Wireless Networks

https://doi.org/10.1007/s11276-021-02872-z (0123456789().,-volV)(0123456789().,-volV)

Energy harvesting in ARQ-based cooperative broadcast and NOMA

networks
Emirhan Özgün1 • Ümit Aygölü1

Accepted: 13 December 2021

 The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2022

Abstract
In this paper, the effects of energy harvesting (EH) on an automatic repeat request based cooperative communication
network comprised of a source, a relay, and two destination nodes, are investigated in Rayleigh fading channels. The
energy of the decode-and-forward relay is provided by EH throughout the radio frequency signal from the source. In the
proposed transmission protocol, the source transmits a data packet to the relay and two destination nodes by direct links. If
the transmission is successful at both destination nodes, these nodes reply by an acknowledgement (ACK) signal to the
source and the relay. Otherwise, there are two alternatives: If the relay has received this data packet successfully and
replied by an ACK signal, then it cooperates and transmits the data packet for a second time to two destination nodes. If the
relay has failed and replied by a non-acknowledgement signal, the source in this case retransmits this packet to the
destination nodes by direct links. First, a broadcast channel is considered where the source transmits the same data packet
to the relay and destination nodes. Second, the non-orthogonal multiple access technique is applied to the same network
where different packets are sent to each destination node by means of superposition coding. In both cases, the system
throughputs are obtained from the derived closed-form expressions for the outage probabilities based on the Markov chain
model. All theoretical results are supported by computer simulation results.

Keywords Energy harvesting  Decode and forward relaying  Cooperative communications  Broadcast channels 
ARQ  NOMA

1 Introduction from RF signals, there are two basic protocols: The time-
switching relaying (TSR) where the relay divides the
The systems that supply energy from an external battery transmission time interval between EH and information
are short-lived due to the limited battery capacity [1]. It is processing (IP), and the power splitting relaying (PSR)
difficult and expensive to replace or refill batteries in where it divides the incoming power for energy acquisition
building structures or for sensors that are embedded within and IP.
the human body, although communication continuing Some recent studies have been considered EH in the
[2, 3]. In these scenarios, collecting energy from the network with a source, a relay, and a destination node. An
environment especially from the incoming radio frequency amplify-forward (AF) relay using the harvested energy
(RF) signals can be a safe and convenient alternative [4]. from the received source signal is considered in [5]. In this
This idea emerges from the fact that RF signals can study, both delay-limited and delay-tolerant transmission
simultaneously carry energy and information. Thus, modes are considered and the throughputs of the PSR and
energy-limited nodes can collect energy and process the TSR protocols for the three-node network are derived. In
information, simultaneously. For energy harvesting (EH) [6], the network structure in [5] is considered for the
decode-forward (DF) relaying case. In [7], a wireless net-
work consisting of M source–destination pairs communi-
& Emirhan Özgün
ozgunem@itu.edu.tr cating by means of a relay capable of accumulating all the
harvested energy is investigated. In this study, the relay
1
Department of Electronics and Communication Engineering, shares its energy among the transmitted signals to the
Istanbul Technical University, Istanbul, Turkey

123
 Wireless Networks

destination nodes. In [8], the energy-constrained relays well studied in the literature, to the best of the authors’
collect energy from the information signal transmitted from knowledge, this paper is the first study applying the ARQ
the source and from the interference signals transmitted technique to a NOMA network with an EH relay, to
through different channels in the environment. A delay- increase the reliability of the system.
tolerant system that transmits information to the destination The rest of the paper is organized as follows: In Sect. 2,
using the DF relaying where the relays harvest energy from the considered ARQ-based cooperative network and the
the incoming signal, is considered. Although using the applied transmission protocols are introduced. The
same system structure in [5] and [6, 8] differs from other throughput analysis of PSR and TSR EH protocols in an
studies by the consideration of co-channel interference. In ARQ-based system is performed for broadcast transmission
[9], an energy-limited relay that provides all of its trans- in Sect. 3. Section 4 comprises the throughput analysis of
mission energy from the harvested energy is taken into the ARQ-based NOMA system. Theoretical and simulation
account. In this multi-user network, the source transmits results are presented for broadcast transmission and
different signals to multiple destinations. These signals are NOMA cases in Sect. 5. Finally, Sect. 6 concludes the
combined at the source and the successive interference paper.
cancellation (SIC) is applied at the relay. The selection of
the appropriate multiple-access technique is one of the
most important issues in system capacity improvement. In 2 System model
[10] and [11], the capacity improvement provided by the
non-orthogonal multiple access (NOMA) technique over The considered ARQ-based cooperative communication
the classical orthogonal multiple access (OMA) techniques network with one source (base station) (S), one relay (R),
which are widely used in 4G, is exposed. In [12], the and two destination nodes (users) (U1, U2) is shown in
outage performance, spectral and energy efficiencies of Fig. 1. It is assumed that the source has an internal power
NOMA are investigated. NOMA enables multiple users in supply and the relay harvests energy from the source. The
the same cell to be served over a single frequency channel DF relay can harvest energy only from the previously sent
simultaneously and provides higher spectral efficiency, data packet by the source. It uses all of its energy when it is
higher performance, and lower transmission latency for activated to send this data packet and resets its energy
users on the edge of a cell. In [13], communication of a when it is not engaged for cooperation. All units are
source with IoT devices by applying NOMA technique is equipped with one antenna and the relay is half-duplex. d1,
considered. The outage analysis of the cooperative dual- d2, dr ; d1 , d2 and h1, h2, gr, g1, g2 denote the distances and
hop relaying scheme is performed assuming linear EH fading channel coefficients of the links S ? U1, S ? U2,
model and time switching protocol. In [14], outage and S ? R, R ? U1, and R ? U2, respectively. h1, h2, gr, g1,
throughput performances of an overlay spectrum sharing g2 are modeled as zero-mean complex Gaussian ran-
scheme with power-constraint full-duplex secondary dom variables with variances r2 ¼ 1=2k ¼ d m for which
transmitters applying cooperative NOMA technique is
investigated. Secondary transmitters apply linear EH model
and power splitting EH mode. In [15], a cooperative
NOMA network where a power-constraint near user
applies piecewise linear EH model is evaluated in terms of
outage performance and throughput in Rayleigh fading
channels. In [16], the rate optimization and fairness prob-
lems of a NOMA network consisting of multiple users and
a base station, are investigated. The optimization problem
is solved by linear and convex optimization methods.
In this paper, a wireless communication network with a
source, an EH relay, and two destination nodes commu-
nicating with each other via Rayleigh fading channels, is
considered. It is assumed that the DF relay transmits using
only the harvested energy from the previously received
source signal. In that sense, there is no energy buffer at the
relay. For both broadcast and NOMA transmission tech-
niques, system throughputs are obtained from the derived
outage probabilities based on the Markov chain model. As
well as the NOMA networks involving EH techniques are Fig. 1 Two-user cooperative ARQ system model with a DF relay

