Massive Wireless Energy Transfer with Multiple Power ...

c© 2021 by the authors; licensee RonPub, Lubeck, Germany. This article is an open access article distributed under the terms and conditions ofthe Creative Commons Attribution license (http://creativecommons.org/licenses/by/4.0/).

Open Access

Open Journal of Internet of Things (OJIOT)Volume 7, Issue 1, 2021

http://www.ronpub.com/ojiotISSN 2364-7108

Massive Wireless Energy Transfer withMultiple Power Beacons

for Very Large Internet of ThingsOsmel Martınez RosabalA, Onel L. Alcaraz LopezA, Hirley AlvesA,

Richard D. SouzaB, Samuel Montejo-SanchezC

ACentre for Wireless Communications, University of Oulu, Pentti Kaiteran katu 1, Oulu, Finland,{osmel.martinezrosabal, onel.alcarazlopez, hirley.alves}@oulu.fi

BFederal University of Santa Catarina (UFSC), Florianopolis, Brazil, [email protected] Institucional de Fomento a la I+D+i, Universidad Tecnologica Metropolitana,

Santiago, Chile, [email protected].

ABSTRACT

The Internet of Things (IoT) comprises an increasing number of low-power and low-cost devices that autonomouslyinteract with the surrounding environment. As a consequence of their popularity, future IoT deployments will bemassive, which demands energy-efficient systems to extend their lifetime and improve the user experience. Radiofrequency wireless energy transfer has the potential of powering massive IoT networks, thus eliminating the need forfrequent battery replacement by using the so-called power beacons (PBs). In this paper, we provide a framework forminimizing the sum transmit power of the PBs using devices’ positions information and their current battery state.Our strategy aims to reduce the PBs’ power consumption and to mitigate the possible impact of the electromagneticradiation on human health. We also present analytical insights for the case of very distant clusters and evaluate theirapplicability. Numerical results show that our proposed framework reduces the outage probability as the number ofPBs and/or the energy demands increase.

TYPE OF PAPER AND KEYWORDS

Regular research paper: clustering, massive IoT, optimal PBs deployment, RF-WET, energy outage

1 INTRODUCTION

The recent advances in wireless communicationstechnologies have brought a plethora of newopportunities for developing the so-called Internetof Things (IoT). In short, IoT defines a network of

This paper is accepted at the International Workshop on VeryLarge Internet of Things (VLIoT 2021) in conjunction with theVLDB 2021 conference in Copenhagen, Denmark. The proceedingsof VLIoT@VLDB 2021 are published in the Open Journal ofInternet of Things (OJIOT) as special issue.

uniquely identified physical objects that autonomouslysense, actuate, compute and exchange informationthroughout the Internet [33]. In general, “an object” canbe anything from a smart ring, electronic appliances,or sensors that feedback environmental measurements.Currently, this paradigm has taken over in many spheresof our daily life and society such as in industrialautomation, smart healthcare, supply chain, smartinfrastructure, and social & business applications [13].Not surprisingly, the number of connected devices isconstantly growing, and by 2030 the estimations target

83

http://creativecommons.org/licenses/by/4.0/

http://www.ronpub.com/ojiot

Open Journal of Internet of Things (OJIOT), Volume 7, Issue 1, 2021

massive deployments of 71 billion IoT devices [15].One of the big challenges of such applications is

extending their lifetime as IoT devices are generallypowered by tiny batteries. Besides, regular maintenancebecomes impractical or too costly in most cases dueto the massive number of devices or the hard-to-reachinstallation conditions, such as in medical implants,waste management, and civil infrastructure monitoringapplications [29]. This issue can cause serviceinterruptions due to power outages in the network, whichdegrades the end-user experience and hence lowers theprofits of the networks’ operators [34]. In multi-hop IoTapplications, for instance, power outages in few devicescan affect the network’s performance, as some of themhave the role of forwarding the packets of the far awaydevices to the central receiver [17].

To extend the lifetime of IoT applications, radio-frequency (RF) wireless energy transfer (WET)envisions a network where the so-called power beacons(PBs) wirelessly replenish devices’ batteries or evensustain the operation of batteryless devices by meansof electromagnetic waves [26]. Moreover, RF-WETexhibits other desirable features such as energybroadcast to power multiple users simultaneously,service provisioning under non-line-of-sight conditions,and the potential to charge moving IoT devices [22, 35].By making RF-WET a ubiquitous service, one candeal with the high heterogeneity of IoT networks andguarantee different levels of quality-of-service (QoS) interms of energy demands. Hence, as RF-WET reducesfrequent maintenance, it could considerably diminishwater and air pollution caused by toxic chemicals thatare released into the environment after inadequatebattery recycling [20].

