Microprocessors and Microsystems 39 (2015) 1271–1278

Contents lists available at ScienceDirect

Microprocessors and Microsystems

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

Impacts of traveling paths on energy provisioning for industrial wireless

rechargeable sensor networks
Guangjie Han a,b,⇑, Aihua Qian a, Li Liu a, Jinfang Jiang a, Chuan Zhu a
a
Department of Information & Communication Engineering, Hohai University, China
b
Changzhou Key Laboratory of Special Robot and Intelligent Technology, Changzhou, China

a r t i c l e i n f o a b s t r a c t

Article history: Traditional Industrial Wireless Sensor Networks (IWSNs) are constrained by limited battery energy.
Received 30 April 2015 Recent breakthroughs in wireless power transfer have inspired the emergence of Industrial Wireless
Revised 7 July 2015 Rechargeable Sensor Networks (IWRSNs). IWRSNs usually contain one or more mobile chargers which
Accepted 12 July 2015
can traverse the network to replenish energy supply for sensor nodes. The essential problem in mobile
Available online 26 July 2015
energy provisioning is to find the optimum path along which the mobile chargers travel to improve
charging performance, prolong the battery lifespan of nodes and reduce the charging latency as much
Keywords:
as possible. In this paper, we introduce and analyze the impacts of four traveling paths, namely, SCAN,
IWRSNs
Mobile charger
HILBERT, S-CURVES(ad) and Z-curve on energy provisioning for IWRSNs. This evaluation aims to embody
Traveling path effective and essential properties that a superior traveling path should possess. Our simulations show
Energy provisioning that S-CURVES(ad) outperforms the other traveling paths in the lifetime of nodes and traveling efficiency.
And at the same time, it has relatively small charging latency.
Ó 2015 Elsevier B.V. All rights reserved.

1. Introduction
Nowadays, WSNs have been widely applied in many fields such
as smart home, disaster aid, environment monitoring and so on
[1,2]. Applying WSNs to industrial systems overcomes the
deficiency that traditional industrial wired automation systems
require expensive communication cables, which need to be
installed and maintained regularly [3–5]. Meanwhile, it as well
brings about many technical challenges due to the unique
characteristics of IWSNs, one of which is energy constraints of
nodes [6–8]. As is universally known, energy is the foundation
and premise to ensure various operation of sensor nodes. In prac-
tical applications, plenty of tiny sensor nodes are always deployed
in complicated scenarios which are powered by batteries, which
makes it impossible to replace all the dead nodes with new ones
especially in large scale sensing area.
Researchers attempt to solve the energy problem from the fol-
lowing two aspects: ‘‘increasing income’’and ‘‘reducing expendi-
ture’’. ‘‘Increasing income’’ means enabling nodes to harvest
energy from environment by converting mechanical, thermal,
⇑ Corresponding author at: Department of Information & Communication Engi-
neering, Hohai University, China.
E-mail addresses: hanguangjie@gmail.com (G. Han), bettyhhuc@gmail.com
(A. Qian), liulihhuc@gmail.com (L. Liu), jiangjinfang1989@gmail.com (J. Jiang),
dr.river.zhu@gmail.com (C. Zhu).
photovoltaic energy into electrical energy [9,10]. But the strength
acquired by this method is typically low and unsteady because it
highly relies on the environment [11]. ‘‘Reducing expenditure’’
usually means adopting efficient routing protocols to reduce the
energy consumption as much as possible [12,13]. Although
energy-efficient routing protocols can prolong the network lifetime
to some extent, the energy constraints problem cannot be solved at
its most fundamental by this way.
Recent advances of wireless energy transfer technology have
inspired the appearance of IWRSNs, in which the mobile chargers
can travel around the network and replenish energy for nodes
without interconnecting wires [14,15]. Since the traveling paths
of mobile chargers certainly have significant impacts on energy
provisioning, in this paper we focus on the impacts of different
traveling paths on energy provisioning for IWRSNs. Specifically,
we compare four traveling paths including SCAN, HILBERT,
S-CURVES(ad) and Z-curve in terms of the number of charging stop
locations, the length of traveling path, energy provisioning distri-
bution, alive nodes over time, charging latency and traveling
efficiency.
The main contribution of this paper is that we introduce and
compare the performance of four traveling paths in charging per-
formance comprehensively, analyze and conclude the characteris-
tics that an effective traveling path should possess from the
simulation results. The purpose of this work is to establish a solid

