Online Single Range-based High No Low
Virtual Ruler [27] Online Multiple Range-based High No Low
Note: NM—Number of MAN; CC—Communication Cost; AC—Area Full Coverage.

2.1. Offline Path Planning Algorithm

At present, most offline path planning algorithms assume an obstacle-free environment [7,12–19].
In [14], MAN comprises three anchor nodes and is based on the SCAN path trajectory to visit the entire
Sensors 2018, 18, 889 4 of 19

sensing field. However, the construction cost of a MAN is substantially high. Han et al. proposed
a localization algorithm with a mobile anchor node based on trilateration (LMAT), where the MAN
moves in the trajectory of an isosceles triangle and sets the three corners of this triangle as the beacon
points; thereby, solving the beacon point collinearity problem [15]. In [7], the three path planning
schemes, namely SCAN, DOUBLE-SCAN, and Hilbert are compared. In the SCAN scheme, a MAN
moves in a straight line along the X-axis (or Y-axis). Thus, the SCAN method covers the entire
sensing field, and is convenient to implement. However, the beacon point collinearity problem remains
unresolved in this scheme. In the DOUBLE-SCAN scheme, a MAN travels in straight lines along
the x- and y-axes. Compared to SCAN, DOUBLE-SCAN has excessively long path lengths, thereby
resulting in high energy consumption in the MAN and unknown nodes. Hence, the Hilbert method
divides a sensing field into 4n sub-squares, and the central point of each sub-square is linked by 4n line
segments, thereby generating non-collinear beacon points for MANs. Compared to the SCAN and
DOUBLE-SCAN schemes, the Hilbert scheme substantially reduces the total path length of a MAN.
The CIRCLES and S-CURVES path-planning schemes were proposed in [12] to address the
collinearity problem. In the CIRCLES scheme, the center of the sensing field is treated as the center of
a circle, and a trajectory of a MAN is constructed thereafter from a few concentric circles of varying
dimensions. The S-CURVES method is based on SCAN, although it alters the linear paths of a MAN into
“S” shaped curves. Compared with SCAN, DOUBLE-SCAN, and Hilbert, the CIRCLES and S-CURVES
methods can provide higher localization accuracy to the unknown nodes. Hu et al. proposed a scheme
called mobile anchor centroid localization (MACL) [13], which is similar to CIRCLES because the
former also uses the center of the sensing field as a center. However, MANs are assigned a spiral
trajectory instead, thereby resulting in shorter total path lengths than those of the CIRCLES method.
The aforementioned offline path planning schemes assume an obstacle-free environment; however,
only a few studies discuss the obstacle-present environment. Although Z-curve [8], perpendicular
intersection (PI) scheme [16], and SCAN-based scheme [17] have been proposed as solutions for the
obstacle-present problem, these schemes cannot guarantee that the unknown nodes in the area can
receive three location-information from the MAN for localization. Guo et al. proposed a path planning
scheme called PI [16], where a MAN is installed on a rotating arm and continuously broadcasts location
information to unknown nodes based on RSSI and the MAN rotation angle. PI uses perpendicular
intersection geometric relationship to calculate the location of the unknown nodes. In this method,
the MAN requires continuous broadcast location information, thereby resulting in the high energy
consumption of the MAN and unknown nodes. In [17], a range-free localization scheme is used,
whereas a new path planning scheme based on SCAN is proposed in [18]. In view of the obstacle
blocking problem, Ou et al. [17] propose the concept of using virtual keys, which can predict the
location of the beacon point. SCAN-based schemes and PI exhibit identical drawbacks, and the MAN
requires continuous broadcast location information. Rezazadeh et al. proposed a path planning method
called Z-curve [8] to construct a “Z”-shaped path trajectory. The Z-curve can solve the beacon point
collinearity problem, and is more effective than SCAN, Hilbert, and CIRCLES in terms of localization
accuracy and time costs.

2.2. Online Path Planning Algorithm

The benefits of the online path planning algorithms are their environmental adaptability and
flexibility because they do not require any prior knowledge of the environment. In addition, in the
online path planning algorithms, the MAN’s path trajectory depends on the density of requests sent
by unknown sensor nodes, and the MAN does not visit the entire area; therefore, online path planning
algorithms are advantageous in terms of rapid positioning and power savings.
However, they are not suitable for applications that need full localization coverage. Online path
planning algorithms are conducive to specific environments where sensors are distributed in
special shapes (e.g., “C”- or “L”-shaped distributions) [19–21] and not isolated by obstacles [21–23].
However, for most applications in which unknown sensors are evenly distributed, online path
Sensors 2018, 18, 889 5 of 19

planning algorithms fail to guarantee full coverage of sensors’ localization because the MAN’s path
trajectory is not deterministic and depends on the density of requests sent by unknown sensor
nodes [21,24,25]. In addition, the location of beacon points is not globally optimal, and therefore
causes the unknown nodes to receive insufficient and inappropriate location messages, resulting in
collinearity problems [26]. Therefore, online and offline path planning algorithms are applied to
different scenarios.
Moreover, the online path planning algorithms are costly on hardware and communication.
Hardware-wise, to implement online path planning algorithms, the MAN requires additional
equipment. In [20], the MAN must configure directional antennas to receive and transmit signals in
different directions and communicate with the unknown nodes within different directions in a region.
In [27], the MAN was fixed with a pair of ultrasound transmitters. Each unknown node was assumed
to be equipped with an ultrasound receiver at the two ends of a vehicle. Thus, the MAN behaved
as a virtual ruler that traveled in the monitoring area to provide distance measurements to pairwise
unknown nodes. The disadvantage of MAN in this case, is that it must configure a pair of ultrasound
transmitters. In [19], a new method of mobile anchor group consisting of five mobile beacons, called
maximum multi-hop of nodes (MMNs) was proposed. However, the MAN group compared to
a single MAN is more expensive owing to the communication cost (CC) because unknown nodes
proactively send a localization “request” message to the MAN in the process of positioning. Therefore,
the unknown nodes consume a higher CC than the offline path planning algorithms [19–21].
Finally, a MAN must have high computing ability because it must make decisions (e.g., move
paths or broadcast timing) during movements to deal with various situations, such as [21], where nodes
would be clustered first, and MAN would visit these cluster heads based on the genetic algorithm.
However, the computing complexity of MBL increases with the number of unknown nodes.
In this paper, we propose an offline path planning algorithm, which ensures that all the unknown
nodes can receive three location-information from the MAN, and which optimizes the number of
beacon points; therefore the MAN broadcast packet times and unknown node received packet times
are reduced.

3. Overview

3.1. Preliminary
Table 2 presents the terminologies used in this paper.

Table 2. Terminologies.

Terminologies Definition and Description

The ordinary sensor nodes are used to finish the sensing data collection.
The ordinary nodes are assumed to have less battery energy and keep silent
Unknown (Ordinary) Node
state unless the broadcast beacon packet from MAN wakes them up, then
estimate their coordinates.
The node with more energy, larger communication and mobility, is aware of
Mobile Anchor Node (MAN) its position by GPS or manually placed to give help to locate the
ordinary nodes.
MANAL Mobile-anchor-node-assisted localization.
Beacon Packet (packet) A packet with position information broadcast by MAN.
Beacon Point The location where the mobile anchor node broadcasts packets.
Basic Square The minimum sensing field in OTPP.
Sub Square The basic unit of basic square
k-covered area An area can receive k non-collinear beacon packets from the MAN
OTPP Obstacle-tolerant path planning
Sensors 2018, 18, 889 6 of 19