A key component in RF energy harvesting (EH)systems is the rectenna, which comprises [6]: i)receiving antenna(s); ii) a matching network; iii) arectifier circuit; and iv) a low-pass filter. At thereceiver, the rectenna turns the microwave energy intoa direct current (DC) used in most of the circuitsof IoT devices. Compared with traditional wirelessinformation decoders, rectennas require a significantamount of incident RF power to provide a non-zeroDC output, typically in the order of −10 dBm [3].However, the channel impairments such as distant-dependent loss, large- and small- scale fading processescannot be often overcome to deliver such incident powerlevels by deliberately increasing the transmit power. Inthis direction, energy beamforming algorithms offer apromising solution as the intention is to steer sharpenergy beams directly to the intended devices avoidingspreading the energy where there is no device [2].However, realizing efficient energy beamforming may bechallenging in practical RF-WET systems. The reasons

are:

• On one hand, accurate CSI is required, which isdifficult/costly to acquire, and its benefits vanish, inmassive deployments [22]. For this reason, someinitiatives have been proposed to prescind frominstantaneous CSI at the price of transmitting withsuboptimal beamforming strategies, e.g., [24, 32],using statistical CSI and received energy feedback,respectively. It is worth noticing, that althoughboth are sub-optimal strategies, their performancebecomes asymptotically optimal in the presence ofstrong line-of-sight components and large channeltraining periods, respectively.

• On the other hand, energy beamforming at hightransmit power increases the RF electromagneticfield radiation (RF-EMF) in relatively smallincident areas whose risk on human health hasconcerned the research community [11, 36, 37]. Toevaluate the impact of the RF signals on the humanhealth, the most frequently used metrics are: thespecific absorption rate (SAR) and the plane-waveequivalent power density (PD) [27].

Hence, the transmission schemes for enabling massiveWET can be designed instead, over the basis of side-information such as devices’ positions, statistical CSI,or even received energy feedback from the IoT devices[25]. At the same time, these methods must ensuresufficient energy to meet stringent QoS requirementswhile guaranteeing the human safety against high RF-EMF levels exposure. Last, but not less important,the algorithms for efficiently deploying the PBs alsoplay a key role to eliminate blind spots in the network,and to distribute the energy according to the specificrequirements.

1.1 Related Work

The research community has put effort into the designof PBs’ deployment algorithms to meet stringent QoSrequirements in the coming IoT use cases, e.g., [9,10, 12, 30, 31]. The authors in [12] proposed a PBs’deployment algorithm that guarantees a network-wideenergy availability at the devices while considering theinstallation/maintenance costs of PBs. They conductedexperiments to show that: i) the complexity of proposedalgorithms decreases as the network densifies sinceeach PB charges more devices simultaneously; ii) PBsinstallation cost gets more expensive as the per-deviceminimum energy requirements increase; and iii) whenusing directional antennas at the PBs, the deploymentcost reduces as the beamwidth of the antennas slimdown at the price of reducing the incident RF power

84

O. M. Rosabal et al.: Massive Wireless Energy Transfer with Multiple Power Beacons for Very Large Internet of Things

in the network. In [10], the authors considered thescenario where a mobile PB charges the service areawhile moving along the area perimeter to minimizethe charging time. Therein, the authors divided theservice area into smaller sub-partitions and focused onthe instantaneous worst position, which changes as thePB moves. In [9], the authors studied the PBs’ placementproblem aimed to maximize the overall harvestedenergy. They proposed an algorithm for optimizingthe positions and orientations of PBs using a piecewiselinear EH model, and evaluated its performance usingboth numerical simulations and field experiments. Inaddition, [31] proposed an algorithm for finding thepositions and orientation of a set of PBs with differenthardware capabilities to maximize the overall harvestedenergy. Finally, the optimal deployment of PBs forcharging a massive IoT network was studied in [30].Therein, the authors optimized the PBs’ positions tomaximize the foreseen minimum average RF energyavailable in the area, since neither instantaneous CSInor devices’ locations information was available. Basedon simulation results, the authors promoted distributeddeployments of PBs over the installation of a centredPB radiating with the same total power to overcome thedominant impact of the distance-dependent loss.

Moreover, other works have also considered safetymetrics in the WET optimization. For instance, in[37] the authors studied a smart healthcare applicationwhere PBs charge wearable devices. Therein, theyminimize the maximum incident power while providingsufficient energy for the wearables. In [36], the authorsproposed a framework for optimizing the position andheight of a distributed antenna system, where eachantenna is connected via underground lines to a centralPB, considering a safe radiation level. In [11], it ismaximized the network-wide harvested energy such thatthe probability of exceeding a certain RF-EMF is met[27].

It is worth noticing that none of these worksexploits simultaneously sensors’ position and batterystate information. Hence, their performance has limitedcapabilities to act proactively to face the imminent“death” of one of its IoT devices after the total dischargeof its battery.

1.2 Contributions

Different from previous works, herein we aim tominimize the sum transmit power of the PBs suchthat the energy available at devices’ batteries sufficesfor proper operation. We exploit sensors’ positioninformation to divide the IoT network into clusterseach headed by a PB; and the current battery state toproactively optimize the power allocation strategy.

Our main contributions are summarized below.

