wireless information and power transfer for iot

IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 2, APRIL 2019 3257

Wireless Information and Power Transfer for IoTApplications in Overlay Cognitive Radio Networks

Devendra S. Gurjar , Member, IEEE, Ha H. Nguyen , Senior Member, IEEE, and Hoang Duong Tuan

Abstract—This paper proposes and investigates an overlayspectrum sharing system in conjunction with the simultaneouswireless information and power transfer to enable communica-tions for the Internet of Things (IoT) applications. Consideredis a cooperative cognitive radio network, where two IoT devices(IoDs) exchange their information and also provide relay assis-tance to a pair of primary users (PUs). Different from mostexisting works, in this paper, both IoDs can harvest energy fromthe radio-frequency signals received from the PUs. By utiliz-ing the harvested energy, they provide relay cooperation to PUsand realize their own communications. For harvesting energy, atime-switching-based approach is adopted at both IoDs. Withthe proposed scheme, one round of bidirectional informationexchange for both primary and IoT systems is performed in fourphases, i.e., one energy harvesting phase and three informationprocessing phases. Both IoDs rely on the decode-and-forwardoperation to facilitate relaying, whereas the PUs employ selec-tion combining technique. For investigating the performance ofthe considered network, this paper first provides exact expres-sions of user outage probability (OP) for the primary and IoTsystems under Nakagami-m fading. Then, by utilizing the expres-sions of user OP, the system throughput and energy efficiencyare quantified together with the average end-to-end transmissiontime. Numerical and simulation results are provided to give use-ful insights into the system behavior and to highlight the impactof various system/channel parameters.

Index Terms—Cooperative cognitive radio network (CCRN),decode-and-forward (DF), Internet of Things (IoT), Nakagami-mfading, outage probability (OP), simultaneous wireless informa-tion and power transfer (SWIPT).

I. INTRODUCTION

SPECTRUM sharing for the Internet of Things (IoT) is oneof the most promising technologies in the fifth-generation

wireless networks, which allows autonomous devices to realizecommunications for IoT applications in the licensed spec-trum [1], [2]. The concept of IoT has been introduced witha vision to accommodate various physical things, such as sen-sors, mobile phones, home appliances, healthcare gadgets, andeven intelligent furniture, that can be connected through a

Manuscript received October 17, 2018; revised November 2, 2018; acceptedNovember 9, 2018. Date of publication November 20, 2018; date of currentversion May 8, 2019. This work was supported by NSERC Discovery Grant249772-2012. (Corresponding author: Devendra S. Gurjar.)

D. S. Gurjar and H. H. Nguyen are with the Department of Electrical andComputer Engineering, University of Saskatchewan, Saskatoon, SK S7N 5A9,Canada (e-mail: [email protected]; [email protected]).

H. D. Tuan is with the Faculty of Engineering and Information Technology,University of Technology Sydney, Ultimo, NSW 2007, Australia (e-mail:[email protected]).

Digital Object Identifier 10.1109/JIOT.2018.2882207

communication network to exchange information about them-selves and their surroundings. From the communications pointof view, all these things, connected through a network, canbe referred to as IoT devices (IoDs). Addressing communi-cation aspects of such autonomous things (electronic devices)is crucial for bonding them together to form the IoT. Manyapplications are emerging to exploit the features and capabili-ties of IoT. For example, instruments can collaborate with eachother in factories and farms to enhance the performance andefficiency of factory and farm operations [3]. Exploiting IoTcan also be useful in refineries where devices and sensors canbe deployed to form automation in various operations with-out changing the core environment. Likewise, smart homesare made possible by implementing IoT-based home appli-ances. Moreover, IoDs are expected to be the critical entityfor improving traffic management and transportation safety inautonomous driving vehicle industry [4]. For a reliable andubiquitous IoT deployment, two fundamental challenges, i.e.,network lifetime and spectrum scarcity, need to be addressedand they are the focus of this paper.

To prolong the lifetime of wireless communicationnetworks, energy harvesting (EH) from the surrounding envi-ronment has been envisioned as one of promising solutions tocounterpoise power limitations of connected wireless devices.Specifically, it has been observed that the conventional sourcesfor EH, such as solar, wind, thermoelectric, etc., could beunreliable to provide perpetual energy supply as these meth-ods rely on location specific climate and environment [5], [6].Consequently, simultaneous wireless information and powertransfer (SWIPT) technology is gaining tremendous atten-tion due to its ability in providing sustainable and ubiquitouscommunications for numerous wireless communication sce-narios, including IoT. This technique exploits the idea that theradio-frequency (RF) signals can be utilized for both wirelesspower transfer and wireless information transfer at the sametime [6], [7]. Specifically, the antenna of a receiving node firstcaptures the transmitted energy in RF radiation. Then, usingan appropriate circuit [7], the captured energy can be storedin the battery of that node after converting it into the directcurrent form. For enabling SWIPT in wireless networks, threepractical receiver designs have been investigated in the liter-ature, namely, time switching (TS), power splitting (PS), andantenna switching (AS) [8]. In TS-based SWIPT, a receivingnode switches in time between information processing (IP)and EH. In PS-based SWIPT, the node splits the power ofthe received signal for IP and EH. The AS is another way toenable SWIPT in a multiantenna-based system, whereby the

2327-4662 c© 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

https://orcid.org/0000-0002-9674-5378

https://orcid.org/0000-0001-6481-0422

https://orcid.org/0000-0003-0292-6061

3258 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 2, APRIL 2019

strongest antennas are exploited for IP, and others are used forEH or vice-versa [9]. Although the amount of harvested energyfrom RF signals is currently less as compared to other conven-tional sources such as solar energy, its ubiquitous availability(indoor, outdoor, day, or night) makes it more promising forfuture IoT networks.

Spectrum scarcity is another critical design constraint inmassive IoT deployments. Enabling IoT communications inthe industrial, scientific, and medical band is not a long-lastingand feasible solution as most of the wireless technologies oper-ating in this band, e.g., ZigBee, Wi-Fi, and Bluetooth cannotprovide seamless communications with the desired quality ofservice [3]. On the other hand, it is not feasible to rely on thelicensed band communications due to the scarcity of spec-trum and the presence of a massive number of devices inIoT. Therefore, a promising solution is to exploit commu-nications over the licensed spectrum without degrading theperformance of legitimate users. Cooperative spectrum sharingis the suitable mechanism for achieving such an attribute in theIoT networks. For enabling spectrum sharing in such systems,three strategies are commonly adopted in the literature, i.e.,interweave, underlay, and overlay [10]. As such, the inter-weave spectrum sharing suffers from traffic pattern errors ofthe primary system, whereas underlay spectrum sharing mustcomply with strict interference threshold constraint based oninstantaneous channel state information which may be difficultto acquire in practice. As a result, the overlay scheme adoptedin this paper is more appealing for such IoT systems. With thisscheme, the IoDs could provide an incentive to the primaryusers (PUs) for spectrum access, i.e., PUs could get bene-fits of improved performance due to relay assistance, whilein return, relaying IoDs can explore their own transmissionopportunities.

To summarize, incorporating the SWIPT technology withcooperative cognitive radio networks (CCRNs) can effectivelysolve technical problems related to lifetime and spectrumscarcity in massive IoT deployment.

A. Prior Arts

In recent years, SWIPT technique has attracted sig-nificant attention for its inclusion in the relay-basedwireless systems [11]–[17]. Specifically, Li et al. [11]and Lee et al. [12] have considered one-way relay networks,where a relay node can harvest energy from the RF signalsreceived from the source node. To improve the spectral effi-ciency, Men et al. [13], Du et al. [14], and Hu and Lok [15]have utilized the concept of SWIPT with amplify-and-forward(AF) relaying strategy for two-way relay systems. In particu-lar, the work in [13] has jointly optimized the problem of relayselection and power allocation for an SWIPT-enabled asym-metric two-way relay system. Du et al. [14] have considereda TS-based SWIPT scheme at the relay node and investigatedthe outage performance, whereas, Hu and Lok [15] have solvedthe optimization problem concerning power splitting factor andrelay processing matrix for such spectral efficient systems.Different from AF-based two-way systems, Peng et al. [16]and Do et al. [17] have focused on decode-and-forward

(DF) relaying scheme for SWIPT-enabled bidirectional relaysystems.

