Received April 1, 2019, accepted April 17, 2019, date of publication April 22, 2019, date of current version

May 9, 2019.
Digital Object Identifier 10.1109/ACCESS.2019.2912661

Non-Orthogonal Multiple Access With

Cooperative Truncated ARQ and
Relay Selection
ZHIYUAN YU 1 , (Member, IEEE), CHAO ZHAI 2,3 , (Member, IEEE),
WAN NI1 , AND DESHENG WANG1
1 School of Journalism and Communication, Shandong University, Jinan 250100, China
2 School of Information Science and Engineering, Shandong University, Qingdao 266237, China
3 National Mobile Communications Research Laboratory, Southeast University, Nanjing 210096, China

Corresponding author: Zhiyuan Yu (yuzhiyuan@sdu.edu.cn)

This work was supported in part by the Fundamental Research Funds of Shandong University under Grant 2018GN051 and Grant
2017TB0011, in part by the Open Research Fund of the National Mobile Communications Research Laboratory, Southeast University,
under Grant 2019D09, and in part by the China Postdoctoral Science Foundation.

ABSTRACT In this paper, we propose a cooperative truncated automatic repeat-request (T-ARQ)-based

non-orthogonal multiple access schemes with relay selection. There are two phases in our proposed scheme.
In the broadcast phase, an access point superimposes the signals of near-user U1 and far-user U2 and then
broadcasts the composite signal to both users and all the relays. All the terminals will try to decode the
signal of U2 by treating the signal of U1 as interference. After that, U1 and relays will further decode the
signal of U1 by canceling the signal of U2 . The relays that can correctly decode the signals of U2 and U1
are classified into two decoding sets. In the cooperative T-ARQ phase, the relay from the decoding set with
the best channel quality toward U2 or U1 is selected to retransmit the signal if necessary. Considering the
various decoding statuses of U1 , U2 , and relays, we analyze the system throughput and reveal the impacts
of key parameters, such as power allocation factor, transmit power, and target rate. The numerical results
show that our proposed scheme can achieve much higher throughput than the T-ARQ-based time-division
multiple access schemes.

INDEX TERMS Non-orthogonal multiple access, truncated ARQ, relay selection, superposition coding,
interference cancelation.

I. INTRODUCTION allocating the transmit powers for the signals of different

Non-orthogonal multiple access (NOMA) is a promising users [16], [17].
radio access technique that can greatly increase the through-
put of 5G networks [1]–[7]. Researchers from both academia A. RELATED WORKS
and industry are working on proposing new NOMA schemes Cooperative NOMA can bring the spatial degree of freedom
for various application scenarios, such as dense networks [8], and diversity gain for a better performance [18]. Coopera-
short-packet [9] and power line [10] communications, multi- tive NOMA with relay selection has recently been studied
carrier systems [11], Internet of Things [12], and full-duplex in [19]–[26]. The two-stage decode-and-forward (DF) and
systems [13], etc. Compared with the traditional orthogonal amplify-and-forward (AF) relay selection schemes were pro-
multiple access (OMA), NOMA can greatly enhance the posed in [19], where the superimposed signal is received by
spectral efficiency via transmitting the signals of multiple multiple relays and the best one will forward the signal to both
users over the same time-frequency channel [14], [15], mean- near and far users according to the NOMA protocol. The two-
while the fairness among users can be guaranteed by properly stage best relay selection can achieve the maximal diversity
gain [20]. Deng et al. proposed to jointly select the best
The associate editor coordinating the review of this manuscript and source-relay pair from multiple sources and relays [21].
approving it for publication was Liang Yang. By using the Almouti coding, Zhao et al. proposed
2169-3536
2019 IEEE. Translations and content mining are permitted for academic research only.
56228 Personal use is also permitted, but republication/redistribution requires IEEE permission. VOLUME 7, 2019
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Z. Yu et al.: NOMA With Cooperative Truncated ARQ and Relay Selection
two dual relay selection schemes with fixed or dynamic
power allocation [22]. Yu et al. proposed a two-phase coop-
erative relaying based NOMA protocol, wherein a decodable
relay is selected to forward the far-user’s signal. The relay
transmission can not only help the far-user to decode its
desired data, but also help the near-user cancel the inter-
ference for the data decoding [23]. Using the partial relay
selection, Lee et al. derived close-form outage probabilities
and revealed that the performance gain can be enhanced with
more relays deployed [24]. If the near-user is powered by
energy harvesting (EH), a best-near best-far user selection
scheme was proposed in [25], which shows that the diver-
sity order depends on the number of far users rather than
that of near users. Based on the users’ distances towards
the base station, Liu et al. proposed three user selection
schemes and proved that the diversity gain is the same as
that of conventional cooperative system without EH [26].
In cognitive radio networks, a decodable secondary user (SU)
can be selected to transmit the superimposed signal to pri-
mary user and unsuccessful SUs [27]. Three SU scheduling
methods were proposed in [28] under the condition of no
instantaneous channel state information (CSI), partial CSI,
and full CSI, respectively. For the secure relaying,
Li et al. aimed to maximize the achievable secrecy sum rate
under the transmit power and EH constraints when the eaves-
dropper is with/without CSI [29]. Yang et al. investigated the
secrecy outage probability and average secrecy capacity in
presence of co-channel interference [30].
Choi combined hybrid automatic repeat-request (H-ARQ)
with NOMA and showed that the outage probabil-
ity decreases dramatically with more retransmissions
performed [31]. Li et al. proposed an enhanced
H-ARQ algorithm for the multiple-input multiple-output
system by adopting the power allocation, user pairing, and
precoding techniques [32]. Cai et al. studied the outage
probability and diversity order for individual users by using
the H-ARQ with chase combining [33]. For the large-scale
device-to-device network, Shi et al. proposed an interference-
aware H-ARQ based NOMA scheme and revealed that the
temporally and spatially correlated interferences can degrade
the outage performance [34]. Longer delay will be introduced
for a larger number of retransmissions, which can hardly
satisfy the low-latency requirement of 5G communications.
Therefore, the truncated ARQ (T-ARQ) with limited retrans-
mission is preferred [35], and it is shown that the relaying
NOMA with T-ARQ can greatly outperform the conventional
OMA scheme in terms of throughput.
B. MOTIVATION AND CONTRIBUTION
In the NOMA system, the information of different users can
be transmitted simultaneously over the same spectrum in
the same time slot. In the T-ARQ system, the data will be
retransmitted only when the original transmission fails and
the limited retransmission can reduce the delay. Thus, the two
schemes can both improve the communication efficiency
by causing less latency. However, it is not clear how much
VOLUME 7, 2019
performance gains can be achieved by integrating the coop-
erative T-ARQ into the NOMA system. Inspired by this fact,
we study a multi-relay scenario to improve the throughput
of NOMA system by jointly considering the cooperative
T-ARQ and the best relay selection based on the DF relaying
protocol.
In this work, we propose a cooperative T-ARQ based
NOMA scheme by selecting the best relay, where the trans-
mission process consists of the broadcast phase and the coop-
erative T-ARQ phase. In the broadcast phase, an access point
(AP) linearly combines the signals of both far-user U2 and
near-user U1 using the superposition coding technique, and
then broadcasts it over the given channel. All the terminals
including U1 , U2 , and relays try to decode the signal of U2
firstly. If U1 or any relay can successfully decode the signal
of U2 , it will cancel this signal from the received composite
signal using the successive interference cancellation (SIC)
technique and then decode the signal of U1 . The relays that
can correctly decode the signals of U1 and U2 are classi-
fied into the decoding sets D1 and D2 , respectively. Since
some relays may correctly decode the signals of both users,
D1 and D2 could have common elements. In the coopera-
tive T-ARQ phase, the selected relay from D1 or D2 will
retransmit the signals if U1 or U2 erroneously decodes their
desired data. Otherwise, if both U1 and U2 can correctly
decode their desired data in the broadcast phase, there is
no data retransmission and AP will continue to transmit a
new signal. We analyze the system throughput considering
various decoding statuses of U1 , U2 , and relays. The time
division multiple access scheme with relay selection and
T-ARQ (TDMA-RT) and the proposed scheme with imper-
fect CSI are also presented as the benchmarks for the com-
parison. Extensive numerical results show that our proposed
scheme can achieve a much higher throughput than the
TDMA-RT scheme.
The rest of this paper is organized as follows: Section II
illustrates the system model. Section III explains the trans-
mission process and the achievable rate. Section IV analyzes
the system throughput for various decoding statues. Numeri-
cal and simulation results are presented in Section V. Finally,
Section VI concludes this paper.
II. SYSTEM MODEL
We consider a network as shown in Fig. 1, which con-
sists an AP, a near-user U1 , a far-user U2 , and K relays.
U1 is close to AP and U2 is far away from AP. The aver-
age channel condition between AP and U1 is better than
that between AP and U2 . In practice, AP can represent a
broadcast tower of the next generation broadcasting wire-
less network (NGB-W), which can provide the signals of
mobile TV, public Wi-Fi, or telecommunication services
using the NOMA protocol. U1 , U2 , and relays may have
different quality of service (QoS) requirements. By exploiting
the cooperative and retransmission diversity achieved in both
space and time domains, the data can be transmitted more
reliably.
56229

