IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 37, NO.

1, JANUARY 2019 131

Hybrid Precoding-Based Millimeter-Wave Massive

MIMO-NOMA With Simultaneous Wireless
Information and Power Transfer
Linglong Dai , Senior Member, IEEE, Bichai Wang , Student Member, IEEE,
Mugen Peng , Senior Member, IEEE, and Shanzhi Chen , Senior Member, IEEE

Abstract— Non-orthogonal multiple access (NOMA) has been I. I NTRODUCTION

recently considered in millimeter-wave (mmWave) massive
MIMO systems to further enhance the spectrum efficiency.
In addition, simultaneous wireless information and power trans-
fer (SWIPT) is a promising solution to maximize the energy
M ILLIMETER-WAVE (mmWave) massive MIMO has
been considered as one of the promising techniques
for 5G wireless communications, since it can provide wider
efficiency. In this paper, for the first time, we investigate bandwidth and achieve higher spectrum efficiency [1], [2].
the integration of SWIPT in mmWave massive MIMO-NOMA
systems. As mmWave massive MIMO will likely use hybrid It is well known that in conventional MIMO systems, each
precoding (HP) to significantly reduce the number of required antenna usually requires one dedicated radio-frequency (RF)
radio-frequency (RF) chains without an obvious performance chain to realize the fully digital signal processing [3], [4].
loss, where the fully digital precoder is decomposed into a high- In this way, the use of a very large number of antennas
dimensional analog precoder and a low-dimensional digital pre- in mmWave massive MIMO systems leads to an equally
coder, we propose to apply SWIPT in HP-based MIMO-NOMA
systems, where each user can extract both information and energy large number of RF chains, which will result in unaffordable
from the received RF signals by using a power splitting receiver. hardware cost and energy consumption [5]. To address this
Specifically, the cluster-head selection algorithm is proposed to issue, hybrid precoding (HP) has been proposed to signifi-
select one user for each beam at first, and then the analog cantly reduce the number of required RF chains in mmWave
precoding is designed according to the selected cluster heads massive MIMO systems without an obvious performance
for all beams. After that, user grouping is performed based on
the correlation of users’ equivalent channels. Then, the digital loss [6]. The key idea of HP is to decompose the fully digital
precoding is designed by selecting users with the strongest precoder into a high-dimensional analog precoder (realized by
equivalent channel gain in each beam. Finally, the achievable the analog circuit) to increase the antenna array gain and a
sum rate is maximized by jointly optimizing power allocation low-dimensional digital precoder (realized by a small number
for mmWave massive MIMO-NOMA and power splitting factors of RF chains) to cancel interference [7]–[10]. Usually, two
for SWIPT, and an iterative optimization algorithm is developed
to solve the non-convex problem. Simulation results show that typical HP architectures are adopted [5]: 1) Fully-connected
the proposed HP-based MIMO-NOMA with SWIPT can achieve architecture, where each RF chain is connected to all antennas;
higher spectrum and energy efficiency compared with HP-based 2) Sub-connected architecture, where each RF chain is con-
MIMO-OMA with SWIPT. nected to only a subset of antennas. For general comparison,
Index Terms— SWIPT, mmWave, massive MIMO, NOMA, the fully-connected architecture can achieve higher spectrum
hybrid precoding, power allocation, power splitting. efficiency, while the sub-connected architecture is expected to
achieve higher energy efficiency [5].
Manuscript received March 12, 2018; revised July 4, 2018; accepted Sep-
tember 6, 2018. Date of publication October 8, 2018; date of current version To further increase the spectrum efficiency, non-orthogonal
December 14, 2018. This work was supported in part by the State Major multiple access (NOMA) has been recently considered in
Science and Technology Special Project under Grant 2017ZX03001025-006, mmWave massive MIMO systems [11]–[14]. It has been
in part by the National Natural Science Foundation of China for Outstanding
Young Scholars under Grant 61722109, in part by the National Natural Science shown that NOMA can significantly improve the spectrum
Foundation of China under Grant 61571270, and in part by the Royal Acad- efficiency compared to the conventional orthogonal multiple
emy of Engineering through the U.K.-China Industry Academia Partnership access (OMA) schemes [15]–[18]. By using NOMA, more
Programme Scheme under Grant U.K.-CIAPP\49. (Corresponding author:
Shanzhi Chen.) than one user can be supported in each beam with the aid of
L. Dai and B. Wang are with the Tsinghua National Laboratory for Informa- intra-beam superposition coding and successive interference
tion Science and Technology and the Department of Electronic Engineering, cancellation (SIC) [15], [16], which is essentially different
Tsinghua University, Beijing 100084, China (e-mail: [email protected];
[email protected]). from conventional mmWave massive MIMO using one beam
M. Peng is with the Key Laboratory of Universal Wireless Communications, to serve only one user at the same time-frequency resources.
Ministry of Education, Beijing University of Posts and Telecommunications, Particularly, NOMA was applied to beamspace MIMO for the
Beijing 100876, China (e-mail: [email protected]).
S. Chen is with the State Key Laboratory of Wireless Mobile Communi- first time in [11], which can be regarded as a low-complexity
cation, China Academy of Telecommunication Technology, Beijing 100191, realization of HP, and power allocation was optimized to max-
China (e-mail: [email protected]). imize the achievable sum rate. In addition, NOMA was also
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. utilized in fully-connected HP architecture in [12], and digital
Digital Object Identifier 10.1109/JSAC.2018.2872364 precoding was designed by modifying the conventional block
0733-8716 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Authorized licensed use limited to: Cornell University Library. Downloaded on September 02,2020 at 03:11:02 UTC from IEEE Xplore. Restrictions apply.
132 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 37, NO. 1, JANUARY 2019