• We propose a framework for optimizing PBs’positions and power allocation based on thedevices’ positions and their current battery state;

• We provide analytical approximations to the powerallocation problem that holds when the contributionof the head PBs dominates the incident RF powerwithin their corresponding cluster;

• Our optimization strategy has two mainimplications: i) to reduce the energy consumptionat the PBs, possibly powered by a limited source,and ii) mitigate the possible impact of the EMFradiation on human health;

• Our results show that the outage probabilitydecreases as the number of deployed PBs increasesand/or the energy demands grow with respect to theminimum battery level for proper operation.

1.3 Organization of the Paper

Next, Section 2 introduces the system model, andSection 3 presents the problem formulation. We discussthe methods for finding the optimal PBs’ positions andcomputing the power allocation in Section 4, while inSection 5 we discuss some practical considerations forimplementing our strategy. In Section 6, we showand analyze numerical results, and finally, Section 7concludes the paper.

Notation: Herein, we use boldface lowercase lettersto denote column vectors, e.g., x = {xi}. The operator| · | can represent either the absolute value for scalars orthe cardinality of a set, while ‖x‖p= (

∑∀i≥1 |x

pi |)1/p

denotes the `p-norm [18, eq. (1)]. Moreover, P[A] is theprobability of event A and E[x] denotes the expectationof the randon variable x, whereas O(·) is the big-Onotation, which specifies worst-case complexity. Table 1lists the symbols used throughout the paper.

2 SYSTEM MODEL

Let us consider that a set B = {PBi|i = 1, 2, . . . , |B|} ofPBs wirelessly powers a set S = {Sj |j = 1, 2, . . . , |S|}of IoT devices deployed in a certain area A as shown inFigure 1. Each PBi is equipped with an omnidirectionalantenna and transmits an energy-carrying signal xi thatsatisfies E[|xixi′ |] = 0 with i 6= i′, and E[|xi|2] = pi,∀PBi,PBi′ ∈ B, where pi denotes the transmit power.We assume that the positions of the IoT devices areknown1 and denoted by uj ∈ R2, while PBs locations1 In many massive IoT applications, devices tend to have fixed

positions or limited mobility, therefore position knowledge can beacquired at deployment or estimated via previous transmissions [23].

85


Table 1: List of symbols

Parameter DefinitionSj jth IoT device in the set SPBi ith PB in the set B%j,i path-loss function of the PBi → EHj linkdj,i distance of the PBi → EHj link{pi} transmit power of PBi

pmax PBs’ maximum transmit powerxi energy-carrying signal transmitted from PBi

τ duration of a charging time slott time slot index{uj} IoT devices’ locations{ui} PBs’ positionsE

(t)j battery level of Sj in the time slot t

α(t)j activation state of Sj in the time slot t

∆E(t)j energy consumed by Sj during the time slot t

ξ(t)j harvested energy of Sj during the time slot tξ(t)j estimated ξ(t)j for the time slot tG,G−1 EH function and its inverse respectivelyGi antenna gain of PBi

Gj antenna gain of Sj

Eth target energy level at the batteriesEmax devices’ battery capacityPactive devices’ power consumption in active modePsleep devices’ power consumption in sleep mode

Figure 1: System model

have coordinates ui ∈ R2, both in the Cartesian system.The distance of the link PBi → Sj is denoted by dj,i =‖uj − ui‖2, while its path loss is given by

ρj,i = F(dj,i). (1)

Each IoT device Sj carries a rechargeable battery andharvests

ξ(t)j = τG

|S|∑i=1

p(t)i ρj,i

, (2)

energy units in a time slot t of duration τ . Herein,we consider that each charging time slot is sufficientlylong so channel fluctuations may approximately averageout, thus, impacting little the performance. The functionG : R+ → R+ characterizes the non-linear energy

harvesting model of the IoT devices and distinguishesthree regions [7]: i) zero-output power, in which theincident RF power is below the sensitivity of the EHcircuit; ii) input power-dependent region, where theoutput DC power is an increasing non-linear function ofthe input RF power; iii) saturation region, in which theoutput DC power becomes constant and independent ofthe input signal.

Finally, the residual battery energy of device Sj isdenoted by Ej , and its current activation state by αj ∈{0, 1}. In idle/sleep mode (αj = 0), the powerconsumption of IoT devices is Psleep, whereas in activemode (αj = 1), it is Pactive. Hence, at the beginning ofthe charging time slot t + 1, the battery level is updatedaccording to

E(t+1)j ← min(E

(t)j + ξ

(t)j −∆E

(t)j , Emax), (3)

where

∆E(t)j =τ

[(1− α(t)

j )Psleep + α(t)j Pactive

], (4)

corresponds to the consumed energy during the time slott, and Emax is the battery capacity.

Moreover, we consider that the devices feedbacktheir current battery state over reliable channels. Thisinformation is used by the PBs to optimize their powerallocation strategy.