FIGURE 1. The cooperative T-ARQ based NOMA with relay selection.
The transmission process can be divided into two phases:
broadcast phase and cooperative T-ARQ phase. The opera-
tions in these two phases are illustrated as follows:
In the broadcast phase, AP linearly combines the signals
of U1 and U2 using the superposition coding technique and
then broadcasts it over the wireless channel. The fraction of
power allocated to the signal s2 of U2 is denoted as θ and the
remaining fraction (1 − θ ) is allocated to the signal s1 of U1 .
Each terminal should firstly decode the signal of U2 , and then
decode the signal of U1 after cancelling the signal of U2 . With
this decoding order, we can set θ > 0.5. In the broadcast
phase, all the relays in the cooperative region can overhear
and try to decode the signals of U2 and U1 . The relays that
can correctly decode s2 and s1 are classified into the decoding
set D2 and D1 , respectively. Here, D2 and D1 may contain
the same relays. The number of decodable relays in D2 is no
less than that of D1 , because each relay decodes s2 first, and
decodes s1 only after correctly decoding and cancelling s2 .
If U1 or U2 correctly decodes its desired data in the broad-
cast phase, an acknowledgement (ACK) frame is fed back;
Otherwise, the negative acknowledgement (NACK) frame is
released to notify the relays for a retransmission. We assume
that the ACK and NACK frames are error-free and the latency
is negligible [36], [37].
In the cooperative T-ARQ phase, if either U1 or U2 or both
of them fail in decoding their desired data, a best relay from
D1 or D2 will be selected to retransmit s1 or s2 . Otherwise,
if both U1 and U2 correctly decode their desired data, new
signals will be transmitted by AP in the following blocks.
In our system, for either user, the maximum retransmission
attempts of respective signal from the best relay is set as
one by considering the delay-limited requirement of users.
That is, a packet will be discarded if it is still incorrectly
received by the user after one retransmission. This assump-
tion is suitable for real-time traffics, e.g., the transmitted
data may be audio or video signals, which can tolerate some
packet loss, but require very small packet delays. In fact,
increasing the number of retransmissions can only marginally
increase the system performance compared with the one-time
56230
Z. Yu et al.: NOMA With Cooperative Truncated ARQ and Relay Selection
retransmission [37], [38]. From this perspective, we pre-
fer to adopt the one-time retransmission. If U1 or U2 still
fails in decoding their desired data after the retransmission,
the outage event occurs. The operations in the cooperative
T-ARQ phase is classified into four cases based on the decod-
ing statuses of U1 and U2 in the broadcast phase.
A. CASE 1: U1 CORRECTLY DECODES s1 ; U2 CORRECTLY
DECODES s2
In the broadcast phase, U1 correctly decodes s1 after success-
fully decoding and canceling s2 , and U2 correctly decodes s2 .
AP will continue to transmit a new composite signal to U1 and
U2 in the next time block.
B. CASE 2: U1 CORRECTLY DECODES s1 ;
U2 ERRONEOUSLY DECODES s2
In the broadcast phase, U1 correctly decodes s1 by suc-
cessfully decoding and canceling s2 . But, U2 erroneously
decodes s2 . If D2 6 = ∅, a relay from D2 will be selected
to retransmit s2 to U2 in the cooperative T-ARQ phase.
U2 combines the signals received in the two time blocks using
the maximal-ratio combining (MRC) technique to further
decode s2 ; Otherwise, if D2 = ∅, no relay can be selected
for the retransmission of s2 , so the outage event occurs to U2 .
C. CASE 3: U1 ERRONEOUSLY DECODES s1 ; U2
CORRECTLY DECODES s2
In the broadcast phase, U2 correctly decodes s2 , but
U1 erroneously decodes s1 . Considering whether U1 correctly
decodes and cancels s2 or not, we have two subcases illus-
trated as follows:
1) SUBCASE 3A (U1 CORRECTLY DECODES s2 BUT
ERRONEOUSLY DECODES s1 ; U2 CORRECTLY
DECODES s2 )
In the broadcast phase, U1 correctly decodes s2 and cancels it
from the received composite signal, but erroneously decodes
s1 from the remaining signal. The transmission process can
be divided into two situations based on whether there exist
decodable relays in D1 or not.
• If D1 6 = ∅, a relay from D1 is selected to retransmit s1
to U1 in the cooperative T-ARQ phase. U1 combines
the signals received in the two time blocks using the
MRC technique, and then tries to decode s1 again.
• If D1 = ∅, there is no available relay for the retransmis-
sion of s1 , so the outage event occurs to U1 .
2) SUBCASE 3B (U1 ERRONEOUSLY DECODES s2 ;
U2 CORRECTLY DECODES s2 )
Given that U1 erroneously decodes s2 in the broadcast phase,
considering the decoding statuses of relays, there are two
situations:
• If D2 6 = ∅, a relay from D2 is selected to retransmit s2
to U1 . Combining the signals from AP and the best relay
by MRC technique, U1 tries to decode s2 again. In this
VOLUME 7, 2019
Z. Yu et al.: NOMA With Cooperative Truncated ARQ and Relay Selection
subcase, there are still two decoding statuses for U1
stated as follows:
- If U1 correctly decodes s2 , then it will try to decode
s1 by canceling s2 . If U1 correctly decodes s1 ,
a new signal will be sent by AP in the next time
block. If U1 erroneously decodes s1 and D1 6 = ∅,
a relay from D1 is selected to retransmit s1 to U1 ;
Otherwise, if D1 = ∅, no relay can be selected to
retransmit s1 , so the outage event occurs to U1 .
- If U1 still fails in decoding s2 , the outage event
occurs to U1 . Since the signal s2 cannot be correctly
decoded and canceled, U1 cannot decode s1 .
• If D2 = ∅, no relay is available for the retransmission
of s2 , so the outage event occurs to U1 . In this case,
AP will transmit a new signal in the next time block.
D. CASE 4: U1 ERRONEOUSLY DECODES s1 ;
U2 ERRONEOUSLY DECODES s2
In this case, both U1 and U2 erroneously decode their desired
signal, based on whether U1 correctly decodes s2 or not in the
broadcast phase, two subcases are summarized as follows.
1) SUBCASE 4A (U1 CORRECTLY DECODES s2 BUT
ERRONEOUSLY DECODES s1 ; U2 ERRONEOUSLY
DECODES s2 )
In the broadcast phase, if U1 successfully decodes and can-
cels s2 but erroneously decodes s1 , while U2 erroneously
decodes s2 , the operations in the cooperative T-ARQ phase
can be summarized as follows:
• If D2 6 = ∅ and D1 6 = ∅, a relay from D1 is selected to
retransmit s1 to U1 and then a relay from D2 is selected
to retransmit s2 to U2 in the successive time blocks.
• If D2 6 = ∅ and D1 = ∅, no relay is available for the
retransmission of s1 to U1 , so the outage event occurs
to U1 , while a relay from D2 is selected to retransmit s2
to U2 .
• If D2 = ∅, which means D1 = ∅ also, no relay is
available for the retransmission of s2 , so the outage event
occurs to U1 and U2 . In this condition, AP will send a
new signal in the next time block.
2) SUBCASE 4B (U1 ERRONEOUSLY DECODES s2 ; U2
ERRONEOUSLY DECODES s2 )
Both U1 and U2 erroneously decode s2 in the broadcast phase,
the retransmission will occur only when there exist decodable
relays in D2 , illustrated as follows:
• If D2 6 = ∅, a relay from D2 with the most favorable
channel condition towards U1 is selected to retransmit
s2 for U1 . Meanwhile, U2 can also overhear this signal.
Both U1 and U2 will combine the signals from AP
and the best relay in the two time blocks using the
MRC technique and decode s2 again.
- If U1 correctly decodes s2 after the retransmission,
then it will try to decode s1 by canceling s2 . If U1
correctly decodes s1 , there is no retransmission
VOLUME 7, 2019
of s1 towards U1 in the next time block. Otherwise,
if U1 erroneously decodes s1 and D1 6 = ∅, a best
relay from D1 is selected to retransmit s1 for U1 ;
Otherwise, if D1 = ∅, no relay can be selected
for the retransmission of s1 , and thus U1 fails to
decode s1 finally.
- If U1 still erroneously decodes s2 after the retrans-
mission of s2 , the outage event occurs to U1 , as it
cannot cancel s2 for the further decoding of s1 .
• If D2 = ∅, no relay from D2 is available for the
retransmission of s2 , so the outage event occurs to both
U1 and U2 . In this case, AP will send a new signal to U1
and U2 in the next time block.
If U2 erroneously decodes s2 in the broadcast phase,
it will listen the signal of s2 retransmitted from the selected
relay towards U1 in the cooperative T-ARQ phase and then
decode s2 again by using the MRC technique.
Since we consider the data packet transmission in each
time block, our proposed NOMA with cooperative T-ARQ is
actually delay-limited. In a time block, AP will first broadcast
the superimposed signal to both the near-user and the far-user
with a fixed transmission rate. Meanwhile, all the relay nodes
will overhear the transmission and try to decode the signals of
both users. When the data decoding is not successful at either
the near-user or the far-user, a suitable relay will be selected
to retransmit the data packet according to various decoding
cases of both users. Since only one-time retransmission is
permissible, the longest time delay of one data packet is
two-block.
III. TRANSMISSION PROCESS AND ACHIEVABLE RATE
A. BROADCAST PHASE
In the broadcast phase, AP sends a superimposed signal to U1 ,
U2 , and all the relays. The received signals at U1 , U2 , and Ri
(i = 1, · · ·, K ) can be expressed as
p 
p
y(1)
u1 = h au1 (1 − θ )P t s1 + θP t s2 + ω u1 ,
p p 
y(1)
u2 = hau2 (1 − θ )Pt s1 + θPt s2 + ωu2 ,
p p 
y(1)
ri = hari (1 − θ)Pt s1 + θ Pt s2 + ωri , (1)
where hau1 , hau2 and hari represent the channel coefficient
from AP to U1 , U2 and Ri , respectively. Pt is the transmit
power of AP. θ is the fraction of power allocated to s2 and the
remaining (1 − θ ) fraction of power is allocated to s1 . ωu1 ,
ωu2 , ωri ∼CN 0, σ 2 denote the additive white Gaussian


noise (AWGN) at U1 , U2 and Ri , respectively. The signals
are assumed to have unit power, i.e., E{|s1 |2 } = E{|s2 |2 } = 1.
As done in many existing works, such as [18]–[21],
we assume that each receiver knows the exact channel state
information (CSI) between the transmitter and itself for the
coherent decoding. The related CSI can be obtained through
the pilot training, where the transmitter can broadcast a pre-
defined pilot signal at the first beginning of each time block,
then all the receivers will try to estimate the CSI based on the
received pilot signal. Since the duration of pilot training is
56231
 Z. Yu et al.: NOMA With Cooperative Truncated ARQ and Relay Selection

much shorter than the duration of data transmission, the time retransmit s1 or s2 . The index of the best relay is given
overhead of CSI estimation is neglected in our performance as:
analysis. After obtaining the CSI, each receiver can perform
the successive interference cancelation to retrieve the desired b1 = arg max |hrm u1 |2 and b2 = arg max |hrn u1 |2 . (5)
m∈D1 n∈D2
data.
Since the received signal s2 is much stronger, U1 and all the The received signals at U1 are respectively given by
relays try to decode s2 firstly by treating s1 as interference.
p
u1 ,s1 = hrb1 u1 Pr s1 + ωu1 , Rb1 selected,
y(2)
The achievable rate of s2 at U1 and Ri are denoted as Cus12 and p
Crsi2 , respectively, given by u1 ,s2 = hrb2 u1 Pr s2 + ωu1 , Rb2 selected.
y(2) (6)