123
Wireless Networks


ðk; dÞ 2 ðkh1 ; d1 Þ; ðkh2 ; d2 Þ; ðkr ; dr Þ; ðkg1 ; d1 Þ;ðkg2 ; d2 Þg If at least one of the nodes U1 and U2 fail to receive the
packet from the source and the relay receives the same
and m denotes the path loss coefficient. Therefore, all
packet successfully, the relay becomes active in the next
random variables jhj2 , h 2 fh1 ; h2 ; gr ; g1 ; g2 g follow the time slot and sends the same packet with the harvested
exponential distribution as jhj2  eðkÞ with probability energy to U1 and U2 from a closer distance compared to
  2
density function (PDF) f jhj2 ¼ kekjhj : the source. If the relay fails to receive this packet too, the
source resends the same packet in the next time slot. In
The source broadcasts a data packet to the relay and
case of success of a transmitted packet in the first time slot
destination nodes during the first time slot. In our scenario,
or after retransmission, regardless of whether the packet
it is assumed that the relay harvests sufficient energy from
was successfully received by U1 and U2, the source
the incoming data packet from the source. This is equiva-
transmits a new packet. A failure after retransmission
lent to assume that the relay is near the source. As well as
means that this packet is dropped. The state transition
the ARQ protocol can be classified as stop-and-wait, the
probabilities in Fig. 2 are given by
waiting time (time-out period) is very short compared to
the transmission period of a data packet. If the source does P1 ¼ ð1Ps1 Þð1Ps2 Þ;
not receive ACK/NACK feedback from the relay within P2 ¼ Psr ð1  P1 Þ;
this short time-out period, the same data packet is trans- P3 ¼ ð1Ps1 Þð1Ps2 Þ;
mitted by itself. Therefore, the long time periods to receive
P4 ¼ 1  P3 ; ð1Þ
ACK/NACK signal from the relay are avoided. Assume
that the fading channel coefficients vary from one frame to P5 ¼ ð1  Psr Þð1  P1 Þ;
another (quasi-static fading). The ACK/NACK signals sent P6 ¼ ð1Pr1 Þð1Pr2 Þ;
by the destination nodes are also detected by the relay. P7 ¼ 1  P6
Thus, the relay intervenes only when the source and the
destination nodes cannot communicate directly, to avoid a Figure 2 shows the Markov chain model of the proposed
permanent loss in spectral efficiency [17]. It is assumed transmission protocol whose flow chart is given in Fig. 3.
that the feedback channels are ideal. The transmission In Fig. 2, S1, S2, and S3 represent the states corresponding
process is performed as follows:
• The source transmits a packet to U1, U2, and R through Table 1 State transitions of the considered network protocol
direct links.
• If both users did not return with ACK, then there are State Relay User 1 User 2 New State
two possibilities:
NACK ACK
• If the relay has received the packet successfully, it
NACK ACK NACK S2
transmits this packet to U1 and U2;
• Otherwise, the source is going to retransmit the NACK NACK
packet. S transmits new
NACK ACK
packet
• One retransmission (from the relay or the source) is ACK ACK NACK S3
allowed for each data packet in case of failure. S1
NACK NACK

NACK /
ACK ACK S1
3 Energy harvesting in ARQ-based ACK
broadcast channel
ACK NACK
S transmits the
In this section, the throughput analysis of the considered same packet NACK ACK
- S1
network is performed for broadcast channels where the NACK NACK
S2
source transmits the same data packet to the relay and two
ACK ACK
destination nodes. Both TSR and PSR EH protocols are
considered and analytical expressions for the outage R transmits the ACK NACK
probabilities are derived. The states and state transitions for same packet NACK ACK
- S1
the proposed transmission protocol with respect to the NACK NACK
S3
success or failure of the receiving nodes are shown in
ACK ACK
Table 1.