3 PROBLEM FORMULATION

Our main goal is to minimize the sum PBs’ transmitpower such that the actual energy in the batteriesEj ,∀Sj ∈ S, is above an energy threshold Eth at theend of the charging time slot. The optimization problemcan be formulated as follows

P1 : min.{p(t)

i },{ui}

B∑b=1

p(t)i (5a)

s.t. E(t−1)j + ξ

(t)j ≥ Eth, ∀Sj ∈ S, (5b)

p(t)i ≤ pmax, ∀PBi ∈ B. (5c)

Notice that the constraint (5b) considers an estimateharvested energy ξ

(t)j given by (2), that should suffice

to replenish devices’ batteries above the threshold Eth,during the next time slot. Herein, we consider F(·) tobe known in advance via prior measurement campaigns[5] and using machine learning/artificial intelligence [1]methods, but in practice this will never be exact, andestimation errors will affect somewhat the performance.Moreover, the constraint (5c) guarantees that the transmitpower doesn’t violate PBs’ hardware specifications orEMF regulations.

86


This problem is non-convex and highly non-linearin general due to the relationship between the distanceand the path-loss function and the non-linearity of thefunction G(·), and to the best of the authors’ knowledgeit cannot be written in convex form. Therefore, a globaloptimum solution is not guaranteed by any solver.

4 OPTIMIZATION FRAMEWORK

To solve P1, we propose a two-step method: i) firstoptimize PBs’ positions by taking advantage of theknowledge about devices’ positions while keeping thetransmit power constant; and then ii) solve the optimalpower allocation problem.

4.1 PBs Deployment Optimization Based onClustering

Herein, we utilize the well-known K-Means clusteringalgorithm [21] to place the PBs as head clusters.Note that since the path-loss function depends on theEuclidean distance between PBs and IoT devices, the K-Means approach seems an appealing choice for solvingour problem. However, we propose a slight modificationof the original algorithm. Our approach aims tominimize the distance from each PB to the farthestsensor in its cluster, which impacts the ultimate incidentpower. Thus, the proposed algorithm re-computes thePBs’ positions as Chebyshev centres.

Let Si ⊂ S denotes the subset of IoT devicesassociated with PBi after running the K-Meansalgorithm, and given by

Si = {uj : ‖uj − ui‖2 ≤ ‖uj − ui′‖2}, (6)

with i′ 6= i, ∀PBi,PBi′ ∈ B. Then, each K-ChebyshevPBs positions is updated according to

ui = argminu∈R2

maxSj∈Si

‖uj − u‖2, (7)

which can be found by solving the following equivalentconvex problem

P2 : min.ui,ri

ri (8a)

s.t. ‖uj − ui‖2 − ri ≤ 0, ∀Sj ∈ Si, (8b)

where√ri is the radius of the minimal-radius

circumference enclosing all the devices in Si. Since P2is now convex, we can solve it using standard solverspackages such as CVX [8, 14].

Algorithm 1 details the steps for determining the PBs’positioning based on clustering. Notice that the step 5determines the cluster each device belongs to, while andstep 6 updates the clusters’ centroids according to the

Algorithm 1: Clustering-based PBs deployment

1: Input: {uj}, µ2: Set k = 03: Initialize {u0

i }4: repeat5: Compute S(k)i according to (6)6: u

(k+1)i = 1

|S(k)i |

∑uj∈Sk

iuj

7: k ← k + 18: until ‖u(k+1)

i − u(k)i ‖2 ≤ µ

9: Solve P210: Output: {u(k+1)

i }

mean of each cluster. The loop ends when the centroids’positions do not change significantly, according to theerror parameter µ. Finally, once the cluster assignmentis done, the algorithm re-computes the clusters’ centroidswith the solutions of problem P2.

4.2 PBs’ Power Optimization

Once we find the PBs’ positions, P1 becomes a linearprogramming (LP) problem in {pi} since (5b) is turnedinto a linear system in the form

|B|∑i=1

p(t)i ρj,i ≥ G−1

(Eth − E(t−1)

j

τ

), ∀Sj ∈ S, (9)

where G−1(·) denotes the inverse of the EH function.Then, the resulting LP with |B| variables and |S| +|B| inequality constraints can be efficiently solvedusing interior-point methods with an accuracy ε inO(√|S|+ |B| log (1/ε)) iterations, where each iteration

demands O((|S|+ |B|)3) arithmetic operations [28].

4.2.1 Approximate Power Allocation Strategy

It is worth noticing that as the clusters are moreseparated, the contribution of the head PB dominatesthe incident RF power in its corresponding cluster. Thismight be the case when the IoT network has a highlysparse deployment of EH IoT devices arranged in veryseparated clusters, or when the distance-dependent lossis sufficiently strong to neglect the power contributionfrom neighbour clusters. In both cases, we can neglectthe contribution of the neighbour PBs, and the linearsystem (9) reduces to

p(t)i ρj,i ≥ G−1

(Eth − E(t−1)

j

τ

), ∀Sj ∈ Si, (10)

p(t)i

(a)= max

j

[ρ−1j,i G

−1

(Eth−E(t−1)

j

τ

)], (11)

87


where (a) comes from considering the device withthe higher product of path loss and energy demands.Observe that if pi, computed as in (11), is not greaterthan pmax, it constitutes a low-complexity solution tothe power allocation problem. Otherwise, the resultingpower allocation is not feasible in the sense that PBi isjust allowed to transmit with a maximum power pmax,which in this case does not suffice the devices’ energyrequirements. In any case, the power allocation can beset as

p∗(t)i = min(p

(t)i , pmax) (12)

to allow satisfying, at least partially, the devices’ energyrequirements. Notice that, to obtain the optimal solution{p∗(t)i } we just require at most O(|S|+ 2|B|) arithmeticoperations.