diagonalization (BD) precoding scheme. Furthermore, a more and power splitting factors design. Specifically, the contribu-
sophisticated digital precoding design, i.e., minorization- tions of this paper can be summarized as follows.
maximization (MM) based precoding, was proposed in [13] 1) We propose to integrate SWIPT in HP-based mmWave
to maximize the achievable sum rate. Besides, power allo- massive MIMO-NOMA systems, including both the
cation was optimized in [14] to maximize the energy effi- fully-connected architecture and sub-connected archi-
ciency of mmWave massive MIMO-NOMA systems, and an tecture. To the best of our knowledge, it is the first
iterative algorithm was proposed to obtain the optimal power time to consider SWIPT in massive MIMO-NOMA
allocation. systems. On the one hand, HP architecture at the base
In addition to improving the spectrum efficiency, energy station (BS) can significantly reduce the number of
efficiency is also one of the key performance indicators (KPIs) required RF chains without an obvious performance
for 5G, which is expected to be 100 times compared with loss, which can largely save the energy consumption
that of current 4G wireless communications [19]. To this at the BS, while guarantee the spectrum efficiency of
end, simultaneous wireless information and power trans- mmWave massive MIMO systems. On the other hand,
fer (SWIPT), which was initially proposed in [20], has by using SWIPT, users can harvest energy from the
attracted great interests in recent years [21]–[25]. The key received RF signals to prolong their life, which makes it
idea of SWIPT is that both information and energy could more energy-efficient at the users. Note that although the
be extracted from the same received RF signals, which can application of SWIPT in conventional MIMO systems
be realized by power splitting receivers in practice [26]. With have been studied [21]–[23], [28], [29], the introduction
the help of SWIPT, the battery-powered wireless communi- of NOMA will result in additional challenges for the
cation devices can harvest energy from the RF signals to joint transceiver and power splitting optimization.
prolong their lifetime, which provides the potential to explore 2) To enable the HP-based mmWave massive MIMO-
more energy-efficient networks, especially for Internet of NOMA systems with SWIPT, we investigate the joint
Things (IoT) with millions of wireless devices [27]. However, transceiver and power splitting optimization. Specif-
the trade-off between information rate and harvested energy ically, the cluster-head selection (CHS) algorithm is
level should be carefully considered to facilitate efficient proposed to select one user for each beam, and then
SWIPT in multi-user systems, since inter-user interferences the analog precoding is designed to obtain the antenna
are usually harmful for the information decoding (ID), while array gain according to the selected cluster heads for
they can be useful for energy harvesting (EH) [21]. all beams. After that, user grouping is performed based
In fact, some efforts have been endeavored to address on the equivalent channel correlation of the remaining
this problem. Particularly, a joint transmit beamforming users and the cluster-heads. Then, the digital precoding
and power splitting optimization was investigated in [28], is designed to cancel the inter-user interference by
where the transmit power was minimized under the signal- selecting users with the strongest equivalent channel
to-interference-and-noise ratio (SINR) and EH quality of gain in each beam. Finally, the achievable sum rate
service (QoS) constraints for multi-user MIMO systems. is maximized by jointly optimizing power allocation
In addition, the joint transceiver and power splitting design as well as power splitting factors, which is very dif-
for downlink multi-user MIMO SWIPT networks was also ficult to obtain the optimal solutions due to the cou-
investigated based on the mean square error (MSE) criterion pling of different users’ power allocation factors as
in [29]. For multi-cell multi-user downlink SWIPT systems, well as the power splitting factors. To address this
the joint transceiver and power splitting design was studied to issue, an iterative optimization algorithm is developed
optimize the energy efficiency in [30]. Although SWIPT has to solve the non-convex problem, and the convergence
the potential to realize energy efficient wireless communica- and computational complexity are also analyzed.
tions, and has been investigated in some multi-user systems, 3) The performance in terms of both spectrum effi-
the application of SWIPT in mmWave massive MIMO-NOMA ciency and energy efficiency of the proposed HP-based
systems has not been considered in the literature to the best of mmWave massive MIMO-NOMA systems with SWIPT
our knowledge, where new challenges will arise for the joint is evaluated by simulations. The convergence of the
transceiver and power splitting optimization. developed iterative optimization algorithm for joint
In this paper, we propose to integrate SWIPT in HP-based power allocation and power splitting is validated, and it
mmWave massive MIMO-NOMA systems to realize the is shown that only 10 times of iteration are required to
spectrum- and energy-efficient wireless communications.1 Par- make it converged. Furthermore, we show that the pro-
ticularly, a power splitting receiver is used for each user to posed mmWave massive MIMO-NOMA systems with
achieve SWIPT by dividing the received signal into two parts SWIPT can achieve higher spectrum and energy effi-
for simultaneous information retrieval and energy storage. ciency than those of mmWave massive MIMO-OMA
In such an system, we investigate the joint optimization of systems with SWIPT.
transceiver for ID and power splitting for EH, including The rest of this paper is organized as follows. The system
user grouping, hybrid precoding design, power allocation, model of HP-based mmWave massive MIMO-NOMA sys-
tem with SWIPT is introduced in Section II. User grouping
1 Simulation codes are provided to reproduce the results presented in this and hybrid precoding design are discussed in Section III.
paper: http://oa.ee.tsinghua.edu.cn/dailinglong/publications/publications.html. In Section IV, the joint power allocation and power splitting

Authorized licensed use limited to: Cornell University Library. Downloaded on September 02,2020 at 03:11:02 UTC from IEEE Xplore. Restrictions apply.
DAI et al.: HYBRID PRECODING-BASED mmWAVE MASSIVE MIMO-NOMA WITH SWIPT 133

optimization problem is formulated to maximize the achiev-

able sum rate, under the achievable rate QoS and EH QoS con-
straints for each user. Furthermore, an iterative optimization
algorithm is proposed to solve the non-convex problem. Sim-
ulation results are provided in Section V. Finally, conclusions
are drawn in Section VI.
Notation: We use upper-case and lower-case boldface letters
to denote matrices and vectors, respectively; (·)T , (·)H , (·)−1 ,
and  · p denote the transpose, conjugate transpose, matrix
inversion, and lp norm operation, respectively. E {·} denotes
the expectation. |Γ| denotes the number of elements in set Γ.
A(i, :)i∈Γ denotes the submatrix of A that consists of the
ith row of A for all i ∈ Γ, while A(:, j)j∈Γ denotes the
submatrix of A that consists of the jth column of A for
all j ∈ Γ. We use the notation CN (m, R) to denote the
complex Gaussian distribution with mean m and covari-
ance R, and U (a, b) to denote the uniform distribution in the
range (a, b). ⊗ denotes the kronecker product. Finally, IN is
the N × N identity matrix, and Φ denotes the empty set.