123
 Wireless Networks

3.1 Outage analysis for PSR EH protocol

In PSR, half of the total transmission time interval T is used

for transmission from the source to the relay and destination
nodes. During this half the relay harvests energy from the
source signal, as well as the relay and the destination nodes
process the incoming information. The other half of the total
transmission time is used for retransmission from the relay or
the source to the destination nodes when at least one of the
destination nodes fails to decode the source’s signal in the
first half. During the first half of the time interval T, the
fraction of the received signal power, qPs is used for energy
harvesting at the relay, where 0 \ q \ 1 is the power split-
ting factor. The remaining power ð1  qÞPs is used for
information processing at the relay.
Fig. 2 Markov chain model of the proposed ARQ-based protocol During the first-time interval of T/2, the received signal
at the destination node Ui, i = 1, 2 is written as [18]
to the cases of ‘‘the source transmits a new packet’’, ‘‘the pffiffiffiffiffi
source retransmits the same packet’’ and ‘‘the relay trans- yd1i ðkÞ ¼ Ps hi sðkÞ þ nd1i ðkÞ; ð5Þ
mits the same packet’’, respectively. The steady-state
where k is the time index of the symbol s sent from the
probabilities of the states S1, S2, and S3 are denoted by
source and Ps stands for the source transmit power. nd1i ðkÞ
P(S1), P(S2), and P(S3).
denotes the AWGN component at Ui. The received signal
In (1); P1, P3, and P6 denote the probabilities that a
at the relay can be given as
packet is received correctly at both destination nodes in pffiffiffiffiffi
states S1 ; S2 and S3 , respectively. Ps1 ; Ps2 ; Pr1 ; Pr2 ; and Psr yr ðkÞ ¼ Ps gr sðkÞ þ nr ðkÞ; ð6Þ
represent the outage probabilities of the links S ? U1,
where nr ðkÞ is the AWGN component at the relay. The
S ? U2, R ? U1, R ? U2, and S ? R respectively.
harvested energy at the relay during the time interval of T/2
From Fig. 2, P3 þ P4 ¼ 1; P6 þ P7 ¼ 1; P1 þ P2 þ P5 ¼ 1
can be expressed as
and the steady-state probabilities satisfy the following
matrix equations EhPS ¼ Ps jgr j2 gqðT=2Þ; ð7Þ
2 3 2 32 3
Pð S 1 Þ P1 P3 þ P4 P 6 þ P7 Pð S 1 Þ where 0\g\1 is the energy conversion efficiency. The
4 Pð S 2 Þ 5 ¼ 4 P2 0 0 54 PðS2 Þ 5: ð2Þ signal received at the destination node Ui, i = 1,2, during
Pð S 3 Þ P5 0 0 Pð S 3 Þ
the second time interval of T/2 when the relay cooperates,
P3
Using PðSi Þ ¼ 1, from (2) we obtain is written as
i¼1 qffiffiffiffiffiffiffi
PðS1 Þ ¼ 1=ð1  P2  P5 Þ; yPS ð Þ ¼ PPS
d2i k r gi s~ðk Þ þ nd2i ðkÞ; ð8Þ
PðS2 Þ ¼ P2 =ð1  P2  P5 Þ; ð3Þ
PðS3 Þ ¼ P5 =ð1  P2  P5 Þ: where s~ðkÞ is the decided symbol at the relay and nd2i ðkÞ is
the AWGN component at Ui during the second time
The outage probability of the broadcast system is given interval. The transmit power of the relay can be expressed
by as
Pout ¼ 1  ½P1 PðS1 Þ þ P3 PðS2 Þ þ P6 PðS3 Þ 2EhPS
PPS
r ¼ ¼ Ps jgr j2 gq: ð9Þ
¼ 1  ½ð1Ps1 Þð1Ps2 ÞPðS1 Þ þ ð1Ps1 Þð1Ps2 Þ T
PðS2 Þ þ ð1Pr1 Þð1Pr2 ÞPðS3 Þ: The probability that Ui fails to decode a signal from the
ð4Þ source, namely, the outage probability of direct links can
be given as
In order to evaluate (4) for PSR and TSR protocols, we    
calculate the corresponding state transition probabilities in Ps
Psi ¼ PrðRsi \RT Þ ¼ Pr log2 1 þ 2 jhi j2 \RT
the sequel. rd1i
ci
¼ Prðjhi j2 \ci Þ ¼ r khi ekhi x dx ¼ 1  ekhi ci ;
0
ð10Þ

123
Wireless Networks

Fig. 3 Flow chart of ARQ-based energy harvesting broadcast protocol

RT 1
where i ¼ 1; 2; ci ¼ P2s =r 2
2 ; RT is the target rate and rd1i is
Pri ¼ PrðRri \RT Þ
d1i ! !
the variance of the AWGN component at Ui during the first gqPs jgr j2 2
time interval of T/2. (10) is valid also for retransmission  Pr log2 1 þ jgi j \RT
r2d2i
from the source.  
The outage probability at the relay can be written as 2 2 r2d2i ð2RT  1Þ ð12Þ
¼ Pr jgr j jgi j \
gqPs
Psr ¼ PrðRsr \RT Þ ð11Þ sffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffi!
! 4b 4bi
Ps ¼ Pr ðjgr j2 jgi j2 \bi ¼ 1  i
k1 ;
¼ Prð log2 1 þ ð1  qÞjgr j2 \RT Þ kr kgi kr kgi
r2nr
! r2 ð2RT 1Þ
2RT  1 where bi ¼ d2i gqPs , K1 (.) is the first-order modified
¼ 1  exp kr ;
ðð1  qÞPs Þ=r2nr Bessel function of the second kind. (12) can be proved by
following the steps given in ‘‘Appendix A’’. Finally, using
where r2nr is the variance of nr ðkÞ. The outage probability at (4), the broadcast system throughput for PSR EH protocol
Ui, i ¼ 1; 2 when it receives a signal from the relay is can be calculated from
written as s ¼ ð1  Pout ÞRT : ð13Þ