5 PRACTICAL CONSIDERATIONS

Herein, we discuss the practicalities for implementingthe proposed strategy. First, notice that clusteringrequires devices’ location information, which canbe provided during the network planning in staticdeployments. However, in quasi-static or mobilescenarios, clustering as a positioning strategy mayrequire that also the PBs can move/fly to update theirposition as the network changes.

Another detail to consider is that battery statefeedback is subject to the channel impairments andhence, not all PBs could reliably receive it all thetime. In this direction, we propose that PBs collectdevices’ battery state and validate this information ina distributed way based on previous updates, statisticsof devices’ activation probabilities, and devices’ powerconsumption. For instance, the PBs can have a highlyreliable link between them, over which they can run ablockchain-based lightweight algorithm to validate allenergy transactions [16].

Finally, due to the inherent characteristics of the EHcircuit the charging period may take more than one timeslot. That is, if the energy requirements of an IoT devicesurpass the amount it can harvest during the time τ ,then the optimization problem will be infeasible and thesystem will have to schedule additional time slots, untilthe device finally meets the energy requirements.

6 NUMERICAL RESULTS

In this section, we present numerical results on how toaddress P1. The default parameters for simulations arelisted in Table 2 unless we establish the contrary.

For the EH circuit of the IoT devices we adopt the

Table 2: Default simulation parameters

Parameter Value Parameter ValueA 30× 15 m2 τ 120 s

Emax 1 J Psleep 10 µWαj ∼ Beta(0.5, 0.5) Pactive 1 mWf 2.4 GHz GiGj 24S 64 pmax 4 W

Figure 2: Clustering-based PBs’ positioningapproach, where the circumferences represents theclusters’ limits and the numbers their correspondingindex.

sigmoidal-based model in [4]

G(x) =$(1− e−c1x)

(1 + e−c1(x−c0)), (13)

whose inverse is

G−1(y) = − 1

c1ln

($ − y

yec0c1 +$

), (14)

where $ = 10.73 mW is the saturation level, x is theincident RF power, y the harvested energy, and c0 =5.365, c1 = 0.2308 are unitless constant obtained bystandard curve fitting using the measurement data in[19].

As a metric of QoS, we consider the energy outageprobability. That is, an IoT device is in outage if theactual battery state at the beginning of the charging slotis insufficient for it to operate for τ time units. That is,

Pj = P[E(t)j < ∆E

(t)j ], (15)

and the average outage probability, i.e., the averageprobability of having one device in outage, is

Pout = E[Pj ]. (16)

Finally, we adopt the log-distance path-loss model

ρj,i = GiGj

(λ

4π

)2

d−2.7j,i , (17)

88


0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

12

13

14

15

16

17

18

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

10-2

10-1

Figure 3: Network performance vs Eth, for |B| = 15,in terms of: i) sum transmit power (top) and ii) outageprobability (bottom).

where Gi and Gj denote the antenna gains of PBi andSj , respectively, and λ is the wavelength of the energy-carrying signals.

Figure 2 depicts the clustering strategy for positioningthe PBs using both the original K-Means algorithm andthe clustering with K-Chebyshev centroids.

The reader can notice that both strategies lead todifferent PBs deployments. Indeed, taking the seventhcluster as an example, we can see that the K-Chebyshevcentroid has a lower maximum intra-cluster distancewith respect to the K-Means solution.

Figure 3 illustrates the impact of the energy thresholdEth on the optimal sum transmit power and theoutage probability for |B| = 15. As the devices’energy demands are tightened, the PBs’ transmit powerincreases and the outage probability decreases. Inparticular, K-Chebyshev centering achieves better resultsin almost all cases, as it maximizes the minimumincident RF power within each cluster. Moreover,the approximation (11) achieves lower values of sumtransmit power at the cost of a poorer performance interms of energy outage probability since it considerseach cluster as an independent network. From theseplots, we can define the optimal energy threshold E∗thas the value from which the outage probability doesn’timprove significantly. For this particular deployment, wehave E∗th = 0.4 J approximately.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

6

8

10

12

14

16

18

20

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 2010

-3

10-2

10-1

100

Figure 4: Network performance vs |B|, for Eth =0.25 J , in terms of: i) sum transmit power (top) andii) outage probability (bottom).

Although the approximation (11) performs poorly fora relatively large number of PBs, we can observe inFigure 4 that when |B| ≤ 6 it holds accurate due tothe very low contribution of neighbor clusters. Thecurve |B|pmax delimits the feasible region over which thePBs can transmit their energy-carrying signals withoutviolating the constraint (5c).