II. S YSTEM M ODEL

In this paper, we consider a single-cell downlink mmWave
massive MIMO-NOMA system, where the BS is equipped
with N antennas and NRF RF chains to simultaneously serve
K single-antenna users [12]–[14], and each user is equipped
with a power splitting receiver for SWIPT.
Fig. 1 shows three architectures of mmWave massive MIMO
systems, i.e., the fully digital MIMO as shown in Fig. 1 (a),
the fully-connected HP architecture as shown in Fig. 1 (b), and
the sub-connected HP architecture as shown in Fig. 1 (c). From
Fig. 1, we can see that for the fully digital MIMO, each
antenna requires one dedicated RF chain, and thus the number
of RF chains is equal to the number of antennas, which results
in unaffordable energy consumption and hardware cost. On the
Fig. 1. System models of mmWave MIMO architectures: (a) Fully
contrary, the number of RF chains in HP architectures is digital MIMO; (b) Fully-connected HP architecture; (c) Sub-connected
less than the number of antennas, which can be realized by HP architecture.
a high-dimensional analog precoder and a low-dimensional

G
digital precoder. Specifically, for the fully-connected HP archi- and we have Si ∩ Sj = Φ for i = j as well as |Sg | = K.
tecture in Fig. 1 (b), each of the NRF RF chain is connected g=1
Then, the received signal at the mth user in the gth beam can
to all N antennas by finite-resolution phase shifters, where
be modeled as
N NRF phase shifters are required, and thus the full array gain
|Si |
can be exploited by every RF chain. For the sub-connected 
G 
√
HP architecture in Fig. 1 (c), each RF chain is connected to yg,m = hH
g,m A di pi,j si,j + vg,m
only a subset of N BS antennas, so only N phase shifters i=1 j=1
√
are required. In general, the sub-connected architecture is = hHg,m Adg pg,m sg,m
easier to be implemented and will likely be more energy   
desired signal
efficient, while it may suffer some performance loss compared ⎛ ⎞
|Sg |
to the fully-connected architecture. In this paper, both of 
m−1
√  √
+ hH
g,m Adg
⎝ pg,j sg,j + pg,j sg,j ⎠
the fully-connected and sub-connected architectures will be
j=1 j=m+1
considered.   
In HP-based mmWave massive MIMO systems, the number intra−beam interferences
of beams cannot exceed the number of RF chains, and each |Si |
 √
beam can only support one user at most [7]. Therefore, to fully + hH
g,m A di pi,j si,j + vg,m , (1)
achieve the multiplexing gain, we assume that the number 
i=g j=1
of beams G is equal to the number of RF chains NRF ,    noise
inter−beam interferences
i.e., G = NRF . On the contrary, each beam can support more
2
than one user by using NOMA. Let Sg for g = 1, 2, · · · , G where sg,m is the transmitted signal with E{|sg,m | } = 1,
denote the set of users served by the gth beam with |Sg | ≥ 1, pg,m is the transmitted power for the mth user in the gth beam,

Authorized licensed use limited to: Cornell University Library. Downloaded on September 02,2020 at 03:11:02 UTC from IEEE Xplore. Restrictions apply.
134 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 37, NO. 1, JANUARY 2019

vg,m is the noise following the distribution CN 0, σv2 , where h̄H H

g,m = hg,m A is the equivalent channel vector, and
dg is the NRF × 1 digital precoding vector for the gth beam, 0 ≤ η ≤ 1 is the energy conversion efficiency. In the
A of size N × NRF is the analog precoding matrix, and meanwhile, the signal for ID at the mth user in the gth beam
we have Adg 2 = 1 for g = 1, 2, · · · , G. Particularly, can be expressed as
for the fully-connected architecture, the analog precoding
ID

matrix A(full) can be expressed as yg,m = βg,m yg,m + ug,m , (8)
 
(full) (full) (full)
A(full) = ā1 , ā2 , · · · , āNRF , (2) where ug,m is the noise caused by the power splitter following
the distribution CN (0, σu2 ).
N ×1
(full)
where the elements of ān ∈√ for n = 1, 2, · · · , NRF By using NOMA in each beam, intra-beam superposition
have the same amplitude 1/ N but different phases [6]. coding at the transmitter and SIC at the receiver are performed.
For the sub-connected architecture, the analog precoding Without loss of generality, we assume that h̄H g,1 dg 2 ≥
matrix A(sub) is h̄H d 
g,2 g 2 ≥ · · · ≥  h̄H
d 
g,|Sg | g 2 for g = 1, 2, · · · , G. Then,
⎡ (sub) ⎤ the mth user in the gth beam can remove the interference from
ā1 0 ··· 0
⎢ (sub) ⎥ the jth user (for all j > m) in the gth beam by performing
⎢ 0 ā2 0 ⎥
A(sub) = ⎢⎢ .. .. .. ⎥
⎥. (3) SIC [15], and the remaining received signal for ID at the mth
⎣ . . . ⎦ user in the gth beam can be rewritten as
0 0 ···
(sub)
āNRF ⎛
 √ √
m−1