123
 Wireless Networks

Pri ¼ PrðRri \RT Þ

! !
3.2 Outage analysis for TSR EH protocol 2gaPs jgr j2 2
¼ Pr log2 1 þ 2 jgi j \RT
r d ð 1  aÞ
In TSR, the total transmission time T for a particular  
information packet sent from the source to the destination 2 2 r2d ð2RT  1Þð1  aÞ
¼ Pr jgr j jgi j \
nodes is divided in two parts. aT is the energy harvesting 2gaPs
sffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffi!
time interval for relay where 0\a\1 represents the time 2 2 4bi 4bi
switching factor. Relay processes the information during ¼ Pr ðjgr j jgi j \bi Þ ¼ 1  K1
kr kgi kr kgi
the half of the remaining time interval of ð1  aÞT while
the other half is used for information transmission from the ð19Þ
relay or the source to the destination nodes. The selection r2d2i ð2RT 1Þð1aÞ
where bi ¼ and r2d2i ; i ¼ 1; 2 is the variance
2gaPs
of a determines the achievable throughput. The received of AWGN sample at the destination node Ui. (19) is
signal from the source at Ui, i ¼ 1; 2 during the first half of derived following the same steps given in ‘‘Appendix A’’.
the time interval of ð1  aÞT can be expressed as in (5). The throughput of the broadcast system for TSR pro-
The baseband signal received at the relay can be given as tocol is given as
follows
pffiffiffiffiffi s ¼ ð1  Pout ÞRT ð1  aÞ: ð20Þ
yr ðkÞ ¼ Ps gr sðkÞ þ nr ðkÞ: ð14Þ
The amount of harvested energy at relay during aT is
written as 4 Energy harvesting in ARQ-based NOMA
2
EhTS ¼ Ps jgr j gaT: ð15Þ
In this section, we apply the NOMA technique to the net-
The relay first decodes the signal from the source then it work considered in the previous section. Figure 4 com-
sends the decided signal to the destination during the time pares the usage of the spectrum in NOMA and OMA for a
interval of ð1  aÞT=2 with its harvested energy. In this two-user case in the downlink channel. Note that, the
case, the signal received at the destination node Ui, i ¼ 1; 2 NOMA performance degrades for more than two users and
when the relay cooperates can be given as follows therefore user pairing techniques are involved in the liter-
qffiffiffiffiffiffiffi
ature to overcome this problem [19, 20]. In the downlink
yTS
d2i ðkÞ ¼ PTS
r gi s~ðkÞ þ nd2i ðkÞ; ð16Þ
for the two-user case, the source transmits the signal si for
h i
where PTSr is the transmit power of the relay which is Ui (i = 1, 2) where E jsi j2 ¼ 1 with power ai Ps related to
defined as the status of the fading channel coefficients h1 and h2. If
EhTS 2
2Ps jgr j ga jh1 j2 [ jh2 j2 more power is allocated to the user of worse
PTS
r ¼ ¼ : ð17Þ channel ða2 [ a1 Þ where a1 þ a2 ¼ 1. Therefore, the
ð1  aÞT=2 ð 1  aÞ
source broadcasts the superposed signal
In the TSR protocol, for the signal transmitted from the pffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffi
xs ¼ a1 Ps s1 þ a2 Ps s2 . For the ARQ-based cooperative
source, the received SNRs at U1 and U2 are the same as in NOMA applied to the network with one source (S), one
the PSR case. Thus (10) is still valid for the calculation of relay (R), and two users (U1, U2) given in Fig. 1, U2
Psi , i ¼ 1; 2: The outage probability at the relay can be decodes s2 assuming the signal for U1 as interference. U1
written as applies first the successive interference cancellation (SIC)
Psr ¼ PrðRsr \RT Þ technique to decode the signal of U2, subtracts it from the
! received signal, and decodes its own signal s1 without
Ps 2
¼ Prð log2 1 þ 2 jgr j \RT Þ ð18Þ interference. The relay harvests energy from the signal xs
rnr during these operations and decodes xs to be transmitted to
 
¼ Pr jgr j2 \c ¼ 1  ekr c ; the users in case of cooperation. The ARQ mechanism and
the channel model are as indicated for the broadcast case.
R
where c ¼ P2 s T=r1 2 The flow chart for the ARQ-based NOMA case is given in
2 and rnr is the variance of the AWGN noise
nr
Fig. 5.
sample at the relay. The outage probability of the link
R ? Ui, i ¼ 1; 2 can be written as

123
Wireless Networks

Fig. 4 NOMA and OMA spectrum usage in downlink

! !
4.1 Outage analysis of ARQ-based EH NOMA   a 1 Ps j h 1 j 2
ð 1Þ ð1Þ
Ps1 ¼ Pr Rs1 \RT ¼ Pr log2 1þ \RT
r2d11
In this subsection, the outage analysis of the ARQ-based
ð 1Þ
ð1Þ
EH cooperative NOMA scheme is derived. The Markov ¼ Prðjh1 j2 \cs1 Þ ¼ 1  ekh1 cs1
chain model given in Fig. 3, the state transition and steady- ð21Þ
state probabilities given in (1) and (3), respectively, are still
valid for the NOMA case. The main difference from the ð1Þ
where cs1 ¼ c0 r2d11 =a1 Ps and c0 ¼ 2RT  1: The outage
broadcast channel is in the calculation of the outage probability of the link S ? U2 can be written as
probabilities Ps1 ; Ps2 ; Pr1 ; Pr2 ; and Psr of the links S ? U1, ! !
S ? U2, R ? U1, R ? U2, and S ? R, respectively. ð1Þ ð1Þ a2 Ps jh2 j2
Ps2 ¼ Pr ðRs2 \RT Þ ¼ Pr log2 1 þ \RT
These probabilities are equal for both TSR and PSR pro- a1 Ps jh2 j2 þr2d12
 
tocols except for Psr which is derived separately for the two ð1 Þ ð1Þ
¼ Pr jh2 j2 \cs2 ¼ 1  ekh2 cs2
EH protocols. Since the SIC process is performed at the
user with better channel condition, the following cases ð22Þ
have to be considered for the two transmission intervals. ð1Þ
where ¼ c0 r2d12 =ða2  c0 a1 ÞPs .
cs2
For the first time interval, the following two cases are
Case 2: If jh1 j2 \ jh2 j2 , the outage probability of the
valid:
link S ? U1 is obtained as
Case 1: If jh1 j2 [ jh2 j2 , the outage probability of the ! !
link S ? U1 can be given as ð2Þ ð2Þ a1 Ps jh1 j2
Ps1 ¼ Pr ðRs1 \RT Þ ¼ Pr log2 1 þ \RT
a2 Ps jh1 j2 þr2d11
  ð2Þ
ð2 Þ
¼ Pr jh1 j2 \cs1 ¼ 1  ekh1 cs1 ;