http://dx.doi.org/10.1016/j.micpro.2015.07.002
0141-9331/Ó 2015 Elsevier B.V. All rights reserved.
1272 G. Han et al. / Microprocessors and Microsystems 39 (2015) 1271–1278
foundation for traveling path planning of mobile chargers and pro-
vide guidance on designing more advanced charging traveling
paths.
The remainder of the paper is organized as follows. Section 2
summarizes the state-of-the-art in charging algorithms of
WRSNs. Section 3 describes the characteristics of SCAN, HILBERT,
S-CURVES(ad) and Z-curve in detail. Section 4 presents detailed
analysises of simulation and evaluation results. Section 5 draws
conclusions and summarizes contributions.
2. Related work
Recent breakthroughs of wireless charging technologies have
made great contributions to the emergence of IWRSNs. In this
paper, we study a IWRSN built from the industrial wireless identi-
fication and sensing platform (WISP) and commercial RFID readers.
The RFID readers serve as wireless chargers and can transmit RF
signals to sensors [14]. In the following, we refer to RFID readers
as chargers for simplicity. According to the wireless charging
model which is based on Friis’ free space equation [14], the rela-
tionship between the transmission power of transmitter P t and
the received power of receiver Pr can be formulized as follows:
 2
Gs Gr gPt k
Pr ¼ ð1Þ
Lp 4pðd þ bÞ
where Gs and Gr denote the antenna gain of the transmitter and the
receiver, respectively. g is rectifier efficiency, Lp refers to polariza-
tion loss, k is the wavelength of the signal, b is a parameter to adjust
the Friis’ free space equation for short distance transmission, and d
is the distance between the transmitter and the receiver. In this
case, the transmitter is a charger and the receiver is a sensor node.
To ease the description of the equation, we formulate formulation
2
a , where a ¼ Gs Gr gPt ð k Þ . According to the hardware
(1) by P r ¼ ðdþbÞ 2 Lp 4p
and environment parameters set and experiments conducted in
[12], a ¼ 4:32  104 ; b ¼ 0:2316. We can see from this formulation
that the received power of nodes decreases rapidly with the
increase of distance between the transmitter and the receiver.
That is, if the charger is too far away from the sensor node, the wire-
less charging power would be too low to be harvested. Therefore,
we assume that there exists a threshold of distance, denoted by
charging range R, beyond which the nodes cannot be charged. We
set R be 5 m in this paper, that is, only the nodes within 5 m of
the charger can be replenished energy.
So far there have been a certain number of researches on charg-
ing strategies in WRSNs [15–21]. Since there has no uniform stan-
dard of classification for these charging algorithms, here we
classify the existing charging algorithms with respect to the mobil-
ity state of chargers into two groups: (1) static chargers such as
methods proposed in [15,16], (2) mobile chargers such as methods
proposed in [17–20]. In consideration of the fact that the available
charging area of each charger is very limited, multiple static charg-
ers are needed to replenish energy for all nodes, which is costly
especially when it comes to large scale sensor networks. Hence
we only focus on algorithms under the scenario of one mobile
charger.
Constantinos Marios Angelopoulos et al. proposed three novel,
alternative protocols for wireless charging in [17]: Global
Knowledge Protocol (GKP), Limited Reporting Protocol (LRP) and
Reactive Trajectory Protocol (RTP). The authors introduced ‘‘criti-
cality’’ to capture a node’s importance in the network, which relies
on the traffic served by the node and the energy consumed by the
node. GKP is an on-line and centralized protocol which requires a
global knowledge of the state of the network, such as the criticality
of all nodes to decide the charging sequence. LRP is a distributed
protocol which selects some representative nodes in the network
to report their criticality and provide guidance for mobile charger.
LRP only uses limited network information. RTP reactively adapts
to energy shortage alerts by forming trees rooted at nodes. GKP
outperforms the other two strategies but also introduces great
communication overhead. Various simulation results demon-
strated that the three charging protocols proposed in this paper
have better performance than existing charging strategies in most
cases.
Zi Li et al. presented a set of heuristics to determine the wireless
charging strategies under various routing schemes in [18]. In con-
sideration of the fact that there exists mutual dependency between
wireless charging and routing protocols, the authors formulated
the problem of maximizing the sensor network lifetime as
ML-JRC by combining the aspects of routing and charging. Then
they proved the NP-hardness nature of the problem and derived
the theoretical maximum sensor network lifetime that could be
achieved with ML-JRC. But it is computationally complicated to
find an optimal solution as the network scale increases. Hence,
the authors proposed low-complexity heuristic solutions to deter-
mine the energy charging strategies for the mobile charger under
different routing schemes, Least Residual Energy First (LRE), Least
Estimated Lifetime First assuming Fixed Routes (LEL) and
Adaptive Energy Allocation with Dynamic Routes (AEA), which give
priority to residual energy, estimated lifetime and adaptive energy
allocation respectively. The heuristics were proved to be superior
in prolonging the sensor network lifetime.
Liguang Xie et al. proposed an energy-renewal approach with
wireless charging technology in [19]. The authors firstly introduced
a new concept called renewable energy cycle and studied how to
maximize the ratio of the charger’s vacation time over the whole
cycle time. Through the mathematical deduction, this paper proved
that the optimal traveling path for the mobile charger in each
renewable cycle is the shortest Hamiltonian cycle. In addition, this
paper took permanent operation of wireless sensor networks into
consideration and the solution proposed can make sensor net-
works immoral. While exploiting this traveling path could be inef-
ficient and lead to high computational complexity when the node
density increases.
In [20], Liang He et al. proposed an on-line path planning
scheme for the mobile charger, which laid theoretical foundation
for the on-demand mobile charging problem. Specifically, in this
work the authors analyzed the on-demand mobile charging prob-
lem using a Nearest-Job-Next with Preemption (NJNP) discipline
for the mobile charger where individual sensor nodes would send
charging requests to the charger when their energy run low. NJNP
takes both spatial and temporal properties of charging requests
from nodes into account, on which that the mobile charger will
select the next to-be-charged node is based. While NJNP is an
on-line protocol which poses considerable communication costs
of the nodes and it is not suitable for large-scale and dense sensor
networks.
While pioneering works on the mobile charging problem
mainly focus on dynamic path planning of the mobile charger,
which have high requirements on the network information and
may cause great communication overhead during the energy pro-
visioning procedure, especially for those on-line charging algo-
rithms. Such charging algorithms could be inefficient in relatively
dense IWRSNs because moving back and forth frequently in the
network brings about long traveling length. We note that energy
provisioning using a mobile charger in WRSNs shares something
in common with localizing ordinary nodes with the help of a
mobile beacon node, in that they both require a mobile object to
traverse the whole network and stop at certain locations to serve
the sensor nodes (replenish energy/localize nodes). In addition,
they maintain the same goal: serve more sensor nodes in the
 G. Han et al. / Microprocessors and Microsystems 39 (2015) 1271–1278 1273