Without loss of generality, we assume that M = N /NRF

ID
ŷg,m = βg,m ⎝h̄H H
g,m dg pg,m sg,m + h̄g,m dg pg,j sg,j
is an integer, and each RF chain is connected to M anten- j=1
⎞
nas in the sub-connected architecture. Then, the elements of |Si |
 √
di pi,j si,j + vg,m ⎠ + ug,m .
(sub)
ān√ ∈ M×1 for n = 1, 2, · · · , NRF have the same amplitude + h̄H
g,m
1/ M [7], [8]. i=g j=1
For the N × 1 channel vector hg,m of the mth user in (9)
the gth beam, we consider the widely used mmWave MIMO
channel model as shown below [6]–[8]: Then, according to (9), the SINR at the mth user in the
 gth beam can be written as
N  (l)  (l) (l) 
Lg,m
hg,m = αg,m a ϕg,m , θg,m , (4)  H 
Lg,m h̄ dg 2 pg,m
l=1 g,m 2
γg,m = , (10)
where Lg,m denotes the number of paths for the mth user ξg,m
(l)
in the gth beam. αg,m is the complex gain of the lth path. where
(l) (l)
ϕg,m and θg,m are the azimuth angle of departure (AoD) and
|Si |
(l) (l)
elevation AoD of the lth path, and a(ϕg,m , θg,m ) presents  H 
2 m−1  2 
ξg,m = h̄ 
g,m dg 2 pg,j + h̄H 
g,m di 2 pi,j
the N × 1 array steering vector. Particularly, for the typical
j=1 i=g j=1
uniform linear array (ULA) with N1 elements in horizon and
N2 elements in vertical, where N = N1 N2 [7], we have σu2
+ σv2 + . (11)
βg,m
a (ϕ, θ) = aaz (ϕ) ⊗ ael (θ) , (5)
  As a result, the achievable rate at the mth user in the
where aaz (ϕ) = √1N ej2πi(d1 /λ) sin(ϕ) i∈J(N1 ) , ael (θ) = gth beam is
 j2πj(d /λ) sin(θ) 1
√1
N2
e 2
j∈J(N2 )
, J (n) = {0, 1, · · · , n − 1}, Rg,m = log2 (1 + γg,m ) . (12)
λ is the signal wavelength, d1 is the horizontal antenna
spacing, and d2 is the vertical antenna spacing. In mmWave Finally, the achievable sum rate is
communications, we usually have d1 = d2 = λ/2 [7].
|Sg |
With the aid of power splitting receiver, the received signal 
G 
at each user will be divided into two parts. One part is Rsum = Rg,m , (13)
forwarded to the information decoder for ID, and the other g=1 m=1
is processed for EH [29]. Let βg,m , where 0 < βg,m < 1,
denote the power splitting factor for the mth user in the gth which can be improved by carefully designing user group-
G
beam, then the signal for EH can be represented as ing, analog precoding matrix A, digital precoding {dg }g=1 ,
G,|S |
 power allocation {pg,m }g=1,m=1
g
, and power splitting factors
EH
yg,m = 1 − βg,m yg,m , (6) G,|Sg |
{βg,m }g=1,m=1 . Since it is very difficult to simultaneously
and the harvested energy is [29] obtain the optimal solutions for all these design parameters,
⎛ ⎞ we consider to design user grouping and hybrid precoding at
|Si |
G 
 H 2 first as described in the next section. Then, joint optimization
Pg,m = η (1−βg,m)⎝
EH h̄g,m di  pi,j + σv ⎠,
2
(7)
2 of power allocation and power splitting will be introduced in
i=1 j=1
Section IV.

Authorized licensed use limited to: Cornell University Library. Downloaded on September 02,2020 at 03:11:02 UTC from IEEE Xplore. Restrictions apply.
DAI et al.: HYBRID PRECODING-BASED mmWAVE MASSIVE MIMO-NOMA WITH SWIPT 135

III. U SER G ROUPING AND H YBRID P RECODING Algorithm 1 Proposed CHS Algorithm
We know that the number of users K is larger than the Input:
number of RF chains NRF in the considered system, while The number of users K, and the number of beams G;
only NRF different analog precoding vectors are available Channel vectors: hk for k = 1, 2, · · · , K;
at the same time [8]. Therefore, to enable hybrid precoding, The initial threshold: δ.
we propose the CHS algorithm to select one user for each Output:
beam (there are G = NRF beams), and then the analog The cluster head set Γ.
precoding is designed to obtain the antenna array gain accord- 1: Λ = [a1 , a2 , · · · , aK ], where ak = hk 2 ;
ing to the selected cluster heads for all beams. After that, 2: h̃k = hk /ak for k = 1, 2, · · · , K;
user grouping is performed based on the equivalent channel 3: [∼,O] = sort (Λ, descend );
correlation between the remaining users and the cluster-heads. 4: Γ = O (1);
Then, the digital precoding is designed to cancel inter-user 5: Γc = O/Γ;
interference by selecting users with the strongest equivalent 6: Ω = Γc ;
channel gain in each beam. 7: g = 2.
8: while g ≤ G do
9: if Ω == Φ then
A. The Proposed CHS Algorithm
10: while Ω == Φ do
To improve the system performance, we propose to select
11: δ = δ+ (1 − δ)/10;  
the cluster head for each beam by minimizing the channel  
correlation of the selected cluster heads. In this way, users in 12: Ω = i ∈ Γc h̃H i h̃j  < δ, ∀j ∈ Γ .
different beams will enjoy low channel correlation, which is 13: end while
beneficial for inter-beam interference cancellation. 14: end if    
 
In the proposed CHS algorithm, an adaptive threshold δ is 15: Ω = i ∈ Ω h̃H i h̃j  < δ, ∀j ∈ Γ ;
introduced to measure the channel correlation of the cluster 16: Γ = Γ ∪ Ω (1);
heads. Specifically, the user with the highest channel gain 17: Γc = O/Γ;
is selected as the cluster head for the first beam, and then 18: g = g + 1.
users whose channel correlation with the first selected user 19: end while
is less than the threshold δ will be considered as the cluster 20: return Γ.
head candidates for other beams. Particularly, the user with
the highest channel gain out of the cluster head candidates is precoding matrix A(full) belong to
selected as the cluster head for the second beam. After that, 1  j 2πn 
√ e 2B : n = 0, 1, · · · , 2B − 1 , (14)
the cluster head candidates will be updated by selecting users N
whose channel correlation with the second selected user is less while the non-zero elements of the sub-connected analog
than the threshold. This procedure is repeated until there is no precoding matrix A(sub) belong to
1  j 2πn 
candidate. Next, the threshold is updated by adding a small
increment, and then the cluster head candidates are obtained by √ e 2B : n = 0, 1, · · · , 2B − 1 . (15)
selecting users whose channel correlations with the previously M
selected cluster heads are less than the threshold. The threshold Based on the cluster head set Γ obtained in the previous
will be adaptively updated until the cluster heads are selected subsection, the analog precoding can be designed according to
for all G beams. The details of the proposed CHS algorithm the channel vectors of users in Γ. More particularly, the analog
are described in Algorithm 1, and the set of the selected precoding vectors can be obtained by maximizing the array
(full) 2
cluster heads is denoted as Γ. gains |hH
Γ(g) āg | for the fully-connected architecture and
The proposed CHS algorithm enjoys the polynomial com- |hH
(sub)
|2 for the sub-connected architecture, separately,
Γ(g) āg
plexity. Specifically, in each iteration, the maximum complex- where g = 1, 2, · · · , G. As a result, the ith element, where i =
ity is (2 + 2(K − 1))(K − 1) from step 9 to step 14, while the 1, 2, · · · , N , of the fully-connected analog precoding vector
maximum complexity is 2(K − 1) from step 15 to step 18. (full)
āg can be expressed as
Therefore, the complexity of Algorithm 1 is O(GK 2 ).
1 2π n̂
ā(full)
g (i) = √ ej 2B , (16)
B. Analog Precoding N
where
In this paper, we consider the typical two-stage HP pro-  
 