θ Pt gau1 `au1 If U1 correctly decodes s2 , while U2 erroneously

 
•
s2
Cu1 = log2 1 + , decodes s2 in the broadcast phase, a relay from D2 will
(1 − θ )Pt gau1 `au1 + σ 2

θ Pt gari `ari
 be selected to retransmit s2 and its index can be
Crsi2 = log2 1 + , (2) expressed as
(1 − θ )Pt gari `ari + σ 2
b02 = arg max |hrm u2 |2 . (7)
where |hau1 |2 = gau1 `au1 and |hari |2 = gari `ari with gau1 n∈D2
(or gari ) and `au1 (or `ari ) denoting small-scale power fading
The received signal at U2 is given by
and large-scale path-loss between AP and U1 (or AP and Ri ),
−γ
respectively. The path-loss is modeled as `au1 = dau1 and
p
−γ u2 = hrb0 u2 Pr s2 + ωu2 ,
y(2) (8)
`ari = dari , where dau1 and dari represent the distance from 2

AP to U1 and Ri , respectively. where |hrbj uj |2 = grbj uj `rbj uj represents the channel

If U1 and Ri can successfully decode and cancel s2 , they power gain between the best relay and U1 or U2
will further decode s1 with the achievable rate given by (j ∈ {1, 2}), respectively. grbj uj and `rbj uj represent
(1 − θ )Pt gau1 `au1 the small-scale power fading and the large-scale path-
 
s1
Cu1 = log2 1 + , loss, respectively. Pr is the transmit power of relay and
σ2
ωuj ∼CN (0, σ 2 ) is the AWGN.
(1 − θ )Pt gari `ari
 
Crsi1 = log2 1 + . (3) Since gri uj is independently and exponentially distributed
σ2
with unit mean, we can derive the probability density function
In the broadcast phase, U2 will try to decode s2 directly. (PDF) of grbj uj as
The achievable rate of s2 at U2 can be expressed as
dPr{Z < z}
θ Pt gau2 `au2
   W−1
fZ (z)= = W exp(−z) 1−exp(−z) (9)
Cus22 = log2 1 + , (4) dz
(1 − θ )Pt gau2 `au2 + σ 2
where Z = grbj uj and the cardinality of D is |D| = W .
where the channel power gain is |hau2 |2 = gau2 `au2 with gau2 Therein, W is an auxiliary parameter, for D1 , W = M and
−γ
and `au2 = dau2 representing the small-scale power fading for D2 , W = N .
and the large-scale path-loss, respectively. In Case 1, both U1 and U2 can correctly decode their
desired data in the broadcast phase. Next, we will present
B. COOPERATIVE T-ARQ PHASE the achievable rates of other cases considering the retrans-
Given that D1 and D2 contain M and N (M , N ∈ {1, ..., K } missions of relays.
and M ≤ N ) decodable relays, respectively. If U1 or U2
erroneously decodes its desired data, it will broadcast a 1) THE ACHIEVABLE RATE OF CASE 2
request-for-retransmission (RER) packet to all the relays. In Case 2, U2 erroneously decodes s2 in the broadcast phase
Each potential relay will estimate the channel quality between and D2 6 = ∅, the best relay from D2 will retransmit s2 in
itself and the destined user by measuring the power of the the next time block. U2 adopts the MRC technique for the
received RER packet, then a timer is set with an initial value decoding of s2 , and the achievable rate is given by
inversely proportional to the channel quality. In a given time-
Pr grb0 u2 `rb0 u2
!
window, the relay whose timer expires first will broadcast a s2 θPt gau2 `au2 2 2
Cr 0 u2 =log2 1+ +
retransmission beacon signal to all the neighboring terminals. b2 (1−θ)Pt gau2 `au2+σ 2 σ2
After hearing the beacon signal, all the other relays will stop
(10)
their timers and keep silent. In this case, the best relay is
selected, otherwise, if there is no relay broadcasting the bea- where the second term and the third term represent the signal-
con signal in the time-window, no relay can be selected [39]. to-interference-plus-noise ratio (SINR) and signal-to-noise
Considering the decoding statues of both users, the relay ratio (SNR) of s2 in the broadcast phase and the cooperative
selection process can be summarized as follows T-ARQ phase, respectively. The small-scale power fading and
• If U1 erroneously decodes s2 or s1 , a decodable relay the path-loss between the selected relay from D2 and U2 are
−γ
with the best channel quality towards U1 is selected to denoted as grb0 u2 and `rb0 u2 = drb0 u2 , respectively.
2 2 2

56232 VOLUME 7, 2019

Z. Yu et al.: NOMA With Cooperative Truncated ARQ and Relay Selection

2) THE ACHIEVABLE RATE OF CASE 3 TABLE 1. Event denotations.

In Subcase 3a, U1 successfully decodes s2 but fails to

decode s1 and D1 6 = ∅, the best relay in D1 will retransmit s1
to U1 with the achievable rate given by
(1−θ )Pt gau1 `au1 Pr grb1 u1 `rb1 u1
 
Crsb1 u1 =log2 1+ + (11)
1 σ2 σ2
where the second term and the third term represent the SNR of
s1 in the broadcast phase and the cooperative T-ARQ phase,
respectively. The small-scale power fading and the large-scale
path-loss between the best relay from D1 and U1 are denoted
−γ
as grb1 u1 and `rb1 u1 = drb1 u1 , respectively.
In subcase 3b, if U1 erroneously decodes s2 in the broad-
cast phase and D2 6 = ∅, the best relay from D2 will retrans-
mit s2 to U1 in the following time block. U1 will perform the
MRC decoding to retrieve s2 with achievable rate given by
θ Pt gau1 `au1 Pr grb2 u1 `rb2 u1
 
Crsb2 u1 = log2 1+ +
2 (1−θ )Pt gau1 `au1+σ 2 σ2
(12)
where the second term and the third term represent the SINR
and SNR of s2 in the broadcast phase and the cooperative achievable rate of respective signal is greater than the target
T-ARQ phase, respectively. The small-scale power fading and rate V1 or V2 , the transmission is deemed to be successful.
the large-scale path-loss between Rb2 and U1 are denoted as Otherwise, the outage event occurs. In our delay-limited
−γ
grb2 u1 and `rb2 u1 = drb2 u1 , respectively. NOMA system with T-ARQ, the transmission time is divided
If U1 successfully decodes and cancels s2 , it will con- into equal-length blocks. Each time block is normalized to be
tinue to decode s1 with the achievable rate Cus11 given in (3). one second. A data packet is transmitted in each time block
If U1 erroneously decodes s1 and D1 6 = ∅, a best relay from with a fixed rate. The system throughput is defined as the
D1 is selected to retransmit s1 in the next time block with the transmission rate multiplied by the data success probability.
achievable rate given in (11). If D1 = ∅, the signal s1 is not The total system throughput is expressed as
retransmitted and the outage event occurs to U1 . T = T1 + T2 + T3 + T4 , (14)
3) THE ACHIEVABLE RATE OF CASE 4 where Tj represents the throughput of Case j.
In subcase 4a, if D1 6 = ∅, the achievable rate of s1 at U1 is
given in (11). In subcase 4b, if D2 6 = ∅, the achievable rate of A. THROUGHPUT OF CASE 1 AND CASE 2
s2 at U1 is given in (12). Given that U1 correctly decodes s2 In Case 1, U1 and U2 can successfully decode their desired
after retransmission but erroneously decodes s1 with D1 6 = ∅, data without retransmission, which means Events A2 and
the retransmission of s1 will occur and the achievable rate is A1 occur at U1 and Event C occurs at U2 , so the throughput
given in (11). of Case 1 is denoted as
If U2 fails to decode s2 in the broadcast phase and D2 6 = ∅,
T1 = (V1 + V2 ) Pr(A2 , A1 )Pr(C ), (15)
U2 will overhear the retransmitted signal when the best relay | {z }
resends s2 to U1 , and then combine the received signals from φ11
AP and relay for the MRC decoding, the achievable rate of s2 The pre-factor 1 is before the transmission rate. Because only
at U2 is given by one time block is used for the data transmission when the
Pr grb2 u2 `rb2 u2 original transmission succeeds. We do not consider the pilot-
θPt gau2 `au2
 
s2 ing training phase, as it lasts a much shorter time compared
Crb u2 =log2 1+ +
2 (1−θ )Pt gau2 `au2+σ 2 σ2 with the data transmission process.
(13) The term φ11 in (15) can be calculated as
where the second term and the third term represent the SINR φ11 = exp (−9) exp (−8) , (16)
and the SNR of s2 in the broadcast phase and the cooperative
ξ2 σ
, (1−θξ1)Pσ t `au , 8 =
 2 2