network while reducing the traveling length as much as possible.

Path planning of the mobile beacon in localization has been well
studied in the past few years [21–25]. Thus, in this paper, we
choose four classic traveling paths which have been designed for
localization in WSNs, namely, SCAN, HILBERT, S-CURVES, and
Z-curve to study their performance in wireless charging.
Although with different traveling trajectories, the four traveling
paths we choose in this paper are all space-filling curves, which
means that they can achieve full charging coverage. Moreover,
we note that there exists remarkable differences between charging Base Station
and localization because the practically limited charging range,
charging time at each stop location and charging latency should
be taken into consideration in wireless charging problems.
It is particularly worth mentioning that while we introduce four
classic traveling paths for mobile chargers in IWRSNs and analyze
their impacts on charging performance, we do not suggest that
they are the best path planning methods for wireless charging.
R
Instead, the purpose of this work is to find the good properties of
those traveling paths which can improve charging performance Service Station
and establish a solid foundation for designing more advanced
charging algorithms.
Next we will introduce the four traveling paths and analyze the mobile charger stop location sensor node
impacts of them on energy provisioning in detail.
traveling path
3. Traveling paths
Fig. 1. Network model.

In this section, we introduce SCAN, HILBERT, S-CURVES(ad), and

Z-curve as traveling paths for a mobile charger to travel along, 80
which are suitable for random-deployment and dense sensor net-
works. Before that, we will introduce the network model firstly. 70
Imagine such a scenario, sensor nodes are deployed and integrated
using radio frequencies to monitor environmental conditions 60
according to requirements of different industrial systems, such as
temperature, sound, pressure and so on. The sensor nodes can col- 50

laborate with each other to perform distributed sensing and deliver

y (m)

40
the data they sensed to Base Station so that people can be aware of
the environmental conditions. As we can see in Fig. 1, nodes are
30
densely-deployed in a uniform way, which can generate sensing
data periodically and send them to base station. To investigate 20
the impacts of different traveling paths on energy provisioning
more visually, we assume the energy consumption rate of nodes 10
is rcon (J/s) like in [14]. A mobile charger starts from a service sta-
tion and its traveling speed is v (m/s). The mobile charger traverses 0
0 10 20 30 40 50 60 70 80
along the predefined traveling path and stops at each stop location
x (m)
to provision energy for nodes within R distance of it. The time that
the charger stays at each stop location is t (s). After the charger’s Fig. 2. Traveling path and stop locations of SCAN.
traversing the whole network for one time, it comes back to the
service station to recharge its battery and gets ready for the next  2
tour. L L2
NSCAN ¼ ; lSCAN ¼ d ð2Þ
d d
3.1. SCAN

SCAN is a simple and easily implemented trajectory, which was
introduced for beacon-assisted localization in WSNs [22]. The
mobile charger traverses the network area along one dimension,
as shown in Fig. 2. The red line refers to the traveling path and a
series of blue ‘‘⁄’’ represent stop locations, similarly hereinafter.
In this figure, the mobile charger travels along the y axis, and we
define the distance between two successive segments of the path,
parallel to the y axis as the resolution of the trajectory. Here we set
the resolution d equal to the charging range R. In addition, the
charger stops every d distance to replenish energy for nodes
around it to guarantee absolute charging coverage. If we denote
by L the side length of the square deployment area, the total num-
ber of stop locations N SCAN and total distance traversed lSCAN in
SCAN trajectory can be calculated as follows:
3.2. HILBERT
HILBERT was first created by D. HILBERT as a space-filling curve
with many turns, which was generated based on bit manipulation,
finite-state diagrams, and recursive construction [22,23]. HILBERT
curve is shown in Fig. 3. A level-n HILBERT curve is generated by
dividing the two-dimensional space into 4n square cells and con-
necting the centers of those cells using 4n line segments, each of
length equal to the length of the side of a square cell. Here we
define the resolution of the HILBERT curve as the length of each
line segment. In this case, the level n is 4 and the relationship
between n, side length L and resolution d is: L ¼ 2n d. Then the total
number of stop locations N HILBERT and distance traversed lHILBERT in
HILBERT trajectory can be calculated as follows:
1274 G. Han et al. / Microprocessors and Microsystems 39 (2015) 1271–1278