Definition 1. Basic square. According to References [7,8], any N × N region can be partitioned into k basic
squares. Each basic square must be divided into four sub-squares, Sqh , with a side length of d; where h ∈ {1, 2,
3, 4}, as shown in Figure 2.
Sensors 2018, 18, x FOR PEER REVIEW 6 of 19
The locations of the beacon points of a MAN in a basic square are denoted by c0 , c1 , c2 , c3 , and c4 ,
in Figure 2. Lemmaof1the
The locations ensures that
beacon all the
points of unknown
a MAN innodes
a basicinsquare
a basicare
square can by
denoted receive
c0, c1,at
c2,least three
c3, and c4,
position
in Figurepackets
2. Lemma from the MAN.
1 ensures that all the unknown nodes in a basic square can receive at least three
position packets from the MAN.
Lemma 1. An unknown node, si , (i ∈ {1,2, . . . , n}), in a basic square can receive at least three position packets
i, (i ∈
p
from the MAN
Lemma 1. An ifunknown if CR s≥
and only node, (5/2 )d, where
{1,2,..,n}), in CR represents
a basic squarethe
cancommunication
receive at leastrange
threeofposition
the MAN [7,8].
packets
from the MAN if and only if ≥ (5/2) , where CR represents the communication range of the MAN
[7,8].
Proof. This lemma originates from reference [7] where the detailed proof can be found. Here, we
give a brief proof. Based on the localization method, an unknown node, s1 , inside Sq3 can be
Proof. This lemma originates from reference [7] where the detailed proof can be found. Here,pwe give
localized only if it can receive three non-collinear position packets from the MAN. As CR ≥ (5/2)d,
a brief proof. p Based on the localization p method, an unknown node, p s1, inside Sq3 can be localized only
dist(c0 , s1 ) ≤ (5/2)d, dist(c3 , s1 ) ≤ (5/2)d, and dist(c4 , s1 ) ≤ (5/2)d, s1 can receive the position
if it can receive three non-collinear position packets from the MAN. As ≥ (5/2) , dist(c0, s1) ≤
packets broadcast by the MAN at locations c1 , c3 , and c4 , thereby being localized, as illustrated in
(5/2) 2., dist(c
Figure 3, s1) ≤
Similarly, all(5/2) , and dist(c
the unknown nodes si (i (5/2)
4, s1) ≤
∈ {1, 2,, s..,
1 can receive the position packets broadcast
n}) that are inside Sq1 , Sq2 , and Sq4 can
by the MAN
be localized. at locations c 1, c3, and c4, thereby being localized, as illustrated in Figure 2. Similarly, all

the unknown
The farthest nodes si (i ∈ {1,
location from2, ..,the
n})center
that are ofinside Sq1, square
the basic Sq2, andisSq 4 can be localized.
the vertex of the square. Assume
The farthest location from the center of the
the unknown node s1 located at the vertex of the square can receive basic square is the vertex of the
at least square.
three Assume
position the
packets
unknown node s 1 located at the vertex of the square can receive at least three position packets from
from the three nearest beacon points c1 , c3 , and c4 , the communication range CR must be larger than
the three nearest beacon
, s )], dist(c points c1, c3, and c4, the communication range CR must be p larger
(5/2)than
Max[dist(c
p 0 1 3 , s1 ), dist(c4 , s1 )] and Max[dist(c1 , s1 )], dist(c3 , s1 ), dist(c4 , s1 )] = d so
Max[dist(c
CR ≥ (5/2)d. 0, s1)], dist(c3, s1), dist(c4, s1)] and Max[dist(c1, s1)], dist(c3, s1), dist(c4, s1)] = (5/2) so ≥
(5/2) . □

Figure 2. Communication
Communication radius,
radius, CR,
CR, must
must bebe larger
larger than
than or
or equal to (5/2)
equal to (5/2)d to
to ensure
ensure that
that all
all the
p
Figure 2. the
unknown
unknown nodes in aa basic
nodes in basic square
square receive
receive at
at least
least three
three position
position packets
packets from
from the
the MAN.
MAN.

3.2.
3.2. Problem
Problem Description
Description
In an obstacle-present
In an obstacle-presentenvironment,
environment,assumeassumethat
thatananobstacle
obstacleexists
exists between
between s1 sand
1 and c1 as well as
c1 as well as s1
sand
1 and c2; then, s1 will receive only the position packets broadcast by the MAN at beacon points c3 and
c2 ; then, s1 will receive only the position packets broadcast by the MAN at beacon points c3 and c4 ,
cas
4, as shown in Figure 3.
shown in Figure 3.
Without loss of
Without loss of generality, in this
generality, in this work,
work, an
an arbitrary
arbitrary convex
convex polygonal obstacle is
polygonal obstacle is converted
converted into
into
aa rectangle,
rectangle, as as shown
shown in in Figure
Figure 4.
4. LLv is
is the
the edge
edge of
of the
the obstacle
obstacle and
and is
is parallel
parallel to
to the
the Y-axis;
Y-axis; Wv is
W is the
the
v v
edge of the obstacle and is parallel to the X-axis. According to the relationship among Lv, Wv, and d,
obstacles may be classified into the following three types:

Type I: < and < .

Type II: < ≤ .
TypeIII: < ≤ .
Sensors 2018, 18, 889 7 of 19

edge of the obstacle and is parallel to the X-axis. According to the relationship among Lv , Wv , and d,
obstacles may be classified into the following three types:

 Type I : Lv < d and Wv < d.

Type II : Lv < d ≤ Wv .
 Type III : W < d ≤ L .

v v
Sensors 2018, 18, x FOR PEER REVIEW 7 of 19
Sensors 2018, 18, x FOR PEER REVIEW 7 of 19

Figure 3.
3. The
Figure 3. The unknown node
node sss111 is
unknown node is blocked
blocked by an obstacle and thus cannot receive
receive aaa minimum of
minimum of three
of three
three
Figure The unknown by an obstacle and thus cannot receive minimum
position packets
positionpackets
position from
packetsfrom the
fromthe MAN.
theMAN.
MAN.

Figure
Figure 4.
4. For
For simplicity, rectangles
simplicity, rectangles are
rectangles are used
are used to
usedto bind
tobind obstacles
bindobstacles in
obstaclesin an
inan obstacle-filled
anobstacle-filled scenario.
obstacle-filledscenario.
scenario.
Figure 4. For simplicity,

k-covered areas
k-covered areas
k-covered are
areas are defined
are defined to
defined to demonstrate
to demonstrate the
demonstrate the effects
the effects of
effects of an
of an obstacle
an obstacle on
obstacle on unknown
on unknown nodes
unknown nodes in
nodes in
in
different areas.
different areas.
different areas.

Definition
Definition2.
Definition 2. k-covered
2. k-covered area.
k-coveredarea. An
area. An area
An area is
area is k-covered
k-covered ififif all
is k-covered all the
all the unknown
the unknown nodes
unknown nodes within
nodes within this
within this area
this area can
area can receive
receive kkk
can receive
non-collinear position
non-collinear position
non-collinear packets
position packets from
packets from the
fromthe MAN.
theMAN.
MAN.

The following
The following
The two
following two scenarios
two scenarios are
scenarios are feasible
are feasible in
in aaa basic
feasible in basic square,
basic square, in
square, in which
in which the
which the unknown
the unknown nodes
unknown nodes that
nodes that are
that are
are
present in
presentin
present areas
inareas where
areaswhere k < 3 cannot
wherekk<< 33 cannot be
cannot be localized:
be localized:
localized:
1.
1. AA two-covered area
area exists
exists inside
inside aaa basic
basic square;
square; the
the unknown
unknown nodesnodes within
within this
this area can receive
1. A two-covered
two-covered area exists inside basic square; the unknown nodes within this area can
area can receive
receive
only
only two position packets from the MAN, as illustrated in Figure 5a.
only two position packets
two position packets from
from the
the MAN,
MAN, as as illustrated
illustrated inin Figure
Figure 5a.
5a.
2.
2. One-covered
One-covered and
and zero-covered
zero-covered areas
areas are
are present
present within
within aa basic
basic square;
square; the unknown
the unknown nodes
unknown nodes within
nodes within
within
2. One-covered and zero-covered areas are present within a basic square; the
these
these areas
areas can
can receive
receive only
only one
one and
and zero
zero packets
packets from
from the
the MAN,
MAN, as
as shown
shown in
in Figure
Figure 5b,c.
5b,c.
these areas can receive only one and zero packets from the MAN, as shown in Figure 5b,c.
Evidently,
Evidently, additional
additional beacon
beacon points
points are
are required
required toto ensure
ensure that
that the
the entire
entire basic
basic square
square isis aa three-
three-
covered area.
covered area.
Sensors 2018, 18, 889 8 of 19

Evidently, additional beacon points are required to ensure that the entire basic square is a
three-covered
Sensors 2018, 18, x area.
FOR PEER REVIEW 8 of 19

(a) (b) (c)

Figure
Figure 5.
5. Obstacle
Obstacleshadowing
shadowingresults
resultsinin
the presence
the of of
presence non-three-covered areas
non-three-covered in ainbasic
areas square:
a basic (a)
square:
two-covered area, (b) one-covered and zero-covered areas where < and W v ≥ d, and (c) one-
(a) two-covered area, (b) one-covered and zero-covered areas where Lv < d and Wv ≥ d, and
covered and zero-covered
(c) one-covered areas where
and zero-covered ≥ and
areas where L ≥ d<and. W < d.
v v

4. Algorithm of OTPP
4. Algorithm of OTPP

4.1.
4.1. Deployment
Deployment of
of Beacon
Beacon Points
Points in
in OTPP
OTPP
Table
Table 33 presents
presentsthe
theflow
flowofofhow
howtotodeploy
deploythe
the beacon
beacon points
points byby OTPP
OTPP which
which is coherent
is coherent withwith
the
the
following lemmas. The following lemmas prove that OTPP guarantees full coverage of localization of
following lemmas. The following lemmas prove that OTPP guarantees full coverage of
localization
unknown nodes.of unknown nodes.