7 SUMMARY AND CONCLUSIONS

In this paper, we studied the sum power minimizationproblem of PBs for powering an IoT deployment. Forsolving this problem, we first used the devices’ positionsinformation to arrange clusters in the network, eachheaded by a PB, and then optimize the power allocationstrategy based on the current devices’ battery states. Inorder to obtain fairer results for the worst-positioneddevice within each cluster, we proposed a modificationof the traditional K-Means algorithm for re-computingthe clusters’ centroids called K-Chebyshev centering.Numerical results show that this approach achievesbetter results in terms of outage probability with less sumtransmit power when there is a large number of PBs.

89


ACKNOWLEDGEMENTS

This research has been financially supported byAcademy of Finland, 6Genesis Flagship (Grantno318927) and EE-IoT (no319008), by CNPq,Print CAPES-UFSC “Automation 4.0”, andRNP/MCTIC (Grant 01245.010604/2020-14) 6GMobile Communications Systems in Brazil, as well asby ANID FONDECYT Iniciacion No. 11200659 inChile.

REFERENCES

[1] S. M. Aldossari and K.-C. Chen, “Machine learningfor wireless communication channel modeling:An overview,” Wireless Personal Communications,vol. 106, no. 1, pp. 41–70, 2019.

[2] Y. Alsaba, S. K. A. Rahim, and C. Y. Leow,“Beamforming in wireless energy harvestingcommunications systems: A survey,” IEEECommunications Surveys Tutorials, vol. 20, no. 2,pp. 1329–1360, 2018.

[3] S. Bi, C. K. Ho, and R. Zhang, “Wireless poweredcommunication: opportunities and challenges,”IEEE Communications Magazine, vol. 53, no. 4,pp. 117–125, 2015.

[4] E. Boshkovska, D. W. K. Ng, N. Zlatanov,and R. Schober, “Practical non-linear energyharvesting model and resource allocation forSWIPT systems,” IEEE Communications Letters,vol. 19, no. 12, pp. 2082–2085, 2015.

[5] G. Caso, O. Alay, L. De Nardis, A. Brunstrom,M. Neri, and M.-G. Di Benedetto, “Empiricalmodels for NB-IoT path loss in an urban scenario,”IEEE Internet of Things Journal, pp. 1–1, 2021.

[6] K. W. Choi, S. I. Hwang, A. A. Aziz, H. H.Jang, J. S. Kim, D. S. Kang, and D. I. Kim,“Simultaneous wireless information and powertransfer (SWIPT) for Internet of Things: Novelreceiver design and experimental validation,” IEEEInternet of Things Journal, vol. 7, no. 4, pp. 2996–3012, 2020.

[7] B. Clerckx and E. Bayguzina, “Waveform designfor wireless power transfer,” IEEE Transactions onSignal Processing, vol. 64, no. 23, pp. 6313–6328,2016.

[8] I. CVX Research, “CVX: Matlab software fordisciplined convex programming, version 2.0,”http://cvxr.com/cvx, Aug. 2012.

[9] H. Dai, X. Wang, A. X. Liu, H. Ma, G. Chen,and W. Dou, “Wireless charger placement for

directional charging,” IEEE/ACM Transactions onNetworking, vol. 26, no. 4, pp. 1865–1878, 2018.

[10] H. Dai, X. Wang, L. Xu, C. Dong, Q. Liu,L. Meng, and G. Chen, “Area charging for wirelessrechargeable sensors,” in 29th InternationalConference on Computer Communications andNetworks (ICCCN), 2020, pp. 1–9.

[11] R. Dai, Y. Zhao, G. Chen, W. Dou, C. Tian, X. Wu,and T. He, “Robustly safe charging for wirelesspower transfer,” in IEEE Conference on ComputerCommunications, 2018, pp. 378–386.

[12] X. Ding, Y. Wang, G. Sun, C. Luo, D. Li,W. Chen, and Q. Hu, “Optimal charger placementfor wireless power transfer,” Computer Networks,vol. 170, p. 107123, 2020.

[13] P. Goyal, A. K. Sahoo, T. K. Sharma, andP. K. Singh, “Internet of Things: Applications,security and privacy: A survey,” Materials Today:Proceedings, vol. 34, pp. 752–759, 2021.

[14] M. Grant and S. Boyd, “Graph implementations fornonsmooth convex programs,” in Recent Advancesin Learning and Control, ser. Lecture Notes inControl and Information Sciences, V. Blondel,S. Boyd, and H. Kimura, Eds. Springer-VerlagLimited, 2008, pp. 95–110, http://stanford.edu/∼boyd/graph dcp.html.

[15] F. Grijpink, E. Kutcher, A. Menard,S. Ramaswamy, D. Schiavotto, J. Manyika,M. Chui, R. Hamill, and E. Okan, “Connectedworld: An evolution in connectivity beyond the 5Grevolution,” McKinsey Glob. Inst., no. 304, p. 2,2020.

[16] L. Jiang, S. Xie, S. Maharjan, and Y. Zhang,“Blockchain empowered wireless power transferfor green and secure Internet of Things,” IEEENetwork, vol. 33, no. 6, pp. 164–171, 2019.