80 80

70 70

60 60

50 50

y (m)
y (m)

40 40

30 30

20 20

10 10

0 0
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
x (m) x (m)

Fig. 3. Traveling path and stop locations of HILBERT. Fig. 4. Traveling path and stop locations of S-CURVES(ad).
 2
L L
N HILBERT ¼ 4n ¼ 4log2 ðdÞ ¼
d 80
ð3Þ
2
L
lHILBERT ¼ ð4n  1Þd ¼ d 70
d
60

3.3. S-CURVES(ad)
50

S-CURVES is based on SCAN, which progressively scans the

y (m)

40
deployment area from left to right [24]. However, S-CURVES takes
an ‘‘S’’ curve instead of going in a straight line, as shown by blue1 30
curves in Fig. 4. The series of ‘‘S’’ curves are connected with short
straight lines. In [24], the authors set the resolution of S-CURVES 20
be d and the radius of each half circle be d=2. To keep consistent
with the other traveling paths in this paper, we set the stop loca- 10
tions every 1/4 circle, as exhibited in Fig. 4. But we find that con-
necting the charging points straightly will make the traveling 0
0 10 20 30 40 50 60 70 80
length shorter than that of S-CURVES proposed in the original x (m)
paper and that will not have any impact on the charging perfor-
mance. Hence we adjust S-CURVES in this paper to improve charg- Fig. 5. Traveling path and stop locations of Z-curve.
ing performance, named as S-CURVES(ad). The red line in Fig. 4
shows the trajectory of S-CURVES(ad) we propose. Then we can
section we firstly give the simulation setup with related parame-
calculate the total number of stop locations N SCURVESðadÞ and dis-
ters, and then compare the four traveling paths in terms of the fol-
tance traversed lSCURVESðadÞ in S-CURVES(ad) trajectory as follows.
lowing five aspects: (1) the number of stop locations and total
length of traveling path, (2) energy provisioning distribution, (3)
3.4. Z-curve alive nodes over time, (4) average charging latency, (5) traveling
efficiency. Average of 50 simulation runs are taken as the final
Z-curve was proposed in [25] as a superior path planning mech- results to reduce accidental error.
anism for mobile beacon-assisted localization in wireless sensor
networks. Z-curve is generated similarly to HILBERT: it is also gen-
4.1. Parameter setting
erated based on recursive construction. As we can see from Fig. 5, a
level-n Z-curve is generated by dividing the two-dimensional space
We assume a uniform deployment of static sensors in a square
into 4n square cells and then connecting centers of those cells and network and a single mobile charger moving around to replenish
some vertexes with line segments by shape ‘‘Z’’. Here we define the energy for them. Specific parameters and their values are listed
resolution d of the Z-curve as the length of each line segment par- in Table 1.
allel to x-axis or y-axis. The relationship between n; L and d is:
L ¼ 2n d. Then the total charging stop locations N Zcurv e and distance 4.2. Results and evaluations
traversed in Z-curve trajectory lZcurv e can be calculated as follows.
The performance of various traveling paths is evaluated using
4. Simulation and evaluation some essential metrics and variables. The simulation results and
the resulting analysis are discussed next.
In this paper, we use MATLAB to simulate the charging perfor-
mance of SCAN, HILBERT, S-CURVES(ad), and Z-curve. In this 4.2.1. The number of stop locations and total length of traveling path
As analyzed above, the number of stop locations (denoted as N)
1
For interpretation of color in Fig. 4, the reader is referred to the web version of and total traveling length of each traveling path (denoted as l)
this article. under the set scenario can be calculated as in Table 2.
 G. Han et al. / Microprocessors and Microsystems 39 (2015) 1271–1278
Table 1
Parameter configuration.
Parameters Symbol Value
Side length L 80 m
Number of nodes n 160, 180, 200, 220, 240, 260
Charger speed v 1 m/s
Charging range R 5m
Charging time t 2 s, 4 s, 5 s, 6 s, 8 s, 10 s
Initial energy E0 1J
Energy threshold Eth 0.7 J, 0.75 J, 0.8 J, 0.85 J, 0.9 J,
Resolution d 5m
Energy consumption rate r con 0.01 J/s
Table 2
Values of N and l of four traveling paths.
Traveling paths Parameters
N l (m)
SCAN 256
HILBERT 256
S-CURVES(ad) 343
Z-curve 320
We can see from Table 2 that SCAN has the same number of
charging stop locations and traveling length as HILBERT. That is,
although the two traveling paths have different charging order,
they achieve the same charging performance after charger’s
traversing over the whole network for one time. S-CURVES(ad)
has the most stop locations and the shortest traveling path, which
can be considered as good characteristics of traveling path in that
the mobile charger can spend less time in moving between stop
locations and more time in delivering energy to sensor nodes,
which is beneficial to reduce the charging latency as well as
improve the traveling efficiency. In the same way, that Z-curve
has the longest traveling path means it wastes more time on trav-
eling between stop locations, and at the same time, it ranks second
only to S-CURVES(ad) in the number of stop locations, which
means it can replenish relatively more energy to nodes so that
lifetime of the nodes can be prolonged. Simulation results in the
following parts also verify our analysis above.
4.2.2. Energy provisioning distribution
In this part we draw contour of energy provisioning distribution
for SCAN, HILBERT, S-CURVES(ad), and Z-curve respectively to gen-