Table
Table 3.
3. Deployment
Deployment of
of beacon
beacon points
points in
in OTPP.

OTPP Beacon Point Deployment Algorithm

OTPP Beacon Point Deployment Algorithm
Input: Given an × sensing field and the communication range of MAN, CR.
Input: Given
Output: an N × of
Deployment N beacon field and the communication range of MAN, CR.
sensingpoints
Output: Deployment of beacon points
1: Compute the maximum side length, d, of a basic square based on CR ≥√ 5/2 .
1: Compute the maximum side length, d, of a basic square based on CR ≥ 5/2d.
2: Compute the minimum number of basic squares to cover the sensing field.
2: Compute the minimum number of basic squares to cover the sensing field.
3:3: Segment
Segmentthe
thesensing
sensingarea
areatotomultiple
multiplebasic
basicsquares.
squares.
4:4: For
Foreach
eachbasic
basicsquare
squaresegmentation
segmentationininthe
thesensing
sensingarea
area
i:i: Analyze the composition (k-covered area) of the
Analyze the composition (k-covered area) of the basic basicsquare.
square.
ii: Add beacon points into the basic square according to k-covered
ii: Add beacon points into the basic square according to k-covered area. area.

Definition 3. Basic square with obstacles. Assume that each basic square contains a maximum of one obstacle
Definition
with 3. Basic
sides of lengths Ly square
and Wy,with obstacles.
and the Assume
center of the that
obstacle lieseach basic
on the square
position contains
of the beaconapoint
maximum
c0. of one
obstacle with sides of lengths Ly and Wy , and the center of the obstacle lies on the position of the beacon point c0 .
To determine the optimal placement of beacon points in a basic square with obstacles, we adopt
To determine the
divide-and-conquer optimal placement
approach. of beacon
We first divide points
a basic in a basic
square to foursquare with obstacles,
sub-squares becausewe it adopt
is the
divide-and-conquer approach. We first divide a basic square to four
maximum number of sub-squares which are identical and with minimum area size. To minimize sub-squares because it is the
the
maximum
number of number of sub-squares
beacon points, which
the optimal are identical
location and with
of beacon pointminimum
should bearea ablesize. To minimize
to cover all the
the number
unknown of beacon
nodes points, theThe
in a sub-square. optimal
unionlocation
of optimal of beacon
locationspoint
are should be As
called reg. ablelong
to cover
as thereall are
the
unknown nodes in a sub-square. The union of optimal locations are called reg.
three non-collinear locations in reg selected as beacon points, all the unknown nodes can be localized As long as there are
three
in non-collinear
a sub-square. locations inwe
Accordingly, regcan
selected
obtainas beacon
the points,
following all the2.unknown nodes can be localized
Lemma
in a sub-square. Accordingly, we can obtain the following Lemma 2.
Lemma 2. If all the unknown nodes si in the sub-square can be localized, then, three non-collinear beacon points
Lemma
need If all theand
to be2.present, unknown nodesbeacon
all the three si in the sub-square
points will liecan be localized,
within reg1. then, three non-collinear beacon points
need to be present, and all the three beacon points will lie within reg1 .
Proof. The localization of an unknown node’s position requires at least three non-collinear beacon
points [2]. Lemma 1 indicates that all the unknown nodes within the basic square can be localized.
Therefore, the communication range (CR) of the MAN and unknown nodes are at least (5/2) .
Since the four sub-squares are identical, we consider sub-square Sq1 as representative. In sub-square
Sq1, the maximum distance is √2 < (5/2) . Hence, three beacon points are required to localize all
the unknown nodes when no obstacles are present. When obstacles are present, to find reg, we extend
the sides of the obstacle and hence the sub-squares are divided into three regions. reg is then bounded
Sensors 2018, 18, 889 9 of 19

Proof. The localization of an unknown node’s position requires at least three non-collinear beacon
points [2]. Lemma 1 indicates that all the unknown nodes within the basic square can be localized.
p
Therefore, the communication range (CR) of the MAN and unknown nodes are at least (5/2)d. Since
the four sub-squares are identical,
√ we consider sub-square Sq1 as representative. In sub-square Sq1 ,
p
the maximum distance is 2d < (5/2)d. Hence, three beacon points are required to localize all the
unknown nodes when no obstacles are present. When obstacles are present, to find reg, we extend
Sensors 2018, 18, x FOR PEER REVIEW
the
9 of 19
sides of the obstacle and hence the sub-squares are divided into three regions. reg is then bounded
by
by the
the two
two lines
lines from
fromthetheextended
extendedsides
sidesofofthe
theobstacle.
obstacle.The Theother
othertwo regions,AA11 and
tworegions, andAA22,, cannot
cannot
guarantee to cover all the unknown nodes in Sq . To cover all unknown nodes
guarantee to cover all the unknown nodes in Sq1. To cover all unknown nodes in A1, the locations of
1 in A 1 , the locations
of three
three beacon
beacon points
points areare bounded
bounded to the
to the areaarea on the
on the left left of obstacle
of the the obstacle
whichwhich
are Aare A1reg
1 and and regcover
1. To 1 . To
cover all unknown
all unknown nodes nodes
in Ain A2 , locations
2, the the locations of three
of three beacon
beacon points
points arebounded
are boundedtotothe
the area
area above
above thethe
obstacle which are A and reg . The intersection of the two areas is reg, as shown
obstacle which are A22 and reg11. The intersection of the two areas is reg, as shown in Figure 6. □ in Figure 6.

Figure 6.
Figure 6. The
The layout
layout of
of beacon
beacon points
points in
in the
the OTPP
OTPP scheme
scheme and
and the
the dots
dots are
are the
the possible
possible locations
locations of
of
beacon points.
beacon points.

We may obtain the following (Corollary 1) to enable the localization of all the unknown nodes si
within
Wethemayentire basic
obtain thesquare.
following (Corollary 1) to enable the localization of all the unknown nodes si
within the entire basic square.
Corollary 1. Given that each sub-square Sqh (h ∈ {1, 2, 3, 4}) requires three beacon points, twelve beacon points
should be added
Corollary to cover
1. Given the entirety
that each of aSq
sub-square basic square.
h (h ∈ {1, 2, 3, 4}) requires three beacon points, twelve beacon points
should be added to cover the entirety of a basic square.
Beacon points should be added in reg1–reg4 owing to obstacle blocking and to ensure that each
sub-square
Beacon is a three-covered
points should be addedarea. in
For
rega1 –reg
two-covered area, we can further obtain the following
4 owing to obstacle blocking and to ensure that each
Lemma 3.
sub-square is a three-covered area. For a two-covered area, we can further obtain the following
Lemma 3.
Lemma 3. When a basic square contains two-covered areas, all the unknown nodes in a basic square can be
localized,3.and
Lemma a minimum
When of fourcontains
a basic square beacon points should be
two-covered added.
areas, all the unknown nodes in a basic square can be
localized, and a minimum of four beacon points should be added.
Proof. When a basic square contains two-covered areas; the basic square can be divided into 12
regions based on the sides of the obstacle (i.e., reg1–reg4 and A1–A8), and two-covered areas lie within
Proof.
Ai, i ∈ {1,When a basic
2, …, square contains
8}. Therefore, addingtwo-covered
a beacon point areas; the1basic
to reg square
results can
in the A1be
anddivided into 12
A2 regions regions
receiving
based
three on the sides
beacon of thethereby
packets, (i.e., reg1 –reg
obstacleenabling and A1 –A8 ), of
the4 localization andthe
two-covered
unknown areas
nodeslieinwithin i , i ∈ {1,
theseAregions.
2, . . . , 8}. Therefore, adding a beacon point to reg results in the A and A regions
Similarly, the addition of a beacon point to each of reg2–reg4 will enable each of the A3–A8 regions
1 1 2 receiving threeto
beacon packets, thereby enabling the localization of the unknown nodes in these
receive three beacon packets and enable the localization of all the unknown nodes in these regions. regions. Similarly,
the addition
Therefore, of abeacon
four beaconpoints
point should
to eachbe reg2 –reg
of added will enable
to4 enable each of theof
the localization A3all
–Athe
8 regions
unknownto receive
nodes
three beacon packets and enable the
within the entirety of the basic square. □ localization of all the unknown nodes in these regions. Therefore,
four beacon points should be added to enable the localization of all the unknown nodes within the
entirety
Lemma of 4. the
Whenbasic square.
a basic square contains zero-covered and one-covered areas, twelve beacon points should be
added to the entirety of the basic square.