T-ARQ phase, respectively. where 9 = max [θ−ξ2 (1−θ)]P t `au 1 1
ξ2 σ 2
[θ−ξ2 (1−θ)]Pt `au2 , ξ2 = 2V2 − 1, and ξ1 = 2V1 − 1. Thus,
IV. SYSTEM THROUGHPUT ANALYSIS
In this section, we will analyze the system throughput of four the condition θ > ξ2ξ+1
2
should be satisfied and then the
cases. The denotations of events are listed in Table 1. If the Events A2 and C will happen in the region of ( ξ2ξ+1
2
, 1).
VOLUME 7, 2019 56233
 Z. Yu et al.: NOMA With Cooperative Truncated ARQ and Relay Selection

Substituting the results of (16) into (15), the throughput of Substituting (65) and (19) into (18), we can obtain the
Case 1 can be obtained. result of φ21 .
In Case 2, U1 can correctly decode s1 after decoding and The term φ22 in (17) can be derived as
cancelling s2 , but U2 fails to retrieve s2 in the broadcast
K  
phase (i.e., C occurs). There are three subcases based on X K
φ22 = exp(−9) exp (−n11 )
the decoding statuses of relays and U2 . The total system n
n=1
throughput considering both U1 and U2 in Case 2 is given as K −n
× 1−exp (−11 )

Q1 (E1 ). (20)
V1 + V2
T2 = Pr(A2 , A1 )Pr(B2 , C , F )
2 | {z } The term φ23 in (17) can be derived as
φ21
K 
φ23 = exp(−9) 1−exp (−11 ) 1−exp(−8) .
 
V1 (21)
+ Pr(A2 , A1 )Pr(B2 , C , F )
2 | {z }
φ22 Plugging the results of (18), (20), and (21) into (17),
+V1 Pr(A2 , A1 )Pr(B 2 )Pr(C ) . (17) the throughput of Case 2 can be obtained.
| {z }
φ23
B. THROUGHPUT OF CASE 3
1
We incorporate the pre-factor before the transmission rate
2 In Case 3, U1 erroneously decodes s1 , while U2 successfully
when either s1 or s2 is retransmitted from the best relay to decodes s2 in the broadcast phase. Based on the decoding
either of user. Because, two time blocks are used for the data statuses of s2 and s1 at U1 , the throughput of Case 3 can be
transmission of both users. expressed as
• If D2 6 = ∅ (Event B2 occurs), the best relay from D2
will be selected to retransmit s2 in the next time block. T3 = T3a + T3b , (22)
U2 combines the signals from both phases in two time
blocks and then tries to decode s2 again. There are two where T3a and T3b represent the throughput of Subcase 3a
possible outcomes: and Subcase 3b, respectively.
- U2 correctly decodes s2 (i.e., Event F occurs).
A new data packet will be sent from AP in the 1) THROUGHPUT OF SUBCASE 3A
next time block. The throughput is expressed as the In Subcase 3a, U1 correctly decodes s2 but erroneously
first line of Eq. (17), where, where we use the pre- decodes s1 in the broadcast phase, i.e., Events A2 and A1
ξ2 σ 2 ξ1 σ 2
factor 12 before V1 +V2 to represent that the spectral occur and the constraint [θ−ξ2 (1−θ )]Pt `au1 < (1−θ )Pt `au1
efficiency is half due to the extra time block used for should be satisfied. Otherwise, the throughput of Subcase 3a
the retransmission. is zero. In the following time block, if D1 6 = ∅, U1 requires
- U2 still fails to decode s2 (i.e., Event F occurs). the retransmission of s1 from a selected relay, based on
AP will transmit a new superimposed signal in the whether U1 correctly decodes s1 or not after retransmis-
following time block. The throughput is expressed sion (i.e., D or D happens), the expressions of data suc-
as the second line of Eq. (17), where the pre-factor cess probability are denoted as φ311 Pr(C ) and φ312 Pr(C ),
1 respectively. Otherwise, if D1 = ∅ or D2 = ∅, no relay
2 before V1 is used to represent that only the data
of near-user is successfully transmitted in two time can retransmit s1 for U1 and the expressions of data suc-
blocks. cess probabilities are given as φ321 Pr(C ) and φ322 Pr(C ),
• If D2 = ∅ (Event B 2 occurs), no relay can be selected respectively. The throughput of Subcase 3a can be expressed
for the retransmission of s2 , so the outage event occurs as
to U2 . The throughput is expressed as the third line 
of Eq. (17), where the pre-factor 1 before V1 is used V1+V2 V2
to represent that the data of near-user is successfully T3a = Pr(A2 , A1 , B1 , D)+ Pr(A2 , A1 , B1 , D)
2 | {z } 2 | {z }
transmitted in one time block. φ311 φ312
The term φ21 in (17) can be expressed as


K +V2 Pr(A2 , A1 , B2 , B1 )+V2 Pr(A2 , A1 , B2 ) Pr(C ),


X  
φ21 = exp(−9) Pr(|D2 | = n)Pr C , F |D2 | = n

(18)
| {z } | {z }
φ321 φ322
n=1
(23)
where the derivation details of Pr{|D2 | = n} in (65) are given
in Appendix A. The joint probability of Event C and Event F where Pr(C ) has been derived in (16).
in (18) is derived as The term φ311 in (23) is given as
 
Pr C , F |D2 | = n = 1 − exp(−8) − Q1 (E1 ), (19)
 K
X  
φ311 = Pr(|D1 | = m)Pr A2 , A1 , D |D1 | = m ,

(24)
where Q1 (E) is given in (66) in Appendix A. m=1

56234 VOLUME 7, 2019

Z. Yu et al.: NOMA With Cooperative Truncated ARQ and Relay Selection

where the first probability is derived in (67) in Appendix A and φ342 , respectively. If |D2 | = 0 or |D1 | = 0, no relay
and the second probability is calculated as is available to retransmit signal for U1 , then the expressions
  of success probability regarding to Event D and D are
Pr A2 , A1 , D |D1 | = m

denoted as φ351 and φ352 . Denoted as T3b , the throughput of
ξ2 σ 2 Subcase 3b is expressed as
 
= exp −
[θ −ξ2 (1−θ)]Pt `au1


ξ1 σ 2  V1 +V2 V2
Pr(A2 , A1 , B2 , E )+ Pr(A2 , B2 , E )
 
T3b = 
−exp − −Q2 (E2 ), (25) 2 2 |
(1−θ )Pt `au1
| {z } {z }
φ331 φ332
ξ2 σ 2 V1 +V2
, ξ1 σ
 2

where the integral range is E2 = [θ−ξ2 (1−θ)]P t `au1 (1−θ)Pt `au1
+ Pr(A2 , A 1 , B2 , B1 , E , D)
3 | {z }
and Q2 (E) is given in (68) in Appendix A. φ341
Substituting (67) and (25) into (24), the result of φ311 can V2
be obtained. + Pr(A2 , A 1 , B2 , B1 , E , D)
3 |
The term φ312 in (23) can be expressed as
{z }
φ342
K V2
X   + Pr(A2 , A 1 , B2 , B 1 , E )
φ312 = Pr(|D1 | = m)Pr A2 , A1 , D ||D1 | = m , (26)

2 | {z }
m=1 φ351

 
where Pr A2 , A1 , D ||D1 | = m = Q2 (E2 ), substituting (67)

+ V2 Pr(A2 , B 2 ) Pr(C ). (30)

into (26), the result of φ312 can be obtained. | {z }
The summation of φ321 and φ322 in (23) can be expressed φ352

as The pre-factor 13 before the transmission rate is used when

 
φ321 + φ322 = Pr A2 , A1 both s1 and s2 are respectively retransmitted by the selected
relays, three time blocks are occupied to transmit the two
K
!
X users’ signals.
× Pr(|D2 | = n, |D1 | = 0)+Pr(|D2 | = 0) , (27) ξ2 σ 2 ξ1 σ 2
n=1
If [θ−ξ2 (1−θ )]Pt `au1 > (1−θ )Pt `au1 , the first term in (30) is given
by
where the joint probability of Events A2 and A1 is derived as K
X
φ331 = Pr{|D2 | = n}Pr(A2 , A1 , E |D2 | = n),

  ξ1 σ 2
 
 (31)
Pr A2 , A1 = −exp −
(1−θ )Pt `au1 n=1

ξ2 σ 2 where the second probability is derived as

 
+ exp − .
[θ −ξ2 (1−θ )]Pt `au1 ξ1 σ 2
  
Pr(A2 , A1 , E |D2 | = n) = exp −


(28) (1−θ )Pt `au1
Thus, if 11 < 13 , the joint probability of Events B2 and ξ2 σ 2
  
−exp − −Q3 (E3 ) . (32)
B1 is derived in (69) in Appendix A. Pr(|D2 | = 0) can be [θ −ξ2 (1 − θ )]Pt `au1
derived as The function Q3 (E) is given in (70) in Appendix  A and the
integral range is E3 = (1−θξ)P
1σ ξ2 σ 2
2
Pr(|D2 | = 0) = [1−exp(−11 )]K . (29) ,
t `au1 [θ−ξ2 (1−θ )]Pt `au1
.
The second term in (30) is given by
Plugging the results of (28), (29), and (69) into (27), the sum-
mation of φ321 and φ322 can be obtained. K
X  
φ332 = Pr {|D2 | = n} Pr A 2 , E |D2 | = n ,