Besides, research works in [18]–[29] have incorporated theconcept of SWIPT in the spectrum sharing-based systems andcellular networks. Specifically, Yin et al. [18] have proposeda cognitive radio network, where the secondary node canextract energy from RF signals and utilize it for transmittingits own message or providing relay cooperation in differenttime slots. In particular, the authors have provided optimalconditions to maximize the system throughput for two sce-narios, namely, cooperative mode and noncooperative mode.Different from [18], the secondary node can transmit bothprimary and secondary signals simultaneously in [19] withthe overlay mode. For this set up, the authors have derivedexact expressions of outage probability (OP) for both primaryand secondary systems over Rayleigh fading channels. Further,research works [20] and [21] have adopted an underlay cogni-tive radio scenario with EH and analyzed the systems’ outageperformance. By extending the system model of [20] and [21],Kalamkar and Banerjee [22] have considered multiple pri-mary transmitters and receivers and evaluated the outage andergodic capacity performance for the secondary system inthe presence of multiple primary interferences. Furthermore,Verma et al. [23] have introduced one-way cooperative cogni-tive radio network (CCRN) with energy assisted DF relayingand investigated the OP and throughput for both systems.With the similar system model as in [23], Yan and Liu [24]have studied opportunistic relaying by employing a dynamicSWIPT protocol. Nguyen et al. [25] have derived the expres-sions of OP for EH-enabled CCRN under Nakagami-m fading.In contrast to the one-way CCRN [18]–[25], a cognitive two-way relay network with EH has been investigated in [26] underRayleigh fading.

Recently, a few works [27]–[29] have exploited the ben-efits of EH for IoT applications in the licensed band. Inparticular, Yang et al. [27] have studied resource allocationfor a machine-to-machine enabled cellular network with EHby focusing on two different strategies, i.e., nonorthogonalmultiple access and time division multiple access. Further,in [28], software-defined networking has been proposed tooptimize network management and to control EH for IoTapplications. Very recently, Yan et al. [29] have developeda mathematical framework for the design and analysis ofrelay-assisted underlay cognitive radio networks with EH andinvestigated the systems’ outage performance.

Most of the works as discussed above either considered EHwith underlay cognitive radio networks or one-way CCRN,where the information exchange for both systems is carriedout unidirectionally. To the authors’ best knowledge, no workhas yet considered the concept of SWIPT in a cognitive radionetwork with bidirectional primary and IoT transmissions.

B. Main Contributions

Focusing on the critical constraints of IoT deployments,this paper proposes a DF-based two-way CCRN with EH toleverage the benefits of both spectrum sharing and SWIPTtechnique. Herein, a pair of IoDs harvests energy from the

GURJAR et al.: WIRELESS INFORMATION AND POWER TRANSFER FOR IoT APPLICATIONS IN OVERLAY COGNITIVE RADIO NETWORKS 3259

RF signals by applying the TS technique and utilizes theaccumulated energy for providing relay cooperation and itsown information exchange. Moreover, the proposed schemeimproves the overall spectrum efficiency and resolves thecrucial reliability issue for primary links by enabling relayassistance from two IoDs in consecutive phases. For relaying,a DF-based operation is considered at the IoDs. With DF relay-ing, a selection combining (SC) technique is employed at thePUs to exploit multiple copies of their intended signals. Themajor contributions of this paper are summarized as follows.

1) This paper introduces an overlay spectrum sharingscheme with EH to enable IoT communications in thelicensed spectrum.

2) With the proposed scheme, exact expressions of userOP for both primary and IoT systems are derived underNakagami-m fading. Then, the expression of systemthroughput is obtained for delay-limited transmissions.

3) This paper also provides an expression of overall energyefficiency for the considered system. Moreover, the crit-ical value of spectrum sharing factor is deduced forwhich the OP of primary links with the proposed schemeexhibits the same OP as of direct communications(without spectrum sharing).

4) To evaluate the delay performance, this paper formulatesan expression for the average end-to-end transmissiontime of the primary system.

5) This paper reveals impacts of different system andchannel parameters through extensive numerical andsimulation results. The obtained results help to addresssome key physical-layer design aspects for practicaldeployments of such complex systems.

The rest of this paper is organized as follows. In Section II,the system model and proposed scheme are described.Specifically, Sections II-A and II-B present the consideredEH model and IP signaling, respectively, and derive end-to-end instantaneous signal-to-noise ratios (SNRs). For eval-uating the system performance, expressions of OP, systemthroughput, and energy efficiency are obtained in Section III.Numerical and simulation results are provided and discussedin Section IV. Finally, Section V concludes this paper.

Notations: Throughout this paper, fX(·) and FX(·) repre-sent the probability density function (PDF) and the cumulativedistribution function (CDF) of a random variable X, respec-tively, and Pr[ · ] denotes probability. �[·, ·], ϒ[·, ·], and �[ · ]represent, respectively, the upper incomplete, the lower incom-plete, and the complete gamma functions [30, eq. (8.350)].E[ · ] and Kv(·) denote expectation operation and vth ordermodified Bessel function of second kind [30, eq. (8.432.1)],respectively, whereas Wu,v(·) represents Whittaker function[30, eq. (9.222)]. Table I lists the fundamental notations andparameters.

II. SYSTEM AND SCHEME DESCRIPTION

Fig. 1 depicts a SWIPT-enabled cognitive radio system con-sidered in this paper. Two primary nodes PUa and PUb wantto communicate to each other, but due to heavy shadowingor blockage, the direct link between them is not good enough

TABLE ILIST OF PARAMETERS AND THEIR PHYSICAL MEANING/EXPRESSION,

WHERE i, i ∈ {1, 2} FOR i �= i, AND j, j ∈ {a, b} FOR j �= j

Fig. 1. System model for SWIPT-enabled bidirectional cognitive radionetwork.

to achieve specified target rates. A pair of proximate IoDs,1

referred to as IoD1 and IoD2, provides relay assistance tothe primary transmissions and gets the opportunity to real-ize its bidirectional communications over the same licensedband. All the participating nodes (primary and IoT) operatein half-duplex mode and are equipped with single antennadevices.

The EH and IP processes can be done in separately ded-icated time slots. To this end, one block duration is dividedinto two phases, i.e., EH phase and IP phase. During the EHphase, both IoDs harvest energy from the RF signals and storethis energy to utilize it for providing relaying to the primarysystem and for their own transmissions. After EH phase, one

1One can consider a generalized scenario of the considered system byassuming the presence of several potential pairs of IoDs. Amongst them, thebest pair can be selected through some appropriate selection process (see [31]and references therein).


Fig. 2. Frame structure of TS-based SWIPT in the proposed cognitive radionetwork.

round of end-to-end information exchanges between two PUsand two IoDs takes three IP phases. In the first IP phase ormultiple access channel (MAC) phase, both PUa and PUb

transmit their signals to both IoD1 and IoD2. After that, bothIoD1 and IoD2 first decode the primary signals. On successfuldecoding, they apply bit-wise XOR operation to generate a re-encoded primary signal for performing DF operation. AmongIoD1 and IoD2, the first relaying IoD is the one who wants tocommunicate first to the other one. In the second IP phase orthe first broadcast channel (BC) phase, the first relaying IoDbroadcasts the encoded primary signal after adding its ownsignal intended for the other IoD. Likewise, in the third IPphase or the second BC phase, the other IoD applies the sameprocedure as done by the first one. If any IoD fails to decodeboth primary signals in the first IP phase, it transmits one-bitnegative acknowledgment in the respective BC phase. At thereceiving PUs, the SC scheme is employed to make use oftwo intended signal copies.

A block fading scenario is considered in this paper, wherechannel gains remain unchanged for one block duration. Thechannel gains of the links from PUj to IoDi and from IoDi toPUj are denoted as hj,i and hi,j, respectively, for i ∈ {1, 2} and

j, j ∈ {a, b}, where j �= j. Likewise, the channel gain of the linkfrom IoDi to IoDi is denoted by hi,i. All the channel gains ofindividual hops are assumed to obey reciprocity, i.e., hi,j = hj,iand hi,i = hi,i. Further, hj,i for i ∈ {1, 2} and j ∈ {a, b} followsNakagami-m distribution with fading severity mij and averagepower �ij. The integer-valued fading parameters are adoptedfor modeling Nakagami-m channels, through which a widevariety of wireless fading scenarios can be represented. It isalso assumed that all the receiving terminals are affected bythe additive white Gaussian noise (AWGN) with zero meanand variance σ 2.

A. Energy Harvesting

In TS-based EH approach as adopted in [6] and [32], onetime slot is dedicated for harvesting energy from the RF sig-nals and another slot for processing the information. In thispaper, one transmission block duration T is divided into twoslots of durations βT and (1 −β)T as shown in Fig. 2, where0 < β < 1. Herein, βT is allocated for harvesting energy,whereas (1 − β)T is dedicated for information exchanges ofprimary and IoT systems. The value of β that reflects theamount of harvested energy at the IoDs has a strong influenceon the system performance in terms of achievable throughput

and link reliability. The IP phase is further divided into threeequal time slots, i.e., one MAC phase and two BC phases.