Proof. The basic square contains zero-covered and one-covered regions, and can be divided into
twelve regions based on the sides of the obstacle (i.e., reg1–reg4 and A1–A8), in which zero- and one-
covered areas are present in A1, A4, A5, and A8. Meanwhile, two-covered regions are present in A2, A3,
Sensors 2018, 18, 889 10 of 19

Lemma 4. When a basic square contains zero-covered and one-covered areas, twelve beacon points should be
added to the entirety of the basic square.

Proof. The basic square contains zero-covered and one-covered regions, and can be divided into
twelve regions based on the sides of the obstacle (i.e., reg1 –reg4 and A1 –A8 ), in which zero- and
one-covered areas are present in A1 , A4 , A5 , and A8 . Meanwhile, two-covered regions are present in A2 ,
A3 , A6 , and A7 . Hence, the addition of three beacon points to reg1 will enable the A1 and A2 regions to
receive three beacon packets and the localization of the unknown nodes within these regions. Similarly,
the addition of three beacon points to each of reg2 –reg4 enables each of A3 –A8 to receive three beacon
packets. Hence, the addition of twelve beacon points enables the localization of all the unknown nodes
in the basic square.

However, in Lemmas 3 and 4, which consider only the number of beacon points that should be
deployed within each sub-square and do not consider the optimal deployment location, the excess
number of beacon points will cause the MAN and unknown nodes to incur high energy cost.

Lemma 5. When a basic square contains two-covered areas, a particular layout will enable two beacon points to
completely cover the entirety of the basic square.

Proof. Lemma 3 indicates that the addition of a beacon point to each sub-square will enable complete
coverage of the entirety of a basic square. However, this placement method does not consider the area
that can be covered by a beacon point and only places a beacon point in each sub-square. Therefore,
p
if we consider CR ≥ (5/2)d, then, A1 , A2 , A3 , and A8 can be covered straightforwardly by placing
a beacon point at the position m1 . In addition, all the unknown nodes in A1 , A2 , A3 , and A8 will be able
to receive at least three beacon packets. Similarly, by placing a beacon point at position m2 , A4 –A7 will
be covered, and all the unknown nodes in A4 –A7 will receive at least three beacon packets. Therefore,
two beacon points should be added to completely cover the entirety of the basic square, as shown in
Figure 7a.

Lemma 6. When a basic square contains zero- and one-covered areas, a particular layout will enable six beacon
points to completely cover the entirety of the basic square.

Proof. Lemma 4 indicates that the addition of three beacon points to each sub-square will cover
the entirety of a basic square. However, this placement method does not consider the area that can
be covered by the beacon point and only places three beacon points in each sub-square. Therefore,
p
if we consider CR ≥ (5/2)d, then, placing a beacon point in each of m4 , m5 , and m6 will cover
reg1 , reg3 , A1 , A2 , A7 , and A8 . In addition, the unknown nodes in these regions will receive at least
three beacon packets. Similarly, adding a beacon point to each of m1 –m3 will cover reg2 , reg4 , and
A3 –A6 ; moreover, the unknown nodes in these regions will also receive at least three beacon packets.
Therefore, the addition of six beacon points will enable complete coverage over the entirety of the basic
square, as shown in Figure 7b.

We can further calculate the total number of beacon points required for OTPP, NBP , as described
by Lemma 7:

Lemma 7. In OTPP, the entire basic square contains only three-covered areas; the total number of beacon points,
NBP , required is five. The basic square contains two-covered areas; the total number of beacon points required is
 coverage of the entirety of a basic square. However, this placement method does not consider the
area that can be covered by a beacon point and only places a beacon point in each sub-square.
Therefore, if we consider CR ≥ (5/2) , then, A1, A2, A3, and A8 can be covered straightforwardly by
placing a beacon point at the position m1. In addition, all the unknown nodes in A1, A2, A3, and A8
will be able to receive at least three beacon packets. Similarly, by placing a beacon point at position
Sensorsm2018,
2, A418,
–A7889
will be covered, and all the unknown nodes in A4–A7 will receive at least three beacon11 of 19
packets. Therefore, two beacon points should be added to completely cover the entirety of the basic
square, as shown in Figure 7a. □
six; the basic square contains zero-covered and one-covered areas, and the total number of beacon points required
is ten.Lemma 6. When a basic square contains
 zero- and one-covered areas, a particular layout will enable six beacon

 5,
points to completely cover the entirety ofwhen three-covered
the basic square.
NBP = 6, when two-covered (1)
Proof. Lemma 4 indicates that the10,
addition of three beacon points
when one-coverd and zero-covered to each sub-square will cover the
entirety of a basic square. However, this placement method does not consider the area that can be
covered by the beacon point and only places three beacon points in each sub-square. Therefore, if we
Proof. In an CR
consider ≥ (5/2) , then,
obstacle-free environment, the total
placing a beacon number
point in eachofof beacon points,
m4, m5, and NBP
m6 will , required
cover reg1, regfor
3, A1OTPP
,
A2,(cA0 7,, cand
is five 1 , c2A, 8c , caddition,
. 3In 4 ). When thethe entire basic
unknown nodessquare
in thesecontains two-covered
regions will areas,
receive at least threewhich
beaconrequire
addingpackets.
two beaconSimilarly, adding
points, thea beacon point to each
total maximum of m1–m
number will cover
of3beacon reg2, required
points reg4, and Ais3–A
six6; (see
moreover,
Figure 7a).
Whenthe theunknown
basic square nodescontains
in these zero-covered
regions will also
andreceive at least areas,
one-covered three beacon
whichpackets.
require Therefore,
adding sixthe beacon
addition of six beacon points will enable complete coverage over the entirety of the basic square, as
points, the total maximum number of beacon points required is ten (see Figure 7b).
shown in Figure 7b. □

Sensors 2018, 18, x FOR PEER REVIEW 11 of 19

Lemma 7. In OTPP, the entire basic square contains only three-covered areas; the total number of beacon
points, , required is five. The basic square contains two-covered areas; the total number of beacon points
required is six; the basic square contains zero-covered and one-covered areas, and the total number of beacon
points required is ten.
5, when three-covered
= 6, when two-covered (1)
10, when one-coverd and zero-covered
(a) (b)

Proof. 7.
Figure In The
an obstacle-free environment,
results of beacon the total number
point simplification of beacon
in scenarios where (a) the, basic
points, required for OTPP
square is
contains
Figure 7. The results of beacon point simplification in scenarios where (a) the basic square contains
five (c0, c1, c2,areas
two-covered c3, c4).and
When
(b) the basic
entiresquare
basic square contains
contains two-covered
zero-covered areas, whichareas.
and one-covered require adding
two-covered
two beacon areas andthe
points, (b)total
the basic square
maximum contains
number zero-covered
of beacon and one-covered
points required areas.
is six (see Figure 7a). When
the can
We basicfurther
squarecalculate
contains zero-covered and one-covered
the total number areas, required
of beacon points which require addingNsix
for OTPP, beacon
BP, as described
4.2. Path points,7:thein
Planning
by Lemma total
OTPPmaximum number of beacon points required is ten (see Figure 7b). □

4.2. Path must

The MAN Planning in OTPP
visit all the positions of the beacon points if the beacon points c1 and c4
were designated as itsmust
The MAN firstvisit
(starting) point and
all the positions of thefinal
beacon(ending)
points if point positions
the beacon points c1forandvisitation.
c4 were In the
designated as its first (starting) point and final (ending) point positions for visitation.
obstacle-free environment, the visit order of the beacon points for MAN is c1 , c2 , c0 , c3 , and c4 . In the obstacle-
free environment, the visit order of the beacon points for MAN is c1, c2, c0, c3, and c4.
1. Basic square containing two-covered areas
When1. theBasic
basic square
squarecontaining
containstwo-covered areas areas, the MAN first visits c and c . From the four
two-covered 1 2
corners of an obstacle,
When the basic the proximate
square contains location
two-covered c2 isthe
fromareas, selected,
MAN firstwhich
visits cis an additional
1 and beacon point
c2. From the four
m1 , and itscorners of anposition
diagonal obstacle, the proximate
is set as m2 .location from c2the
Thereafter, is selected, which is
MAN visits man1
additional
and m 2 . beaconthe
Lastly, point
MAN visits
m1, and its diagonal position is set as m2. Thereafter, the MAN visits m1 and m2. Lastly, the MAN visits
c3 and c4 , ci.e., the MAN completes the travel on the entire basic square, as shown in Figure 8.
3 and c4, i.e., the MAN completes the travel on the entire basic square, as shown in Figure 8.