posed in [9].2 The key idea of this scheme is to divide the n̂ = arg min angle hΓ(g) (i) − 2πn  . (17)
HP design into two step, i.e., analog precoding and digital  2 
B
n∈{0,1,··· ,2B −1}
precoding. Particularly, for analog precoding, only quantized
Similarly, the ith element of the sub-connected analog pre-
phase changes can be applied due to the practical constraints (sub)
coding vector āg , where i = (g − 1)M + 1, (g − 1)M +
of phase shifters [10]. Considering B bits quantized phase
2, · · · , gM , is
shifters, the non-zero elements of the fully-connected analog 1 2π n̂
ā(sub)
g (i) = √ ej 2B , (18)
2 Note that more sophisticated HP schemes can be considered to further M
enhance the performance of mmWave massive MIMO-NOMA systems. where n̂ is the same as that in (17).

Authorized licensed use limited to: Cornell University Library. Downloaded on September 02,2020 at 03:11:02 UTC from IEEE Xplore. Restrictions apply.
136 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 37, NO. 1, JANUARY 2019

C. User Grouping massive MIMO-NOMA systems with SWIPT, since there

After obtaining the analog precoding, the equivalent channel exists not only the coupling of power allocation factors from
vectors for all K users can be written as different users, but also the coupling of power allocation and
power splitting factors. On the other hand, the existing opti-
h̄H H
k = hk A, (19) mization methods for solving the joint optimization problem
where k = 1, 2, · · · , K. Then, user grouping can be realized of power allocation and power splitting in MIMO systems with
according to the correlation of equivalent channels. Specifi- SWIPT cannot be directly used in MIMO-NOMA systems
cally, user m (m ∈/ Γ) can be classed as the ĝth beam, where with SWIPT, where there are multiple groups and multiple
 H  users in each group, since both inter-group and intra-group
h̄m h̄Γ(g) 
ĝ = arg max    
h̄m  h̄Γ(g)  . (20) interferences exist. Therefore, it is very difficult to obtain the
g∈{1,2,··· ,G} 2 2 optimal solutions. To solve this intractable problem, an iter-
In this way, users in the same beam will enjoy high correlation ative optimization algorithm is developed in this Section to
of equivalent channels, while the equivalent channels of users obtain the sub-optimal solutions. Specifically, the joint power
in different beams have low correlation owing to the proposed allocation and power splitting optimization problem can be
CHS algorithm, which is conducive to the inter-beam interfer- formulated as
|Sg |
ence cancellation and thus the improvement of multiplexing 
G 

gains. max Rg,m

{pg,m },{βg,m }
g=1 m=1

D. Digital Precoding s.t. C1 : pg,m ≥ 0, ∀g, m,

|Sg |
After analog precoding and user grouping, the equivalent 
G 

channel vector for the mth user in the gth beam can be denoted C2 : pg,m ≤ Pt ,
as h̄H
g,m as introduced in Section II. Then, the design of digital
g=1 m=1

precoding actually becomes a conventional MIMO-NOMA C3 : Rg,m ≥ Rg,m

min
, ∀g, m,
precoding problem to eliminate inter-beam interference. With- EH
C4 : Pg,m ≥ Pg,m
min
, ∀g, m, (24)
out loss of generality, the low-complexity zero-forcing (ZF)
precoding is adopted for digital precoding, according to where Rg,m is the achievable rate of the mthe user in the
the equivalent channel vectors of users having the highest gth beam as defined in (12), the constraint C1 indicates that
equivalent channel gain in each beam [11], [12]. the power allocated to each user must be positive, C2 is the
Specifically, assuming that the mg th user has the highest transmitted power constraint with Pt being the maximum total
equivalent channel gain in the gth beam, we have transmitted power by the BS, C3 is the data rate constraint for
  each user with Rg,mmin
being the minimum data rate for the
H̄ = h̄m1 , h̄m2 , · · · , h̄mG . (21) mth user in the gth beam, and C4 is the EH QoS constraint
min
Then, the digital precoding matrix of size NRF × NRF can be for each user with Pg,m being the minimum harvested energy
generated by for the mthe user in the gth beam. Note that the optimization
  −1
problem (24) is non-convex due to the non-convexity of the
D̄ = d̄1 , d̄2 , · · · , d̄G = H̄ H̄H H̄ (22) objective function, and the constraints C3 as well as C4 .
To solve the non-convex problem (24), an iterative opti-
After normalizing, the digital precoding vector for the
mization algorithm is developed. Particularly, according to the
gth beam can be written as
extension of the Sherman-Morrison-Woodbury formula [37],
d̄g i.e.,
dg =  
Ad̄g  . (23)
−1 −1
2 (A + BCD) = A−1 −A−1 B I+CDA−1 B CDA−1 ,
Afterwards, the users in each beam will be reordered such (25)
that h̄H g,1 dg 2 ≥ h̄g,2 dg 2 ≥ · · · ≥ h̄g,|Sg | dg 2 for g =
H H

1, 2, · · · , G, which is assumed in Section II for SIC. we have

Up to now, user grouping and hybrid precoding have been (1 + γg,m )−1
carefully designed to obtain antenna array gains and multi-  2   H  −1
plexing gains. In the next Section, the joint optimization of = 1 − pg,m h̄H 
g,m dg 2 pg,m
h̄g,m dg 2 + ξg,m
2
, (26)
power allocation and power splitting will be investigated to
maximize the achievable sum rate in (13). where g = 1, 2, · · · , G and m = 1, 2, · · · , |Sg |.
On the other hand, let
IV. J OINT O PTIMIZATION OF P OWER A LLOCATION √
m−1
AND P OWER S PLITTING √
ỹg,m = h̄H d
g,m g p s
g,m g,m + h̄H
d
g,m g pg,j sg,j
Although power allocation has been studied in existing j=1
MIMO-NOMA systems [17], [34]–[36], the joint optimization |Si |
 √ 1
of power allocation and power splitting has not been consid- + h̄H
g,m di pi,j si,j + vg,m +  ug,m .
ered. The introduction of power splitting factors will result in i=g j=1
β g,m

additional challenges for the joint optimization in mmWave (27)