Substituting the results of φ311 , φ312 , φ321 , φ322 into (23), (33)
the throughput of Subcase 3a is obtained. n=1
 
where Pr A 2 , E |D2 | = n = Q3 (E4 ) and the integral range

2) THROUGHPUT OF SUBCASE 3B
ξ2 σ 2
 
In Subcase 3b, U1 fails to decode s2 in the broadcast phase. is E4 = 0, [θ−ξ2 (1−θ)]P t `au1
. Substituting (65) into (33),
Based on the decoding statuses of relays, the retransmission the result of φ332 is obtained.
of s2 for U1 will happen when |D2 | 6 = 0. Corresponding to The term φ341 in (30) is expressed as
Event E or Event E , the data success probabilities are denoted K
as φ331 and φ332 , respectively. Given that U1 successfully
X
φ341 = Pr(|D2 | = n, |D1 | = m)
decodes and cancels s2 after a retransmission, it will decode n=1
s1 directly. If U1 erroneously decodes s1 and |D1 | 6 = 0,  
×Pr A2 , A 1 , B2 , B1 , E , D |D2 | = n, |D1 | = m ,

the best relay will retransmit s1 towards U1 in the next time
block, then the data success probabilities are denoted as φ341 (34)

VOLUME 7, 2019 56235

 Z. Yu et al.: NOMA With Cooperative Truncated ARQ and Relay Selection

where the joint probability of Event B2 and B1 is derived where T4a and T4b represent the throughput of Subcase 4a
in (71) in Appendix A. The second probability in (34) can be and Subcase 4b, respectively.
calculated as
1) THE THROUGHPUT OF SUBCASE 4A
 
Pr A2 , A 1 , B2 , B1 , E , D |D2 | = n, |D1 | = m

In Subcase 4a, U1 correctly decodes s2 but erroneously
= 1−exp(−)−Q2 (E5 )−Q3 (E5 )−Q4 (E5 ), (35) decodes s1 , while U2 erroneously decodes s2 . If D1 6= ∅,

ξ2 σ 2 ξ1 σ 2
 a best relay is selected to retransmit s1 for U1 in the next time
wherein  = min [θ−ξ2 (1−θ )]Pt `au , (1−θ )Pt `au , the function
1 1 block. After the retransmission of s1 , a best relay in D2 will
Q4 (E) is given in (72) in Appendix A and the integral range be selected to retransmit s2 for U2 . Based on the decoding
is E5 = (0, ). statuses of U1 and U2 , Event D (or D) and Event F (or F )
Substituting (71) and (35) into (34), we can obtain the will occur, where the corresponding data success probabilities
result of φ341 . are denoted as φ411 , φ412 , or φ413 , respectively. If D1 = ∅,
The term φ342 in (30) is expressed as no relay can resend s1 , i.e., Event B 1 happens, and the
ξ2 σ 2 ξ1 σ 2
K X
X n expression is given as φ42 . If [θ −ξ2 (1−θ )]Pt `au1 < (1−θ )Pt `au1 ,
φ342 = Pr {|D2 | = n, |D1 | = m} the throughput of Subcase 4a is expressed as
n=1 m=1
V1 +V2
Pr(A2 , A1 , B2 , B1 , D, C , F )
 
×Pr A 2 , E , A 1 , D |D2 | = n, |D1 | = m , T4a =

(36) 3 | {z }
φ411
where the second probability can be derived as
V1
+ Pr(A2 , A1 , B2 , B1 , D, C , F )
 
Pr A 2 , E , A 1 , D |D2 | = n, |D1 | = m

3 | {z }
φ412
= Q2 (E5 ) − Q4 (E5 ). (37)
V2
Substituting (71) and (37) into (36), we can obtain the result + Pr(A2 , A1 , B2 , B1 , D, C , F )
3 |
of φ342 .
{z }
φ413
If 11 < 13 , the φ351 in (30) can be derived as
V2
K + Pr(A2 , A1 , B2 , B 1 , C , F ) . (41)
X 2 |
φ351 =
{z }
Pr {|D2 | = n, |D1 | = 0} φ42
n=1
  The term φ411 in (41) can be expressed as
×Pr A 2 , E , A 1 |D2 | = n, |D1 | = 0

n
K X
K   X
X K φ411 = Pr(|D2 | = n, |D1 | = m)
= exp(−n11 )[1−exp(−11 )]K −n
n n=1 m=1
n=1  
×Pr A2 , A1 , D, C , F |D2 | = n, |D1 | = m ,

×[1−exp(−13 )]n 1−exp (−)−Q3 (E5 ) .
 

(38) (42)

Otherwise, 11 ≥ 13 the term φ351 = 0. where the second probability is derived as

The term φ352 in (30) is calculated as  
Pr A2 , A1 , D, C , F |D2 | = n, |D1 | = m

φ352 = [1−exp(−11 )]K
 
ξ2 σ 2
 = [1−exp(−8)−Q1 (E1 )]
× 1 −exp − . (39)  
ξ2 σ 2

[θ −ξ2 (1−θ )]Pt `au1 × exp −
[θ −ξ2 (1 − θ )]Pt `au1
Substituting the results of φ331 , φ332 , φ341 , φ342 , φ351 , and
ξ1 σ 2
  
φ352 into (30), the throughput of Subcase 3b is obtained. −exp − −Q2 (E2 ) . (43)
Adding up the throughput of Subcase 3a in (23) and (1−θ )Pt `au1
Subcase 3b in (30), the total throughput of Case 3 can be
Substituting (71) and (43) into (42), the term φ411 is
obtained.
obtained.
The term φ412 in (41) can be expressed as
C. THROUGHPUT OF CASE 4
In Case 4, both U1 and U2 erroneously decode their desired K X
X n
data in the broadcast phase. Considering whether U1 correctly φ412 = Pr(|D2 | = n, |D1 | = m)
decodes s2 or not, i.e., Event A2 occurs or not, The throughput n=1 m=1
 
of Case 4 can be expressed as ×Pr A2 , A1 , D, C , F |D2 | = n, |D1 | = m ,


T4 = T4a + T4b , (40) (44)

56236 VOLUME 7, 2019

Z. Yu et al.: NOMA With Cooperative Truncated ARQ and Relay Selection

where the second probability is derived as are denoted as φ441 , φ442 , and φ443 . Otherwise, if |D1 | = 0,
  no relay can resend s1 for U1 and the expression of success
Pr A2 , A1 , D, C , F |D2 | = n, |D1 | = m

probability is denoted as φ451 . If U1 fails to decode s2 after
 
ξ2 σ 2
 the retransmission, the expression is denoted as φ452 . The
= Q1 (E1 ) exp − throughput of Subcase 4b can be expressed as (50), as shown
[θ − ξ2 (1 − θ )]Pt `au1
at the top of the next page.
ξ1 σ 2
  
−exp − − Q2 (E2 ) . (45) The term φ431 in (50) is given by
(1 − θ )Pt `au1
K
Substituting (71) and (45) into (44), the term φ412 can be
X
φ431= Pr{|D2 |=n}Pr(A2 , A1 , E , C , G |D2 |=n).

(51)
obtained. n=1
The term φ413 in (41) can be derived as
ξ2 σ
> (1−θξ)P
1σ
2 2
K X
X n If [θ−ξ2 (1−θ)]P t `au1 t `au1
, the second probability can be
φ413 = Pr(|D2 | = n, |D1 | = m) derived as
n=1 m=1
Pr(A2 , A1 , E , C , G |D2 | = n)

 
×Pr A2 , A1 , D, C , F |D2 | = n, |D1 | = m ,

!
(46) ξ2 σ 2
= exp − ϒ(E1 )
Pr `rb2 u2
where the second probability can be calculated as
ξ1 σ 2
    
Pr A2 , A1 , D, C , F |D2 | = n, |D1 | = m

× exp − − Q3 (E3 )
(1−θ )Pt `au1
= [1 − exp(−8) − Q1 (E1 )]Q2 (E2 ). (47)
ξ2 σ 2
 
Substituting (71) and (47) into (46), the term φ413 can be −exp − . (52)
[θ − ξ2 (1−θ )]Pt `au1
obtained.
The term φ42 in (41) is expressed as where the function ϒ(E) is given in Appendix A.
K
Substituting (65) and (52) into (50), the result of φ431 is
φ42 =
X
Pr(|D2 | = n, |D1 | = 0) obtained.
The term φ432 in (50) is expressed as
n=1
 
×Pr A2 , A 1 , C , F |D2 | = n, |D1 | = 0 .
 K
X
φ432 = Pr{|D2 | = n}
(48)
n=1
If 11 < 13 , the joint probability of B2 and B 1 is derived ×Pr(A2 , A1 , E , C , G |D2 | = n).

(53)
in (69) and the second probability in (48) can be derived
ξ2 σ 2 ξ1 σ 2
as If [θ −ξ2 (1−θ )]Pt `au1 > (1−θ)Pt `au1 , the second probability can
be calculated as
 
Pr A2 , A 1 , C , F |D2 | = n, |D1 | = 0


Pr(A2 , A1 , E , C , G |D2 | = n)

= [1 − exp(−8) − Q1 (E1 )]
ξ2 σ 2
   " ! #
× exp − ξ2 σ 2
[θ − ξ2 (1 − θ )]Pt `au1 = 1 − exp(−8) − exp − ϒ(E1 )
Pr `rb2 u2
ξ1 σ 2
 
−exp − . (49)
(1 − θ )Pt `au1 ξ1 σ 2
  
× exp − − Q3 (E3 )
Substituting (69) and (49) into (48), φ42 is obtained. (1 − θ)Pt `au1
Plugging the results of φ411 , φ412 , φ413 , and φ421 into (41), 
ξ2 σ 2

the throughput of Subcase 4a, i.e., T4a can be obtained. −exp − . (54)
[θ − ξ2 (1 − θ )]Pt `au1
2) THROUGHPUT OF SUBCASE 4B Substituting (65) and (54) into (53), the result of φ432 is
In Subcase 4b, both U1 and U2 erroneously decode s2 in the obtained.
broadcast phase. If D2 6 = ∅, a best relay is selected to resend The term φ441 in (50) can be expressed as
s2 to U1 and meanwhile U2 overhears the signal. Both users K n
try to decode s2 again by using the MRC technique. If U1 cor- V1 +V2 X X
φ441 = Pr(|D2 | = n, |D1 | = m)
rectly decodes s2 and then can decode s1 directly (i.e., Event 3
n=1 m=1
A1 happens), the expressions are given as φ431 and φ432 .  
×Pr A2 , A 1 , E , D, C , G |D2 | = n, |D1 | = m ,