Figure 8. The trajectories of OTPP with basic square containing two-covered area. The red arrows are
Figure 8. The trajectories of OTPP with basic square containing two-covered area. The red arrows are
the trajectories while the shadow area is the obstacle.
the trajectories while the shadow area is the obstacle.
2. Basic square containing zero- and one-covered areas
When the basic square contains zero-covered and one-covered areas, the MAN verifies whether
the path from c1 to c2 is obstructed by an obstacle or if c1 to c2 has a straight path without being
obstructed by an obstacle. The MAN first visits c1 and c2. Among the additional beacon points mq (q =
1, 2, …, 6), the additional beacon point proximate to c2 is selected as the next beacon point to be visited
following the visit to m1–m6. Lastly, the MAN visits c3 and c4, as shown in Figure 9.
In another case, if the straight-line path from c1 to c2 is blocked by an obstacle, then, the positions
Sensors 2018, 18, 889 12 of 19

2. Basic square containing zero- and one-covered areas

When the basic square contains zero-covered and one-covered areas, the MAN verifies whether
the path from c1 to c2 is obstructed by an obstacle or if c1 to c2 has a straight path without being
obstructed by an obstacle. The MAN first visits c1 and c2 . Among the additional beacon points mq
(q = 1, 2, . . . , 6), the additional beacon point proximate to c2 is selected as the next beacon point to be
visited2018,
Sensors following
18, x FORthe visit
PEER to m1 –m6 . Lastly, the MAN visits c3 and c4 , as shown in Figure 9.
REVIEW 12 of 19

Sensors 2018, 18, x FOR PEER REVIEW 12 of 19

Figure 9. The straight-line path from c11 to

to c22 without
without blockage
blockage by
by an
an obstacle.
obstacle. The
The red
red arrows
arrows are the
trajectories while the shadow area is the obstacle.

In another case, if the straight-line path from c1 to c2 is blocked by an obstacle, then, the positions
of c2 and c3 are interchanged. The MAN first visits c1 and c2 . Thereafter, additional beacon points
mq , proximate to c2 are selected as the beacon points to be visited next. The sequential visits to the
additional
Figurebeacon points follow
9. The straight-line the
path order
from c1 tomc52, without
m6 , m4 , blockage
m3 , m1 , and m2obstacle.
by an . Lastly,The
the red
MAN arec3the
visits
arrows and c4 ,
trajectories
as shown while10.
in Figure theFigure
shadow 10area is the obstacle.
is considered as the result of the rotation of Figure 9 by 90◦ .

Figure 10. The straight-line path from c1 to c2 is blocked by an obstacle; therefore, swap the positions
of c2 and c3. The red arrows are the trajectories while the shadow area is the obstacle.

5. Performance Evaluation
The performance of OTPP is evaluated by three metrics (1) number of beacon points, (2) average
localization errors, and (3) localization coverage. In Section 5.1, we describe the three metrics and
Figure 10. The straight-line path from c11 to c22 is is blocked
blocked by
by an
an obstacle; therefore, swap the positions
compare three common localization techniques. The settings of experimental parameters are
of c22 and
and c3 . The red arrows are the trajectories while the shadow area
c3. The red arrows are the trajectories while the shadow area is
is the
the obstacle.
obstacle.
introduced in Section 5.2. Finally, we change the shapes of obstacles and the communication range
to see the impact of blocking rate and the impact of resolution. The results are presented in Section
5. Performance
5. Performance Evaluation
Evaluation
5.3.
The performance
The performance of of OTPP
OTPP isis evaluated
evaluated by by three
three metrics (1) number
metrics (1) number of of beacon
beacon points,
points, (2)
(2) average
average
localization
5.1. Evaluation errors,
Metricsandand(3) localization
Localization coverage.
Techniques In Section 5.1, we describe
localization errors, and (3) localization coverage. In Section 5.1, we describe the three metrics the three metrics and
compare three common localization techniques. The settings of experimental
and compare three common localization techniques. The settings of experimental parameters are parameters are
1. Number
introduced in of beacon
in Section points:
Section 5.2. The
5.2. Finally,
Finally,wenumber
wechange of
changethebeacon
theshapespoints stand
shapesofofobstacles
obstaclesfor the
and number of times that the
introduced and thethe communication
communication range
range to
to seeMAN
the broadcasts
impact of packets.rate
blocking Theandmorethe the timesofofresolution.
impact MAN broadcasting,
The results the
are more the packets
presented in the
Section
see the impact of blocking rate and the impact of resolution. The results are presented in Section 5.3.
5.3. unknown nodes receive. An increase in the number of packets received by an unknown node,
leads to a corresponding increase in power consumption per unknown node so that the
performance
5.1. Evaluation of entire
Metrics sensor networks,
and Localization decreases.
Techniques
2. Average localization error: Localization error is defined as the difference between the estimated
1. locations
Number ofand beacon points:
actual The number
locations of unknownof beacon
nodes.points stand
Equation (2)for the number
presents of times that
the localization errorthe
of
MAN broadcasts packets. The more the times of MAN broadcasting,
an unknown node , where ( , ) is the coordinate of the actual location and ( the more the packets
, )the
is
Sensors 2018, 18, 889 13 of 19

5.1. Evaluation Metrics and Localization Techniques

1. Number of beacon points: The number of beacon points stand for the number of times that the MAN
broadcasts packets. The more the times of MAN broadcasting, the more the packets the unknown
nodes receive. An increase in the number of packets received by an unknown node, leads to
a corresponding increase in power consumption per unknown node so that the performance of
entire sensor networks, decreases.
2. Average localization error: Localization error is defined as the difference between the estimated
locations and actual locations of unknown nodes. Equation (2) presents the localization error
of an unknown node si , where ( xi , yi ) is the coordinate of the actual location and (xesti , yesti ) is
the coordinate of the estimated location. Average localization error is presented in Equation (3)
where n is the total number of unknown nodes. The lower the average localization errors, the
better the performance of sensor networks:
q
error (si ) = ( xi − xesti )2 + (yi − yesti )2 (2)

∑in=1 error (si )

AvgError = (3)
n
3. Localization coverage: An unknown node can successfully do self-positioning as long as it receives
three or more packets from the MAN. The more the unknown nodes successfully self-positions,
the better the performance of sensor networks. Therefore, we define the localization coverage as
the ratio of the number of unknown nodes that receive three or more MAN packets to the total
number of unknown nodes [8].

There are three common localization techniques, weighted centroid localization (WCL) [28],
accuracy priority trilateration (APT) [8], and time priority trilateration (TPT) [8]. The descriptions are
as follows:

(A) Weighted Centroid Localization (WCL): WCL defines a weight function RSSij which is the RSS value
of the MAN at different beacon points to indicate the impact of different beacon point coordinates
bj (x, y). Equation (4) expresses how an unknown sensor, si calculates its own positions, P(si ) based
on averaging the coordinates of all Nr received beacon points.