Authorized licensed use limited to: Cornell University Library. Downloaded on September 02,2020 at 03:11:02 UTC from IEEE Xplore. Restrictions apply.
DAI et al.: HYBRID PRECODING-BASED mmWAVE MASSIVE MIMO-NOMA WITH SWIPT 137

(t)
If the minimum mean square error (MMSE) detection is used of {cg,m } in the tth iteration can be obtained according to (31),
to solve sg,m from ỹg,m in (27), this detection problem can i.e.,
be formulated as c(t)
g,m
" !∗  −1
cog,m = arg min eg,m ,  H 
cg,m
(28) =
(t−1) H
pg,m h̄g,m dg p(t−1) h̄ dg 2 + ξ (t−1) ,
g,m g,m 2 g,m

where   (38)
2
eg,m = E |sg,m − cg,m ỹg,m | (29) where
  (t−1)
2 m−1
is the mean square error (MSE), and cg,m is the channel ξg,m (t−1) = h̄H 
g,m dg 2 pg,j
equalization coefficient. Substituting (27) into (29), we have j=1
√  |Si |
eg,m = 1 − 2Re cg,m pg,m h̄H dg 2 
 g,m 2  + h̄H d 
g,m i 2
(t−1)
pi,j
+ |cg,m | pg,m h̄H 
2
g,m dg 2 + ξg,m . (30) i=g j=1
σu2
Then, by solving the partial derivatives of (28) based on (30), + σv2 + (t−1)
. (39)
the optimal equalization coefficient cog,m can be obtained by βg,m

√ ∗
  H 2 −1 In the meanwhile, the optimal solution of {ag,m } in the tth
(t)
cog,m = pg,m h̄H d p h̄ d  + ξ . (31)
g,m g g,m g,m g 2 g,m iteration can be calculated by
1
Substituting (31) into (30), the MMSE can be written as a(t)
g,m = o(t) , (40)
eg,m
 2   H  −1
eog,m = 1 − pg,m h̄H 
g,m dg 2 pg,m
h̄ dg 2 + ξg,m
g,m 2
, where
(32) eo(t)
g,m
 H 2  (t−1)  H 2 −1
−1 = 1 − p(t−1) h̄ d  p h̄ d  + ξ (t−1)
.
which is equal to (1 + γg,m ) in (26), i.e., we have g,m g,m g 2 g,m g,m g 2 g,m

−1
(41)
(1 + γg,m ) = min eg,m . (33)
cg,m Then, the optimization problem (37) can be simplified as
|Sg |
Then, the achievable rate of the mth user in the gth beam can 
G 

be rewritten as  
min  a(t) (t)
g,m eg,m
(t) (t)
pg,m , βg,m g=1 m=1
Rg,m = log2 (1 + γg,m ) = max (−log2 eg,m ) . (34) (t)
cg,m s.t. C1 (t)
: pg,m ≥ 0, ∀g, m,
|Sg |
Note that in (34), the polynomial division has been removed (t)

G 

by using eg,m rather than γg,m , which significantly simplifies C2 : g,m ≤ Pt ,

p(t)
the objective function. Furthermore, to remove the log function g=1 m=1
(t)
in (34), Proposition 1 is introduced [11], [34]. C3 (t)
: Rg,m ≥ Rg,m
min
, ∀g, m,
Proposition 1: Let f (a) = − lnab2 + log2 a + ln12 and a be a (t)
C4 EH(t)
: Pg,m ≥ Pg,m
min
, ∀g, m, (42)
positive real number, we have
where " !
max f (a) = −log2 b, (35) (t)
a>0 e(t)
g,m = 1 − 2Re c(t)
g,m
H
pg,m h̄g,m dg
where the optimal value of a is a = o 1  2   H  
b.   h̄g,m dg 2 + ξg,m
According to Proposition 1, we can rewrite (34) as g,m 
+ c(t) p(t)
g,m 2
(t)
. (43)
! (t)
ag,m eg,m 1 Furthermore, we introduce the variables {τg,m } such that
Rg,m = max max − + log2 ag,m + . (36) (t)
cg,m ag,m >0 ln 2 ln 2 τg,m ≥ (t)
1
, and rewrite the objective function in (42) as
βg,m
As a result, the optimization problem (24) can be reformu- |Sg |

G 
lated as  
min  a(t) (t)
g,m ẽg,m , (44)
(t) (t)
pg,m , βg,m
 |Sg |
G   a e  g=1 m=1
g,m g,m
max max max − + log2 ag,m where
{pg,m },{βg,m }
g=1 m=1
cg,m ag,m >0 ln 2  "
(t)
ẽ(t)
g,m = 1 − 2Re c(t) H
g,m h̄g,m dg pg,m
s.t. C1 , C2 , C3 , C4 . (37) ⎛
  2 
 (t) 2 ⎝
m
To solve (37), the iterative optimization algorithm is intro- +cg,m  h̄H 
g,m dg 2
(t)
pg,j
duced to optimize {cg,m }, {ag,m }, and {pg,m } as well j=1
⎞
as {βg,m }, separately. Specifically, given the optimal power |Si |
   (t)
(t−1)
allocation solution {pg,m } and the power splitting solution + h̄H
g,m di
2
p + σu2 τg,m
(t)
+ σv2 ⎠ , (45)
(t−1) 2 i,j
{βg,m } in the (t − 1)th iteration, the optimal solution i=g j=1