erally show the available charging power in each area. To make the
charging characteristics more distinct, we only show the contour of
energy distribution in a 20 m ⁄ 20 m area in Fig. 6.
We can conclude from the figures that the charging stop loca-
tions have great impacts on energy provisioning and the energy
is focused in the vicinity of charging stop locations of chargers,
which is decided by the wireless charging model: the received
power decreases rapidly as the distance between nodes and char-
ger increases. The less is the distance between the node and the
charger, the more energy the node can be replenished. We can
see from Fig. 6 that only the nodes very close to the charger can
get relatively much energy. SCAN and HILBERT have the same con-
tour of energy provisioning contribution because they have the
same charging stop locations. That is, they have the same charging
performance after the charger’s traversing the network for one
time. As mentioned above, S-CURVES(ad) has the most stop loca-
tions, which means charger can deliver more power to sensor
nodes along the S-CURVES(ad) traveling path.
4.2.3. Alive nodes over time
In this part, we set the number of nodes 200 and charging time
8 s to see the number of alive nodes during the process of the
experiment. Here we define alive nodes as the nodes with residual
1275
energy to operate. As we can see in Fig. 7, the number of alive
nodes under S-CURVES(ad) is always lager than that of the other
traveling paths. That is, S-CURVES(ad) performs the best in pro-
longing the lifetime of nodes, and Z-curve takes the second place.
That is reasonable because S-CURVES(ad) and Z-curve have more
stop locations of the mobile charger, which means nodes in the
network can obtain more energy from the charger under the same
situation. Meanwhile, we find a common phenomenon that most
of the nodes will die during the time slot between 10,000 s and
10,040 s, which is corresponded to the energy provisioning distri-
bution displayed in subSection 4.2.2. Only a small number of nodes
can get relatively much energy from the charger, which causes
unbalanced energy distribution and remains a issue for us to solve
in the future.
4.2.4. Average charging latency
Average charging latency is defined as the average time since
1275 the energy of node drops to a threshold Eth to the time the charger
1275 stops to deliver power to it [26]. Charging latency is what the
1248.8 nodes care about most because if the charging latency is long,
1473.8
the nodes cannot get charged in time, which will lead to decreasing
lifetime of the nodes. In this part, we analyze the impacts of three
factors on charging latency: the number of nodes, energy threshold
and charging time. In the first experiment, we set Eth be 0.8 J,
charging time be 5 s and change the number of nodes from 160
to 260, step 20, to see how average charging latency changes with
increasing number of nodes. In the second experiment, we set the
number of nodes be 200, charging time be 5 s and change Eth from
0.7 J to 0.9 J, step 0.05 J. In the third experiment, we keep the num-
ber of nodes and energy threshold Eth be 200 and 0.8 J respectively,
and change the charging time from 2 s to 10 s, step 2 s.
As we can see from Fig. 8(a), average charging latency of
S-CURVES(ad) and Z-curve are longer than those of SCAN and
HILBERT. Z-curve has the longest latency because the charger takes
more time to traverse the whole network than other paths. As we
have analyzed above, by the same token, SCAN and HILBERT almost
have the same average charging latency because they have the
same charging stop locations and the same traveling length, and
at the same time, the nodes are uniformly deployed and we
assume that the energy consumption rate of each node is a certain
value, thus the charging order of the nodes does not have much
effect on the charging latency.
As shown in Fig. 8(b), average charging latency increases with
the increase of energy threshold and the increasing trends of
charging latency of the four traveling paths are almost the same.
As we expected, the increasing energy threshold certainly results
in increasing charging latency because it proposes higher require-
ment for charger to deliver energy to nodes in time. The simulation
results give us two inspirations in respect to shortening the charg-
ing latency: (1) improve the moving speed of the mobile charger,
(2) introduce more than one mobile charger in the network so that
the energy threshold could be set lower.
Fig. 8(c) depicts the relationship between average charging
latency and charging time. As we can see from this picture, average
charging latency increases with the increase of charging time. On
one hand, increasing charging time means delivering more power
to nodes, thus it will take longer time for nodes to drop to the
energy threshold. On the other hand, it costs the mobile charger
more time to traverse the whole network, which results in longer
average charging latency. Obviously under this scenario, the trav-
eling time has more significant effects on the charging latency
because the increasing received power caused by increasing charg-
ing time is tiny. We can conclude from this picture that the
increasing charging time has the most significant impact on aver-
age charging latency of Z-curve and has relatively less impact on
SCAN and HILBERT.
1276 G. Han et al. / Microprocessors and Microsystems 39 (2015) 1271–1278