N
∑ j=r 1 RSSij × b j
P ( si ) = N
(4)
∑ j=r 1 RSSij

(B) Time-Priority Trilateration (TPT): TPT calculation starts as soon as three localization packets are
received. This enables sensors to save power and the time they need to spend waiting. With these
localization packets, the following system of equations needs to be solved to obtain the localization
result. For computing the position of si (xi , yi ) in a 2-D space, the following system of equations
must be solved:  q
 dist(r1 , si ) = ( x1 − xi )2 + (y1 − yi )2


 q
P ( si ) = dist r , s
( 2 i) = ( x − x i )2 + ( y2 − y i )2 (5)

 q 2
 dist(r , s ) = ( x − x )2 + (y − y )2

3 i 3 i 3 i

(C) Accuracy-Priority Trilateration (APT): Accuracy-Priority Trilateration (APT) is very similar to

TPT. The difference between APT and TPT is that APT checks whether to start the trilateration
calculation within a period of time. If a sensor has received at least three localization packets,
it starts the trilateration calculation with three of the nearest keys.
 ( )= ( , )= ( − ) +( − ) (5)
( , )= ( − ) +( − )
(C) Accuracy-Priority Trilateration (APT): Accuracy-Priority Trilateration (APT) is very similar to TPT.
The difference between APT and TPT is that APT checks whether to start the trilateration
Sensorscalculation
2018, 18, 889 within a period of time. If a sensor has received at least three localization packets,
14 of 19
it
starts the trilateration calculation with three of the nearest keys.
In Figure
Figure 11,
11,the
theexperimental
experimentalresults
resultsshow
showthat the
that localization
the errors
localization of APT,
errors TPT,
of APT, andand
TPT, WCL are
WCL
0.265, 0.339,
are 0.265, and 0.434
0.339, CR, respectively
and 0.434 while while
CR, respectively their corresponding localization
their corresponding coverages
localization are similar.
coverages are
Therefore, we adoptwe
similar. Therefore, theadopt
APT the
technique in the following
APT technique experiments
in the following because APT
experiments is more
because APTaccurate
is more
than TPTthan
accurate and TPT
WCL.andTheWCL.
experimental parameterparameter
The experimental settings are shownare
settings in shown
Table 4.in Table 4.

Figure
Figure 11.
11. Comparison
Comparison of
of APT,
APT, TPT, and WCL techniques
techniques in
in terms
terms of
of localization
localization error
errorand
andcoverage.
coverage.

5.2. Simulation Setup

The first experiment aims to investigate the impact of blocking rate which is the ratio of the
obstacle area to the entire area. We increase the rate from 0% to 30%. The second experiment aims
to investigate the impact of the resolution, which is CR/d. We assume that the entire area size is
128 m × 128 m, and d is fixed at 16 m. The number of MAN is one. √ The total number of unknown nodes
is 500, and the communication range CR of the first experiment is 5 × d. The wireless channel model
is the log-normal model [8,16,29], as shown in Equations (6) and (7), where, Prec denotes the received
power, while Ptrans and Ploss denote the transmission power of the MAN and the power attenuation
during transmission, respectively. Xσ is the Gaussian random variable where the standard deviation
and mean are set to 2 and 0 according to the environment, which has different set parameters [16].
All the experiments are implemented in C++ programming language and all the results are the average
of 100 simulations. Table 4 shows the experimental parameters that reference the Z-curve [8], which is
the compared target in this paper:
Prec = Ptrans − Ploss (6)
 
d
Ploss (d) = Ploss (d0 ) + 10γ log + Xσ (7)
d0

Table 4. Experimental parameters.

Parameter Value
Field Size 32 m × 32 m, 128 m × 128 m, 1000 m × 1000 m
Number of Unknown Nodes (NS ) 500
Number of MAN √ 1
Communication range (CR) 5 d, d, 1.25d, 1.5d, 1.75d, 2d, 2.25d
Transmission Power (Ptrans ) 250 dBM
Path Loss (γ) 3.3
Power Loss PL (d0 ) 55 dB
d0 1
Simulation Run 100
Sensors 2018, 18, 889 15 of 19

5.3. Simulation Results

5.3.1. Impact of Blocking Rate

In an obstacle-free environment, OTPP performs as the Z-curve. The MAN travels the entire
basic square with five beacon points. However, in the obstacle-present environment, the number of
beacon points varies with the shapes of obstacles. Without loss of generality, shapes of obstacles are
categorized into three types which are described in Section 3.2. Table 5 displays the experimental
results which show the average number of beacon points needed by OTPP and Z-curve for different
types of obstacles. The average number of beacon points needed by OTPP is observed to be less than
that needed by the Z-curve in all obstacle-present environments.

Table 5. Average number of beacon points.

Types of Obstacle OTPP Z-Curve

Obstacle-free 5 5
Type I 6 8
Type II 10 16
Type III 10 16

Figure 12a is the result of comparing the Z-curve and OTPP, with respect to the impact of blocking
rate on
Sensors localization
2018, error.
18, x FOR PEER Both the localization errors of OTPP and Z-curve increase with an increase
REVIEW in
15 of 19
blocking rate. The higher the blocking rate, the higher the difference between the localization errors of
the Z-curve errors
localization and OTPP schemes.
of the Z-curve The difference
and is more obvious
OTPP schemes. when theisblocking
The difference rate is higher
more obvious when than
the
10%. When
blocking theisblocking
rate rate is10%.
higher than 30%, When
the difference between
the blocking theislocalization
rate errors reaches
30%, the difference a valuethe
between of
0.072 CR. errors reaches a value of 0.072 CR.
localization
Figure12b
Figure 12b is the
theresult
resultofofcomparing
comparing thethe
Z-curve and and
Z-curve OTPP, with respect
OTPP, to the impact
with respect to the of blocking
impact of
rate on localization
blocking coverage.coverage.
rate on localization The coverage of the Z-curve
The coverage of the decreases with an increase
Z-curve decreases in the blocking
with an increase in the
rate while
blocking OTPP
rate maintain
while almost fullalmost
OTPP maintain coverage. When the When
full coverage. ratio isthe
30%, the is
ratio difference
30%, theindifference
localization
in
coverage between
localization coveragethebetween
Z-curve theandZ-curve
OTPP reaches
and OTPP15%.reaches 15%.

(a) (b)

Figure
Figure12.
12.Impact
Impactof
ofblocking
blockingrate
rateon
on(a)
(a)localization
localizationerror
errorand
and(b)
(b)coverage
coverageratio.
ratio.

We
We then
then analyzed
analyzed the
the OTPP
OTPP in in terms
terms ofof the
the localization
localization accuracy
accuracy and
and coverage
coveragewith
withaadifferent
different
communication range by fixing d as 16 m and changing the resolutions (CR/d). Figure
communication range by fixing d as 16 m and changing the resolutions (CR/d). Figure 13a,b illustrate 13a,b illustrate
the
thelocalization
localizationaccuracy
accuracyand
andcoverage
coverage with
withdifferent resolution
different resolution values (CR/d
values = 1.25d,
(CR/d 1.5d, 1.5d,
= 1.25d, 1.75d,1.75d,
and
2d,
andrespectively), in OTPP.
2d, respectively), in OTPP.
In an obstacle-free environment
In an obstacle-free environment which
which is blocking
is blocking rate =rate = 0,
0, the the localization
localization accuracyaccuracy and
and coverage
coverage
ratio withratio with resolution
resolution = 1.25d is=lower
1.25d than
is lower than the
the others others its
because because its communication
communication range
range is less is
than
less which1.6
1.6d than which
violates violates
lemma lemmain1,insufficient
1, resulting resulting in insufficient
location messageslocation messages
received received
by unknown by
nodes.
unknown
When thenodes. Whenisthe
resolution resolution
higher than 1.6is higher
d, the than 1.6 d, theerror
localization localization error then
then decreases to decreases
0.2 CR and to 0.2
the
CR and the coverage ratio is 100%. As the blocking rate of obstacles increases, the beacon points are
added in OTPP according to the obstacle types and therefore shorten the distance between unknown
nodes and beacon points so the localization error with resolution 1.25d and blocking rate decreases
and the coverage ratio increases when blocking rate is under 5%. As the blocking rate increases to
more than 5%, the localization accuracy and coverage both increase. However, for those whose
communication range by fixing d as 16 m and changing the resolutions (CR/d). Figure 13a,b illustrate
the localization accuracy and coverage with different resolution values (CR/d = 1.25d, 1.5d, 1.75d, and
2d, respectively), in OTPP.
In an obstacle-free environment which is blocking rate = 0, the localization accuracy and
coverage ratio with resolution = 1.25d is lower than the others because its communication range is
Sensors 2018, 18, 889 16 of 19
less than 1.6 which violates lemma 1, resulting in insufficient location messages received by
unknown nodes. When the resolution is higher than 1.6 d, the localization error then decreases to 0.2
CR and theratio
coverage coverage ratio
is 100%. Asisthe
100%. As therate
blocking blocking rate of increases,
of obstacles obstacles increases,
the beacon the beacon
points arepoints
addedare in
added
OTPP in OTPP according
according to the obstacle
to the obstacle types andtypes and therefore
therefore shortenshorten the distance
the distance between between
unknown unknown
nodes
nodes and beacon
and beacon points points so the localization
so the localization errorresolution
error with with resolution 1.25d
1.25d and and blocking
blocking rate decreases
rate decreases and the
and the coverage
coverage ratio increases
ratio increases when blocking
when blocking rate 5%.
rate is under is under
As the 5%. As the rate
blocking blocking rate to
increases increases
more thanto
more
5%, thethan 5%, the accuracy
localization localization
andaccuracy
coverageandbothcoverage
increase. both increase.
However, However,
for those whose for those whose
resolution > 1.6d,
resolution > 1.6d, their
their localization errorlocalization
and coverage error
are and coverage
almost are almost
not affected not affected
by blocking rates.by blocking rates.
In
In summary,
summary, compared
compared to to Z-curve,
Z-curve, OTPP
OTPP hashas less
less number
number of of beacon
beacon points,
points, lower
lower localization
localization
errors and
errors and higher localization coverage. OTPP therefore performs better in obstacle-present
OTPP therefore performs better in obstacle-present
range CR≥≥ (5/2)
p
environments.
environments. When Whenthe thecommunication
communicationrange (5/2,)100%
d, 100% localization
localizationcoverage
coverage ratio is
ratio
achieved.
is achieved.