Authorized licensed use limited to: Cornell University Library. Downloaded on September 02,2020 at 03:11:02 UTC from IEEE Xplore. Restrictions apply.
138 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 37, NO. 1, JANUARY 2019

with the additional constraint function in (37) [34]. As a result, the proposed iterative
(t) 1 optimization algorithm for joint power allocation and power
(t)
C5 : τg,m ≥ (t)
, ∀g, m. (46) splitting will converge to at least a local optimal solution.
βg,m
In the meanwhile, the proposed joint optimization algorithm
Intuitively, the objective function (44) becomes convex for enjoys a polynomial complexity. Specifically, in each iteration,
(t) (t)
the optimization variables {pg,m } and {τg,m }. At the same the complexity for obtaining the optimal {cm,n } in (38) and
(t)
time, the constraint C3 in (42) can be also transformed into {am,n } in (40) is linear to the number of users, i.e., O(K). The
a convex constraint as convex optimization problem (52) can be solved with a worst-
 2 (t)  H  (t)
 m−1 case complexity of O(Tmax K 4.5 log2 (1/ε)) given a solution
: h̄H  h̄g,m dg 2
(t)
C̃3 g,m dg 2 pg,m − ωg,m 2
pg,j accuracy ε > 0 [39]. Therefore, the computational complexity
j=1 of the proposed iterative optimization algorithm is at most
   (t) |Si | O(Tmax K 4.5 log2 (1/ε)).
− ωg,m h̄H di 2 pi,j − ωg,m σu2 τg,m
(t)
g,m 2
i=g j=1
V. S IMULATION R ESULTS
≥ ωg,m σv2 , (47)
The performance in terms of spectrum efficiency and energy
min
where ωg,m = 2Rg,m − 1. Nevertheless, the constraints efficiency of the mmWave massive MIMO-NOMA systems
(t) (t)
C4 and C5 in (42) are still non-convex due to the with SWIPT, including both the fully-connected HP and
multi-variable coupling. To make it solvable, another variables the sub-connected HP, proposed in this paper is evaluated
(t)
{μg,m } are introduced such that via simulations. Specifically, the simulation parameters are
described as follows: The system bandwidth is assumed to
min
Pg,m be 1 Hz, which coincides to the achievable rate in (12). The
(t)
g,m ≥
C6 : μ(t)   , ∀g, m. (48)
η 1 − βg,m
(t) BS is equipped with an ULA of N = 64 antennas and NRF =
4 RF chains to simultaneously serve K ≥ NRF users. All the
Then, the constraint C4
(t)
in (42) can be rewritten as K uses are grouped into G = NRF = 4 beams, and there are
more than one user in each beam. For the mth user in the
|Si |

G 
 H  gth beam, the channel vector is generated based on (4), where
(t)
C̃4 : h̄ di 2 p(t) + σ 2 ≥ μ(t) , ∀g, m, (49)
g,m 2 i,j v g,m we assume: 1) Lg,m = 3, including one line-of-sight (LoS)
i=1 j=1
component and two non-line-of-sight (NLoS) components;
2) αg,m ∼ CN (0, 1), and αg,m ∼ CN (0, 10−1 ) for 2≤
(1) (l)
which becomes a convex constraint.
(t) (l) (l)
To deal with the non-convex constraints C5 in (46) and l ≤ Lg,m ; 3) ϕg,m and θg,m follow the uniform distribution
(t)
C6 in (48), we transform them into matrix form according U(−π, π) for 1 ≤ l ≤ Lg,m . B = 4 bits quantized phase
to the Schur complement lemma [29], i.e., shifters are adopted,
% and the signal-to-noise ratio (SNR) is
# $ defined as Pt σv2 [8]. The maximum transmitted power Pt =
(t)
(t) τg,m 1
C̃5 : (t) ≥ 0, ∀g, m, (50) 30 mW, the minimal achievable rate for each user is Rf m /10,
1 βg,m where Rf m is the minimal achievable rate among all users by
and using fully digital ZF precoding, and the minimal harvested
⎡ " % ⎤ energy for each user is 0.1 mW.
(t)
min η
Pg,m
μg,m In this paper, the spectrum efficiency is defined as the
: ⎣" ⎦ ≥ 0, ∀g, m.
(t)
C̃6 % (51)
achievable sum rate in (13), and the energy efficiency is
min η 1 − β (t)
Pg,m g,m
defined as the ratio between the achievable sum rate and the
As a result, the optimization problem (42) can be reformu- total power consumption [7], i.e.,
lated as
Rsum
|Sg | EE = (bps/Hz/W),

G 
Ptr + NRF PRF + NPS PPS + PBB
 
min  a(t) (t)
g,m ẽg,m (53)
(t) (t)
pg,m , βg,m g=1 m=1
(t) (t) (t) (t) (t) (t)  g |
G |S
s.t. C1 , C2 , C̃3 , C̃4 , C̃5 , C̃6 , (52) where Ptr = pg,m is the total transmitted power,
g=1 m=1
which is a standard convex optimization problem, and can be PRF is the power consumed by each RF chain, PPS is
solved by numerical convex program solvers [38]. the power consumption of each phase shifter, and PBB is
By iteratively solving the optimal values of {cg,m }, {ag,m }, the baseband power consumption. Particularly, we adopt the
and {pg,m } as well as {βg,m } via (38), (40), and (52), typical values PRF = 300 mW, PPS = 40 mW (4-bit phase
separately, we can obtain the final power allocation solution shifter), and PBB = 200 mW [8]. NPS is the number of phase
{pog,m } and power splitting solution {βg,m
o
} with the max- shifters, which is equal to N NRF for the fully-connected HP
imum iteration times Tmax . Particularly, since the obtained and N for the sub-connected HP.
(t) (t) (t) (t)
{cm,n }, {am,n }, and {pm,n } as well as {βm,n } are opti- In the simulations, we consider the following five typical
mal solutions in the tth iteration, iteratively updating these mmWave massive MIMO systems with SWIPT for compar-
variables will increase or maintain the value of the objective ison: (1) “SWIPT-Fully digital ZF Precoding”, where each

Authorized licensed use limited to: Cornell University Library. Downloaded on September 02,2020 at 03:11:02 UTC from IEEE Xplore. Restrictions apply.
DAI et al.: HYBRID PRECODING-BASED mmWAVE MASSIVE MIMO-NOMA WITH SWIPT 139

Fig. 2. Spectrum efficiency of fully-connected architecture against the Fig. 4. Spectrum efficiency against SNR.
number of iterations for the joint power allocation and power splitting
optimization.