20 0.04 20 0.04

18 18
0.035 0.035
16 16
0.03 0.03
14 14

0.025 0.025
12 12
y (m)

y (m)
10 0.02 10 0.02

8 8
0.015 0.015
6 6
0.01 0.01
4 4

0.005 0.005
2 2

0 0 0 0
0 5 10 15 20 0 5 10 15 20
x (m) x (m)

20 0.04 20 0.04

18 18
0.035 0.035
16 16
0.03 0.03
14 14

0.025 0.025
12 12
y (m)

y (m)

10 0.02 10 0.02

8 8
0.015 0.015
6 6
0.01 0.01
4 4

0.005 0.005
2 2

0 0 0 0
0 5 10 15 20 0 5 10 15 20
x (m) x (m)

Fig. 6. Contour of energy provisioning distribution of SCAN, HILBERT, S-CURVES(ad), and Z-curve.

200 Etotal
SCAN g¼ ð4Þ
HILBERT
T
180
S−CURVES(ad)
160 Z−curve
We can see from formulation (6) that more energy delivering to
nodes and less traversing time will lead to higher traveling effi-
Number of alive nodes

140
20 ciency. The higher the traveling efficiency is, the better performance
120
the traveling path has. In this part, we design two experiments to
100 15 study the traveling efficiency of four traveling paths. Firstly, we
change the number of nodes from 160 to 260 and keep the charging
80 10 time 5 s to study how traveling efficiency changes as the number of
60 nodes increases. Secondly, we change the charging time from 2 s to
5
10 s while keep the number of nodes 200 to see the impact of charg-
40 1.004 1.006 1.008 1.01
4
ing time on traveling efficiency.
x 10
20 Fig. 9(a) depicts the relationship between traveling efficiency
and the number of nodes. S-CURVES(ad) outperforms the other
0
1 1.001 1.002 1.003 1.004 1.005 1.006 1.007 1.008 1.009 1.01 three traveling paths in traveling efficiency when the number of
Time (s) 4
x 10 nodes increases. Traveling efficiency certaninly increases when
the number of nodes becomes larger because of the charging
Fig. 7. Alive nodes over time for SCAN, HILBERT, S-CURVES(ad) and Z-curve. model: a charger can deliver power to multiple nodes within the
charging range of it at the same time. But the traveling efficiency
4.2.5. Traveling efficiency cannot increase infinitely because deploying too many nodes in
In this paper, we define traveling efficiency g as the ratio of the the network is unnecessary and impractical.
total received energy of all nodes Etotal and the time of the mobile Fig. 9(b) shows how traveling efficiency changes as the increase
charger for traversing over the whole network for one time, which of charging time. When charging time increases, both the total
is denoted by a charging circle T. It can be calculated as follows: energy that nodes can receive and T increase correspondingly.
 G. Han et al. / Microprocessors and Microsystems 39 (2015) 1271–1278 1277

210 700
200 SCAN SCAN SCAN
700
190 HILBERT HILBERT HILBERT
600
180 S-CURVES(ad) S-CURVES(ad) S-CURVES(ad)
Z-curve Z-curve 600
170 Z-curve
Average charging latency (s)

Average charging latency (s)

500
160

Charging latency (s)

500
150
140 400
130 400
120 300
110 300
100
200
90
200
80
100
70
60 100
50 0
40 0
160 170 180 190 200 210 220 230 240 250 260 0.70 0.75 0.80 0.85 0.90 2 4 6 8 10

The number of nodes Energy threshold (J) Charging time (s)

Fig. 8. (a) Charging latency VS the number of nodes; (b) Charging latency VS the energy threshold; (c) Charging latency VS charging time.

0.00014 0.00014
SCAN SCAN
HILBERT 0.00013 HILBERT
0.00013
S-CURVES(ad) S-CURVES(ad)
Z-curve 0.00012 Z-curve
0.00012
0.00011
Traveling efficiency

0.00011
Traveling efficiency 0.00010

0.00010 0.00009

0.00008
0.00009
0.00007
0.00008
0.00006

0.00007 0.00005

160 170 180 190 200 210 220 230 240 250 260 2 4 6 8 10

The number of nodes Charging time (s)

(a) (b)
Fig. 9. (a) Charging efficiency VS charging time; (b) Charging efficiency VS number of nodes.