ð23Þ
ð2Þ
where cs1 ¼ c0 r2d11 =ða1  c0 a2 ÞPs : The outage probability
of the link S ? U2 can be written as

123
 Wireless Networks

Fig. 5 Flow chart of ARQ-based energy harvesting NOMA protocol

! ! ! !
ð2Þ ð2Þ a2 Ps jh2 j2 ð 1Þ ð1Þ b1 PR jg1 j2
Ps2 ¼ PrðRs2 \RT Þ ¼ Pr log2 1þ \RT Pr1 ¼ Pr ðRr1 \RT Þ ¼ Pr log2 1þ \RT
r2d12 r2d21
ð 2Þ ð 2Þ
ð2Þ ð1Þ
¼ Prðjh2 j2 \cs2 Þ ¼ 1ekh2 cs2 ¼ Pr ðjg1 j2 \cr1 Þ ¼ 1ekg1 cr1 ;
ð24Þ ð25Þ
ð2Þ
where ¼ c0 r2d12 =a2 Ps .
cs2 ð1Þ
where cr1 ¼ c0 r2d21 =b12 PR . The outage probability of the
For the second time interval, there are again two cases: link R ? U2 is given as
Case 3: If jg1 j2 [ jg2 j2 , the outage probability of the !
link R ? U1 is obtained as ð 1Þ ð1Þ b2 PR j g 2 j 2 c
Pr2 ¼ PrðRr2 \RT Þ ¼ Prðlog2 1 þ \RT Þ
b1 PR jg2 j2 þr2d22
ð 2Þ
ð1Þ
¼ Prðjg2 j2 \cr2 Þ ¼ 1ekg2 cr2 ;
ð26Þ

123
Wireless Networks

ð1Þ
where cr2 ¼ c0 r2d22 =ðb2  c0 b1 ÞPr . 5 Performance analysis
Case 4: If jg1 j2 \ jg2 j2 , the outage probability of the
In this section, the throughput performance of cooperative
link R ? U1 is obtained as
! broadcast and NOMA networks are numerically evaluated
ð2Þ ð2Þ b1 PR jg1 j2 in the presence of EH at the relay. The computer simulation
Pr1 ¼ PrðRr1 \RT Þ ¼ Prðlog2 1 þ \RT Þ
b2 PR jg1 j2 þr2d21 results obtained in the MATLAB environment by using the
ð2Þ ð 2Þ Monte-Carlo technique, are compared in each case with the
¼ Prðjg1 j2 \cr1 Þ ¼ 1ekg1 cr1 ; theoretical results derived in the previous sections for both
ð27Þ TSR and PSR protocols.
ð2Þ
where cr1 ¼ c0 r2d21 =ðb1  c0 b2 ÞPR . The outage probability 5.1 Energy harvesting in ARQ-based broadcast
of the link R ? U2 is calculated as channel
! !
ð2Þ ð2Þ b2 P R j g2 j 2
Pr2 ¼ PrðRr2 \RT Þ ¼ Pr log2 1 þ \RT Throughout the theoretical results and computer simula-
r2d22 tions, the distances between nodes were taken as
ð 2Þ
2 ð2Þ
¼ Prðjg2 j \cr2 Þ ¼ 1e kg2 cr2
; d1 ¼ d2 ¼ 1:93; d1 = d2 = dr = 1. For both TSR and PSR
protocols, the noise variance at all nodes was fixed to 0.02
ð28Þ
except the noise variance at relay for PSR protocol which
ð2Þ
where cr2 = c0 r2d22 =b2 Pr . The outage probabilities of the was taken as 0:01ð1  qÞ þ 0:01 because the total noise at
the relay is comprised of the baseband AWGN due to the
link S ? R for the PSR and TSR protocols are given as pffiffiffiffiffiffiffiffiffiffiffi
   receiving antenna which is proportional to 1  q and the
Ps ð1  qÞ
PPSR PSR
sr ¼ PrðRsr \RT Þ ¼ Pr log2 1 þ jgr j2 \RT sampled AWGN due to the baseband signal conversion.
r2r The path loss coefficient was set to m ¼ 4. The targeted
2 PSR kr cPSR data transmission rate RT and the energy conversion effi-
¼ Pr(jgr j \csr Þ ¼ 1  e sr

ciency g was taken equal to 3 and 1, respectively.