[17] P. S. Lakshmi, M. G. Jibukumar, and V. S. Neenu,“Network lifetime enhancement of multi-hopwireless sensor network by RF energy harvesting,”in International Conference on InformationNetworking (ICOIN), 2018, pp. 738–743.

[18] M. Lange and T. Villmann, “Derivatives oflp-norms and their approximations,” MachineLearning Reports, vol. 7, pp. 43–59, 2013.

[19] T. Le, K. Mayaram, and T. Fiez, “Efficient far-fieldradio frequency energy harvesting for passivelypowered sensor networks,” IEEE Journal of Solid-State Circuits, vol. 43, no. 5, pp. 1287–1302, 2008.

[20] Q. Liu, K. S. Yildirim, P. Pawełczak, andM. Warnier, “Safe and secure wireless power

90

http://cvxr.com/cvx

http://stanford.edu/~boyd/graph_dcp.html

http://stanford.edu/~boyd/graph_dcp.html


transfer networks: challenges and opportunitiesin RF-based systems,” IEEE CommunicationsMagazine, vol. 54, no. 9, pp. 74–79, 2016.

[21] S. Lloyd, “Least squares quantization in PCM,”IEEE Transactions on Information Theory, vol. 28,no. 2, pp. 129–137, 1982.

[22] O. L. A. Lopez, H. Alves, R. D. Souza, S. Montejo-Sanchez, E. M. G. Fernandez, and M. Latva-aho, “Massive wireless energy transfer: Enablingsustainable IoT towards 6G era,” IEEE Internet ofThings Journal, pp. 1–1, 2021.

[23] O. L. A. Lopez, N. H. Mahmood, H. Alves, C. M.Lima, and M. Latva-aho, “Ultra-low latency, lowenergy, and massiveness in the 6G era via efficientCSIT-limited scheme,” IEEE CommunicationsMagazine, vol. 58, no. 11, pp. 56–61, 2020.

[24] O. L. A. Lopez, F. A. Monteiro, H. Alves,R. Zhang, and M. Latva-Aho, “A low-complexitybeamforming design for multiuser wireless energytransfer,” IEEE Wireless Communications Letters,vol. 10, no. 1, pp. 58–62, 2021.

[25] O. L. Lopez, H. Alves, S. Montejo-Sanchez,R. D. Souza, and M. Latva-aho, “CSI-free rotaryantenna beamforming for massive RF wirelessenergy transfer,” arXiv preprint arXiv:2104.12536,2021.

[26] N. H. Mahmood, S. Bocker, A. Munari, F. Clazzer,I. Moerman, K. Mikhaylov, O. Lopez, O.-S. Park,E. Mercier, H. Bartz et al., “White paper on criticaland massive machine type communication towards6G,” arXiv preprint arXiv:2004.14146, 2020.

[27] I. Nasim and S. Kim, “Adverse impacts of 5Gdownlinks on human body,” in SoutheastCon,2019, pp. 1–6.

[28] Y. Nesterov and A. Nemirovskii, Interior-pointpolynomial algorithms in convex programming.SIAM, 1994.

[29] S. J. Ramson, S. Vishnu, and M. Shanmugam,“Applications of Internet of Things (IoT) - Anoverview,” in 5th International Conference onDevices, Circuits and Systems (ICDCS), 2020, pp.92–95.

[30] O. M. Rosabal, O. L. A. Lopez, H. Alves,S. Montejo-Sanchez, and M. Latva-aho, “On theoptimal deployment of power beacons for massivewireless energy transfer,” IEEE Internet of ThingsJournal, vol. 8, no. 13, pp. 10 531–10 542, 2021.

[31] X. Wang, H. Dai, W. Wang, J. Zheng, N. Yu,G. Chen, W. Dou, and X. Wu, “Practicalheterogeneous wireless charger placement

with obstacles,” IEEE Transactions on MobileComputing, vol. 19, no. 8, pp. 1910–1927, 2020.

[32] J. Xu and R. Zhang, “Energy beamforming withone-bit feedback,” IEEE Transactions on SignalProcessing, vol. 62, no. 20, pp. 5370–5381, 2014.

[33] M. Younan, E. H. Houssein, M. Elhoseny,and A. A. Ali, “Challenges and recommendedtechnologies for the Industrial Internet of Things:A comprehensive review,” Measurement, vol. 151,p. 107198, 2020.

[34] S. Zeadally, F. K. Shaikh, A. Talpur, andQ. Z. Sheng, “Design architectures for energyharvesting in the Internet of Things,” Renewableand Sustainable Energy Reviews, vol. 128, p.109901, 2020.

[35] Y. Zeng, B. Clerckx, and R. Zhang,“Communications and signals design for wirelesspower transmission,” IEEE Transactions onCommunications, vol. 65, no. 5, pp. 2264–2290,2017.

[36] C. Zhang and G. Zhao, “On the deployment ofdistributed antennas of power beacon in wirelesspower transfer,” IEEE Access, vol. 6, pp. 7489–7502, 2018.

[37] Y. Zhu, X. Tian, K. Chi, C. Wen, andY. Zhu, “Real-time power control of wirelesschargers in battery-free body area networks,”in IEEE Global Communications Conference(GLOBECOM), 2019, pp. 1–6.