In the EH phase, the harvested energy at IoDi can beformulated as [32]

Ei = ηiβT(

Pa|ha,i|2 + Pb|hb,i|2)

(1)

for i ∈ {1, 2}, where 0 < ηi < 1 represents the energy conver-sion efficiency of the EH circuit at IoDi, and Pa and Pb aretransmit powers at PUa and PUb, respectively. By using (1),the transmit power at IoDi can be expressed as

Pi = 3ηiβ

1 − β

(Pa|ha,i|2 + Pb|hb,i|2

). (2)

Without loss of generality, this paper assumes that all the har-vested energy will be used for broadcasting the informationsignals at IoDs [33].

B. Information Processing

After the EH phase, PUa and PUb transmit unit-energysymbols xa and xb in the first IP phase (MAC phase), respec-tively. Thereby, the received signals at IoD1 and IoD2 can beexpressed, respectively, as

y1 = √Paha,1xa +√

Pbhb,1xb + n1 (3)

and

y2 = √Paha,2xa +√

Pbhb,2xb + n2 (4)

where n1 ∼ CN (0, σ 2) and n2 ∼ CN (0, σ 2) are AWGN com-ponents at IoD1 and IoD2, respectively. After receiving theconcurrent primary transmissions, both IoDs first decode xa

and xb and then broadcast the combined primary signal usingthe DF operation.

1) Decode-and-Forward Operation: The IoDs can performDF operation only when they successfully decode both the pri-mary signals in the first IP phase. After decoding xa and xb, theIoDs obtain a re-encoded symbol by performing bit-wise XOR

operation (xa ⊕ xb) and utilize it for further transmissions. Asin some practical applications, e.g., video streaming, gaming,and file transfer, the required data rates may be asymmetric fortwo opposite traffic flows. Therefore, the bit sequences corre-sponding to the primary signals may have different lengths.For ensuring the same bit sequence length, zero padding canbe done on the shorter sequence [34].

Let IoDi be the first relaying node and it broadcasts thesymbol xa ⊕ xb by adding its own symbol xi intended for theother IoD. If μi represents the power splitting factor (resourceallocation factor for primary transmissions) at IoDi, then thesignal transmitted from IoDi in the second IP phase (first BCphase) can be expressed as

xBCi = √

μiPi(xa ⊕ xb) +√(1 − μi)Pixi (5)

where Pi is the transmit power at IoDi. Further, the receivedsignal at PUj in the first BC phase can be given as

yi,j = √μiPihi,j(xa ⊕ xb) +√

(1 − μi)Pihi,jxi + nj (6)

where nj is AWGN variable at PUj. As both PUs know theirrespective transmitted signals, they can extract the desiredinformation from the combined primary signal.


2) End-to-End Instantaneous SNRs: Considering the IoTsignal (interference to PUs) as additional noise and invok-ing the expression of Pi from (2) into (6), the end-to-endinstantaneous SNR at PUj can be expressed as

γi,j = μiζij|hi,j|4 + μiζij|hi,j|2|hi,j|2(1 − μi)ζij|hi,j|4 + (1 − μi)ζij|hi,j|2|hi,j|2 + 1

(7)

where ζij = [(3ηiρjβ)/(1 − β)] and ζij = [(3ηiρjβ)/(1 − β)]

with ρj = Pj/σ2 and ρj = Pj/σ

2, for i ∈ {1, 2}, j, j ∈ {a, b},j �= j. On the other hand, the received signal at IoDi in thesecond IP phase can be expressed as

yi,i = √μiPihi,i(xa ⊕ xb) +√

(1 − μi)Pihi,ixi + ni (8)

where ni is AWGN variable at IoDi. Since both IoDs can havethe knowledge of the primary signals after decoding xa andxb, they can remove the primary interference from the receivedsignal. Thereby, the effective instantaneous SNR at IoDi in thefirst BC phase can be given as

γi,i = (1 − μi)|hi,i|2(ζij|hi,j|2 + ζij|hi,j|2

). (9)

Similar to the second IP phase, IoDi also broadcasts thecombined primary signal (xa ⊕ xb) in the third IP phase (sec-ond BC phase) by adding its own signal xi intended for IoDi.Likewise, the effective instantaneous SNR at PUj in the thirdIP phase can be given as

γi,j = μiζij|hi,j|4 + μiζij|hi,j|2|hi,j|2(1 − μi

)ζij|hi,j|4 + (

1 − μi

)ζij|hi,j|2|hi,j|2 + 1

(10)

where ζij = [(3ηiρjβ)/(1 − β)] and ζij = [(3ηiρjβ)/(1 − β)]

with ρj = Pj/σ2 and ρj = Pj/σ

2, for i ∈ {1, 2}, j, j ∈ {a, b},j �= j. Following similar steps as applied to obtain (9), theeffective instantaneous SNR at IoDi can be give as

γi,i = (1 − μi

)|hi,i|2(ζij|hi,j|2 + ζij|hi,j|2

). (11)

III. PERFORMANCE ANALYSIS

This section first obtains closed-form expressions for userOP of primary and IoT systems under Nakagami-m fadingenvironment. Using these OP results, expressions of systemthroughput and energy efficiency are then provided for theconsidered system.

A. Outage Probability of Primary System

The OP is an important performance metric to quantifythe link reliability of a wireless system over fading chan-nels. With the proposed scheme, PUs can have two copiesof their intended signals received from two IoDs in consecu-tive IP phases. Consequently, the outage event takes place atany PU if its instantaneous data rate achieved by exploitingboth signal copies falls below a predefined target data rate.Mathematically, the user OP for the primary system can be

computed as [35]

Pout,j = Pr[Qi] Pr[Qi

]Pr[Rsc,j < rth

]

+ Pr[Qi](1 − Pr

[Qi

])Pr[Ri,j < rth

]

+ (1 − Pr[Qi]) Pr[Qi

]Pr[Ri,j < rth

]

+ (1 − Pr[Qi])(1 − Pr

[Qi

])(12)

for j ∈ {a, b}, i, i ∈ {1, 2}, i �= i. Here, Pr[Qi] denotesthe probability of successful decoding of xa and xb in thefirst IP phase at IoDi. In (12), the first term accounts forthe case when both IoDs successfully decode both signals(xa and xb). On the other hand, the second and third termsrepresent the cases when only one IoD decodes the primarysignals successfully. The forth term corresponds to case whenboth IoDs fail to decode the primary signals. Furthermore,rth = max(ra, rb), where ra and rb denote the target datarates at PUa and PUb, respectively. When the SC techniqueis employed at PUs to select the best signal copy (based onthe maximum SNR), the instantaneous data rate can be givenas R sc,j = ((1 − β)/3) log2(1 + max(γi,j, γi,j)). Similarly, theinstantaneous data rate related to individual signal copy at PUscan be expressed as Ri,j = ((1−β)/3) log2(1+γi,j). In the firstIP phase, a nonorthogonal multiple access scenario is consid-ered, where both PUs transmit their signals xa and xb to IoDsover the same frequency band [36]. As such, the expression forcorrect decoding of both primary signals at IoDi is providedin Lemma 1 by following the procedure described in [37] fordecoding of simultaneously received signals.

Lemma 1: The expression of Pr[Qi], for i ∈ {1, 2}, in (12)can be given as

Pr[Qi] =⎧⎨⎩PQi , for mij

�ijρj�= mij

�ijρj

PQi , for mij�ijρj

= mij�ijρj

(13)

where PQi and PQi are given by (14) and (15), shown at thebottom of the next page, with Ci = (ϕjϕj − ϕj)/ρj, Di =(ϕjϕj − 1)/ρj, ϕj = 23rj/(1−β) and ϕj = 23rj/(1−β).

Proof: Consider Yi � |hj,i|2 and Zi � |hj,i|2 for

i ∈ {1, 2}, j, j ∈ {b, m} with j �= j, which areGamma-distributed random variables with PDFs fYi(yi) =(mij/�ij)

mij[(ymij−1i )/(�[mij])]e−[(mijyi)/(�ij)], yi ≥ 0, and

fZi(zi) = (mij/�ij)mij[(z

mij−1

i )/(�[mij])]e−[(mijzi)/(�ij)], zi ≥ 0.