(a) (b)

Figure 13. Impact of blocking rate on (a) localization error and (b) coverage ratio.
13. Impact of blocking rate on (a) localization error and (b) coverage ratio.
Sensors 2018, 18, xFigure
FOR PEER REVIEW 16 of 19

5.3.2. Impact of Resolution

Figure 14a,b
Figure 14a,b show
show the
the impact
impact ofof resolution
resolution(CR/d)
(CR/d) on localization errors and coverage.
coverage. The field
size is
is 128
128 m
m× × 128 m and
and the
the communication
communicationrange rangeincreases
increasesfrom fromd dtoto2.25d.
2.25d.According
According toto Lemma
Lemma 1,
than (5/2)
(5/2)d ≅ ∼
= 1.6
p
1, the
the minimum
minimum communication
communication range,CR
range, CRmust
mustbebeequal
equaltotoororlarger
largerthan 1.6d toto ensure
ensure
that all unknown
unknown nodes
nodes ininaabasic
basicsquare
squarecancanreceive
receiveatatleast
leastthree
threeposition
position packets
packetsfrom
from thethe
MANMAN in
anan
in obstacle-free environment.
obstacle-free environment. In In
Figure
Figure14a, with
14a, anan
with increase
increase in in
resolution,
resolution,thethe
localization
localization errors of
errors
thethe
of OTPP
OTPPdecreases
decreases together
together with
withthat
thatofofthe
theZ-curve,
Z-curve,with
withbothbothstabilizing
stabilizing when
when the resolution is
than 1.6d.
larger than 1.6d. OTPP performs better than Z-curve with respect to localization errors even even though
though
enough to fully
the resolutions are not large enough fully cover
cover aa basic
basic square.
square. In Figure 14b, with an increase increase in
resolution, the
resolution, the coverage
coverage ofof the
the OTPP
OTPP increases
increases together
together with
with that
that of
of the
the Z-curve.
Z-curve. The coverage of the
OTPP stabilizes
OTPP stabilizeswhen
whenresolution
resolutionis sufficient (∼
is sufficient (≅ 1.6
= 1.6d) but)thebutcoverage
the coverage of the Z-curve
of the Z-curve does notdoes not
stabilize
stabilize
until is ≥ 2d. is ≥ 2 .
until the resolution
the resolution

(a) (b)

Figure 14. Impact of resolution on (a) localization error and (b) coverage ratio.
Figure 14. Impact of resolution on (a) localization error and (b) coverage ratio.

5.3.3. OTPP vs. Online Path Planning Algorithm

Figure 15a,b show the OTPP compared with online path planning algorithms in terms of
localization errors and coverage. The online path planning algorithms of interest are MMN [19], Li
[20], and MBL [21], with unknown nodes randomly deployed over a square sensing field of 1000 m ×
1000 m, and communication range, set to 35 m, 50 m, 70 m, and 90 m, respectively. It is observed that
 (a) (b)
Sensors 2018, 18, 889 17 of 19
Figure 14. Impact of resolution on (a) localization error and (b) coverage ratio.

5.3.3.
5.3.3. OTPP
OTPP vs.
vs. Online
OnlinePath
Path Planning
Planning Algorithm
Algorithm
Figure
Figure 15a,b
15a,b show
show the
the OTPP
OTPP compared
compared with with online
online path
path planning
planning algorithms
algorithms in in terms
terms of of
localization
localization errors
errorsandandcoverage.
coverage.The Theonline
onlinepath
pathplanning
planning algorithms
algorithms of interest areare
of interest MMN MMN [19],[19],
Li
[20], andand
Li [20], MBLMBL[21], with
[21], unknown
with unknownnodes randomly
nodes randomlydeployed over a square
deployed over sensing
a squarefield of 1000 m
sensing field of×
1000 m,
1000 m × and communication
1000 range, set torange,
m, and communication 35 m, set
50 m,to 70
35 m,m, and
50 m, 90 70
m, m,
respectively.
and 90 m, Itrespectively.
is observed that
It is
the localization
observed error
that the localization error≅ 0.2CR
of the OTPP is less∼
of the OTPP =than0.2CRthatis of MBL,
less thanMMN,
that ofand
MBL,Li. MMN,
AlthoughandtheLi.
localization accuracy of MMN and Li are close to that of OTPP, the localization
Although the localization accuracy of MMN and Li are close to that of OTPP, the localization coverage coverage of these
online
of thesepath planning
online algorithms
path planning is lessisthan
algorithms 50%,50%,
less than whereas
whereas OTPP OTPPcancan achieve≅∼
achieve 95%.
= 95%.AsAs the
the
communication range increases,
communication range increases,the
thecoverage
coverageareaareais is also
also improved
improved (e.g.,
(e.g., MBL);
MBL); however,
however, the MAN
the MAN will
will determine
determine the direction
the direction andofpath
and path of the trajectory
the trajectory based on based on the of
the number number of request
location locationmessages
request
messages from unknown
from unknown nodes, andnodes,
there isand there is nothat
no guarantee guarantee that all the
all the unknown unknown
nodes nodesthree
can receive can packets
receive
three packets
from the MAN. from the MAN.

(a) (b)
Figure 15. OTPP compared with online path planning in terms of (a) localization error and (b)
OTPPThe
Figure 15.ratio.
coverage compared with online pathrange
unit of communication planning in terms
is meter (m).of (a) localization error and (b) coverage
ratio. The unit of communication range is meter (m).

6. Summary and Future Work

Mobile-anchor-node-assisted localization (MANAL) has becomes more popular because it not
only saves the power consumption of unknown nodes in sensor networks, but also adapts to various
environments. The key problem of MANAL is the path planning of the MAN. Many path planning
schemes have been proposed for MANAL; however, most of them only adapt to obstacle-free
environments, which is not practical. In this paper, we propose an obstacle-tolerant path planning
algorithm, called OTPP. Compared to the other path planning algorithms for MANAL, OTPP ensures
that all unknown nodes are covered by at least three beacon points by using minimum beacon points
in a unit square area. The experimental results show that OTPP performs better than the Z-curve
as the total number of beacon points is reduced, localization errors lowered and coverage increased.
This implies that OTPP is more adaptive to obstacle-present environments. In the future, we will
consider more practical environments, such as obstacles with mobility or 3D spaces in path planning
for application in unmanned aerial vehicles or detection of marine resources.

Author Contributions: R.-G. Tsai and P.-H. Tsai conceived and designed the experiments; R.-G. Tsai performed
the experiments; P.-H. Tsai analyzed the data; and R.-G. Tsai and P.-H. Tsai wrote the paper.
Conflicts of Interest: The authors declare no conflict of interest.
Sensors 2018, 18, 889 18 of 19

Appendix

Table A1. Values of Figure 12 (Average localization error/Standard deviation).

Blocking Rate (%) OTPP Z-Curve

0% 0.072642/0.002473 0.072548/0.002266
5% 0.07423/0.002835 0.080395/0.003239
10% 0.075144/0.003157 0.080409/0.003066
15% 0.075517/0.003261 0.088176/0.002725
20% 0.082042/0.003323 0.109765/0.002742
25% 0.092433/0.004709 0.142201/0.005814
30% 0.099022/0.004542 0.162859/0.007393

Table A2. Values of Figure 13 (Average localization error/Standard deviation).