ð29Þ
The effect of a and q on the throughput is depicted in
and Fig. 6 for an SNR value of 20 dB. From this figure, the
    PSR and TSR protocols provide the maximum throughput
Ps 2
PTSR
sr ¼ PrðR TSR
sr \R T Þ ¼ Pr log 2 1 þ j g r j \R T values for q ¼ 0:75 and,a ¼ 0:25, respectively.
r2r
In Fig. 7 and Fig. 8, the throughput at different trans-
kr cTSR
= Pr ðjgr j2 \cTSR
sr Þ ¼ 1  e
sr ;
mission rates is presented for PSR and TSR protocols,
ð30Þ respectively. Since the increase in SNR does not drastically
change the value of the power splitting factor that maxi-
where cPSR 2
sr ¼ c0 rr =ð1  qÞPr and cTSR 2
sr ¼ c0 rr /Pr , mizes the throughput, PSR protocol approaches the nomi-
respectively. nal throughput value with increasing SNR for each
Finally, the average outage probabilities of the links S transmission rate. However, since the value of the time
! Ui and R ! Ui, are calculated for i ¼ 1; 2 fromand switching factor a maximizing the throughput is highly
ð1Þ ð2Þ dependent on the SNR value and since a is taken constant
Psi ¼ pPsi þ ð1  pÞPsi
in Fig. 8, the throughput of TSR protocol does not
ð1Þ ð2Þ
Pri ¼ qPri þ ð1  qÞPri ; ð31Þ approach its nominal values with increasing SNR.
 
2 2 According to Fig. 7 and Fig. 8, the PSR protocol is more
respectively, where p¼Pr jh1 j [jh2 j ¼kh2 =ðkh1 þkh2 Þ
   efficient than the TSR protocol.
and q¼Pr jg1 j2 [jg2 j2 ¼kg2 = kg1 þkg2 . In Fig. 9, the effect of the relay on system efficiency is
The overall outage probability of the ARQ-based EH examined. Direct transmission refers to the case in which
NOMA system can be derived from (4) where the steady- the source communicates with the users without any relay
state probabilities PðSi Þ; i ¼ 1; 2; 3 remain the same as in assistance. While the direct communication provides
the broadcast case. The throughputs for PSR and TSR approximately the throughput value of 0.13 when the SNR
protocols in the ARQ-based NOMA can be calculated from is equal to 18 dB, the PSR and TSR protocols for optimum
(13) and (20), respectively. q and a values, provide throughput values of approximately
0.63 and 0.53, respectively. According to Fig. 9, it can be
said that the cooperation of the EH relay with TSR and
PSR protocols significantly contributes to the efficiency of

123
 Wireless Networks

Fig. 8 Throughput for different transmission rates in TSR protocol

Fig. 6 Effects of a and q on throughput at 20 dB SNR (a ¼ 0:25; g ¼ 1; m ¼ 4)

Fig. 7 Throughput for different transmission rates in PSR protocol Fig. 9 Comparison between TSR and PSR protocols with direct
(q ¼ 0:75; g ¼ 1; m ¼ 4) transmission (RT ¼ 3; m ¼ 4; q ¼ 0:75; a ¼ 0:25)

the system compared to the direct transmission, especially considered equal to 1. The fading parameters of the Ray-
at low SNR values. leigh channels were taken as indicated in Section II. The
power splitting coefficients used in NOMA were consid-
5.2 Energy harvesting in ARQ-based NOMA ered as a1 ¼ 0:2; b1 ¼ 0:2; a2 ¼ 0:8; b2 ¼ 0:8.
In Fig. 10, the throughputs of two ARQ-based NOMA
The theoretical results for the throughput of ARQ-based systems are compared. The first one has 10 dB fixed
cooperative NOMA network are compared with the com- transmit SNR at both source and relay, the second one has
puter simulations, in this subsection. The results are pro- fixed 20 dB transmit SNR at the source and an EH relay.
vided for the distances between nodes taken as Although the total power consumed by both systems is
d1 ¼ d2 ¼ 1:74; d1 = dr = 1 and d2 ¼ 2:4 to make mean- equal, the system using the EH relay is superior compared
ingful the utilization of the NOMA technique. For both to the system with fixed power at the relay.
TSR and PSR protocols, the noise variance on all nodes The effect of q and a on the system efficiency was
was taken equal to 0.02, except the noise variance on the investigated and the results are shown in Fig. 11. The PSR
relay for PSR protocol which was assumed to change as protocol provides again better throughput performance in
0:01ð1  qÞ þ 0:01. The path loss coefficient m and the this case. At an SNR value of 20 dB, PSR and TSR pro-
targeted data transmission rate RT between nodes were both tocols provide the maximum throughput values at q ¼ 0:83
taken as equal to 2. Energy conversion efficiency g was and a ¼ 0:34, respectively.

123
Wireless Networks

Fig. 12 Optimum a=q values for TSR and PSR protocols

(RT ¼ 1; g ¼ 1; m ¼ 2)
Fig. 10 Comparison between EH-Based and fixed power relays
(RT ¼ 2; m ¼ 2; q ¼ 0:83; a ¼ 0:34)

Fig. 13 Throughput in the case of using optimum a/q values for TSR
and PSR protocols (RT ¼ 1; g ¼ 1; m ¼ 2)

Fig. 11 Effects of a and q on throughput at 20 dB SNR case, the maximum throughput can be reached for both
protocols.
According to Fig. 12, the ARQ-based cooperative Figure 14 depicts the throughput of PSR protocol for
NOMA system operating with the PSR protocol achieves different values of the target rate. The PSR protocol
the maximum throughput for the values of q between 0.8 reaches the maximum throughput value for each rate with
and 0.9, almost independently of the SNR value. However, increasing values of SNR when q ¼ 0:83. However,
the value of a maximizing the throughput decreases dras- according to Fig. 15, the TSR protocol cannot show the
tically in TSR protocol with increasing values of SNR. same performance when a fixed time switching factor value
Therefore, for the TSR protocol, the ARQ-based coopera- a ¼ 0:34, optimum for 20 dB SNR is applied.
tive NOMA system cannot work in full performance for Finally, Fig. 16 shows the effect of energy conversion
constant a values. During the computer simulations, the efficiency g on the throughput for both PSR and TSR
values of a and q have been changed in small increments to protocols. For g ¼ 0:6, the TSR protocol provides
obtain their optimum values which maximize the approximately a throughput value of 0.4 while for the PSR
throughput. protocol the throughput is equal to 0.55 at the same g value.
Figure 13 shows the throughput of the ARQ-based For all g values, PSR protocol provides higher throughput
cooperative NOMA system when the optimum a and q than TSR protocol, and the throughput increases for both
values from Fig. 12 are used for each corresponding SNR protocols with increasing values of g:
value. Accordingly, it is concluded from Fig. 13 that in this