When Event A 1 occurs and then the best relay from D1 will
retransmit s1 , which the expressions of success probability (55)

VOLUME 7, 2019 56237

 Z. Yu et al.: NOMA With Cooperative Truncated ARQ and Relay Selection

V1 + V2 V1
T4b = Pr(A2 , A1 , B2 , E , C , G )+ Pr(A2 , A1 , B2 , E , C , G )
2 | {z } 2| {z }
φ431 φ432
V1 + V2 V1
+ Pr(A2 , A 1 , B2 , B1 , E , D, C , G )+ Pr(A2 , A 1 , B2 , B1 , E , D, C , G )
3 | {z } 3| {z }
φ441 φ442
V2 V2 V2
+ Pr(A2 , A 1 , B2 , B1 , E , D, C , G )+ Pr(A2 , A 1 , B2 , B 1 , E , C , G )+ Pr(A2 , B2 , E , C , G ) . (50)
3 | {z } 2 | {z } 2 | {z }
φ443 φ451 φ452

where the second probability can be calculated as If 11 < 13 , the term φ451 in (50) can be derived as
 
K
Pr A2 , A 1 , E , D, C , G |D2 | = n, |D1 | = m

V1 +V2 X
φ451 = Pr {|D2 | = n, |D1 | = 0}
! 2
ξ2 σ 2 n=1
= exp − ϒ(E1 )  
Pr `rb2 u2 ×Pr A 2 , E , A 1 , C , G |D2 | = n, |D1 | = 0 ,


×[1 − exp(−) − Q2 (E5 ) − Q3 (E5 ) − Q4 (E5 )]. (61)

(56)
where the first probability is given in (69) and the second
Substituting (71) and (56) into (55), we can obtain the result probability is derived as
of φ441 .  
Pr A 2 , E , A 1 , C , G |D2 | = n, |D1 | = 0

The term φ442 in (50) is expressed as
!
K
V1 X X
n ξ2 σ 2
= 1−exp (−)−Q3 (E5 ) exp − ϒ(E1).
 
φ442 = Pr(|D2 | = n, |D1 | = m) Pr `rb2 u2
3
n=1 m=1
  (62)
×Pr A2 , A 1 , E , D, C , G |D2 | = n, |D1 | = m ,

The term φ452 in (50) can be derived as
(57)
K
where the second probability can be calculated as V2 X  
φ452 = Pr {|D2 | = n} Pr A 2 , E , C , G |D2 | = n

  2
Pr A2 , A 1 , E , D, C , G |D2 | = n, |D1 | = m n=1

K  
" ! # V2 X K
ξ2 σ 2 = exp(−n11 )[1 − exp(−11 )]K −n
= 1 − exp(−8) − exp − ϒ(E1 ) 2 n
Pr `rb2 u2 n=1
!
×[1 − exp(−) − Q2 (E5 ) − Q3 (E5 ) − Q4 (E5 )]. ξ2 σ 2
×Q3 (E4 )exp − ϒ(E1 ). (63)
(58) Pr `rb2 u2

Substituting (71) and (58) into (57), we can obtain the result Substituting φ431 , φ432 , φ441 , φ442 , φ443 , φ451 , and φ452
of φ442 . into (50), the throughput of Subcase 4b, i.e., T4b can be
The term φ443 in (50) is expressed as obtained.
Summing the throughput of Subcase 4a in (41) and Sub-
K n
V2 X X case 4b in (50), the total throughput of Case 4 can be obtained.
φ443 = Pr {|D2 | = n, |D1 | = m}
3
n=1 m=1
  V. NUMERICAL AND SIMULATION RESULTS
×Pr A 2 , E , A 1 , D, C , G |D2 | = n, |D1 | = m .

(59)
where the second probability can be derived as
 
In the simulations, unless stated otherwise, the system param-
eters are set as: the number of relays K = 10, the power
allocation factor θ = 0.85, the distance between AP and U2
dau2 = d0 = 100 m, the distance between AP and U1 dau1 =
d1 = 30 m, the distance between the selected relay and U1
Pr A 2 , E , A 1 , D, C , G |D2 | = n, |D1 | = m

(or U2 ) drbi u1 = (µd0 − d1 ) (or drb2 u2 = (1 − µ)d0 ), the dis-
tance factor µ = 0.6, the transmit SNR of AP Pt = 70 dB,
!
ξ2 σ 2
= [Q2 (E5 ) − Q4 (E5 )] exp − ϒ(E1 ). (60) and the transmit SNR of relay Pr = 60 dB. In order to high-
Pr `rb2 u2
light the performance gain, we compare the proposed scheme
Substituting (71) and (60) into (59), we can obtain the result with the TDMA-RT strategy and the proposed scheme in
of φ443 . imperfect CSI, which are illustrated as follows

56238 VOLUME 7, 2019

Z. Yu et al.: NOMA With Cooperative Truncated ARQ and Relay Selection

FIGURE 2. System throughput w.r.t. power allocation coefficient θ2 for FIGURE 3. System throughput w.r.t. target rate V2 for different V1 .
different target rates, and Pr = 50 dB.

• TDMA-RT scheme: AP transmits s1 and s2 to U1 and U2

in successive time blocks.
- If both U1 and U2 correctly decode their desired
data, no retransmission is needed and a new signal
will be transmitted by AP in the following time
block.
- If either U1 or U2 erroneously decodes its desired
signal and D1 6 = ∅ or D2 6 = ∅, a decodable relay
in D1 or D2 is selected to retransmit the respective
signal in the following time block(s) according to
the T-ARQ protocol. Otherwise, if D1 = ∅ or
D2 = ∅, no relay is available for the retransmission.
After the retransmission, if U1 and U2 still erroneously
decode s1 and s2 , AP will send a new signal in the FIGURE 4. System throughput w.r.t. transmit SNR of AP Pt for different
V = V2 = V1 .
following time block.
• The proposed scheme with imperfect CSI (ICSI): Both
near-user and far-user as well as all the relays will esti- Fig. 3 shows the system throughput w.r.t. the target rate V2
mate the channel coefficient hx,y as ĥx,y . For a certain for different V1 . Given the target rate V1 , the system through-
receiver y, we have [40] put improves first and then deteriorates with the increase
of V2 . The higher V2 , the more likely the transmission is
hx,y = ĥx,y + y , (64) unsuccessful, but more information can be delivered once the
where ĥx,y is the estimated channel coefficient with transmission succeeds, as a compromise, the throughput gets
−γ larger first and then smaller. Moreover, the performance of
ĥx,y ∼ CN 0, dx,y − σ2 , the channel estimation TDMA-RT scheme is better than our proposed scheme for
error y ∼ CN 0, σ2 and the variance is fixed as


a larger V2 . Because the interference introduced by NOMA
σ2 = 0.5 × 10−6 . strategy in the broadcast phase will affect the success proba-
bility compared with the TDMA-RT scheme. The throughput
A. SYSTEM THROUGHPUT performance of the NOMA system is severely degraded with
Fig. 2 shows the system throughput with respect to (w.r.t.) the imperfect CSI.
the power allocation factor θ for different target rates V1 = Fig. 4 shows the system throughput w.r.t. the transmit SNR
V2 = V . We can see that the system throughput increases of AP for different V . With the increase of transmit SNR,
dramatically starting from zero when θ > ξ2ξ+1 2
. The reason is the system throughput increases rapidly and finally becomes
that the Events A2 in U1 and C in U2 occur more likely. After constant in the higher SNR region. In the lower SNR
reaching the maximal value, the system throughput decreases region, for each curve, thecTDMA-RT scheme achieves a
gradually. In the region of ( ξ2ξ+1
2
, 1), our proposed scheme higher throughput than our proposed scheme with/without
can achieve a much higher throughput than the TDMA-RT perfect CSI.
scheme and ICSI scheme, and the optimal θ can be deter- Fig. 5 shows the impact of the number of relays K on
mined by one-dimension search for each target rate. the system throughput. We can see that when K goes up

VOLUME 7, 2019 56239


FIGURE 5. System throughput w.r.t. number of relays K for different V2 .
V1 =1 bits/s/Hz, Pt = 65 dB, and Pr = 50 dB.
FIGURE 6. Maximal system throughput w.r.t. transmit SNR of AP Pt for
different V2 .
from 1 to 20, the throughput grows gradually and then
remains unchanged after K = 4. By adopting the T-ARQ
scheme, the throughput goes up fast and reaches the maximal
value with limited number of relays.
From Fig. 2 to Fig. 5, we can see that the numerical results
agree exactly with the simulation results, which can verify the
correctness of our analysis.
B. MAXIMAL SYSTEM THROUGHPUT
Fig. 6 shows the maximal system throughput w.r.t. the
transmit SNR of AP for different target rates. With the
increase of Pt , the maximal throughput can be improved
and then keeps invariant after reaching the maximal value
for a given target rate. For the proper settings, our pro-
posed scheme can achieve a much higher throughput than the
TDMA-RT scheme.
Fig. 7 shows the maximal system throughput w.r.t. the
number of relays K for different Pt . For relative lower
transmit SNR, the more relays in the cooperative region,
the more likely there exists the decodable relays in decoding
sets D2 or D1 . The maximal system throughput goes up first
56240
Z. Yu et al.: NOMA With Cooperative Truncated ARQ and Relay Selection
FIGURE 7. Maximal system throughput w.r.t. the number of relays K for
different Pt , and V2 = V1 = 1 bits/s/Hz.
FIGURE 8. Maximal system throughput w.r.t. distance factor µ for
different Pt .
and then tends to invariant with the increase of the K . How-
ever, for the higher transmit SNR, the maximal throughput
can be reached fast with limited relays. The reason is that
if the relay(s) can retrieve the respective signal and then the
retransmission will occur when U1 or U2 is needed.
Fig. 8 shows the maximal system throughput w.r.t. the
distance factor µ. The target rate is set as V1 = V2 = 1
bits/s/Hz. When the cooperative region gets away from AP,
the maximal throughput decreases gradually for the relative
lower SNR. The severe path-loss will adversely impact on the
system throughput. However, our proposed scheme adopts
the retransmission from the selected relay(s) to improve the
throughput performance of both users. In addition, the max-
imal system throughput is also sensitive to the transmit SNR
of AP. For the higher transmit SNR, the maximal throughput
nearly keeps invariant when µ becomes larger. The higher
transmit SNR is beneficial to combat the path-loss. By adjust-
ing the transmit SNR and the distance factor, our proposed
scheme outperforms the TDMA-RT scheme.
Fig. 9 shows the mutual impacts of the target rates V1 and
V2 on the maximal system throughput. With the increase of V1
VOLUME 7, 2019
Z. Yu et al.: NOMA With Cooperative Truncated ARQ and Relay Selection