S-CURVES(ad) has the highest traveling efficiency among the four charing coverage, which are more suitable for randomly deployed
traveling paths. It is reasonable because S-CURVES(ad) has the and dense IWSNs.
most stop locations and the shortest traveling length. SCAN and In the future, instead of simply assuming that the energy con-
HILBERT have almost the same performance in traveling efficiency. sumption rates of nodes are equal, we plan to combine routing pro-
We can see from the picture that traveling efficiency increases tocols with wireless charging to further prolong the lifetime of
slowly when charging time reaches to 8 s, that is, the improvement nodes since routing strategy has a significant impact on the energy
of the traveling efficiency due to the increase of the charging time consumption of nodes. In addition, nonuniform energy distribution
will be very limited especially when the charging time reaches a due to the traveling paths will be the key problem to solve in the
certain value. At the same time, the increase of charging time will follow-on work.
lead to longer charging latency.
Acknowledgements
5. Conclusion
The work is supported by ‘‘Qing Lan Project’’, ‘‘the Foundation of
Using a mobile charger to replenish energy to nodes in the net- Changzhou Key Laboratory of Special Robot and Intelligent
work is considered as an alternative to solving the energy problem Technology, No. CZSR2014004, P.R. China’’ and ‘‘Natural Science
in IWSNs. In this paper, we introduce four traveling paths: SCAN, Foundation of JiangSu Province of China, BK20140248’’.
HILBERT, S-CURVES(ad) and Z-curve and analyze the impacts of
these paths on energy provisioning in IWRSNs. Results show that
S-CURVES(ad) outperforms Z-curve in charging latency and travel- References
ing efficiency. And its performance in improving the lifetime of [1] Jianqi Liu, Qinruo Wang, Jiafu Wan, Jianbin Xiong, Bi Zeng, Towards key issues
nodes and traveling efficiency is better than SCAN and HILBERT. of disaster aid based on wireless body area networks, KSII Trans. Internet
It benefits from the fact that S-CURVES(ad) has the largest number Inform. Syst. 7 (5) (2013) 1014–1035.
[2] Min Chen, Jiafu Wan, Sergio González, Xiaofei Liao, Leung, Victor C.M. Leung, A
of stop locations the shortest traveling length. SCAN and HILBERT
survey of recent developments in home M2M networks, IEEE Commun.
perform almost the same because they have the same stop loca- Surveys Tutorials 16 (1) (2014) 98–114.
tions and the nodes are deployed uniformly in the network and [3] Vehbi C. Gungor, Gerhard P. Hancke, Industrial wireless sensor networks:
the energy consumption rates of nodes are assumed to be the challenges, design principles, and technical approaches, IEEE Trans. Indust.
Electron. 56 (10) (2009) 4258–4265.
same. These four traveling paths are all static path planning and [4] José A. Gutiérrez, David B. Durocher, Bin Lu, Applying wireless sensor networks
can bestrew the whole network to guarantee a hundred percent in industrial plant energy evaluation and planning systems, in: Pulp and Paper
1278 G. Han et al. / Microprocessors and Microsystems 39 (2015) 1271–1278