123
 Wireless Networks

6 Conclusion

In this paper, the end-to-end throughput performance of

ARQ-based cooperative broadcast and NOMA networks
has been investigated when a DF EH relay between a
source and two destination nodes is employed. Based on
the ARQ technique, the relay intervenes only when the
channels between the source and the destination nodes are
in an outage to avoid a permanent loss in spectral effi-
ciency. Both PSR and TSR protocols have been considered
for the EH process in both broadcast channels and NOMA.
Theoretical closed-form expressions have been derived for
the outage probabilities from a Markov chain model to
obtain the system throughputs in all considered cases. The
theoretical results have been validated by computer simu-
Fig. 14 Throughput for different transmission rates in PSR protocol lations. It has been shown that ARQ-based cooperative
(q ¼ 0:83) broadcast and NOMA networks outperform the direct
transmission system without a relay. An important result is
that the throughput of the TSR protocol is highly sensitive
to the time switching factor a since its optimum value is
highly dependent on the SNR value in NOMA. However,
the optimal power splitting factor q maximizing the
throughput of PSR protocol is almost independent of the
SNR in NOMA. Therefore, PSR protocol has been shown
to be superior almost at every transmission rate and SNR
value to the TSR protocol in terms of throughput. This
result is also valid for broadcast channels. The only
exception for this is that at relatively low SNR values, TSR
can provide better throughput than PSR and this result has
been also stated in [2, 5, 6] for broadcast channels. Another
important result is that the TSR protocol has been nega-
tively affected by the increase of transmission rate at higher
Fig. 15 Throughput for different transmission rates in TSR protocol SNRs in both broadcast channels and NOMA. However,
(a ¼ 0:34Þ when the transmission rate increases, the SNR interval in
the low SNR region for which TSR outperforms PSR,
increases in NOMA. For future works, this study can be
extended to broadcast channels with more than two users.
On the other hand, for multi-user NOMA structures, user
grouping techniques can be applied.

7 Appendıx A

7.1 Proof of (12)

In order to find out the probability in (12), the analytical

expression of the probability density function (PDF) of the
multiplied two exponential random variables is required.
Let Z ¼ XY be the product of two exponentially distributed
random variables X and Y described with PDFs
Fig. 16 Relation between throughput and energy conversion effi-
ciency in TSR and PSR protocols
(RT ¼ 2; a ¼ 0:34; q ¼ 0:83; m ¼ 2)

123
Wireless Networks

fX ð xÞ ¼ kx ekx x ðA:1aÞ energy accumulation. Available: http://arxiv.org/abs/1608.