In the first IP phase, for decoding the primary signals at IoDi,the following three conditions should be satisfied [34], [38]:⎧⎪⎪⎨

⎪⎪⎩

1−β3 log2

(1 + ρjYi

) ≥ rj1−β

3 log2

(1 + ρjZi

)≥ rj

1−β3 log2

(1 + ρjYi + ρjZi

)≥ rj + rj

(16)

where ρj = Pj/σ2. Based on (16), one can formulate the

expression of Pr[Qi] as

Pr[Qi] =∫ ∞

ϕj−1

ρj

fYi(yi)

∫ ∞ϕj−1ρ

j

fZi(zi)dzidyi

−∫ Ci

ϕj−1

ρj

fYi(yi)

∫ Di−�jyi

ϕj−1ρ

j

fZi(zi)dzidyi (17)


where Ci and Di are defined after (13) and �j = ρj/ρj. Onrearranging the limits of (17), Pr[Si] can be represented as

Pr[Qi] =∫ ∞

ϕj−1

ρj

fYi(yi)

∫ ∞ϕj−1ρ

j

fZi(zi)dzidyi

−∫ Ci

ϕj−1

ρj

fYi(yi)

∫ Di−�jyi

0fZi(zi)dzidyi

+∫ Ci

ϕj−1

ρj

fYi(yi)

∫ ϕj−1ρ

j

0fZi(zi)dzidyi. (18)

By inserting the respective PDFs in (18) and utilizing[30, eq. (3.381)], one can express it as

PQi = 1 −ϒ

[mij,

mij�ij

(ϕj−1ρj

)]

�[mij

]⎛⎝1 −

ϒ[mij,

mij�ij

Ci

]

�[mij]

⎞⎠

−ϒ[mij,

mij�ij

Ci

]

�[mij] +

(mij�ij

)mij

�[mij] e

− mij

�ijDi

×mij−1∑k=0

(mij�ij

)k

k!

k∑q=0

(k

q

)Dq

i

(−�j)k−q

×∫ Ci

ϕj−1

ρj

ymij−1+k−qi e

−(

mij�ij

− �jmij�

ij

)yi

dyi. (19)

On solving the last integration of (19) with the help of[30, eq. (2.321)], the expression of Pr[Qi] can be given asin (13).

Next, the probability term Psc,j � Pr[R sc,j < rth] in (12)can be formulated as

Psc,j = Pr[max

(γi,j, γi,j

)< γth

]=

2∏i=1

Fγi,j(γth) (20)

where γth = 23rth/(1−β)−1. Further, the expression of Fγi,j(γth)

is given in the following lemma.Lemma 2: The CDF Fγi,j(γth) can be expressed as

Fγi,j(γth) ={

Fγi,j , for γth1+γth

< μi < 11, for 0 < μi <

γth1+γth

(21)

where

Fγi,j =ϒ

[mij,

mij�ij

√γthζijφi

]

�[mij] −

mij−1∑k=0

(mij

�ijφiζij

)k

k!

×(

mij�ij

)mij

�[mij]

k∑n=0

(k

n

)γ n

th

(−ζijφi)k−n

Im (22)

and φi = μi − (1 − μi)γth. Further, the expression of Im isgiven for two cases, as

Im =⎧⎨⎩

I1, formijζij

�ijζij�= mij

�ij

I2, formijζij

�ijζij= mij

�ij

(23)

where I1 and I2 are given in (24) and (25), shown at the bottomof the next page.

Proof: The CDF Fγi,j(γth) can be formulated using (7) as

Fγi,j(γth) = Pr

⎡⎣ μiζijY2

i + μiζijYiZi

(1 − μi)(ζijY2

i + ζijYiZi

)+ 1

< γth

⎤⎦

= Pr

[Zi <

((1 − μi)γth − μi)ζijY2i + γth

(μi − (1 − μi)γth)ζijYi

]. (26)

When the term (μi − (1−μi)γth) ≤ 0, the CDF Fγi,j(γth) = 1.On the other hand, when (μi−(1−μi)γth) > 0, the expression

PQi = 1 −ϒ

[mij,

mij�ij

(ϕj−1ρj

)]

�[mij

]⎛⎝1 −

ϒ[mij,

mij�ij

Ci

]

�[mij]

⎞⎠−

ϒ[mij,

mij�ij

Ci

]

�[mij] +

(mij�ij

)mij

�[mij] e

− mij

�ijDi

mij−1∑k=0

(mij�ij

)k

k!

×k∑

q=0

(k

q

)Dq

i (−�j)k−q

mij+k−q−1∑p=0

(mij+k−q−1p

)(−1)pp!

(ρjmijρj�ij

− mij�ij

)p+1

⎛⎝e

(ρjmijρ

j�

ij− mij

�ij

)CiCmij+k−q−1−p

i − e

(ρjmijρ

j�

ij− mij

�ij

)(ϕ

j−1

ρj

)

×(

ϕj − 1

ρj

)mij+k−q−1−p⎞⎠ (14)

PQi = 1 −ϒ

[mij,

mij�ij

(ϕj−1ρj

)]

�[mij

]⎛⎝1 −

ϒ[mij,

mij�ij

Ci

]

�[mij]

⎞⎠−

ϒ[mij,

mij�ij

Ci

]

�[mij] +

(mij�ij

)mij

�[mij] e

− mij

�ijDi

×mij−1∑k=0

(mij�ij

)k

k!

k∑q=0

(k

q

)Dqi

(−�j)k−q

mij + k − q

⎛⎝Cmij+k−q

i −(

ϕj − 1

ρj

)mij+k−q⎞⎠ (15)


of Fγi,j(γth) can be formulated as

Fγi,j(γth) =∫ √

γthζijφi

yi=0fYi(yi)

∫ γth−ζijφiy2i

φiζijyi

zi=0fZi(zi)dzi dyi (27)

where φi is defined after (22). After inserting the respectivePDFs in (27) and applying [30, eqs. (3.381.1) and (8.352.1)],one obtains

Fγi,j(γth)

=ϒ[mij,

mij�ij

√γthζijφi

]

�[mij] −

mij−1∑k=0

(mij

�ijφiζij

)k

k!

×(

mij�ij

)mij

�[mij]

k∑n=0

(k

n

)γ n

th

(−ζijφi)k−n

×∫ √

γthζijφi

yi=0y

mij+k−2n−1i e

−m

ijγth

�ijφiζij

yi e

(m

ijζij

�ijζij− mij

�ij

)yi

dyi. (28)

As the solution of (28) for the general case is mathematicallyintractable, this paper provides the solutions for two cases.

Case 1: For [(mijζij)/(�ijζij)] �= (mij/�ij), after apply-ing Maclaurin series expansion for the last exponential termof (28), one has

Fγi,j(γth) =ϒ[mij,

mij�ij

√γthζijφi

]

�[mij] −

mij−1∑k=0

(mij

�ijφiζij

)k

k!

(mij�ij

)mij

�[mij]

×k∑

n=0

(k

n

)γ n

th

(−ζijφi)k−n

∞∑p=0

(mijζij

�ijζij− mij

�ij

)p

p!

×∫ √

γthζijφi

yi=0y

p+mij+k−2n−1i e

− mijγth

�ijφiζij

yi dyi. (29)

By applying change of variables with t =[(mijγth)/(�ijφiζijyi)], and then utilizing [30, eq. (3.381.6)],the solution is obtained as given in (21) and (24).

Case 2: For [(mijζij)/(�ijζij)] = (mij/�ij), making changeof variables t = [(mijγth)/(�ijφiζijyi)] in (28), and then using[30, eq. (3.381.6)], one obtains the solution as given in (21)and (25).

The same derivation steps can be followed to obtain expres-sions of other probabilities of (12) by replacing i with i andvice-versa in Lemma 2. On inserting (13)–(25) into (12), onecan get the desired OP expression for the primary system.

B. Outage Probability of IoT System

An outage event occurs at the IoD if any IoDs fail to decodethe primary signals or the instantaneous rate achieved at thatnode falls below a predefined target rate. Following this, theuser OP of the IoT system can be formulated as

Pout,i = 1 − Pr[Qi] Pr[Qi

]Pr[Ri,i > ri

](30)

for i, i ∈ {1, 2} and i �= i, where ri is the target rate at IoDi.Moreover, Ri,i = ((1 − β)/3) log2(1 + γi,i) is the instanta-neous rate at IoDi. The decoding probabilities are alreadyderived in Lemma 1 and the remaining term can be calculatedas Pr[Ri,i > ri] = 1 − Pr[Ri,i < ri] � 1 − Fγi,i

(γ i). Further,the expression of Fγi,i

(γ i) is given in the following lemma.Lemma 3: The CDF Fγi,i

(γ i) can be expressed as

Fγi,i

(γ i

) = 1 −mij−1∑q=0

(mijγ i

�ij(1−μi)ζij

)q

�[mii

]q!

(mii

�ii

)mii

× 2

(mijγ i�ii

�ijmii(1 − μi)ζij

)mii−q

2

× Kmii−q

(2

√mijγ imii

�ij(1 − μi)ζij�ii

)− I3 (31)

where γ i = 23ri/(1−β) − 1 and the expression of I3 is given as(32), shown at the bottom of the next page.

Proof: X � |hi,i|2 is a Gamma dis-tributed random variable with PDF as fX(x) =[(mii)/(�ii)]

mii[(xmii−1)/(�[mii])]e−[(miix)/(�ii)], x ≥ 0.

On utilizing (9), the CDF Fγi,i(γ i) can be formulated as

Fγi,i

(γ i

) = Pr[(1 − μi)ζijYiX + (1 − μi)ζijZiX < γ i

]

= Pr

[Zi <

γ i − (1 − μi)ζijYiX

(1 − μi)ζijX

]. (33)

I1 =∞∑

p=0

(mijζij

�ijζij− mij

�ij

)p

p!

(mijγth

�ijφiζij

)p+k−2n+mij(

mijγth

�ijφiζij

√ζijφi

γth

)− (p+k−2n+mij+1)2

e− m

ijγth

2�ijφiζij

√ζijφiγth

× W− (p+k−2n+mij+1)

2 ,1−(p+k−2n+mij+1)

2

(mijγth

�ijφiζij

√ζijφi

γth

)(24)

I2 =(

mijγth

�ijφiζij

)k−2n+mij(

mijγth

�ijφiζij

√ζijφi

γth

)− (k−2n+mij+1)2

e−

mijγth

2�ijφiζij

√ζijφiγth

× W− (k−2n+mij+1)

2 ,1−(k−2n+mij+1)

2

(mijγth

�ijφiζij

√ζijφi

γth

)(25)


Based on (33), Fγi,i(γ i) can be formulated in integration

form as

Fγi,i

(γ i

) =∫ ∞

x=0fX(x)

∫ γi

(1−μi)ζijx

yi=0fYi(yi)

× FZi

(γ i − (1 − μi)ζijyix

(1 − μi)ζijx

)dyi dx. (34)

Now, after applying some mathematical formulations andutilizing [30, eq. (3.471.9)], (34) can be expressed as

Fγi,i

(γ i

) = 1 −mij−1∑q=0

(mijγ i

�ij(1−μi)ζij

)q

�[mii

]q!

(mii

�ii

)mii

× 2

(mijγ i�ii

�ijmii(1 − μi)ζij

)mii−q

2

× Kmii−q

(2

√mijγ imii


)−(

mij�ij

)mij

�[mij]

×∫ ∞

x=0fX(x)e

− mijγ

i�

ij(1−μi)ζijx

mij−1∑k=0

(mij

�ij(1−μi)ζijx

)k

k!

×k∑

s=0

(k

s

)γ s

i(−1)k−s((1 − μi)ζijx

)k−s

×∫ γ

i(1−μi)ζijx

yi=0y

mij−1+k−si e

(m

ijζij

�ijζij− mij

�ij

)yi

dyidx. (35)

On applying Maclaurin series expansion to the terme[(mijζij)/(�ijζij)] in (35) and using [30, eqs. (3.471.9)], thedesired solution is obtained as given in Lemma 3.

On inserting (13) and (31) in (30), one can obtain the desiredOP expression for the IoT system.

C. System Throughput

For a delay-limited scenario, the system throughput for theconsidered cognitive radio network can be defined as thesum of average target rates of two primary and two IoT

transmissions that can be successfully achieved over fadingchannels [32], [33]. Therefore, one can express the systemthroughput as

ST = Sp + SIoT (36)

where Sp and SIoT are throughputs of the primary and IoTsystems, respectively. By utilizing the expressions of individ-ual OP of primary and IoT links, the expressions of Sp andSIoT are given as [33]

Sp = (1 − β)

3

[(1 − Pout,a

)ra + (

1 − Pout,b)rb]

(37)

and

SIoT = (1 − β)

3

[(1 − Pout,1

)r1 + (

1 − Pout,2)r2]

(38)

where Pout,a and Pout,b are given in (12) and Pout,1 and Pout,2are defined in (30).

D. Energy Efficiency

Designing energy efficient wireless networks is gettingmuch attention nowadays and it proceeds toward relying ongreen communication technologies [39]. Consequently, it isrelevant to examine the energy efficiency and investigate theimpact of different parameters. The energy efficiency of theconsidered EH-based cognitive radio can be defined, based onthe classical definition, as the ratio of total amount of datadelivered to the total amount of consumed energy [32]. For adelay-limited scenario, the expression of energy efficiency is

ηTSEE = ST(

1+2β3

)(Pa + Pb)

(39)

where ST is given in (36).

E. Effective Spectrum Sharing

When both PUs communicate to each other directly withoutreceiving relay cooperation of IoDs, the achievable rate RD

j,jat the primary nodes can be expressed as

RDj,j

= 1

2log2

(1 + Pj|hj,j|2

σ 2

). (40)

I3 =(

mii�ii

)mii

�[mii

]mij−1∑k=0

(mij

�ij(1−μi)ζij

)k

k!

k∑s=0

(k

s

)γ s

i(−1)k−s((1 − μi)ζij

)k−s∞∑

p=0

(mijζij

�ijζij

)p

p!

(mij

�ij

)−p−k+s

× 2�[p + mij + k − s

]⎛⎜⎝(

mijγ i�ii

�ij(1 − μi)ζijmii

)mii−s

2

Kmii−s

(2

√mijγ imii


)−

p+mij+k−s−1∑l=0

(mijγ i

�ij(1−μi)ζij

)l

l!

×((

mijγ i

�ij(1 − μi)ζij

+ mijγ i


)�ii

mii

)mii−s−l

2

× Kmii−s−l

⎛⎝2

√√√√(

mijγ i


+ mijγ i


)mii

�ii

⎞⎠⎞⎟⎠ (32)


Here, the prelog factor 1/2 appears due to the fact that twosuccessive phases are required to realize end-to-end transmis-sions. Hereby, the OP for this direct transmission link can begiven as

PDout,j = Pr

[Pj|hj,j|2

σ 2< γj

]=

ϒ

[mjj,

mjjγjσ2

�jjPj

]

�[mjj

] (41)

where γj = 22rj − 1. In fact, the spectrum sharing for IoDscan be permissible until it does not affect the required outageperformance of the primary system. However, for effectivespectrum sharing, the primary links of the considered EH-enabled system should attain equal or lower OP than that of thedirect transmissions (without spectrum sharing) of the primarysystem for the same predefined target rate [35], [40], i.e.,

Pout,a ≤ PDout,a and Pout,b ≤ PD

out,b. (42)

On utilizing (42), one can obtain the critical value of powersplitting factor (say μ�) for which the system can offer effec-tive spectrum sharing, where μ� ≤ μi. Although the analyticalevaluation of μ� directly from (42) appears mathematicallyintractable, numerical methods can be used to obtain thedesired value.

F. Average End-to-End Transmission Time

The estimation of end-to-end transmission time for a packetto reach the destination is useful in the design of cognitiveradio networks to meet latency requirements. According toShannon’s third theorem, the transmission time is inverselyproportional to the achievable transmission rate of the corre-sponding channel [41]. Therefore, the time taken by a packet toarrive at the destination Sm after transmitting from the sourceSl is given by

�l,m = LW log2

(1 + γl,m

) = Lln(1 + γl,m

) (43)

where L is the length of the packet, W is the bandwidthof the channel, and L = L ln(2)/W . Further, it is assumedthat the transmission time and processing time of feed-back/acknowledgment signals are negligible as compared tothe packet transmission time and the transmitted packet arrivesat the destination node before time-out [42], [43]. With theconsidered system, the total transmission time is given as

�j,j = TEH + TIP (44)

where TEH denotes the time taken for EH and TIP representsthe time taken for IP and broadcasting. Since TEH = βT andTIP = (1−β)T , the relationship between TEH and TIP is TEH =(β/1 − β)TIP. Now, (44) can be expressed as

�j,j = 1

1 − βTIP. (45)

For the considered relaying scheme, the average end-to-endtransmission time from PUj to PUj is determined as

�j,j = 1

1 − β

×(

Pr[Qi](1 − Pr

[Qi

])(E[max

(�j,i, �j,i

)]+ E

[�i,j

])

+ Pr[Qi

](1 − Pr[Qi])

(E[max

(�j,i, �j,i

)]+ E

[�i,j

])

+ Pr[Qi]Pr[Qi

](E[max

(�j,i,�j,i

)]+ E

[�i,j

]

+ E[�i,j

]))(46)

where i, i ∈ {1, 2} and j, j ∈ {a, b}, with i �= i, j �=j. Furthermore, �j,i, �j,i, �i,j, and �i,j can be obtainedfrom (43) by inserting respective expressions of the instan-taneous SNRs. If any relaying device fails to decode primarysignals in the MAC phase, then it aborts broadcasting of thecombined signal in the assigned BC phase. As a result, alimited feedback signal will be transmitted by that device toacknowledge all the corresponding nodes. For brevity, it isassumed that the time taken in this process is negligible [43].For the case when both relaying IoDs successfully decode theprimary signals and broadcast in two consecutive BC phases,the average end-to-end transmission time will predominantlydepend on the receiving time of both signal copies at thedestination node for exploiting the SC technique. Note that,the exact derivation of (46) is highly intractable due to theinvolved complexity. As such, (46) is computed with MonteCarlo simulation in MATLAB. On the other hand, the averageend-to-end transmission time for the direct transmission fromPUj to PUj can be formulated as

�(D)

j,j= E

⎡⎣ L

ln(

1 + γ Dj,j

)⎤⎦. (47)

For numerical results, (46) and (47) are used to compare theend-to-end transmission time of our proposed scheme and thedirect-link transmission time.

IV. NUMERICAL AND SIMULATION RESULTS

This section presents numerical and simulation results anddiscusses the impact of key system/channel parameters onthe performance of the considered SWIPT-enabled cognitiveradio system. For all the numerical results, it is assumed thatPa = Pb = P and define P/σ 2 as the transmit SNR. Further,this paper adopts a path-loss model where the variances ofchannel gains are defined in terms of the corresponding dis-tances between two nodes and the path-loss exponent. As such,for links PUa → IoD1 and PUa → IoD2, the variances ofchannel coefficients are defined as �1a = d−ν

1a and �2a = d−ν2a ,

respectively. Similarly, for PUb → IoD1, PUb → IoD2, andIoD2 → IoD1 links, the variances of channel coefficientsare �1b = d−ν

1b , �2b = d−ν2b , and �12 = d−ν

12 , respec-tively. All simulation results were obtained by considering thatd1a = d2a = 1, d1b = d2b = 0.9, and d12 = 1 with path-lossexponent ν = 3. Moreover, fading severity parameters are setas m1a = m2a = ma and m1b = m2b = mb. The values of othersystem/channel parameters vary in different figures and are


TABLE IINUMBER OF TERMS REQUIRED IN INFINITE SERIES OF (12) AND (30) FOR

ACHIEVING ACCURACY UP TO FIRST SEVEN DECIMAL PLACES

Fig. 3. OP versus SNR curves for PUb → PUa link.

specified therein. For obtaining numerical values in Table II,the parameters are set as ma = mb = m12 = 2, β = 0.2,η1 = η2 = 0.7, and ra = rb = r1 = r2 = 1/3. As depictedin Table II, the infinite series involved in (23) and (31) aretruncated to include the first 15 terms for achieving the suffi-cient accuracy (first seven decimal places) in all the analyticalresults.

A. Outage Probability With SNR

In obtaining Fig. 3, the system parameters are set asη1 = η2 = 0.7, β = 0.2, and ra = rb = rth = 1/3 bps/Hz. Thisfigure shows the OP versus SNR curves for the primary linkPUb → PUa of the considered system with various fading

scenarios. All the analytical curves are in good consonancewith the simulation results, which confirms the accuracy ofthe derived analytical expressions. It can be seen From Fig. 3that when the value of ma and/or mb increases from 1 to 2,the user OP performance of the primary system improves.From this, one can infer that the system experiences better OPperformance with comparatively less severe fading conditions.Given that the OP performance improves with higher SNR

Fig. 4. OP versus SNR curves for IoD2 → IoD1 link.

Fig. 5. OP versus β curves for PUb → PUa link.

values, a required SNR value can be identified to achieve adesired link reliability. Moreover, Fig. 3 also shows OP curvesfor different values of power splitting factors μ1 and μ2. Itcan be observed that as the value of μi increases, the OP ofthe primary link also improves. This behavior is in agreementwith the modeling of spectrum sharing system, where a highervalue of μi represents that more power is assigned for primarytransmissions.

Fig. 4 plots the OP curves versus SNR of the IoT linkIoD2 → IoD1 for different fading scenarios. For this figure,the target rate is set as r1 = 1/3 and all other parametersare the same as in Fig. 3. As can be seen, all the simulationpoints are in perfect match with the corresponding analyticalcurves. Similar to Fig. 3, as the value of fading parametersincreases, the OP of the IoT link also improves. On the otherhand, when the value of μi increases, the corresponding OPof the IoT link degrades. This is because the power allocatedfor IoT transmissions is scaled by (1 − μi) term.

B. Outage Probability With TS Factor

For numerical investigation in Fig. 5, the parameter μ1 =μ2 = 0.9 and SNR = 15 dB. This figure plots OP curves


Fig. 6. OP versus β curves for IoD2 → IoD1 link.

versus the TS factor β for various fading scenarios and differ-ent values of energy conversion efficiency at the IoDs. FromFig. 5, one can see that for a given set of parameters, the pri-mary system achieves the lowest OP at a certain value of β.If the value of β increases or decreases from that value, thesystem OP performance degrades. For lower values of β, OPincreases because less time is allocated for EH at IoDs andhence less transmit power available at IoDs. On the other hand,when the value of β increases after a certain value, the OPalso increases due to a drastic rise in the target SNR with thefactor 23rth/(1−β) − 1. Therefore, it is crucial to set an appro-priate value of β to get optimal OP performance. Moreover,when the target rate at PUs increases, the OP performanceof the primary link degrades. This behavior shows the trade-off between link reliability and achievable data rate for theprimary system. Energy conversion efficiency is another keyfactor in determining the OP performance of the primary link.Lower values of η1 and η2 lead to lower OP performance.

Fig. 6 shows OP versus β curves of the IoT link for variousfading scenarios and SNR values. In this figure, the value ofenergy conversion efficiency is fixed as η1 = η2 = 0.7 andpower splitting factor is set as μ1 = μ2 = 0.7. From thisfigure, one can see that as the target rate increases from 1/9 to1/3, the OP performance of the IoT link degrades. However,this degradation in OP performance can be recovered if thevalue of SNR increases from 15 to 20 dB.

C. Outage Probability With Spectrum Sharing Factor

Section III-E highlights that for effective spectrum sharing,the value of power splitting factor μi should be chosen care-fully. In addition, Figs. 3 and 4 also demonstrate that thevalue of power splitting factor has crucial impacts on theperformance of both primary and IoT systems. Hereby, OPversus μ curves are plotted in Figs. 7 and 8 for the primary andIoT systems, respectively, to show the feasible range and offersome insightful observations. In Fig. 7, the system parametersare set as β = 0.2, η1 = η2 = 0.7, and SNR = 25 dB.From this figure, one can observe that for the effective spec-trum sharing, the value of μi should be greater than a certain

Fig. 7. OP versus μ curves for PUb → PUa link.

Fig. 8. OP versus μ curves for IoD2 → IoD1 link.

value. For determining that critical value of μi, the solutionof (42) is obtained using a numerical method. In Fig. 7, thecritical value of μi (μ�) can be referred as a point at whichthe OP of the primary link with the proposed scheme showsthe same OP of direct transmission (shown by horizontal lines)curves. As such, for μ� < μi, the primary system exhibitsbetter outage performance than that of the direct transmission.Consequently, the effective range for spectrum sharing can begiven as μ� < μi < 1. On the other hand, with the setting ofrth = 1/3 bps/Hz and β = 0.2, the feasible range of powersplitting factor will be 0.58 < μi < 1 for enabling spectrumsharing. Below this value, the OP of the primary link becomesunity as also highlighted by Lemma 2. From here, one cannote that the minimum possible value of μi depends only onthe TS parameter β and target rates of the corresponding pri-mary links. Further, in Fig. 8, the system parameters are setas β = 0.2, rth = 1/3, and SNR = 20 dB. It can be seenfrom this figure that, as the value of μi increases, the OPperformance of the IoT link degrades. Different from the pri-mary link, the IoT link shows considerable OP performancefor the entire range of μi.


Fig. 9. System throughput versus SNR curves with different target rates.

Fig. 10. System throughput versus β curves with different parameters.

D. System Throughput

Fig. 9 plots the system throughput versus SNR curves.Herein, the parameters are set as ma = mb = m12 = 1,β = 0.2, μ1 = μ2 = 0.8, and η1 = η2 = 0.7. Observethat, for low SNR values, the curves corresponding to highertarget rates show lower system throughput as compared to thecurves corresponding to lower target rates. This is due to thefact that in the low SNR region, as the value of target rateincreases, the corresponding target SNR also increases, whichdegrades the OP performance of both systems. When the OP ofboth systems become higher, the system throughput decreases.On the contrary, in medium to high SNR region, the impactof degradation in OP performance is less as compared to theenhancement due to higher target rates. After a particular SNRvalue, the system throughput curves attain a saturation pointthat can be referred to as the maximum achievable throughputfor the considered set of parameters.

The setting of TS factor is also crucial for system throughputperformance. Fig. 10 shows the system throughput versus β

curves. Herein, the parameters are set as ma = mb = m12 = 2,

Fig. 11. Energy efficiency versus SNR and target rate.

η1 = η2 = 0.7, μ1 = μ2 = 0.8, and SNR = 10 dB. Asexpected, the curves corresponding to higher target rates attainthe maximum achievable throughput in the range 0 < β < 0.5.For the case when target rate is rth = 1/3, the system achievesthe maximum throughput at β = 0.18 for the considered set ofparameters. When the target rates decrease, the value of β atwhich the system attains the maximum throughput also shiftstoward lower values.

E. Energy Efficiency

To reveal the impact of different parameters on the overallenergy efficiency of the considered system, Fig. 11 plots theenergy efficiency versus SNR and target rate. Here, the param-eters are set as β = 0.2, η1 = η2 = 0.7, and μ1 = μ2 = 0.8,and the target rates of both primary and IoT systems areassumed to be equal. From this figure, one can see that with alower target rate, the system achieves significant energy effi-ciency at lower SNR values. For example, when the target rateis 0.1 bps/Hz, the maximum energy efficiency is achieved at0 dB. On the other hand, when the target rate is higher, thesystem attains better energy efficiency from medium to highSNR regime. Based on this observation, one can infer that themaximum energy efficiency can be attained at specific valuesof SNRs only, and that depends on the required target rates.As such, when the target rate increases from a lower value,the SNR value for which the system achieves the maximumenergy efficiency also shifts toward the higher value.

F. Average Transmission Time

Fig. 12 plots the average end-to-end transmission timesversus the transmit SNR for the proposed relaying schemeand direct transmission (without spectrum sharing). For theresults in this figure, the parameters are set as L = 4096,W = 1 MHz, η1 = η2 = 0.7, μ1 = μ2 = 0.9, r1 = r2 = 1/6bps/Hz, and ma = mb = mab = 1. As naturally expected, theend-to-end transmission time of the proposed scheme is higherthan the transmission time of direct transmission. However,the absolute transmission times and their difference quicklydecrease as the SNR increases. This small drawback should be


Fig. 12. Average end-to-end transmission time versus SNR curves forPUa → PUb link.

easily outweighted by the superiority of the proposed schemewith regard to other important performance metrics, includingspectral efficiency, link reliability, and energy efficiency.

V. CONCLUSION

This paper proposed a SWIPT-based spectrum sharingscheme to enable IoT communications in the licensed spec-trum and to realize the primary communications with improvedlink reliability. A pair of SWIPT-based IoDs has been consid-ered for providing relay assistance to primary transmissionby applying DF operation. First, this paper analyzed the out-age performance of both primary and IoT systems with theproposed scheme under Nakagami-m fading. Then, it formu-lated the expressions of energy efficiency and system through-put. Further, it discussed the condition for spectrum sharingfor which the OP performance of the proposed scheme is equalor lower than that of the direct transmission. Numerical andsimulation results elucidated the accuracy of all the derivedexpressions and highlighted the impacts of some critical designparameters, e.g., power splitting factor and TS factor, on thesystem performance. Above all, this paper incorporated theconcept of the cognitive radio system, SWIPT, and spectralefficient relaying for the deployment of future IoT systems.

REFERENCES

[1] A. A. Khan, M. H. Rehmani, and A. Rachedi, “Cognitive-radio-basedInternet of Things: Applications, architectures, spectrum related func-tionalities, and future research directions,” IEEE Wireless Commun.,vol. 24, no. 3, pp. 17–25, Jun. 2017.

[2] A. Ö. Ercan, M. O. Sunay, and I. F. Akyildiz, “RF energy harvestingand transfer for spectrum sharing cellular IoT communications in 5Gsystems,” IEEE Trans. Mobile Comput., vol. 17, no. 7, pp. 1680–1694,Jul. 2018.

[3] X. Liu and N. Ansari, “Green relay assisted D2D communications withdual batteries in heterogeneous cellular networks for IoT,” IEEE InternetThings J., vol. 4, no. 5, pp. 1707–1715, Oct. 2017.

[4] M. Gharbieh, H. ElSawy, A. Bader, and M.-S. Alouini, “Spatiotemporalstochastic modeling of IoT enabled cellular networks: Scalability andstability analysis,” IEEE Trans. Commun., vol. 65, no. 8, pp. 3585–3600,Aug. 2017.

[5] L. R. Varshney, “Transporting information and energy simultaneously,”in Proc. IEEE ISIT, Jul. 2008, pp. 1612–1616.

[6] A. A. Nasir, X. Zhou, S. Durrani, and R. A. Kennedy, “Wireless-poweredrelays in cooperative communications: Time-switching relaying proto-cols and throughput analysis,” IEEE Trans. Commun., vol. 63, no. 5,pp. 1607–1622, May 2015.

[7] X. Zhou, R. Zhang, and C. K. Ho, “Wireless information and powertransfer: Architecture design and rate-energy tradeoff,” IEEE Trans.Commun., vol. 61, no. 11, pp. 4754–4767, Nov. 2013.

[8] L. Wang, F. Hu, Z. Ling, and B. Wang, “Wireless information and powertransfer to maximize information throughput in WBAN,” IEEE InternetThings J., vol. 4, no. 5, pp. 1663–1670, Oct. 2017.

[9] F. Benkhelifa, K. Tourki, and M.-S. Alouini, “Proactive spectrum sharingfor SWIPT in MIMO cognitive radio systems using antenna switch-ing technique,” IEEE Trans. Green Commun. Netw., vol. 1, no. 2,pp. 204–222, Jun. 2017.

[10] A. Goldsmith, S. Jafa, I. Maric, and S. Srinivasa, “Breaking spectrumgridlock with cognitive radios: An information theoretic perspective,”Proc. IEEE, vol. 97, no. 5, pp. 894–914, May 2009.

[11] T. Li, P. Fan, and K. B. Letaief, “Outage probability of energy harvestingrelay-aided cooperative networks over Rayleigh fading channel,” IEEETrans. Veh. Technol., vol. 65, no. 2, pp. 972–978, Feb. 2016.

[12] H. Lee, C. Song, S.-H. Choi, and I. Lee, “Outage probability analysisand power splitter designs for SWIPT relaying systems with direct link,”IEEE Commun. Lett., vol. 21, no. 3, pp. 648–651, Mar. 2017.

[13] J. Men, J. Ge, C. Zhang, and J. Li, “Joint optimal power allocationand relay selection scheme in energy harvesting asymmetric two-way relaying system,” IET Commun., vol. 9, no. 11, pp. 1421–1426,Jul. 2015.

[14] G. Du, K. Xiong, Y. Zhang, and Z. Qiu, “Outage analysis andoptimization for time switching-based two-way relaying with energyharvesting relay node,” KSII Trans. Internet Inf. Syst., vol. 9, no. 2,pp. 545–563, Feb. 2015.

[15] R. Hu and T.-M. Lok, “Power splitting and relay optimization fortwo-way relay SWIPT systems,” in Proc. IEEE ICC, Kuala Lumpur,Malaysia, May 2016, pp. 1–6.

[16] C. Peng, F. Li, and H. Liu, “Optimal power splitting in two-way decode-and-forward relay networks,” IEEE Commun. Lett., vol. 21, no. 9,pp. 2009–2012, Sep. 2017.

[17] T. P. Do, I. Song, and Y. H. Kim, “Simultaneous wireless trans-fer of power and information in a decode-and-forward two-wayrelaying network,” IEEE Trans. Wireless Commun., vol. 16, no. 3,pp. 1579–1592, Mar. 2017.

[18] S. Yin, E. Zhang, Z. Qu, L. Yin, and S. Li, “Optimal cooperation strategyin cognitive radio systems with energy harvesting,” IEEE Trans. WirelessCommun., vol. 13, no. 9, pp. 4693–4707, Sep. 2014.

[19] Z. Wang, Z. Chen, B. Xia, L. Luo, and J. Zhou, “Cognitive relaynetworks with energy harvesting and information transfer: Design, anal-ysis, and optimization,” IEEE Trans. Wireless Commun., vol. 15, no. 4,pp. 2562–2576, Apr. 2016.

[20] G. Im and J. H. Lee, “Outage probability of underlay cognitive radionetworks with SWIPT-enabled relay,” in Proc. IEEE VTC-Fall, Boston,MA, USA, 2015, pp. 1–5.

[21] Z. Yang, Z. Ding, P. Fan, and G. K. Karagiannidis, “Outage performanceof cognitive relay networks with wireless information and powertransfer,” IEEE Trans. Veh. Technol., vol. 65, no. 5, pp. 3828–3833,May 2016.

[22] S. S. Kalamkar and A. Banerjee, “Interference-aided energy harvest-ing: Cognitive relaying with multiple primary transceivers,” IEEE Trans.Cogn. Commun. Netw., vol. 3, no. 3, pp. 313–327, Sep. 2017.

[23] D. K. Verma, R. Y. Chang, and F.-T. Chien, “Energy-assisted decode-and-forward for energy harvesting cooperative cognitive networks,”IEEE Trans. Cogn. Commun. Netw., vol. 3, no. 3, pp. 328–342,Sep. 2017.