The probability of Event B1 is derived as

 
K  m
Pr(|D1 | = m) = Pr{Crsi2 > V2 , Crsi1 > V1 }
m
K−m
× Pr{Cri < V2 }+Pr{Crsi2 > V2 , Crsi1 < V1 }
s2

 K

m exp [−m12 ] {1−exp (−13 )}
K−m
, for 11 < 13
= K

m exp [−m1 2 ] {1−exp (−1 1 )}K−m
, for others
(67)
ξ2 σ ξ1 σ
 2 2

where 12 = max [θ −ξ2 (1−θ )]Pt `ar , (1−θ )Pt `ar , and 13 =
i i
ξ1 σ 2
(1−θ )Pt `ari .
The function Q2 (E) is given by
Z
FIGURE 9. Maximal system throughput w.r.t. target rates V1 and V2 . The
upper layer and lower layer represent the performance of our proposed Q2 (E) = exp(−gau1 )
scheme and the TDMA-RT scheme, respectively. E" !#m
ξ1 σ 2 −(1−θ )Pt gau1 `au1
× 1−exp − dgau1
Pr `rb1 u1
and V2 , the throughput gets better first and then slightly dete-
riorates. But the maximal system throughput of our proposed (68)
scheme maintains at a certain level and greatly outperforms The joint probability of Events B2 and B1 is derived as
the TDMA-RT scheme for the higher target rates.
Pr(|D2 | = n, |D1 | = 0)
VI. CONCLUSION
 
K K −n
exp [−n11 ] 1−exp (−11 )

In this paper, we have proposed a cooperative T-ARQ based =
n
NOMA scheme with the best relay selection. The trans- n
× 1−exp (−13 ) .

(69)
mission process includes two phases: broadcast phase and
cooperative T-ARQ phase. In the broadcast phase, the AP The function Q3 (E) is given by
transmits the superimposed signal to both users (U1 and U2 ) Z
and meanwhile all the relays in cooperative region overhear Q3 (E) = exp(−gau1 )
this signal. After the broadcast phase, if either U1 or U2 fails E" !#n
to decode its desired signal, a best relay will be selected to (ξ2 − X )σ 2
retransmit the signal in the following time block(s). Numeri- × 1 − exp − dgau1 . (70)
Pr `rb2 u1
cal results show that our proposed scheme can achieve a much
higher throughput compared with the TDMA-RT scheme by θPt gau `au
wherein X = (1−θ )P g 1 ` 1 +σ 2 .
properly setting the system parameters. t au1 au1
The joint probability of Events B2 and B1 is derived as
APPENDIX A: DERIVATION OF THE OCCURRENCE OF
Pr(|D2 | = n, |D1 | = m)
EVENTS B2 OR B1 AND THE Q FUNCTION  
K  n  K −n
The Pr{|D2 | = n} can be derived as = Pr{Crsi2 > V2 } 1 − Pr{Crsi2 < V2 }
  n
K 
 
n  K−n n  m  n−m
Pr{|D2 | = n} = Pr{Crsi2 > V2 } 1−Pr{Crsi2 > V2 } × Pr{Crsi1 > V1 } Pr{Crsi1 < V1 }
n m
 
K
 
K−n K K −n
exp (−n11 ) 1−exp (−11 )

exp (−n11 ) 1 − exp (−11 )

= (65) =
n n
 
ξ2 σ
n n−m
exp (−m13 ) 1 − exp (−13 ) .
2 
where 11 = [θ−ξ2 (1−θ)]Pt `ari
. ×
m
(71)
The function Q1 (E) is given by
  n The function Q4 (E) is given by
Z  (ξ −Y )σ 2  " !#n
2
(ξ2 −X )σ 2
Z
Q1 (E) = exp(−gau2 ) 1−exp −  dgau2
E Pr `rb0 u2  Q4 (E)= exp(−gau1) 1−exp −
Pr `rb2 u1

2 E
(66) " !#m
ξ1 σ 2 −(1−θ )Pt gau1 `au1
θ Pt gau2 `au2 × 1−exp − dgau1 .
where Y = , and the integral interval Pr `rb1 u1
(1−θ)Pt gau2 `au2 +σ 2
E1 = (0, 8). (72)