04165v2
and 2. Ran, T., Salem, A., & Ashour, H. K. (2016). Adaptive relaying
protocol for wireless power transfer and information processing.
fY ð yÞ ¼ ky eky y ; ðA:1bÞ IEEE Communications Letters, 20(10), 2027–2030.
3. Zhou, X., Zhang, R., & Ho, C. K. (2013). Wireless information
respectively. Let’s continue with the CDF of Z as and power transfer: Architecture design and rate-energy tradeoff.
IEEE Transactions on Wireless Communications, 61(11),
FZ ðzÞ,Pr ðZ  zÞ ðA:2aÞ 4754–4767.
4. Mheich, Z., & Savin, V. (2017). Cooperative communication
¼ Pr ðXY  zÞ ðA:2bÞ protocols with energy harvesting relays. IEEE 2017 Wireless
Days, 205, 60–65.
¼ Pr ðXY  z; X  0Þ þ Pr ðXY  z; X  0Þ ðA:2cÞ 5. Nasir, A. A., Zhou, X., Durrani, S., & Kennedy, R. A. (2013).
Relaying protocols for wireless energy harvesting and informa-
¼ Pr ðY  z=X; X  0Þ þ Pr ðY  z=X; X  0Þ ðA:2dÞ
tion processing. IEEE Transactions on Wireless Communications,
1 z=x 0 1 12(7), 3622–3636.
¼ r fX ð xÞ r fY ð yÞdxdy þ r fX ð xÞ r fY ð yÞdxdy: ðA:2eÞ 6. Nasir, A. A., Zhou, X., Durrani, S., & Kennedy, R. A. (2014).
0 1 1 z=x Throughput and ergodic capacity of wireless energy harvesting
based DF relaying network. In IEEE international conference on
The desired PDF is obtained by taking the derivative of communications-selected areas in communications symposium
both sides with respect to z as (pp. 4066–4071).
7. Ding, Z., Perlaza, S. M., Esnaola, I., & Poor, H. V. (2014). Power
1 1 0 1 allocation strategies in energy harvesting wireless cooperative
fZ ðzÞ ¼ r fX ð xÞfY ðz=xÞ dx þ r fX ð xÞfY ðz=xÞ dx: networks. IEEE Transactions on Wireless Communications,
0 j xj 1 j xj
13(2), 846–860.
ðA:3Þ 8. Gu, Y., Aissa, S., & Esnaola, I. (2014). RF-Based energy har-
vesting in decode-and-forward relaying systems: Ergodic and
Finally, the PDF of Z can be expressed as outage capacities. IEEE Transactions on Wireless Communica-
1 z 1 tions, 14(11), 6425–6434.
fZ ðzÞ ¼ r fX ð xÞfY dx: ðA:4Þ 9. Du, G., Xiong, K., & Qiu, Z. (2015). Outage analysis of coop-
1 x j xj erative transmission with energy harvesting relaying: Time
switching versus power splitting. Mathematical Problems in
In (A.2b), considering random variables X and Y as jgr j2 Engineering. Hindawi Publishing Corporation.
and jgi j2 , respectively, the CDF of multiplied two expo- 10. Islam, M. R., Avazov, N. K., Dobre, O. A., & Kwak, K. (2017).
nential variables can be obtained as Power-domain non-orthogonal multiple access (NOMA) in 5G
! systems: Potentials and challenges. IEEE Communications Sur-
b c veys & Tutorials, 19(2), 721–742.
Fcri ðcÞ ¼ Pr jgr j2 \ i 2 ðA:5aÞ 11. Higuchi, K., & Benjebbour, A. (2015). Non-orthogonal multiple
j gi j access (NOMA) with successive interference cancellation for
1  bi c
 future radio access. IEICE Transactions Communications, 98(3),
¼ r fjg j2 ð xÞ 1  exkx dx ðA:5bÞ 403–414.
0
i 12. Varshney, L. R. (2008). Transporting information and energy
sffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffi! simultaneously. In Proceedings of IEEE international symposium
4bi c 4bi c on information theory.
¼1 K1 ; ðA:5cÞ 13. Li, X., Wang, Q., Liu, M., Li, J., Peng, H., Piran, J., & Li, L.
kX ky kX ky (2020). Cooperative wireless-powered NOMA relaying for 5G
networks with hardware impairements and channel estimation
where kX ¼ E jg1 j2 , ky ¼ E jg1 j2 , fjg j2 ð xÞ ¼ ky eky y is the errors. IEEE Internet of Things Journal. https://doi.org/10.1109/
½ r  ½ i  i
PDF of exponential random variable jgi j2 and JIOT.2020.3029754
r2 ð2RT 1Þ 14. Le, Q. N., Yadav, A., Nguyen, N. P., Dobre, O. A., & Zhao, R.
bi ¼ d2i gqPs . Here, K1 (.) is the first-order modified (2020). Fuu-Duplex NOMA Cooperative overlay spectrum shar-
Bessel function of the second kind [21], and (A.5c) follows ing networks with SWIPT. IEEE Transactions in Green Com-
1 munications and Networking. https://doi.org/10.1109/TGCN.
b
from (A.5b) by using the equality r e4xcx dx ¼ 2020.3036026
0 15. Liu, Y., Ye, Y., Ding, H., Gao, F., & Yang, H. (2020). Outage
qffiffi pffiffiffiffiffi performance analysis for SEIPT-based incremental cooperative
b
c K1 bc [21, §3.324.1]. NOMA networks with nonlinear harvester. IEEE Communica-
tions Letters, 24(2), 287–291.
16. Diamantoulakis, P. D., Pappi, K. N., Ding, Z., & Karagiannidis,
G. K. (2016). Optimal design of non-orthogonal multiple access
with wireless power transfer. In IEEE international conference
References on communications (ICC).
17. Upadhyay, P. K., & Prakriya, S. (2010). Bidirectional cooperative
1. Ziyi, L., Chen, H., Li, Y., & Vucetic, B. (2016). Wireless energy relaying. In Alessandro, B. (Ed.). InTech radio communications
harvesting cooperative communications with direct link and (pp. 281–306). ISBN: 978-953-307-091-9. Available from: http://

123
 Wireless Networks

www.intechopen.com/books/radiocommunications/bidirectional- Department of Telecommunication Engineering from Istanbul Tech-

cooperative-relaying nical University.
18. Garg, J., Priyanka, M., & Kapil, G. (2013). A review on coop-
erative communication protocols in wireless world. International Ümit Aygölü received his B.S.,
Journal of Wireless & Mobile Networks (IJWMN), 5(2), M.S. and Ph.D. degrees, all in
107–126. electrical engineering, from
19. Zhu, L., Zhang, J., & Cao, X. (2019). Optimal user pairing for Istanbul Technical University,
downlink non-orthogonal multiple access (NOMA). IEEE Wire- Istanbul, Turkey, in 1978, 1984
less Communications Letters, 8(2), 328–331. and 1989, respectively. He was
20. Ding, Z., Fan, P., & Poor, H. V. (2016). Impact of user pairing on a Research Assistant from 1980
5G non-orthogonal multiple-access downlink transmissions. to 1986 and a Lecturer from
IEEE Transactions on Vehicle Technology, 65(8), 6010–6023. 1986 to 1989, at Yildiz Tech-
21. Gradshteyn, I. S., & Ryzhik, I. M. (1980). Table of integrals, nical University, Istanbul, Tur-
series and products (4th ed.). Academic Press Inc. key. In 1989, he became an
Assistant Professor at Istanbul
Publisher’s Note Springer Nature remains neutral with regard to Technical University where he
jurisdictional claims in published maps and institutional affiliations. became an Associate Professor
and Professor in 1992 and 1999,
respectively. His current research interests include MIMO systems,
cooperative communications, cognitive radio, energy harvesting and
Emirhan Özgün was born in spatial modulation.
Turkey, in 1993. He received
his M.S. degree in Telecommu-
nication Engineering from
Istanbul Technical University,
Turkey, in 2018 and his B.S.
degree (with honors) in Elec-
tronics and Communications
Engineering from Yildiz Tech-
nical University, Turkey, in
2015. His research interest
include wireless networks,
cooperative communication and
5G. He is currently pursuing the
Ph.D. degree with the

123