AUTHOR BIOGRAPHIES

Osmel Martınez Rosabalwas born in Camaguey, Cuba,in 1989. He received theB.Sc. (with distinction, 2013)degree in Telecommunicationsand Electronic Engineeringfrom the Central Universityof Las Villas and the M.Sc.

(with distinction, 2021) in Wireless CommunicationsEngineering from the University of Oulu. From 2013-2018 he served as a Specialist in Telematics at the CubanTelecommunication Company (ETECSA), and afterthat, he started his Master’s studies at the Universityof Oulu. He is currently a PhD student in the WirelessCommunications degree program at the University ofOulu, Finland.

91


Onel L. A. Lopez (S’17-M’20)was born in Sancti-Spıritus,Cuba, in 1989. He receivedthe B.Sc. (1st class honors,2013), M.Sc. (2017) andD.Sc. (with distinction, 2020)degree in Electrical Engineeringfrom the Central University ofLas Villas (Cuba), the Federal

University of Parana (Brazil) and the University ofOulu (Finland), respectively. From 2013-2015 heserved as a specialist in telematics at the Cubantelecommunications company (ETECSA). He is a co-recipient of the 2019 IEEE European Conference onNetworks and Communications (EuCNC) Best StudentPaper Award and a collaborator to the 2016 ResearchAward given by the Cuban Academy of Sciences. Hecurrently holds an Assitant Professorship (tenure track)in sustainable wireless communications engineering inthe Centre for Wireless Communications (CWC), Oulu,Finland.

Hirley Alves (S’10-M’14)received the B.Sc. and M.Sc.degrees from the FederalUniversity of Technology -Parana (UTFPR), Brazil, in2010 and 2011, respectively,both in electrical engineering,and the dual D.Sc. degree fromthe University of Oulu and

UTFPR, in 2015. In 2017, he was an Adjunct Professorin machine-type wireless communications with CWC,University of Oulu, Finland. In 2019, he joined CWCas an Assistant Professor and is currently the Head ofthe Machine-type Wireless Communications Group. Heis actively working on massive connectivity and ultra-reliable low latency communications for future wirelessnetworks, 5G and 6G, full-duplex communications, andphysical-layer security. He leads the URLLC activitiesfor the 6G Flagship Program. He is a co-recipientof the 2017 ISWCS Best Student Paper Award anda co-recipient of the 2016 Research Award from theCuban Academy of Sciences. He has been the organizer,chair, TPC, and tutorial lecturer for several renownedinternational conferences. He was the General Chairof the ISWCS’2019 and the Co-Chair of the 1st 6GSummit, Levi 2019.

Richard Demo Souza (IEEES’01-M’04-SM’12) receivedthe D.Sc. degree in ElectricalEngineering from the FederalUniversity of Santa Catarina(UFSC), Brazil, in 2003.From 2004 to 2016 he waswith the Federal University ofTechnology - Parana (UTFPR),Brazil. Since 2017 he hasbeen with UFSC, where he is

a Professor. His research interests are in the areas ofwireless communications and signal processing. He hasserved as Editor or Associate Editor for the SBrT Journalof Communications and Information Systems, the IEEECommunications Letters, the IEEE Transactions onVehicular Technology, and the IEEE IoT Journal. Heis a co-recipient of the 2014 IEEE/IFIP Wireless DaysConference Best Paper Award, the supervisor of theawarded Best PhD Thesis in Electrical Engineering inBrazil in 2014, and a co-recipient of the 2016 ResearchAward from the Cuban Academy of Sciences.

Samuel Montejo-Sanchez(M’17) was born in Camaguey,Cuba, in 1979. He received theB.Sc., M.Sc., and D.Sc. degreesin telecommunications fromthe Central University of LasVillas (UCLV), Cuba, in 2003,2007, and 2013, respectively.From September 2003 to May2017, he was an AssociateProfessor with the Department

of Telecommunications, UCLV. From February 2011to July 2011, he was a Visiting Ph.D. Student with theFederal University of Technology (UTFPR), Brazil. In2017, he held a post-doctoral position at the Universityof Chile. Since 2018, he has been with the InstitutionalProgram for the Promotion of Research, Development,and Innovation, Universidad Tecnologica Metropolitana(UTEM). He was co-recipient of the 2016 ResearchAward from the Cuban Academy of Sciences. He leadsthe FONDECYT Iniciacion No. 11200659 (TowardHigh Performance Wireless Connectivity for IoT andBeyond-5G Networks) Project. He is co-researcher ofthe FONDECYT Regular 1201893 (IoT goes to Space:Wireless Networking protocols and architectures forIoT networks served by LEO satellite constellations)Project. His research interests are in the areas ofwireless communications, cognitive radio, networkcoding, energy efficiency, vehicular ad hoc networks,physical layer security, machine type communication,and Internet of Things connectivity.

92

Massive Wireless Energy Transfer with Multiple Power ...

Documents

Transcript of Massive Wireless Energy Transfer with Multiple Power ...