Fig. 5. Energy efficiency against SNR.

Fig. 3. Spectrum efficiency of sub-connected architecture against the number
of iterations for the joint power allocation and power splitting optimization. Fig. 4 shows the spectrum efficiency against SNR of
the considered five schemes, where the number of users is
antenna is connected to one RF chain, and ZF precod- K = 6. We can find that the proposed mmWave massive
ing is adopted; (2) “SWIPT-Fully-Connected HP-NOMA”, MIMO-NOMA systems with SWIPT can achieve higher spec-
where fully-connected HP architecture is used in the pro- trum efficiency than that of mmWave massive MIMO-OMA
posed mmWave massive MIMO-NOMA systems with SWIPT; systems with SWIPT, either for the fully-connected HP or the
(3) “SWIPT-Fully-Connected HP-OMA”, where the system sub-connected HP, since NOMA can achieve higher spectrum
model is similar with “SWIPT-Fully-Connected HP-NOMA”, efficiency than that of OMA [15]. It is intuitive that the fully
while OMA is performed for users in each beam. Particularly, digital MIMO can achieve the best spectrum efficiency as
we represent OMA with FDMA, where users in the same shown in Fig. 4, since N RF chains are used to serve all
beam are allocated with equal bandwidth; (4) “SWIPT-Sub- users to fully exploit the multiplexing gains. On the other hand,
Connected HP-NOMA”, where sub-connected HP architecture the fully-connected HP can achieve higher spectrum efficiency
is used in the proposed mmWave massive MIMO-NOMA than that of the sub-connected HP as discussed in Section II,
systems with SWIPT; (5) “SWIPT-Sub-Connected HP-OMA”, since the full array gain can be exploited by every RF chain
where the system model is similar with “SWIPT-Sub- in the fully-connected HP.
Connected HP-NOMA”, while OMA is performed for users Fig. 5 shows the energy efficiency against SNR, where the
in each beam. number of users is also K = 6. We can find that the proposed
Fig. 2 and Fig. 3 show the convergence of the proposed mmWave massive MIMO-NOMA systems with SWIPT can
iterative algorithm for joint power allocation and power split- achieve higher energy efficiency than both mmWave massive
ting optimization in Section IV for the fully-connected HP and MIMO-OMA systems with SWIPT and fully digital MIMO
sub-connected HP, separately, where the number of users is set systems with SWIPT. Particularly, the number of RF chains
as K = 6, and SNR = 0 dB. As shown in Fig. 2 and Fig. 3, is equal to the number of BS antennas in fully digital
the spectrum efficiency tends to be stable after 10 times MIMO systems, which leads to very high energy consumption,
of iteration, which verifies the convergence of the proposed e.g., 300 mW for each RF chain. On the contrary, the number
iterative algorithm as discussed in Section IV. In the following of RF chains is much smaller than the number of antennas
simulations, the number of iteration times for the proposed in the proposed mmWave massive MIMO-NOMA systems
iterative optimization algorithm is set as 10. with SWIPT. Therefore, the energy consumption caused by

Authorized licensed use limited to: Cornell University Library. Downloaded on September 02,2020 at 03:11:02 UTC from IEEE Xplore. Restrictions apply.
140 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 37, NO. 1, JANUARY 2019

[4] H. Xie, F. Gao, S. Zhang, and S. Jin, “A unified transmission strategy for
TDD/FDD massive MIMO systems with spatial basis expansion model,”
IEEE Trans. Veh. Technol., vol. 66, no. 4, pp. 3170–3184, Apr. 2017.
[5] R. W. Heath, N. González-Prelcic, S. Rangan, W. Roh, and
A. M. Sayeed, “An overview of signal processing techniques for
millimeter wave MIMO systems,” IEEE J. Sel. Topics Signal Process.,
vol. 10, no. 3, pp. 436–453, Apr. 2016.
[6] O. El Ayach, S. Rajagopal, S. Abu-Surra, Z. Pi, and R. W. Heath, Jr.,
“Spatially sparse precoding in millimeter wave MIMO systems,” IEEE
Trans. Wireless Commun., vol. 13, no. 3, pp. 1499–1513, Mar. 2014.
[7] X. Gao, L. Dai, S. Han, I. Chih-Lin, and R. W. Heath, “Energy-efficient
hybrid analog and digital precoding for MmWave MIMO systems with
large antenna arrays,” IEEE J. Sel. Areas Commun., vol. 34, no. 4,
pp. 998–1009, Apr. 2016.
[8] X. Gao, L. Dai, Y. Sun, S. Han, and C.-L. I, “Machine learning
inspired energy-efficient hybrid precoding for mmWave massive MIMO
systems,” in Proc. IEEE Int. Conf. Commun. (IEEE ICC), May 2017,
pp. 1–6.
Fig. 6. Energy efficiency against the number of users. [9] A. Alkhateeb, G. Leus, and R. W. Heath, “Limited feedback hybrid
precoding for multi-user millimeter wave systems,” IEEE Trans. Wireless
the RF chains can be significantly reduced compared to the Commun., vol. 14, no. 11, pp. 6481–6494, Nov. 2015.
fully digital MIMO systems. In addition, we can see from [10] X. Zhu, Z. Wang, L. Dai, and Q. Wang, “Adaptive hybrid precoding
for multiuser massive MIMO,” IEEE Commun. Lett., vol. 20, no. 4,
Fig. 5 that the sub-connected HP can achieve higher energy pp. 776–779, Apr. 2016.
efficiency than that of the fully-connected HP, since a less [11] B. Wang, L. Dai, Z. Wang, N. Ge, and S. Zhou, “Spectrum and energy-
number of phase shifters is adopted in the sub-connected HP. efficient beamspace MIMO-NOMA for millimeter-wave communica-
tions using lens antenna array,” IEEE J. Sel. Areas Commun., vol. 35,
The performance comparison in terms of energy efficiency no. 10, pp. 2370–2382, Oct. 2017.
against the number of users is shown in Fig. 6, where SNR [12] W. Yuan, V. Kalokidou, S. M. D. Armour, A. Doufexi, and M. A. Beach,
“Application of non-orthogonal multiplexing to mmWave multi-user
is set as 10 dB. We can see that when the sub-connected HP systems,” in Proc. IEEE Veh. