Outage Probability and Throughput Analysis of SWIPT Enabled Cognitive Relay Network With Ambient Backscatter

Page created by Bernard Contreras
 
CONTINUE READING
3198                                                                                     IEEE INTERNET OF THINGS JOURNAL, VOL. 5, NO. 4, AUGUST 2018

       Outage Probability and Throughput Analysis of
         SWIPT Enabled Cognitive Relay Network
                 With Ambient Backscatter
                Syed Tariq Shah, Kae Won Choi , Senior Member, IEEE, Tae-Jin Lee , Member, IEEE,
                                      and Min Young Chung , Member, IEEE

   Abstract—In this paper, we propose an ambient backscatter                    and more flexible solution for the powering-up of energy-
(AB)-enabled decode-and-forward (DF) cognitive relay network                    constrained network nodes. Besides reliability and flexibility,
with wireless energy-harvesting capabilities. In our proposed                   another major advantage of the RF-EH technology is the abil-
scheme, a source node communicates with its destination node via
a radio frequency-powered DF relay. It is assumed that the relay                ity of the RF signals to simultaneously carry both information
node is equipped with two different interfaces and can concur-                  and energy [5], [6]. This technology is known as simultaneous
rently harvest/decode and backscatter the received source signals.              wireless information and power transfer (SWIPT) [7].
A power-splitting-based approach is adopted for the information                    For SWIPT operation, Varshney [5] has considered an ideal
processing and the energy harvesting at the relay. The analytical               receiver design which has the ability to simultaneously har-
expressions for the outage probability at all of the receiving nodes
are derived. It has been shown that the analytical results match                vest energy and process information from received RF signals.
the simulation results. It has also been shown that using the AB                However, this ideal receiver design is not practical because the
for secondary communications can significantly improve the over-                existing RF EH circuits cannot directly decode the informa-
all network performance in terms of the achievable throughput                   tion present in the received RF signals. To enable SWIPT,
and the energy efficiency.                                                      Zhou et al. [8] have proposed two practical EH receivers, i.e.,
  Index Terms—Ambient backscatter (AB), cooperative relay                       time switching (TS) and power-splitting (PS). In TS approach
networks, simultaneous wireless information and power transfer                  the receiver circuit switches between EH and information pro-
(SWIPT), wireless energy harvesting (WEH).                                      cessing in the time domain. On the other hand, in PS approach
                                                                                the receiver circuits split the power of the received signal
                                                                                into two portions: one for EH and the other for information
                          I. I NTRODUCTION
                                                                                processing. Based on the TS and PS receiving architectures,
       IRELESS energy harvesting (WEH)-based communi-
W      cation networks have emerged as a new networking
paradigm. In WEH networks, the energy constrained nodes
                                                                                Nasir et al. [9] proposed two relaying protocols namely, TS
                                                                                relaying (TSR) protocol and PS relaying (PSR) protocol.
                                                                                They further studied the performances of both protocols in
in the network can be remotely replenished via wireless                         an amplify-and-forward (AF) relay-based SWIPT network and
power transfer technology. Unlike legacy battery powered                        they concluded that at high transmission rates and low signal-
networks, the WEH networks do not require any manual                            to-noise ratio (SNR) the TSR protocol outperforms the PSR
recharging/replacement of batteries, leading to an enhanced                     protocol. For a decode-and-forward (DF) relay network, the
network lifetime with a significantly reduced operational                       TSR and PSR performances were analyzed in [10], reveal-
cost [1], [2], [4]. Furthermore, unlike other conventional                      ing that the PSR protocol achieved a performance that is
energy harvesting (EH) techniques such as thermoelectric,                       superior to that of the TSR. A relay-selection scheme for large-
wind, and solar energy, which are not very reliable and are                     scale EH-based Internet-of-Things (IoT) networks is proposed
highly dependent on the surrounding environments, ambient                       in [11]. The proposed scheme selects the relay based on
radio frequency (RF) EH has newly emerged as a reliable                         its residual-energy level and the channel quality. Here, the
                                                                                proposed scheme significantly improved the outage-probability
   Manuscript received January 9, 2018; revised April 20, 2018; accepted
May 8, 2018. Date of publication May 16, 2018; date of current                  performance of the EH relays.
version August 9, 2018. This work was supported by the National                    The performance of SWIPT in cognitive relay networks
Research Foundation of Korea through the Korean Government under Grant          has been evaluated in various research works [12]–[15].
2014R1A5A1011478. (Corresponding author: Min Young Chung.)
   S. T. Shah is with the College of Information and Communication              Wang et al. [12] proposed an SWIPT-enabled AF relay-based
Engineering, Sungkyunkwan University, Suwon 16419, South Korea, and also        cognitive network. In their proposed scheme, the relay node
with the Department of Telecommunication Engineering, FICT, Balochistan         first harvests the energy from the primary-communication sig-
University of Information Technology, Engineering and Management
Sciences, Quetta, Pakistan (e-mail: syed.tariq@skku.edu).                       nals using a PSR protocol, and it then utilizes the harvested
   K. W. Choi, T.-J. Lee, and M. Y. Chung are with the College of               energy to send its own secondary information along with
Information and Communication Engineering, Sungkyunkwan University,             the amplified primary signal. The secondary communication
Suwon 16419, South Korea (e-mail: kaewonchoi@skku.edu; tjlee@skku.edu;
mychung@skku.edu).                                                              in [12] not only acts as an interference to the primary commu-
   Digital Object Identifier 10.1109/JIOT.2018.2837120                          nication but it also consumes a portion of the scarce harvested
                      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.

       Authorized licensed use limited to: Sungkyunkwan University. Downloaded on August 25,2020 at 07:27:11 UTC from IEEE Xplore. Restrictions apply.
SHAH et al.: OUTAGE PROBABILITY AND THROUGHPUT ANALYSIS OF SWIPT ENABLED COGNITIVE RELAY NETWORK WITH AB                                                3199

energy. A similar approach for a TSR-protocol-based DF relay                   metrics are outage probability, throughput, and energy effi-
was studied in [14]. Unlike [12], the network model of [14]                    ciency. For the remainder of this paper, the communications
consists of multiple relays, and an optimal relay-selection                    between the source-to-destination (S-to-D) nodes and the relay
scheme was proposed accordingly. Similar to [12], however, a                   (R)-to-neighboring nodes are labeled as primary communi-
two-way AF relay-based cognitive SWIPT network is studied                      cation and secondary communication, respectively. The main
in [15], wherein a portion of the harvested power is used to                   contributions of the present paper are summarized as follows.
transmit the secondary information of the network.                                1) An AB-enabled SWIPT-based DF relay network is
   The use of an ambient RF signal for WEH has been thor-                            proposed, where the relay node is enabled to concur-
oughly studied in [1] and [3]. Beside EH, another interesting                        rently perform both SWIPT and AB operations.
aspect of ambient RF signals is their use in the infor-                           2) Since the proposed network model adopts the PSR pro-
mation transmission that is via the backscattering technol-                          tocol for EH and information processing at the relay, an
ogy [16], [17]. Ambient backscatter (AB) communication is                            adaptive PS mechanism has been used. For each com-
a new and promising communication method for low data                                munication block time, the considered PS mechanism
rate networks. The AB nodes do not require any specific                              dynamically adjusts the PS ratio based on the received
power storage/supply infrastructure. When data transmission                          SNR at the relay.
is required, the AB transmitter (AB-Tx) backscatters the RF                       3) For both the primary and secondary communications,
signals that are received from an ambient source [16]. For                           the closed-form expressions for both the throughput and
the backscattering operation, the AB-Tx switches its mode                            the outage probability are derived. We show that our
between the nonreflect and reflect modes that correspond                             simulation results match our analytical results, which
to the bits “0” and “1,” respectively. On the other hand,                            not only verify our simulation and analytical models but
the AB receiver (AB-Rx) uses the envelope detection and                              also provide thorough practical insights into the effect
averaging techniques to decode the information from the                              of different system parameters on the overall network
received backscattered signals [17]. Since AB can be imple-                          performance.
mented without complex circuitry and encoding/decoding
mechanisms, it can be easily integrated into any wireless                      B. Organization
communication node [18], [19]. A prototype example of a                           The rest of this paper is organized as follows. Section II
full duplex AB has been reported in [18], where a WiFi AP                      presents the proposed system model and a detailed discus-
transmits data to normal WiFi clients and at the same time                     sion on EH, information decoding (ID), and AB operations.
decodes the backscatter signals reflected by different IoT sen-                Section III derives analytical expressions for the outage
sors. However, due to the ambient nature of this technology,                   probability and the achievable throughput. The performance
the performance of the backscatter communication is highly                     analysis and the conclusion of this paper are presented in
dependent upon the availability of the ambient RF signals.                     Sections IV and V, respectively.

A. Paper Objectives and Contribution                                                    II. S YSTEM M ODEL AND P ROPOSED S CHEME
   The idea of relays for coverage extension in wireless sensor                A. System Model
networks has been well established and widely accepted [20].                      We consider a network consisting of a primary source node
In cooperative wireless sensor networks, the battery power of                  S, a destination node D, a DF relay node R, and a secondary
the cooperating nodes (such as relay nodes) is usually lim-                    backscatter (BS) node C. The node S communicates with the
ited, and to actively perform their role in the network, these                 node D via node R, where R is considered as an energy-
nodes may need to rely on an additional charging mecha-                        constrained node with EH and AB capabilities. In addition to
nism [21], [22]. Moreover, besides the sole action of relaying,                relying S-to-D primary communication, node R also communi-
these energy-constrained sensor nodes also comprise their own                  cates with node C using AB technology. We consider that node
sensed information that they must transmit to another neigh-                   R has two different interfaces for both SWIPT and backscat-
boring node. Therefore, an efficient relaying mechanism that                   tering operations. In summary, node R utilizes the SWIPT and
not only recharges the relay nodes but also aids them to                       backscatter interfaces for S → D primary communication and
transmit their own information is required.                                    the R ↔ C secondary communication, respectively.
   In this paper, we introduce an AB-enabled SWIPT DF                             Our considered network model is depicted in Fig. 1(a),
relay network where a source node transmits its informa-                       where g(i,j) and d(i,j) (i, j ∈ {S, R, C, D}) denote the chan-
tion to a destination node via an energy constrained relay.                    nel coefficients and the distance between nodes, respectively.
The relay node harvests a portion of the received source-                      The channel gains between nodes are modeled as Rayleigh
signal power using the PS approach, and it then utilizes this                  block-fading channels and for each time block the chan-
harvested energy to forward the received information to its                    nels are independent and identically distributed. A path
destination. In addition to the relaying operation, the relay                  loss between communicating nodes has been modeled as a
node also simultaneously communicates with a neighboring                       distance-dependent path loss model with rate d−ε , where d is
node using AB communication. Such a network model can                          the distance between nodes and ε is the path-loss exponent.
widely be used in emerging energy-constrained IoT-based                        The use of such channel and path-loss models is motivated by
relay networks [11]. In the proposed scheme, the performance                   previous research work done in this area [9], [10], [12]–[15].

      Authorized licensed use limited to: Sungkyunkwan University. Downloaded on August 25,2020 at 07:27:11 UTC from IEEE Xplore. Restrictions apply.
3200                                                                                     IEEE INTERNET OF THINGS JOURNAL, VOL. 5, NO. 4, AUGUST 2018

                                                                                using the harvested energy, node R forwards the decoded
                                                                                source signal to node D. Meanwhile, C also backscatters its
                                                                                information to node R.
                                                                                   More specifically, node S transmits its information signals
                                                                                to node R during m1 (phase I). The received source signal
                                                                                (Y(S,R) ) and SNR at node R during m1 can, respectively, be
                                                                                expressed as
                                                                                                         
                                                                                                            PS
                                     (a)                                                        Y(S,R) =    ε     g(S,R) xS + n0           (1)
                                                                                                           d(S,R)
                                                                                                                PS |g(S,R) |2
                                                                                                        γR =      ε                                      (2)
                                                                                                                 d(S,R) σ2

                                                                                where xS , PS , n0 , and σ 2 are the normalized information signal
                                                                                from node S, source Tx power, additive white Gaussian noise
                                                                                (AWGN), and the variance of the AWGN, respectively. The
                                                                                relay node splits the signal √   Y(S,R) that is received
                                                                                                                                 √       via EH/ID
                                     (b)                                        interface
                                                                                √         into  two   portions,    αY (S,R) and    1 − αY
                                                                                                                                      √ (S,R) . The
                                                                                  αY(S,R) portion of signal is used for EH and 1 − αY(S,R)
Fig. 1. System model. (a) System model AB enabled SWIPT relay network.          portion is used for ID. After a successful EH the received SNR
(b) Time-block structure for proposed network model.
                                                                                at node R can be expressed as
                               TABLE I
                                                                                                                (1 − α)PS |g(S,R) |2
                          TABLE OF N OTATIONS                                                          γ̃R =          ε              .                   (3)
                                                                                                                     d(S,R) σ2
                                                                                In the proposed DF relay network, the received source signal
                                                                                at the node R can only be successfully decoded if its SNR is
                                                                                greater than or equal to the minimum required decoding-SNR
                                                                                γ0 , where γ0 = 22U − 1 and U is the source transmission rate.
                                                                                Therefore, based on (3) and γ0 the value of the PS factor can
                                                                                be calculated as
                                                                                                                              ε
                                                                                                                      γ0 σ 2 d(S,R)
                                                                                                          α = 1−                      .                  (4)
                                                                                                                      PS |g(S,R) |2
                                                                                In our proposed DF relay network, the PS strategy that is
                                                                                provided by (4) is ideal. More specifically, if the value of PS
                                                                                factor α is greater than (4), a small portion of the signal power
                                                                                is used for EH, and an unnecessarily-high power is allocated
                                                                                for the signal decoding, resulting in a waste of the valuable
                                                                                power resource. Alternatively, if the value of the PS factor α
                                                                                is less than (4), the relay node utilizes more power for the EH,
                                                                                thereby resulting in the decoding failure of the source signal.
As shown in Fig. 1(b), the transmission block time T is divided                 Based on PS factor α derived in (4), the harvested power (PR )
into two equal slots m1 and m2 and a PS-based relaying proto-                   at node R from the received source signal is defined as PR =
col (PSR) is used for EH and information processing at node                           ε
                                                                                (η/d(S,R)                       ε
                                                                                          )(PS |g(S,R) |2 − γ0 d(S,R) σ 2 ), where η is the energy
R [10]. The α in Fig. 1(b) is the PS factor and its value is                    conversion efficiency and its value is within the range of (0, 1].
real in (0, 1]. For the sake of readers’ convenience, all the                      Meanwhile, node R also transfers its own secondary infor-
notations used in this paper are summarized in Table I.                         mation to node C by backscattering the received signal Y(S,R)
                                                                                using its AB interface. As a result of this backscattering oper-
B. Energy Harvesting, Information Decoding, and Ambient                         ation at node R, the received SNR at node C (BS receiver)
Backscatter Operations                                                          can be expressed as [23]
   As shown in Fig. 1, the whole communication procedure is                                                    ζ PS |g(S,R) |2 |g(R,C) |2
divided into two phases. In phase I, node S transmits its infor-                                    γCBS =           ε     ε                             (5)
                                                                                                                  d(S,R) d(R,C)   σ2
mation to node R during time slot m1 . Node R harvests energy
and decodes the received signal while concurrently backscat-                    where ζ is the fraction of power scattered back by node R and
ter it to transmit its own information to node C. The relay                     its value is within the range of (0, 1].
node performs both of these operations using two separate                          In the phase II of the proposed scheme, the node R forwards
interfaces, i.e., AB interface and EH/ID interface. In phase II,                the decoded primary signal using the PR . The signal broadcast

       Authorized licensed use limited to: Sungkyunkwan University. Downloaded on August 25,2020 at 07:27:11 UTC from IEEE Xplore. Restrictions apply.
SHAH et al.: OUTAGE PROBABILITY AND THROUGHPUT ANALYSIS OF SWIPT ENABLED COGNITIVE RELAY NETWORK WITH AB                                                3201

by node R during m2 time slot can be written as SR = PR x̃S ,                       Proposition 1: According to (9) the outage probability of
where x̃S is the decoded version of source-information signal                    primary communication can be expressed as
xS . Note that during m2 , node C acts as an AB transmit-
ter and communicates with the node R by backscattering the                                                    out = J1 + J2
                                                                                                             PD                                         (10)
received ambient signal (i.e., SR ). It can be observed that, dur-               where
ing both time-slots m1 and m2 the relay node operates in a
                                                                                                                                  l
full-duplex mode and simultaneously performs both primary                                            J1 = 1 − exp −                                     (11)
communication and backscatter operations. The backscatter                                                                       (S,R)
signal transmitted (received) from (at) relay node in both time                  and
slots is a modulated version of the transmitted (received) pri-                                                                                        
                                                                                                    a             n exp(n)E1 (n)            4β     
mary radio signal. Therefore, it can effectively be canceled out                 J2 = exp −                    1−                              K1 ( 4βτ ) .
using recent self-interference cancellation techniques proposed                                PS    (S,R)           PS (S,R)                τ
in [24]–[28].                                                                                                                                           (12)
   In case of primary communication, the secondary AB com-
                                                                                 The K1 (.) in (12) is the first-order modified Bessel function
munication, however, might act as an interference. Therefore,
                                                                                 of the second kind [29]. Other variables are
the signal that is received at the destination node D (Y(R,D) )
during m2 can be expressed as                                                                                           (R,D)
                                                                                                       n=
                                                                                                               w (R,C) (C,D)
                          PR                                                                                          ε
            Y(R,D) =      ε     g(R,D) x̃S + IBS + n0            (6)                                            ζ γ0 d(R,D)
                         d(R,D)                                                                         w=      ε      ε
                                                                                                              d(R,C) d(C,D)
                   √                               ε     ε
where IBS  [( PR ζ |g(R,C) |2 |g(C,D) |2 )/( d(R,C)    d(C,D) )] is                                            ε
                                                                                                        a = γ0 d(S,R) σ2
the interference that is
                        caused by the BS signal reflected by                                                      ε
                                                                                                               γ0 d(S,R)  ε
                                                                                                                         d(R,D) σ2
node C and d(C,D) = d(R,C) 2     + d(R,D)
                                     2     − 2d(R,C) d(R,D) cos(θ ),                                    β=
                                                                                                                    η     (R,D)
where θ is the angle between nodes C, R, and D (∠CRD).
                                                                                                                    1
   Based on the received Y(R,D) signal in (6), the signal-                                              τ =
to-interference-plus-noise ratio (SINR) at node D can be                                                       PS    (S,R)
expressed as                                                                     and
              PR |g(R,D) |2                                                                                           ε
                                                                                                                  γ0 d(S,R) σ2
   γD =     ε
                                                                                                           l=                   .
           d(R,D)   σ 2 + IBS                                                                                         PS
                                            ε
                              ηq|g(R,D) |2 d(R,C)  ε
                                                  d(C,D)
       =
                                                                                      Proof: Proof of Proposition 1 is provided in
                                      ε
           ζ ηq|g(R,C) |2 |g(C,D) |2 d(R,D)    ε
                                            + d(S,R)  ε
                                                     d(R,C)  ε
                                                            d(R,D)  ε
                                                                   d(C,D) σ2     Appendix A.
                                                                           (7)      Based on PD out derived in Proposition 1 the achievable
                                   ε                                             throughput of the primary communication is given by
where q = PS |g(S,R) |2 − γ0 d(S,R)     σ 2 . Similarly, as a result of                                         
backscattering operation at node C (i.e., secondary communi-                                     CD = 1 − PD  out U(T/2)/T             (13)
cation during m2 ), the received SNR at node R can be written
                                                                                 where T/2 is the effective S-to-D communication time.
as
                                                                                 Similar to the primary communication, the outage in the
                                                       ε
         ζ η|g(R,C) |2 |g(C,R) |2 PS |g(S,R) |2 − γ0 d(S,R) σ2                   secondary communication occurs when the received SNR
 γRBS =                    ε      ε      ε                       . (8)           of a backscattered signal at receiving node is less than a
                         d(S,R)  d(R,C) d(C,R) σ2
                                                                                 predefined threshold. Thus, the outage probability of the
                 III. O UTAGE P ROBABILITY AND                                   R-to-C secondary communication link can be defined as
                                                                                                                      
                     T HROUGHPUT A NALYSIS
                                                                                                     out = Pr γC ≤ γ1
                                                                                                   PRC           BS
                                                                                                                                        (14)
   The outage of a communication link occurs when the
received SNR/SINR at receiving node is less than a predefined                    where γ1 = 22U − 1 is the threshold SNR for the AB com-
threshold SNR/SINR. The outage probability of primary com-                       munication and U is the BS transmission rate. The analytical
munication (i.e., S-to-D link) can be calculated as                              expression for the outage probability of node C (PBC     out ) can be
                                                                                 obtained using Proposition 2.
         out = Pr(γ
        PD
                R
                    < γ0 ) + Pr(γR > γ0 , γD < γ0 ) .
                             
                                                                           (9)
                                                                                    Proposition 2: For secondary communication during m1 ,
                         J1                       J2                             the outage probability at node C can be obtained as
The terms J1 and J2 in (9) show that the outage of primary                                                                             
communication link occurs when the received SNR/SINR at                                                 ζ PS |g(S,R) |2 |g(R,C) |2
                                                                                            Pout = Pr
                                                                                              RC
                                                                                                              ε     ε              ≤ γ1
any receiving node (i.e., node R and node D) is less than                                                  d(S,R) d(R,C)   σ2
a decoding threshold SNR (γ0 ). The analytical expression                                     = 1 − νK1 (ν)                                             (15)
for outage probability of the primary-communication link is                               
provided in the following proposition.                                           where v  ([4hγ1 ]/[ (S,R)             (R,C) ]).