Industry Technical Conference, 2006, Conference Record of Annual, 2006, pp. computing, and multimedia communication and security. He has served as a
1–7. Cochair for more than 30 international conferences/workshops and as a Technical
[5] Zhao Gang, Wireless sensor networks for industrial process monitoring and Program Committee member of more than 100 conferences. He has served on the
control: a survey, Netw. Protocols Algor. 3 (1) (2011) 46–63. Editorial Boards of up to 14 international journals, including the International
[6] L. He, Z. Yang, J. Pan, L. Cai, J. Xu, Evaluating service disciplines for mobile Journal of Ad Hoc and Ubiquitous Computing, Journal of Internet Technology and KSII
elements in wireless ad hoc sensor networks, in: INFOCOM, 2012 Proceedings Transactions on Internet and Information Systems. He has served as a Reviewer of
IEEE, 2012, pp. 576–584.
more than 50 journals. He had been awarded the ComManTel 2014, ComComAP
[7] Fachang Jiang, Shibo He, Peng Cheng, Jiming Chen, On optimal scheduling in
2014 and Chinacom 2014 Best Paper Awards. He is a member of IEEE and ACM.
wireless rechargeable sensor networks for stochastic event capture, in: IEEE
8th International Conference on MASS, October 2011, pp. 69–74.
[8] Lingkun Fu, Peng Cheng, Yu Gu, Jiming Chen, Tian He, Minimizing charging
delay in wireless rechargeable sensor networks, in: INFOCOM, 2013 Aihua Qian is currently persuing her Masters degree
Proceedings IEEE, April 2013, pp. 2922–2930. from Department of Information & Communication
[9] Kris Lin, Jennifer Yu, Jason Hsu, Sadaf Zahedi, David Lee, Jonathan Friedman, Engineering at the Hohai University, China. She received
Aman Kansal, Vijay Raghunathan, Mani Srivastava, Heliomote: enabling long- her B.S. degree in Information & Communication
lived sensor networks through solar energy harvesting, in: Proceedings of the Engineering from Hohai University, China, in 2013. Her
3rd International Conference on Embedded Networked Sensor Systems, ACM, current research interests are wireless charging for
2005. 309–309. Wireless Sensor Networks.
[10] Chulsung Park, P.H. Chou, AmbiMax: autonomous energy harvesting platform
for multi-supply wireless sensor nodes, Sensor Ad Hoc Commun. Netw. 1
(2006) 168–177.
[11] Miao Zhao, Ji Li, Yuanyuan Yang, Joint mobile energy replenishment and data
gathering in wireless rechargeable sensor networks, in: Proceedings of the
23rd International Teletraffic Congress, 2011, pp. 238–245.
[12] N.A. Pantazis, S.A. Nikolidakis, D.D. Vergados, Energy-efficient routing
protocols in wireless sensor networks: a survey, IEEE Commun. Surveys
Tutorials 15 (2) (2012) 551–559.
[13] Getsy S. Sara, R. Kalaiarasi, Energy efficient mobile wireless sensor network Li Liu is currently persuing his Masters degree from
routing protocol, Commun. Comput. Inform. Sci. 90 (2010) 642–650. Department of Information & Communication
[14] Shibo He, Jiming Chen, Fachang Jiang, David K.Y. Yau, Guoliang Xing, Youxian Engineering at the Hohai University, China. He received
Sun, Energy provisioning in wireless rechargeable sensor networks, IEEE Trans. his B.S. degree in Information & Communication
Mobile Comput. 12 (2013) 1931–1942. Engineering from Hohai University, China, in 2014. His
[15] Richard Beigel, Jie Wu, Huangyang Zheng, On optimal scheduling of multiple current research interests are coverage and connectivity
mobile chargers in wireless sensor networks, in: Proceedings of the First for Wireless Sensor Networks.
International Workshop on Mobile Sensing, Computing and Communication,
2014, pp. 1–6.
[16] Ji-Hau Liao, Chi-Ming Hong, Jehn-Ruey Jiang, An adaptive algorithm for
charger deployment optimization in wireless rechargeable sensor networks,
in: ICS, 2014.
[17] c. Constantinos Marios Angelopoulos, Sotiris Nikoletseas, Theofanis P. Raptis,
Wireless energy transfer in sensor networks with adaptive, limited knowledge
protocols, Comput. Netw. (2014) 113–141.
[18] Zi Li, Yang Peng, Wensheng Zhang, Daji Qiao, Study of joint routing and
wireless charging strategies in sensor networks, Wireless Algor. Syst. Appl. Jinfang Jiang received the BS degree in information &
6221 (2010) 125–135. communication engineering from Hohai University,
[19] Liguang Xie, Yi Shi Member, Y. Thomas Hou Senior Member, Hanif D. Sherali, China, in 2009. She is currently working toward the
Making sensor networks immortal: an energy-renewal approach with wireless Ph.D. degree from the Department of Information &
power transfer, IEEE/ACM Trans. Netw. 20 (6) (2012) 1748–1761. Communication System at Hohai University, China. Her
[20] Liang He, Yu Gu, Jianping Pan, Ting Zhu, On-demand charging in wireless
current research interests are security and localization
sensor networks: theories and applications, in: IEEE 10th International
for sensor networks.
Conference on Mobile Ad-Hoc and Sensor Systems, 2013, pp. 28–36.
[21] Guangjie Han, Huihui Xu, Jinfang Jiang, Lei Shu, Takahiro Hara, Shojiro Nishio,
Path planning using a mobile anchor node based on trilateration in wireless
sensor networks, Wireless Commun. Mobile Comput. (2013) 1324–1336.
[22] Dimitrios Koutsonikolas, Saumitra M. Das, Y. Charlie Hu, Path planning of
mobile landmarks for localization in wireless sensor networks, Comput.
Commun. (2007) 2577–2592.
[23] D. Hilbert, Über die, ‘‘stetige Abbildung einer Linie auf Flächenstück’’, Math.
Ann. (1891) 459–460.
[24] Rui Huang, Gergely V. Zaruba, Static path planning for mobile beacons to Chuan Zhu is a lecturer of Department of Information &
localize sensor networks, in: Proceedings of the Fifth Annual IEEE International Communication System at Hohai University, China.
Conference on Pervasive Computing and Communications Workshops, 2007.
Currently, he is also a post doctor of Hohai University.
[25] Javad Rezazadeh, Marjan Moradi, Abdul Samad Ismail, Eryk Dutkiewicz,
He received his Ph.D. degree in Department of Computer
Superior path planning mechanism for mobile beacon-assisted localization in
Science from Northeastern University, Shenyang, China,
wireless sensor networks, IEEE Sensors J. (2014).
[26] Wen Yao, Minglu Li, Min-You Wu, Inductive charging with multiple charger in 2009. He has published over 10 papers in related
nodes in wireless sensor networks, Adv. Web Netw. Technol. Appl. (2006) international conferences and journals. His current
262–270. research interests are Sensor Networks, Cloud
Computing, and Computer Networks.

Guangjie Han is currently a Professor with the

Department of Information and Communication
System, Hohai University, Changzhou, China. He
received the Ph.D. degree from Northeastern University,
Shenyang, China, in 2004. From 2004 to 2006, he was a
Product Manager for the ZTE Company. In February
2008, he finished his work as a Postdoctoral Researcher
with the Department of Computer Science, Chonnam
National University, Gwangju, Korea. From October
2010 to 2011, he was a Visiting Research Scholar with
Osaka University, Suita, Japan. He is the author of over
150 papers published in related international confer-
ence proceedings and journals, and is the holder of 65 patents. His current research
interests include sensor networks, computer communications, mobile cloud