[24] J. Yan and Y. Liu, “A dynamic SWIPT approach for cooperative cog-nitive radio networks,” IEEE Trans. Veh. Technol., vol. 66, no. 12,pp. 11122–11136, Dec. 2017.

[25] B. V. Nguyen, H. Jung, D. Har, and K. Kim, “Performance analysis of acognitive radio network with an energy harvesting secondary transmitterunder Nakagami-m fading,” IEEE Access, vol. 6, pp. 4135–4144, 2018.

[26] Z. Wang, Z. Chen, Y. Yao, B. Xia, and H. Liu, “Wireless energy har-vesting and information transfer in cognitive two-way relay networks,”in Proc. IEEE GLOBECOM, Austin, TX, USA, 2014, pp. 3465–3470.

[27] Z. Yang, W. Xu, Y. Pan, C. Pan, and M. Chen, “Energy efficient resourceallocation in machine-to-machine communications with multiple accessand energy harvesting for IoT,” IEEE Internet Things J., vol. 5, no. 1,pp. 229–245, Feb. 2018.


[28] X. Huang, R. Yu, J. Kang, Z. Xia, and Y. Zhang, “Software definednetworking for energy harvesting Internet of Things,” IEEE InternetThings J., vol. 5, no. 3, pp. 1389–1399, Jun. 2018.

[29] Z. Yan, S. Chen, X. Zhang, and H.-L. Liu, “Outage performance analysisof wireless energy harvesting relay-assisted random underlay cogni-tive networks,” IEEE Internet Things J., vol. 5, no. 4, pp. 2691–2699,Aug. 2018.

[30] I. Gradshteyn and I. Ryzhik, Table of Integrals, Series, and Products,7th ed. San Diego, CA, USA: Academic, 2007.

[31] Y. Pei and Y.-C. Liang, “Resource allocation for device-to-devicecommunications overlaying two-way cellular networks,” IEEE Trans.Wireless Commun., vol. 12, no. 7, pp. 3611–3621, Jul. 2013.

[32] Y. Liu, L. Wang, M. Elkashlan, T. Q. Duong, and A. Nallanathan,“Two-way relay networks with wireless power transfer: Design andperformance analysis,” IET Commun., vol. 10, no. 14, pp. 1810–1819,Sep. 2016.

[33] A. A. Nasir, X. Zhou, S. Durrani, and R. A. Kennedy, “Relaying proto-cols for wireless energy harvesting and information processing,” IEEETrans. Wireless Commun., vol. 12, no. 7, pp. 3622–3636, Jul. 2013.

[34] R. Wang, M. Tao, and Y. Liu, “Optimal linear transceiver designs forcognitive two-way relay networks,” IEEE Trans. Signal Process., vol. 61,no. 4, pp. 992–1005, Feb. 2013.

[35] Q. Li, S. H. Ting, A. Pandharipande, and Y. Han, “Cognitive spectrumsharing with two-way relaying systems,” IEEE Trans. Veh. Technol.,vol. 60, no. 3, pp. 1233–1240, Mar. 2011.

[36] P. Popovski and H. Yomo, “Physical network coding in two-way wire-less relay channels,” in Proc. IEEE ICC, Glasgow, U.K., Jun. 2007,pp. 701–712.

[37] T. M. Cover and J. A. Thomas, Elements of Information Theory.New York, NY, USA: Wiley-Intersci., 1991.

[38] X. Zhang et al., “Exact outage analysis in cognitive two-way relaynetworks with opportunistic relay selection under primary user’sinterference,” IEEE Trans. Veh. Technol., vol. 64, no. 6, pp. 2502–2511,Jun. 2015.

[39] Y. Huang, M. Liu, and Y. Liu, “Energy-efficient SWIPT in IoT dis-tributed antenna systems,” IEEE Internet Things J., vol. 5, no. 4,pp. 2646–2656, Aug. 2018.

[40] P. K. Sharma and P. K. Upadhyay, “Cooperative spectrum sharing intwo-way multi-user multi-relay networks,” IET Commun., vol. 10, no. 1,pp. 111–121, Apr. 2016.

[41] F. A. Khan, K. Tourki, M.-S. Alouini, and K. A. Qaraqe, “Delayperformance of a broadcast spectrum sharing network in Nakagami-m fading,” IEEE Trans. Veh. Technol., vol. 63, no. 3, pp. 1350–1364,Mar. 2014.

[42] N. B. Mehta, V. Sharma, and G. Bansal, “Performance analysis of acooperative system with rateless codes and buffered relays,” IEEE Trans.Wireless Commun., vol. 10, no. 4, pp. 1069–1081, Apr. 2011.

[43] S. Solanki, P. K. Sharma, and P. K. Upadhyay, “Adaptive link utilizationin two-way spectrum sharing relay systems under average interference-constraints,” IEEE Syst. J., vol. 12, no. 4, pp. 3461–3472, Dec. 2018.

Devendra S. Gurjar (S’13–M’17) received theB.Tech. degree in electronics and communica-tions engineering from Uttar Pradesh TechnicalUniversity, Lucknow, India, in 2011, the M.Tech.degree in wireless communications and computingfrom the Indian Institute of Information TechnologyAllahabad, Allahabad, India, in 2013, and the Ph.D.degree in electrical engineering from the IndianInstitute of Technology Indore, Indore, India, in2017.

He was with the Department of Electrical andComputer Engineering, University of Saskatchewan, Saskatoon, SK, Canada,as a Post-Doctoral Research Fellow. He is currently an Assistant Professorwith the Department of Electronics and Communication Engineering, NationalInstitute of Technology Silchar, Silchar, India. His current research interestsinclude MIMO communication systems, cooperative relaying, device-to-device communications, and simultaneous wireless information and powertransfer.

Dr. Gurjar is a member of the IEEE Communications Society and the IEEEVehicular Technology Society.

Ha H. Nguyen (M’01–SM’05) received the B.Eng.degree in electrical engineering from the HanoiUniversity of Technology, Hanoi, Vietnam, in 1995,the M.Eng. in electrical engineering degree from theAsian Institute of Technology, Bangkok, Thailand, in1997, and the Ph.D. degree in electrical engineeringfrom the University of Manitoba, Winnipeg, MB,Canada, in 2001.

He joined the Department of Electrical andComputer Engineering, University of Saskatchewan,Saskatoon, SK, Canada, in 2001, where he became

a Full Professor in 2007 and currently holds the position of NSERC/CiscoIndustrial Research Chair in Low-Power Wireless Access for SensorNetworks. He co-authored, with Ed Shwedyk, the textbook A First Course inDigital Communications (Cambridge Univ. Press, 2009). His current researchinterests include communication theory, wireless communications, and statis-tical signal processing.

Dr. Nguyen was an Associate Editor of the IEEE TRANSACTIONS ON

WIRELESS COMMUNICATIONS from 2007 to 2011 and IEEE WIRELESS

COMMUNICATIONS LETTERS from 2011 to 2016. He currently servesas an Associate Editor for the IEEE TRANSACTIONS ON VEHICULAR

TECHNOLOGY. He was the Co-Chair for the Multiple Antenna Systemsand Space-Time Processing Track, IEEE Vehicular Technology Conferences(Ottawa, ON, Canada, 2010, and Quebec, QC, Canada, 2012), the LeadCo-Chair for the Wireless Access Track, IEEE Vehicular TechnologyConferences (Vancouver, BC, Canada, 2014), the Lead Co-Chair and theCo-Chair for the Multiple Antenna Systems and Cooperative CommunicationsTrack, IEEE Vehicular Technology Conference (Montreal, QC, Canada, 2016,and Porto, Portugal, 2018), and the Technical Program Co-Chair for CanadianWorkshop on Information Theory (St. John’s, NL, Canada, 2015). He is aFellow of the Engineering Institute of Canada and is a Registered Member ofthe Association of Professional Engineers and Geoscientists of Saskatchewan.

Hoang Duong Tuan received the Diploma (Hons.)and Ph.D. degrees in applied mathematics fromOdessa State University, Odessa, Ukraine, in 1987and 1991, respectively.

He was an Assistant Professor with theDepartment of Electronic-Mechanical Engineering,Nagoya University, Nagoya, Japan, from 1994 to1999, and then an Associate Professor with theDepartment of Electrical and Computer Engineering,Toyota Technological Institute, Nagoya, from 1999to 2003. He was a Professor with the School of

Electrical Engineering and Telecommunications, University of New SouthWales, Kensington, NSW, Australia, from 2003 to 2011. He is currently aProfessor with the School of Electrical and Data Engineering, University ofTechnology Sydney, Ultimo, NSW, Australia. His current research interestsinclude optimization, control, signal processing, wireless communication,and biomedical engineering.

wireless information and power transfer for iot

Documents