      Authorized licensed use limited to: Sungkyunkwan University. Downloaded on August 25,2020 at 07:27:11 UTC from IEEE Xplore. Restrictions apply.
3202                                                                                     IEEE INTERNET OF THINGS JOURNAL, VOL. 5, NO. 4, AUGUST 2018

                                                                                                                                            Leg
      Proof: Proof of Proposition 2 is provided in                              In legacy scheme, the SNR at relay node is γR = γR and
                                                                                                            Leg
Appendix B.                                                                     SNR at destination node γD can be defined as
    Unlike that of the R-to-C, the C-to-R secondary communi-                                                                           
                                                                                                                               ε
                                                                                              |g(R,D) |2 η PS |g(S,R) |2 − γ0 d(S,R) σ2
cation during the m2 slot is highly dependent on the received                            Leg
SNR (γR ) of the primary source signal (Y(S,R) ) at node R.                            γD =                  ε      ε                     . (22)
                                                                                                            d(S,R) d(R,D) σ2
It is because if the received SNR is below than a predefined
threshold value (i.e., γ0 ), the relay will not be able to decode               Note that, unlike (7) the SNR expression in (22) does not
it. As a result of this decoding failure, there will be no primary              include any interference term. The analytical expression for
communication during m2 which also means that there will be                     outage probability of S-to-D communication link in legacy
no ambient signal available for node C to perform backscat-                     scheme is derived in Proposition 4.
ter communication. Hence, the outage probability of C-to-R                         Proposition 4: Based on (21) the outage probability expres-
secondary communication link can be expressed as                                sion for legacy scheme can be expressed as
                                                                                                                       
                                                                                                  Leg         Leg
                                                                                  PLout = 1 − Pr γR ≥ γ0 , γD ≥ γ0                            (23)
         out = Pr(γ
        PCR          R < γ0 ) + Pr γR > γ0 , γR < γ1 .
                                                BS
                                                              (16)                                           
                               
                       L1                                                                     exp − PS a(S,R)
                                                 L2
                                                                                        = 1−                                                  (24)
                                                                                                  PS (S,R)
The outage probability expression for C-to-R AB communi-
                                                                                            ∞                               ε      ε
                                                                                                                                              
cation during m2 is derived in Proposition 3.                                                              q              γ0 d(S,R) d(R,D) σ2
                                                                                          × exp −                  exp −
   Proposition 3: Based on (16) the outage probability expres-                                          PS (S,R)              η (R,D) q
sion for secondary communication between node C-to-R can                                       0
be expressed as                                                                                                                                          (25)
                                                                                                               a
                                                                                        = 1 − exp −                                            (26)
  out = L1 + L2
 PCR
                           
                                                     (17)                                             PS (S,R)
                                                                                                                      ⎛                      ⎞
            exp − PS a(S,R) 
                             ∞                                                                      ε      ε                   ε      ε
                                        q                                                      4γ0 d(S,R) d(R,D) σ2       4γ0 d(S,R) d(R,D) σ2
      = 1−                     exp −          oK1 (o)dq                                   ×                         K1 ⎝                       ⎠.
               PS (S,R)              PS (S,R)                                                  PS η (R,D) (S,R)           PS η (R,D) (S,R)
                                     0
                                                                                                                                                         (27)
where o            =        (4j/[q (C,R) (R,C) ]) and j   =
      ε
[(γ1 d(S,R)  ε
            d(R,C)  ε
                   d(C,R) σ 2 )/ηζ ].                                                 Proof: The proof of Proposition 4 follows the similar
      Proof: Proof of Proposition 3 is provided in                              steps as provided in Appendix C and therefore it is omitted
Appendix C.                                                                     here.
   Similar to the primary communication, the achievable                            The energy efficiency in both our proposed and legacy
throughput of the secondary communication can also be                           schemes can be defined as the achievable sum-throughput
calculated using the outage probabilities that are derived                      under the unit-energy consumption [30]. Thus, energy efficien-
using (15) and (17). More specifically, the achievable                          cies of the proposed and the legacy networks can, respectively,
throughput of the AB communications at node C (CCBS )                           be calculated as
during m1 and at node R (CRBS ) during m2 can be                                                                CT
                                                                                                           P =                            (28)
calculated as                                                                                                   PS
                                                                                                              CLeg
                                   RC(or CR)                                                               L =       .                    (29)
                    R) = 1 − Pout
               BS
             CC(or                            U(T/2)/T. (18)                                                     PS
Finally, the sum-throughput of the overall network can be
obtained as
                                                                                                   IV. P ERFORMANCE E VALUATION
                      CT =   CD + CCBS     + CRBS .                    (19)        In this section, the analytical results that are derived in
                                                                                Section III are used to provide a detailed insight into the effect
  In order to investigate the efficiency of our proposed                        of the different systemic parameters on the overall network
scheme, we compare our proposed scheme with a legacy                            performance. Unless otherwise stated, for the performance
DF SWIPT relay scheme without backscatters [10]. In legacy                      analysis, the values of the different systemic parameters are
scheme, the throughput of primary S-to-D communication link                     set to PS = 1 W, η = 1, ζ = 0.35, and σ 2 = 10−3 W.
can be calculated as                                                            The distance between source-to-destination is set to 3 m
                                                                              (i.e., dS,R = dR,D = 1.5 m) and the relay-to-secondary
                 CL = 1 − PLout U(T/2)/T               (20)                     node distance is set to 2 m (i.e., dR,C = dC,R = 2 m).
                                                                                The values of transmission rates for both secondary and pri-
where PLout is the outage probability of S-to-D communication                   mary communications are set to U = 2 bits/s/Hz and
link in legacy scheme and it can be calculated as                               U = 1 bits/s/Hz, respectively; this is because the trans-
                                                                            mission rate of radio communication is generally higher than
                               Leg    Leg
              PLout = Pr min γR , γD        < γ0 .       (21)                   that of AB communication.

       Authorized licensed use limited to: Sungkyunkwan University. Downloaded on August 25,2020 at 07:27:11 UTC from IEEE Xplore. Restrictions apply.
SHAH et al.: OUTAGE PROBABILITY AND THROUGHPUT ANALYSIS OF SWIPT ENABLED COGNITIVE RELAY NETWORK WITH AB                                                 3203

Fig. 2. Outage probability with varying values of transmission rates (U and     Fig. 4. Outage probability with varying values of transmission power (PS )
U) for PS = 1, ζ = 0.35, and η = 1.                                             for ζ = 0.35, U = 2 bits/s/Hz, U = 1 bits/s/Hz, and η = 1.

Fig. 3. Throughput with varying values of transmission rates (U and U) for
PS = 1, ζ = 0.35, and η = 1.                                                    Fig. 5. Throughput with varying values of transmission power (PS ) for
                                                                                ζ = 0.35, U = 2 bits/s/Hz, U = 1 bits/s/Hz, and η = 1.

   The outage probability and the achievable throughput with                    at transmission rate less than a certain value, the through-
varying source-transmission rates are shown in Figs. 2 and 3,                   put decreases. Nonetheless, for transmission rate larger than a
respectively. It is evident that the analytical results of this paper           certain value, the throughput again decreases, it is because the
match the corresponding simulation results, thereby verify-                     receiving node is unable to successfully decode a large amount
ing the accuracy of the our analysis. It is shown in Fig. 2                     of received data in limited time. It can also be observed from
that as the value of transmission rate increases the outage                     Fig. 3 that the overall network sum-rate of proposed scheme
probabilities at destination nodes also increase. It is because                 is significantly higher than legacy scheme. However, in legacy
the decoding threshold values (i.e., γ0 and γ1 ) in (9), (14),                  scheme the achievable throughput at primary destination node
and (16) are increasing functions of U and U. Therefore,                        D is slightly higher than the proposed scheme. It is because,
higher transmission rates result in larger values of decoding                   in the proposed scheme, the secondary backscatter commu-
thresholds, which further leads to higher outage probabilities                  nications during time slot m2 [see (7)] causes interference to
at destinations. On the other hand, the achievable through-                     the primary communication. This tradeoff between improved
put of both primary and secondary communications in our                         network sum-rate and decreased primary throughput is part
proposed scheme increases as the transmission rate increases,                   of the underlay cognitive networks and cannot be completely
but then after a certain point (i.e., U 2 bits/s/Hz, and U                      avoided. Similar trends between proposed and legacy schemes
1 bits/s/Hz during m1 and 2 bits/s/Hz during m2 ) it starts                     can be observed in rest of this paper.
declining. This is because the achievable throughput depends                       Figs. 4 and 5 depict the outage probability and achiev-
on the transmission rate [see (13) and (18)] and therefore,                     able throughput with varying source transmission power (PS )

       Authorized licensed use limited to: Sungkyunkwan University. Downloaded on August 25,2020 at 07:27:11 UTC from IEEE Xplore. Restrictions apply.
3204                                                                                     IEEE INTERNET OF THINGS JOURNAL, VOL. 5, NO. 4, AUGUST 2018

Fig. 6. Outage probability with varying values of BS reflection coefficient     Fig. 8. Outage probability with varying values of energy conversion efficiency
(ζ ) for Ps = 1, U = 2 bits/s/Hz, U = 1 bits/s/Hz, and η = 1.                   (η) for Ps = 1, U = 2 bits/s/Hz, U = 1 bits/s/Hz, and ζ = 0.35.

Fig. 7. Throughput with varying values of BS reflection coefficient (ζ ) for    Fig. 9. Throughput with varying values of energy conversion efficiency (η)
Ps = 1, U = 2 bits/s/Hz, U = 1 bits/s/Hz, and η = 1.                            for Ps = 1, U = 2 bits/s/Hz, U = 1 bits/s/Hz, and ζ = 0.35.

values, respectively. The outage probability plot in Fig. 4                     the primary and secondary communications are provided in
shows that as the value of transmission power increases the                     Figs. 6 and 7, respectively. It has been shown that as the
outage probabilities of both primary and secondary commu-                       value of ζ increases, the outage probability of secondary nodes
nication decreases. It is because the higher values of the PS                   decreases and their throughput increases. It is due to the fact
result in improved SNR/SINR at receiving nodes which ulti-                      that at higher values of ζ , a larger portion of received signal
mately decreases their outage probability. Furthermore, it can                  power is reflected back by the AB-Tx, which leads to improved
also be observed from Fig. 4 that the outage probability of                     SNR and consequently results in reduced outage probability
secondary communication during m1 (i.e., PRC  out ) is lower than               and improved throughput at AB-Rx. On the other hand, unlike
others. It is because, unlike SWIPT operation where a por-                      secondary communication, the increasing values of ζ severely
tion of the received signal is used for ID, the AB operation                    affects the outage probability and throughput of primary com-
at the relay is directly performed by backscattering the whole                  munication. It is because, during m2 , the larger values of ζ
received signal to node C using relays backscatter interface.                   results in more severe interference to primary communication
Similar to the outage probability, the throughput results that                  [see (7)].
are plotted in Fig. 5 also show the same trend, where, as the                      The analysis of the outage probability and the throughput
value of the transmission power is increased, the achievable                    of the system with varying values of the energy conversion
throughputs at the destination nodes also increase.                             efficiency, η, are shown in Figs. 8 and 9. It can be observed
   The effects of the backscattering reflection coefficient ζ on                that the throughputs of the primary (i.e., S-R-D) and sec-
the outage probability and the achievable throughput of both                    ondary communications during the m2 (i.e., C-to-R) increase

       Authorized licensed use limited to: Sungkyunkwan University. Downloaded on August 25,2020 at 07:27:11 UTC from IEEE Xplore. Restrictions apply.
SHAH et al.: OUTAGE PROBABILITY AND THROUGHPUT ANALYSIS OF SWIPT ENABLED COGNITIVE RELAY NETWORK WITH AB                                                 3205

Fig. 10. Outage probability with varying values of source-to-relay distance     Fig. 12. Energy efficiency with varying values of transmission power (PS )
(dSR ) for Ps = 1, U = 2 bits/s/Hz, U = 1 bits/s/Hz, ζ = 0.35, and η = 1.       for ζ = 0.35, U = 2 bits/s/Hz, U = 1 bits/s/Hz, and η = 1.

                                                                                further leads to weak transmit power at relay PR which eventu-
                                                                                ally results in poor SINR/SNR for both primary and secondary
                                                                                communications [see (7) and (8)]. The energy efficiency of the
                                                                                proposed network with varying values of transmission power
                                                                                PS is plotted in Fig. 12. It can be observed that compare to
                                                                                legacy scheme the proposed scheme significantly increases the
                                                                                network energy efficiency. It is also shown that there exists a
                                                                                point where the systems energy efficiency is maximum i.e.,
                                                                                PS 0.1.

                                                                                                            V. C ONCLUSION
                                                                                   In this paper, we have studied the performance of an AB
                                                                                enabled SWIPT-based DF relay network. Analytical expres-
                                                                                sions for outage probability and achievable throughput at all
                                                                                destination nodes are derived. With the help of analytical and
                                                                                numerical results, the impact of different system parameters
Fig. 11. Throughput with varying values of source-to-relay distance (dSR )
                                                                                on overall network performance has been studied. We show
for Ps = 1, U = 2 bits/s/Hz, U = 1 bits/s/Hz, ζ = 0.35, and η = 1.              that the proposed AB enabled SWIPT DF relay network can
                                                                                significantly improve both the network sum-throughput and
                                                                                energy efficiency. It is also shown that the performance of the
                                                                                proposed scheme is highly affected by the source transmission
whereas their outage-probability values decrease as the value                   power and transmission rates.
of η increases. However, the throughput and outage probability
of secondary communication between nodes R and C during
                                                                                                              A PPENDIX A
m1 are not affected by the varying values of η. The reason is
that the AB-Tx at node R backscatters the source signal Y(S,R)                    This Appendix derives the outage probability of primary
during m1 , whose power is independent of η. Whereas, during                    communication link provided in (10). By substituting the
m2 , the node C reflects back the signal SR transmitted by node                 values of γR from (2) in J1 , the first term of (9) becomes
R, whose power PR is an increasing function of η.                                                                        
                                                                                                 J1 = Pr |g(S,R) |2 < l
   The impact of the S-to-R distance, dSR , on the outage
probability and throughput of primary and secondary commu-                                                 = F|g(S,R) |2 (l)
nications are depicted in Figs. 10 and 11. The results show that                                                                l
the performance of both the outage probability and throughput                                              = 1 − exp −                                   (30)
                                                                                                                               (S,R)
decreases as the value of dSR increases. This is because the
                                                     ε
larger values of dSR result in increased path loss d(S,R) which                                  ε
                                                                                where l = [(γ0 d(S,R) σ 2 )/Ps ], F|g(S,R) |2 (l), and (S,R) are the
eventually causes lower SNR and harvested power at relay                        cumulative distribution function and mean of the exponen-
[see (4) and (7)]. Moreover, this decline in harvested power                    tial random variable |g(S,R) |2 , respectively. Similarly, after

       Authorized licensed use limited to: Sungkyunkwan University. Downloaded on August 25,2020 at 07:27:11 UTC from IEEE Xplore. Restrictions apply.
3206                                                                                                      IEEE INTERNET OF THINGS JOURNAL, VOL. 5, NO. 4, AUGUST 2018

inserting γD from (7) into (9), the second term of (9) becomes                                    where
                                                                                                  ∞      n = [( (R,D) )/(w (R,C) (C,D) )] and E1 (x) =
                                                                                                   x  [exp(−t)/t]dt is the expectational integral. By substituting
J2                                                                                                the value of I1 derived in (34) into J2 , the integral in (32)
          PS |g(S,R) |2                                                                           can be solved as
 = Pr       ε           > γ0
           d(S,R) σ2                                                                                    ∞
                                                                                                                            b
                                                        ε
                                          ηq|g(R,D) |2 d(R,C)  ε
                                                              d(C,D)                              J2 =       1 − exp −            n exp(n)E1 (n) fq (Q)dq
                                                                                                                         q (R,D)
                                     ε
          ζ ηq|g(R,C) |2 |g(C,D) |2 d(R,D)    ε
                                           + d(S,R)  ε      ε      ε     σ2                             0
                                                    d(R,C) d(R,D) d(C,D)                                              
                                                                                                        exp − PS a(S,R)  
                                                                                                                           ∞
                                                                                                                                        q
          < γ0                                                                                        =                       exp −            dq
                                                                                                           PS (S,R)                  PS (S,R)
                                                                                                                                        0
                          ε                                                                                                                                           ∞
 = Pr Ps |g(S,R) |2 − γ0 d(S,R) σ2 > 0
                                                                                                                                            − n exp(n)E1 (n)
                     b
          |g(R,D) | < + w|g(R,C) |2 |g(C,D) |2
                        2
                                                                                           (31)                                                                       0
                     q                                                                                                                                                                         
                                                                                                                                                                  q                 b
                                                                                                                                         × exp −            −          dq
where b          =              ε
                         ([γ0 d(S,R)  ε
                                     d(R,D) σ 2 )/η] and w    =                                                                                   PS (S,R)     q (R,D)
                                                                                                                                                                        
                                                                                                                                             n exp(n)E1 (n) 4β         
        ε         ε      ε
[(ζ γ0 d(R,D) )/(d(R,C) d(C,D) )]. Conditioning J2 on q the (31)                                                           a
can be expressed as                                                                                  = exp −                             1−                     K1 4βτ
                                                                                                                  PS        (S,R)               PS (S,R)      τ
          ∞                                                                                                                                                                                  (35)
                                     b
  J2 =             1 − Pr |g(R,D) | > + w|g(R,C) |2 |g(C,D) |2
                                                  2
                                                                                                  where β = (b/ (R,D) ) and τ = (1/[PS (S,R) ]). The sec-
                                     q
                                                                                                 ond
                                                                                                   ∞ integral   part in√above equation is solved according to
                                                                                                        −(β/4z)−τ z dz = (β/τ )K (√βτ ) [29] [where K (.) is the
          0
                                                               I1                                      e
                                                                                                    0                            1                      1
          × fq (Q)dq                                                                       (32)   first order modified Bessel function of second kind]. Thus, the
                                                                                                  outage probability of primary communication can be obtained
where fq (Q) = (1/[PS (S,R) ]) exp(−([q + a]/[PS (S,R) ]))                                        by substituting the J1 [from (30)] and J2 [from (35)] in (10).
is the probability density function (PDF) of q and a =                                            This completes the proof of Proposition 1.
    ε
γ0 d(S,R) σ 2 . After conditioning on |g(R,C) |2 |g(C,D) |2 , the term
I1 of (32) can be expressed as                                                                                                             A PPENDIX B
               ∞ ∞                                                                                 The proof provided in this Appendix is for Proposition 2
                                                          b
       I1 =                 Pr |g(R,D) |2 >                 + wz1 z2                              where the analytical expression for outage probability of R-
                                                          q                                       to-C secondary communication link is derived. Based on (15),
               0    0
            × f|g(R,C) |2 (z1 )f|g(C,D) |2 (z2 )dz1 dz2                                           the PRC
                                                                                                       out can also be expressed as
                                                                                                                                                    
            ∞ ∞                                                                                                                            hγ1
                                   b            wz1 z2                                                          PRC
                                                                                                                  out = Pr  |g(S,R) | 2
                                                                                                                                        ≤                 (36)
          =       exp −                     +           f|g(R,C) |2 (z1 )                                                                 |g(R,C) |2
                              q (R,D)            (R,D)                                                             ε     ε
               0    0                                                                             where h = [(d(S,R)    d(R,C) σ 2 )/ζ PS ]. Conditioning the (36) on
               × f|g(C,D) |2 (z2 )dz1 dz2                                                  (33)   |g(R,C) | , the Pout can be expressed as
                                                                                                           2       RC

                                                                                                                  ∞
where f|g(R,C) |2 (z1 ) = [1/ (R,C) ] exp(−[z1 / (R,C) ]) and                                                                                       hγ1
f|g(C,D) |2 (z2 ) = [1/ (C,D) ] exp(−[z2 / (C,D) ]) are the PDF                                        PRC
                                                                                                        out   =            1 − exp −                              f|g(S,R) |2 (z)dz
                                                                                                                                                z   (R,C)
of |g(R,C) |2 and |g(C,D) |2 , respectively. Similarly, (R,C) and                                                 0
                                                                                                                               ⎡
                                                                                                                                   ∞
   (C,D) are their mean values. After inserting the PDF values                                                         1                                  z
in I1 , the integrals in (33) can be solved as                                                                =                ⎣        exp −                     dz
                                                                                                                      (S,R)                           (S,R)
                                                                                                                                 0
                                                                                                                                                                                         ⎤
                   exp − q (R,D)
                           b       ∞                                                                                                      ∞
                                               z2                                                                                                             z               hγ1
             I1 =                     exp −                                                                                            −        exp −                 −                 dz⎦
                            (R,C)     (C,D)                            (C,D)                                                                                  (S,R)       z   (R,C)
                                                      0                                                                                     0
                        ∞                                                                                  = 1 − vK1 (v)                                          (37)
                                              −z1             wz1 z2
                    ×           exp                       −             dz1 dz2                                     
                                              (R,D)            (R,C)                              where ν             (4hγ1 /[ (S,R) (R,C) ]), and f|g(S,R) |2 (z) =
                            0
                                                                                                [1/ (S,R) ] exp(−[z/ (S,R) ]) is the PDF of exponential random
                                                          ∞ exp −        z2
                                                                                                  variable |g(S,R) |2 . This completes the proof of Proposition 2.
                                              b                          (C,D)
               = n exp −                                                             dz2
                                      q       (R,D)            z2 + n    (C,D)
                                                          0                                                                                A PPENDIX C
                                          b                                                          The outage probability expression for secondary commu-
               = exp −                                n exp(n)E1 (n)                       (34)
                                 q        (R,D)                                                   nication during m2 (PCR
                                                                                                                       out ) is derived in this Appendix. Note

       Authorized licensed use limited to: Sungkyunkwan University. Downloaded on August 25,2020 at 07:27:11 UTC from IEEE Xplore. Restrictions apply.
SHAH et al.: OUTAGE PROBABILITY AND THROUGHPUT ANALYSIS OF SWIPT ENABLED COGNITIVE RELAY NETWORK WITH AB                                                 3207

that the solution for first term L1 of (16) is similar to J1                     [6] P. Grover and A. Sahai, “Shannon meets Tesla: Wireless information
derived in (30). After substituting the values of γR and γRBS                        and power transfer,” in Proc. IEEE Int. Symp. Inf. Theory, Austin, TX,
                                                                                     USA, 2010, pp. 2363–2367.
from (2) and (8) into (16), the second term L1 of (16) becomes                   [7] R. Zhang, R. G. Maunder, and L. Hanzo, “Wireless information and
                                                                                     power transfer: From scientific hypothesis to engineering practice,” IEEE
                PS |g(S,R) |2                                                        Commun. Mag., vol. 53, no. 8, pp. 99–105, Aug. 2015.
  L2 = Pr         ε           > γ0                                               [8] X. Zhou, R. Zhang, and C. K. Ho, “Wireless information and power
                 d(S,R) σ2
                                                                                   transfer: Architecture design and rate-energy tradeoff,” IEEE Trans.
                                                             ε                       Commun., vol. 61, no. 11, pp. 4754–4767, Nov. 2013.
                ζ η|g(R,C) |2 |g(C,R) |2 PS |g(S,R) |2 − γ0 d(S,R) σ2            [9] A. A. Nasir, X. Zhou, S. Durrani, and R. A. Kennedy, “Relaying proto-
                                 ε      ε      ε                                     cols for wireless energy harvesting and information processing,” IEEE
                                d(S,R) d(R,C) d(C,R) σ2                              Trans. Wireless Commun., vol. 12, no. 7, pp. 3622–3636, Jul. 2013.
                                                                                [10] A. A. Nasir, X. Zhou, S. Durrani, and R. A. Kennedy, “Throughput
                < γ1                                                                 and ergodic capacity of wireless energy harvesting based DF relay-
                                                                                     ing network,” in Proc. IEEE Int. Conf. Commun. (ICC), Sydney, NSW,
                                                                                     Australia, Jun. 2014, pp. 4066–4071.
                                          j
      = Pr q > 0, |g(C,R) |2 <                                           (38)   [11] H. Kawabata, K. Ishibashi, S. Vuppala, and G. T. F. de Abreu, “Robust
                                      q|g(R,C) |2                                    relay selection for large-scale energy-harvesting IoT networks,” IEEE
                                                                                     Internet Things J., vol. 4, no. 2, pp. 384–392, Apr. 2017.
                 ε     ε      ε                                                 [12] Z. Wang, Z. Chen, B. Xia, L. Luo, and J. Zhou, “Cognitive relay
where j = ([γ1 d(S,R) d(R,C) d(C,R) σ 2 ]/ηζ ). After conditioning                   networks with energy harvesting and information transfer: Design, anal-
on |g(R,C) | and q the (38) can be expressed as
            2                                                                        ysis, and optimization,” IEEE Trans. Wireless Commun., vol. 15, no. 4,
                                                                                     pp. 2562–2576, Apr. 2016.
       ∞ ∞                                                                    [13] Z. Wang et al., “Outage analysis of cognitive relay networks with energy
                             j                                                       harvesting and information transfer,” in Proc. IEEE Int. Conf. Commun.
L2 =            1 − exp −           f|g(R,C) |2 (z)dzfq (Q)dq                        (ICC), Sydney, NSW, Australia, 2014, pp. 4348–4353.
                         zq (C,R)                                               [14] Y. Wang, W. Lin, R. Sun, and Y. Huo, “Optimization of relay selection
      0 0
                                                                                   and ergodic capacity in cognitive radio sensor networks with wire-
      ∞ exp − q+a         ∞                          ∞                           less energy harvesting,” Pervasive Mobile Comput., vol. 22, pp. 33–45,
               PS (S,R)                  z
    =                          exp −              dz −                               Sep. 2015.
         PS (S,R) (R,C)                (R,C)                                    [15] Z. Wang, Z. Chen, Y. Yao, B. Xia, and H. Liu, “Wireless energy har-
        0                        0                                 0                 vesting and information transfer in cognitive two-way relay networks,”
                                                                                    in Proc. IEEE Glob. Commun. Conf., Austin, TX, USA, 2014,
                                                z              j
                               × exp −                −                dz dq         pp. 3465–3470.
                                              (R,C)       zq   (C,R)            [16] A. Liu et al., “Ambient backscatter: Wireless communication out of thin
                                                                                   air,” in Proc. ACM SIGCOMM, Hong Kong, Aug. 2013, pp. 39–50.
      ∞ exp − q+a                                                            [17] A. N. Parks, A. Liu, S. Gollakota, and J. R. Smith, “Turbocharging ambi-
               PS (S,R)
    =                      1 − ok1 (o) dq                                            ent backscatter communication,” in Proc. ACM SIGCOMM, Chicago, IL,
            PS (S,R)                                                                 USA, Aug. 2014, pp. 619–630.
      0
                                                                              [18] D. Bharadia, K. R. Joshi, M. Kotaru, and S. Katti, “BackFi: High
                                                                                     throughput WiFi backscatter,” in Proc. ACM SIGCOMM, London, U.K.,
                a           exp − PS a(S,R)
                                                                                     Aug. 2015, pp. 283–296.
    = exp −              −                                                      [19] D. Darsena, G. Gelli, and F. Verde, “Modeling and performance analysis
             PS (S,R)          PS (S,R)
                                                                                     of wireless networks with ambient backscatter devices,” IEEE Trans.
        ∞                                                                           Wireless Commun., vol. 65, no. 4, pp. 1797–1814, Apr. 2017.
                      q
      × exp −                 oK1 (o)dq                                  (39)   [20] J. Yick, B. Mukherjee, and D. Ghosal, “Wireless sensor network survey,”
                  PS (S,R)                                                           Comput. Netw., vol. 52, no. 12, pp. 2292–2330, Apr. 2008.
            0                                                                   [21] B. Medepally and N. B. Mehta, “Voluntary energy harvesting relays
                                                                                     and selection in cooperative wireless networks,” IEEE Trans. Wireless
where (C,R) is the mean of exponential random variable                               Commun., vol. 9, no. 11, pp. 3543–3553, Nov. 2010.
                     ε
|g(C,R) |2 , a = γ0 d(S,R) σ 2 , and o = (4j/[q (C,R) (R,C) ]).                 [22] P. T. Venkata, S. N. A. U. Nambi, R. V. Prasad, and I. Niemegeers,
                                                                                     “Bond graph modeling for energy-harvesting wireless sensor networks,”
To the best of our knowledge, the integral in the last term                          IEEE Comput. Soc., vol. 45, no. 9, pp. 31–38, Sep. 2012.
of (39) does not submit to any closed form solution. However,                   [23] K. Han and K. Huang, “Wirelessly powered backscatter communica-
it can easily be numerically solved using any well known                             tion networks: Modeling, coverage, and capacity,” IEEE Trans. Wireless
                                                                                     Commun., vol. 16, no. 4, pp. 2548–2561, Apr. 2017.
mathematical programs.                                                          [24] D. Bharadia, K. R. Joshi, and S. Katti, “Full duplex backscatter,” in
                                                                                     Proc. 12th ACM Workshop Hot Topics Netw., College Park, MD, USA,
                                                                                     2013, pp. 1–7.
                             R EFERENCES                                        [25] E. Everett, A. Sahai, and A. Sabharwal, “Passive self-interference sup-
                                                                                     pression for full-duplex infrastructure nodes,” IEEE Trans. Wireless
 [1] X. Lu, P. Wang, D. Niyato, D. I. Kim, and Z. Han, “Wireless networks            Commun., vol. 13, no. 2, pp. 680–694, Feb. 2014.
     with RF energy harvesting: A contemporary survey,” IEEE Commun.            [26] E. Aryafar, M. A. Khojastepour, K. Sundaresan, S. Rangarajan, and
     Surveys Tuts., vol. 17, no. 2, pp. 757–789, 2nd Quart., 2015.                   M. Chiang, “MIDU: Enabling MIMO full duplex,” in Proc. ACM
 [2] S. S. Anjum et al., “Energy management in RFID-sensor networks:                 MobiCom, Istanbul, Turkey, 2012, pp. 257–268.
     Taxonomy and challenges,” IEEE Internet Things J., to be published,        [27] J. I. Choi, M. Jain, K. Srinivasan, P. Levis, and S. Katti, “Achieving
     doi: 10.1109/JIOT.2017.2728000.                                                 single channel, full duplex wireless communications,” in Proc. ACM
 [3] S. Ulukus et al., “Energy harvesting wireless communications: A review          MobiCom, Chicago, IL, USA, 2010, pp. 1–12.
     of recent advances,” IEEE J. Sel. Areas Commun., vol. 33, no. 3,           [28] D. Bharadia and S. Katti, “Fastforward: Fast and constructive full duplex
     pp. 360–381, Mar. 2015.                                                         relays,” in Proc. ACM Conf. SIGCOMM, Chicago, IL, USA, 2014,
 [4] O. B. Akan, O. Cetinkaya, C. Koca, and M. Ozger, “Internet of hybrid            pp. 199–210.
     energy harvesting things,” IEEE Internet Things J., vol. 5, no. 2,         [29] I. S. Gradshteyn and I. M. Ryzhik, Tables of Integrals, Series, and
     pp. 736–746, Apr. 2018, doi: 10.1109/JIOT.2017.2742663.                         Products, 4th ed. New York, NY, USA: Acad. Press, 1980.
 [5] L. R. Varshney, “Transporting information and energy simultaneously,”      [30] Y. Wu, J. Wang, L. Qian, and R. Schober, “Optimal power control for
     in Proc. IEEE Int. Symp. Inf. Theory, Toronto, ON, Canada, Jul. 2008,           energy efficient D2D communication and its distributed implementa-
     pp. 1612–1616.                                                                  tion,” IEEE Commun. Lett., vol. 19, no. 5, pp. 815–818, May 2015.

       Authorized licensed use limited to: Sungkyunkwan University. Downloaded on August 25,2020 at 07:27:11 UTC from IEEE Xplore. Restrictions apply.
3208                                                                                      IEEE INTERNET OF THINGS JOURNAL, VOL. 5, NO. 4, AUGUST 2018

                         Syed Tariq Shah received the M.S. and Ph.D.                                      Tae-Jin Lee (M’99) received the B.S. and M.S.
                         degrees in electronic and electrical engineering from                            degrees in electronics engineering from Yonsei
                         Sungkyunkwan University, Seoul, South Korea, in                                  University, Seoul, South Korea, in 1989 and 1991,
                         2015 and 2018, respectively.                                                     respectively, the M.S.E. degree in electrical engi-
                           He is currently an Assistant Professor with the                                neering and computer science from the University
                         Department of Telecommunications Engineering,                                    of Michigan, Ann Arbor, MI, USA, in 1995, and the
                         Balochistan University of Information Technology,                                Ph.D. degree in electrical and computer engineering
                         Engineering and Management Sciences, Quetta,                                     from the University of Texas at Austin, Austin, TX,
                         Pakistan. His current research interests include 5G                              USA, in 1999.
                         networks, LTE-Advanced networks, wireless energy                                    In 1999, he joined the Corporate Research and
                         harvesting, and device-to-device communications.                                 Development Center, Samsung Electronics, Suwon,
                                                                                 South Korea, as a Senior Engineer. Since 2001, he has been a Professor with
                                                                                 the College of Information and Communication Engineering, Sungkyunkwan
                                                                                 University, Suwon. From 2007 to 2008, he was a Visiting Professor with
                                                                                 Pennsylvania State University, University Park, PA, USA. His current research
                                                                                 interests include performance evaluation, resource allocation, medium access
                                                                                 control, and design of communication networks and systems, wireless
                                                                                 LANs/PANs, vehicular networks, energy-harvesting networks, Internet of
                                                                                 Things, ad hoc/sensor/RFID networks, and next-generation wireless commu-
                                                                                 nication systems.
                         Kae Won Choi (M’08–SM’15) received the B.S.                Dr. Lee has been a voting member of the IEEE 802.11 WLAN Working
                         degree in civil, urban, and geosystem engineering       Group and a member of the IEICE.
                         and M.S. and Ph.D. degrees in electrical engi-
                         neering and computer science from Seoul National
                         University, Seoul, South Korea, in 2001, 2003, and
                         2007, respectively.
                           From 2008 to 2009, he was with the                                               Min Young Chung (M’04) received the B.S.,
                         Telecommunication        Business   of     Samsung                                 M.S., and Ph.D. degrees in electrical engineering
                         Electronics Company Ltd., Suwon, South Korea.                                      from the Korea Advanced Institute of Science and
                         From 2009 to 2010, he was a Post-Doctoral                                          Technology, Daejeon, South Korea, in 1990, 1993,
                         Researcher with the Department of Electrical and                                   and 1999, respectively.
Computer Engineering, University of Manitoba, Winnipeg, MB, Canada.                                            From 1999 to 2002, he was a Senior Member
From 2010 to 2016, he was an Assistant Professor with the Department of                                     of Technical Staff with the Electronics and
Computer Science and Engineering, Seoul National University of Science                                      Telecommunications Research Institute, Daejeon,
and Technology, Seoul. In 2016, he joined the faculty of Sungkyunkwan                                       where he was engaged in research on the devel-
University, Seoul, where he is currently an Associate Professor with the                                    opment of multiprotocol label switching systems.
College of Information and Communication Engineering. His current research                                  In 2002, he joined the faculty of Sungkyunkwan
interests include RF energy transfer, visible light communication, device-to-    University, Suwon, South Korea, where he is currently a Professor with
device communication, cognitive radio, and radio resource management.            the College of Information and Communication Engineering. His current
   Dr. Choi has been serving as an Editor of IEEE C OMMUNICATIONS                research interests include D2D communications, software-defined networking,
S URVEYS       AND    T UTORIALS      since      2014,   IEEE     W IRELESS      5G wireless communication networks, and wireless energy harvesting.
C OMMUNICATIONS L ETTERS since 2015, and the IEEE T RANSACTIONS                     Dr. Chung was an Editor of the Journal of Communications and Networks
ON W IRELESS C OMMUNICATIONS since 2017.                                         from 2005 to 2011. He is a member of the IEICE, KICS, KIPS, and KISS.

       Authorized licensed use limited to: Sungkyunkwan University. Downloaded on August 25,2020 at 07:27:11 UTC from IEEE Xplore. Restrictions apply.
You can also read