VOLUME 7, 2019 56241


The function ϒ(E) in (52) is given by
! for cooperative NOMA,’’ IEEE Trans. Veh. Technol., vol. 66, no. 11,
Yσ2
Z
ϒ(E) = exp −gau2 + dgau2 . (73)
E Pr `rb2 u2
REFERENCES
[1] Y. Saito, A. Benjebbour, Y. Kishiyama, and T. Nakamura, ‘‘System-
level performance evaluation of downlink non-orthogonal multiple
access (NOMA),’’ in Proc. IEEE 24th Annu. Int. Symp. Pers.,
Indoor, Mobile Radio Commun. (PIMRC), London, U.K., Sep. 2013,
pp. 611–615.
[2] Y. Liu, Z. Qin, M. Elkashlan, Z. Ding, A. Nallanathan, and L. Hanzo,
‘‘Nonorthogonal multiple access for 5G and beyond,’’ Proc. IEEE,
vol. 105, no. 12, pp. 2347–2381, Dec. 2017.
[3] C.-L. I, S. Han, Z. Xu, S. Wang, Q. Sun, and Y. Chen, ‘‘New paradigm
of 5G wireless Internet,’’ IEEE J. Sel. Areas Commun., vol. 34, no. 3,
pp. 474–482, Mar. 2016.
[4] S. M. R. Islam, N. Avazov, O. A. Dobre, and K.-S. Kwak, ‘‘Power-domain
non-orthogonal multiple access (NOMA) in 5G systems: Potentials and
challenges,’’ IEEE Commun. Surveys Tuts., vol. 19, no. 2, pp. 721–742,
2nd Quart., 2016.
[5] W. Shin, M. Vaezi, B. Lee, D. J. Love, J. Lee, and H. V. Poor,
‘‘Non-orthogonal multiple access in multi-cell networks: Theory, perfor-
mance, and practical challenges,’’ IEEE Commun. Mag., vol. 55, no. 10,
pp. 176–183, Oct. 2017.
[6] Z. Ding, X. Lei, G. K. Karagiannidis, R. Schober, J. Yuan, and V. Bhargava,
‘‘A survey on non-orthogonal multiple access for 5G networks: Research
challenges and future trends,’’ IEEE J. Sel. Areas Commun., vol. 35, no. 10,
pp. 2181–2195, Oct. 2017.
[7] L. Dai, B. Wang, Y. Yuan, S. Han, C.-L. I, and Z. Wang, ‘‘Non-
orthogonal multiple access for 5G: solutions, challenges, opportuni-
ties, and future research trends,’’ IEEE Commun. Mag., vol. 53, no. 9,
pp. 74–81, Sep. 2015.
[8] Z. Zhang, H. Sun, and R. Q. Hu, ‘‘Downlink and uplink non-orthogonal
multiple access in a dense wireless network,’’ IEEE J. Sel. Areas Commun.,
vol. 35, no. 12, pp. 2771–2784, Dec. 2017.
[9] Y. Yu, H. Chen, Y. Li, Z. Ding, and B. Vucetic, ‘‘On the per-
formance of non-orthogonal multiple access in short-packet com-
munications,’’ IEEE Commun. Lett., vol. 22, no. 3, pp. 590–593,
Mar. 2018.
[10] K. M. Rabie, B. Adebisi, E. H. Yousif, H. Gacanin, and A. M. Tonello,
‘‘A comparison between orthogonal and non-orthogonal multiple access
in cooperative relaying power line communication systems,’’ IEEE Access,
vol. 5, pp. 10118–10129, 2017.
[11] Y. Sun, D. W. K. Ng, Z. Ding, and R. Schober, ‘‘Optimal joint power and
subcarrier allocation for full-duplex multicarrier non-orthogonal multiple
access systems,’’ IEEE Trans. Commun., vol. 65, no. 3, pp. 1077–1091,
Mar. 2017.
[12] Z. Ding, L. Dai, and H. V. Poor, ‘‘MIMO-NOMA design for small
packet transmission in the Internet of Things,’’ IEEE Access, vol. 4,
pp. 1393–1405, 2016.
[13] M. S. ElBamby, M. Bennis, W. Saad, M. Debbah, and M. Latva-Aho,
‘‘Resource optimization and power allocation in in-band full
duplex-enabled non-orthogonal multiple access networks,’’
IEEE J. Sel. Areas Commun., vol. 35, no. 12, pp. 2860–2873,
Dec. 2017.
[14] J.-B. Kim and I.-H. Lee, ‘‘Capacity analysis of cooperative relaying sys-
tems using non-orthogonal multiple access,’’ IEEE Commun. Lett., vol. 19,
no. 11, pp. 1949–1952, Nov. 2015.
[15] Z. Ding, Z. Zhao, M. Peng, and H. V. Poor, ‘‘On the spectral effi-
ciency and security enhancements of NOMA assisted multicast-unicast
streaming,’’ IEEE Trans. Commun., vol. 65, no. 7, pp. 3151–3163,
Jul. 2017.
[16] S. Timotheou and I. Krikidis, ‘‘Fairness for non-orthogonal multiple access
in 5G systems,’’ IEEE Signal Process. Lett., vol. 22, no. 10, pp. 1647–1651,
Oct. 2015.
[17] J. Men and J. Ge, ‘‘Non-orthogonal multiple access for multiple-antenna
relaying networks,’’ IEEE Commun. Lett., vol. 19, no. 10, pp. 1686–1689,
Oct. 2015.
[18] Z. Ding, M. Peng, and H. V. Poor, ‘‘Cooperative non-orthogonal multiple
access in 5G systems,’’ IEEE Commun. Lett., vol. 19, no. 8, pp. 1462–1465,
Aug. 2015.
56242
Z. Yu et al.: NOMA With Cooperative Truncated ARQ and Relay Selection
[19] Z. Yang , Z. Ding, Y. Wu, and P. Fan, ‘‘Novel relay selection strategies
pp. 10114–10123, Nov. 2017.
[20] Z. Ding, H. Dai, and H. V. Poor, ‘‘Relay selection for cooperative
NOMA,’’ IEEE Wireless Commun. Lett., vol. 5, no. 4, pp. 416–419,
Aug. 2016.
[21] D. Deng, L. Fan, X. Lei, W. Tan, and D. Xie, ‘‘Joint user and relay selection
for cooperative NOMA networks,’’ IEEE Access, vol. 5, pp. 20220–20227,
2017.
[22] J. Zhao, Z. Ding, P. Fan, Z. Yang, and G. K. Karagiannidis, ‘‘Dual relay
selection for cooperative NOMA with distributed space time coding,’’
IEEE Access, vol. 6, pp. 20440–20450, 2018.
[23] Z. Yu, C. Zhai, J. Liu, and H. Xu, ‘‘Cooperative relaying based non-
orthogonal multiple access (NOMA) with relay selection,’’ IEEE Trans.
Veh. Technol., vol. 67, no. 12, pp. 11606–11618, Dec. 2018.
[24] S. Lee, D. B. da Costa, Q.-T. Vien, T. Q. Duong, and R. T. de Sousa,
‘‘Non-orthogonal multiple access schemes with partial relay selection,’’
IET Commun., vol. 11, no. 6, pp. 846–854, Apr. 2017.
[25] N. T. Do, D. B. da Costa, T. Q. Duong, and B. An, ‘‘A BNBF user selection
scheme for NOMA-based cooperative relaying systems with SWIPT,’’
IEEE Commun. Lett., vol. 21, no. 3, pp. 664–667, Mar. 2017.
[26] Y. Liu, Z. Ding, M. Elkashlan, and H. V. Poor, ‘‘Cooperative non-
orthogonal multiple access with simultaneous wireless information and
power transfer,’’ IEEE J. Sel. Areas Commun., vol. 34, no. 4, pp. 938–953,
Apr. 2016.
[27] L. Lv, J. Chen, and Q. Ni, ‘‘Cooperative non-orthogonal multiple access
in cognitive radio,’’ IEEE Commun. Lett., vol. 20, no. 10, pp. 2059–2062,
Oct. 2016.
[28] L. Lv, J. Chen, Q. Ni, and Z. Ding, ‘‘Design of cooperative non-
orthogonal multicast cognitive multiple access for 5G systems: User
scheduling and performance analysis,’’ IEEE Trans. Commun., vol. 65,
no. 6, pp. 2641–2656, Jun. 2017.
[29] Q. Li, Q. Zhang, and J. Qin, ‘‘Secure relay beamforming for SWIPT in
amplify-and-forward two-way relay networks,’’ IEEE Trans. Veh. Technol.,
vol. 65, no. 11, pp. 9006–9019, Nov. 2016.
[30] L. Yang, M. O. Hasna, and I. S. Ansari, ‘‘Physical layer security
for TAS/MRC systems with and without co-channel interference over
η–µ fading channels,’’ IEEE Trans. Veh. Technol., vol. 67, no. 12,
pp. 12421–12426, Dec. 2018.
[31] J. Choi, ‘‘H-ARQ based non-orthogonal multiple access with successive
interference cancellation,’’ in Proc. IEEE Global Telecommun. Conf.,
New Orleans, LO, USA, Nov./Dec. 2008, pp. 1–5.
[32] A. Li, A. Benjebbour, X. Chen, H. Jiang, and H. Kayama, ‘‘Inves-
tigation on hybrid automatic repeat request (HARQ) design for
NOMA with SU-MIMO,’’ in Proc. IEEE 26th Annu. Int. Symp. Pers.,
Indoor, Mobile Radio Commun. (PIMRC), Hong Kong, Aug./Sep. 2015,
pp. 590–594.
[33] D. Cai, Z. Ding, P. Fan, and Z. Yang, ‘‘On the performance of
NOMA with hybrid ARQ,’’ IEEE Trans. Veh. Technol., vol. 67, no. 10,
pp. 10033–10038, Oct. 2018.
[34] Z. Shi, S. Ma, H. E. Sawy, G. Yang, and M.-S. Alouini, ‘‘Cooperative
HARQ-assisted NOMA scheme in large-scale D2D networks,’’ IEEE
Trans. Commun., vol. 66, no. 9, pp. 4286–4302, Sep. 2018.
[35] Z. Yu, C. Zhai, and J. Liu, ‘‘Non-orthogonal multiple access relay-
ing with truncated ARQ,’’ IET Commun., vol. 11, no. 4, pp. 514–521,
Mar. 2017.
[36] I. Krikidis, ‘‘Distributed truncated ARQ protocol for cooperative
diversity networks,’’ IET Commun., vol. 1, no. 6, pp. 1212–1217,
Dec. 2007.
[37] Q. Liu, S. Zhou, and G. B. Giannakis, ‘‘Cross-layer combining of
adaptive modulation and coding with truncated ARQ over wireless
links,’’ IEEE Trans. Wireless Commun., vol. 3, no. 5, pp. 1746–1755,
Sep. 2004.
[38] Y. Yang, H. Ma, and S. Aïssa, ‘‘Cross-layer combining of adaptive
modulation and truncated ARQ under cognitive radio resource require-
ments,’’ IEEE Trans. Veh. Technol., vol. 61, no. 9, pp. 4020–4030,
Nov. 2012.
[39] A. Bletsas, A. Khisti, D. P. Reed, and A. Lippman, ‘‘A simple cooperative
diversity method based on network path selection,’’ IEEE J. Sel. Areas
Commun., vol. 24, no. 3, pp. 659–672, Mar. 2006.
[40] Z. Yang, Z. Ding, P. Fan, and G. K. Karagiannidis, ‘‘On the perfor-
mance of non-orthogonal multiple access systems with partial chan-
nel information,’’ IEEE Trans. Commun., vol. 64, no. 2, pp. 654–667,
Feb. 2016.
VOLUME 7, 2019
Z. Yu et al.: NOMA With Cooperative Truncated ARQ and Relay Selection
ZHIYUAN YU (S’17–M’18) received the B.Sc.
degree in electronic information science and tech-
nology and the Ph.D. degree in information and
communication engineering from Shandong Uni-
versity, China, in 2013 and 2018, respectively,
where he is currently an Assistant Professor with
the School of Journalism and Communication.
From 2012 to 2013, he was an Intern with the
Digital Multimedia Technology National Key Lab-
oratory, Hisense Co. Ltd., Qingdao, China. From
2016 to 2017, he was a Graduate Visiting Student with the Research Lab-
oratory of Electronics, Department of Electrical Engineering and Computer
Science, Massachusetts Institute of Technology, Cambridge, MA, USA. His
research interests include information inference, big data analysis, media
convergence, protocol design and analysis for cooperative non-orthogonal
multiple access, energy harvesting communications, and spectrum sharing
in cognitive radio networks.
CHAO ZHAI (S’12–M’15) received the B.Sc.
degree in communication engineering and the
M.Sc. degree in communication and information
systems from Shandong University, Jinan, China,
in 2007 and 2010, respectively, and the Ph.D.
degree in electrical engineering from the Uni-
versity of New South Wales, Sydney, Australia,
in 2013. From 2011 to 2012, he was a Visiting
Ph.D. Student with The Chinese University of
Hong Kong. In 2013, he was a Visiting Ph.D.
Student with The Hong Kong University of Science and Technology. From
2014 to 2016, he held a postdoctoral position at Shandong University. Then,
he joined the School of Information Science and Engineering, Shandong
University, as an Associate Professor. His research interests include protocol
design and performance analysis for wireless stochastic networks, spec-
trum sharing in cognitive radio networks, and energy harvesting wireless
communications.
VOLUME 7, 2019
WAN NI received the B.Sc. degree in elec-
tronic and information system, the B.A. and M.A.
degrees in journalism and communication, and the
Ph.D. degree in arts and literature from Shandong
University, Jinan, China, in 1997, 2000, 2003, and
2009, respectively, where he is currently a Full
Professor and the Vice Dean of the School of
Journalism and Communication. He was a Visit-
ing Scholar with the School of Communication,
Hong Kong Baptist University, in 2005. From
2016 to 2017, he was the Vice Director of the Center of Media Convergence
in Shandong Radio and TV Station, Jinan, China. His research interests
include big data analytic, communication systems, network convergence and
management, online public opinion, and crisis communication.
DESHENG WANG received the B.A. degree in
philosophy and the M.Sc. degree in manage-
ment from Shandong University, Jinan, China,
in 1987 and 2002, respectively, and the Ph.D.
degree in management from Tianjin University,
Tianjin, China, in 2010. He is currently a Full
Professor with the School of Journalism and Com-
munication, Shandong University, where he is also
the Director of the Center of Online Public Opin-
ion. His research interests include online public
opinion and network management.
56243