Blocking Rate (%) CR = 1.25d CR = 1.5d CR = 1.75d CR = 2d

0% 0.522736/0.011586 0.229271/0.004582 0.196305/0.002656 0.198529/0.002546
5% 0.288306/0.006262 0.243143/0.004956 0.260096/0.003309 0.265602/0.002699
10% 0.333439/0.006373 0.269417/0.004646 0.275515/0.003111 0.272391/0.002974
15% 0.374983/0.007359 0.304929/0.005416 0.290966/0.004587 0.270626/0.003400
20% 0.488731/0.009895 0.341111/0.005022 0.311428/0.004127 0.281862/0.003496
25% 0.658398/0.011234 0.383832/0.006595 0.316821/0.005444 0.286928/0.004905
30% 0.913244/0.013579 0.449443/0.007821 0.346859/0.006346 0.300728/0.005595

Table A3. Values of Figure 14 (Average localization error/ Standard deviation).

Resolution (CR/d) OTPP Z-Curve

1d 1.279855/0.021720 1.465825/0.025145
1.25d 0.463241/0.011367 0.766297/0.018753
1.5d 0.245316/0.007551 0.455239/0.013257
1.75d 0.262951/0.006284 0.283528/0.011040
2.d 0.242452/0.005415 0.259893/0.010880
2.25d 0.215318/0.005089 0.219514/0.008451

References
1. Akyildiz, I.F.; Su, W.; Sankarasubramaniam, Y.; Cayirci, E. A survey on sensor networks. IEEE Commun. Mag.
2002, 40, 102–114. [CrossRef]
2. Han, G.; Jiang, J.; Zhang, C.; Duong, T.; Guizani, M.; Karagiannidis, G. A Survey on Mobile Anchor Node
Assisted Localization in Wireless Sensor Networks. IEEE Commun. Surv. Tutor. 2016, 18, 2220–2243.
[CrossRef]
3. Isbitiren, G.; Akan, O.B. Three-Dimensional Underwater Target Tracking with Acoustic Sensor Networks.
IEEE Trans. Veh. Technol. 2011, 60, 3897–3906. [CrossRef]
4. Chaurasiya, V.K.; Jain, N.; Nandi, G.C. A novel distance estimation approach for 3D localization in wireless
sensor network using multi dimensional scaling. Inf. Fusion 2014, 15, 5–18. [CrossRef]
5. Shao, H.; Zhang, X.; Wang, Z. Efficient closed-form algorithms for AOA based self-localization of sensor
nodes using auxiliary variables. IEEE Trans. Signal Process. 2014, 62, 2580–2594. [CrossRef]
6. Xiong, H.; Chen, Z.; An, W.; Yang, B. Robust TDOA Localization Algorithm for Asynchronous Wireless
Sensor Networks. Int. J. Distrib. Sens. Netw. 2015, 2015, 1–10. [CrossRef]
7. Koutsonikolas, D.; Das, S.M.; Hu, Y.C. Path planning of mobile landmarks for localization in wireless sensor
networks. Comput. Commun. 2007, 30, 2577–2593. [CrossRef]
8. Rezazadeh, J.; Moradi, M.; Ismail, A.S.; Dutkiewicz, E. Superior Path Planning Mechanism for Mobile
Beacon-Assisted Localization in Wireless Sensor Networks. IEEE Sens. J. 2014, 14, 3052–3064. [CrossRef]
9. Garey, M.R.; Johnson, D.S. Computers and Intractability: A Guide to the Theory of NP-Completeness; W.H.
Freeman: New York, NY, USA, 1979; ISBN 0-7167-1045-5.
Sensors 2018, 18, 889 19 of 19

10. DeLeon, M. A study of sufficient conditions for Hamiltonian cycles. Rose-Hulman Undergrad. Math. J. 2000, 1, 6.
11. Garey, M.R.; Johnson, D.S.; Stockmeyer, L. Some simplified NP-complete problems. In Proceedings of the
6th Annual ACM Symposium on Theory of Computing, Seattle, WA, USA, 30 April–2 May 1974; pp. 47–63.
12. Huang, R.; Zaruba, G.V. Static path planning for mobile beacons to localize sensor networks. In Proceedings
of the 5th Annual IEEE International Conference on Pervasive Computing and Communications Workshops,
White Plains, NY, USA, 19–23 March 2007; pp. 323–330.
13. Hu, Z.; Gu, D.; Song, Z.; Li, H. Localization in wireless sensor networks using a mobile anchor node.
In Proceedings of the IEEE/ASME International Conference on Advanced Intelligent Mechatronics, Xi’an,
China, 2–5 July 2008; pp. 602–607.
14. Zhang, B.; Yu, F.; Zhang, Z. Collaborative localization algorithm for wireless sensor networks using mobile
anchors. In Proceedings of the 2nd Asia-Pacific Conference on Computational Intelligence and Industrial
Applications, Wuhan, China, 28–29 November 2009; pp. 309–312.
15. Han, G.; Xu, H.; Jiang, J.; Shu, L.; Hara, T.; Nishio, S. Path planning using a mobile anchor node based on
trilateration in wireless sensor networks. Wirel. Commun. Mob. Comput. 2011, 14, 1324–1336. [CrossRef]
16. Guo, Z.; Guo, Y.; Hong, F.; Yang, X.; He, Y.; Feng, Y.; Liu, Y. Perpendicular intersection: Locating wireless
sensors with mobile beacon. IEEE Trans. Veh. Technol. 2010, 59, 3501–3509. [CrossRef]
17. Ou, C.; He, W. Path planning algorithm for mobile anchor-based localization in wireless sensor networks.
IEEE Sens. J. 2013, 13, 466–475. [CrossRef]
18. Ssu, K.F.; Ou, C.H.; Jiau, H.C. Localization with mobile anchor points in wireless sensor networks. IEEE Trans.
Veh. Technol. 2005, 54, 1187–1197. [CrossRef]
19. Yang, Y.; Zhang, L. Dynamic path planning of mobile beacon for localization in wireless sensor network.
In Proceedings of the International Conference on Wireless Communications & Signal Processing (WCSP),
Hangzhou, China, 24–26 October 2013.
20. Li, J.; Yang, K. Improvement of path planning in mobile beacon assisted positioning. In Proceedings of
the 7th International Conference on Intelligent Computing, Liverpool, UK, 7–10 August 2011; Springer:
Heidelberg, Germany, 2011; pp. 309–316.
21. Zhao, F.; Luo, H.; Lin, Q. A mobile beacon-assisted localization algorithm based on network-density
clustering for wireless sensor networks. In Proceedings of the 5th International Conference on Mobile
Ad-hoc and Sensor Networks, Fujian, China, 14–16 December 2009; pp. 304–310.
22. Kim, K.; Lee, W. MBAL: A mobile beacon-assisted localization scheme for wireless sensor networks.
In Proceedings of the 16th International Conference on Computer Communications and Networks, Honolulu,
HI, USA, 13–16 August 2007; pp. 57–62.
23. Li, X.; Mitton, N.; Simplot-Ryl, I.; Simplot-Ryl, D. Dynamic beacon mobility scheduling for sensor localization.
IEEE Trans. Parallel Distrib. Syst. 2012, 23, 1439–1452. [CrossRef]
24. Rashid, A.T.; Ali, A.A.; Frasca, M.; Fortuna, L. Path planning with obstacle avoidance based on visibility
binary tree algorithm. Robot. Auton. Syst. 2013, 61, 1440–1449. [CrossRef]
25. Li, H.; Wang, J.; Li, X.; Ma, H. Real-time path planning of mobile anchor node in localization for wireless
sensor networks. In Proceedings of the International Conference on Information and Automation, Changsha,
China, 20–23 June 2008; pp. 384–389.
26. Mazinani, S.M.; Farnia, F. Localization in wireless sensor network using a mobile anchor in obstacle
environment. Int. J. Comput. Commun. Eng. 2013, 2, 438–441. [CrossRef]
27. Ding, Y.; Wang, C.; Xiao, L. Using mobile beacons to locate sensors in obstructed environments. J. Parallel
Distrib. Comput. 2010, 70, 644–656. [CrossRef]
28. Blumenthal, J.; Grossmann, R.; Golatowski, F.; Timmermann, D. Weighted centroid localization in
zigbee-based sensor networks. In Proceedings of the IEEE International Symposium on Intelligent Signal
Processing, Alcala de Henares, Spain, 3–5 October 2007; pp. 1–6.
29. Rappaport, T.S. Wireless Communications: Principles and Practice, 2nd ed.; Prentice-Hall: Upper Saddle River,
NJ, USA, 2001.

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).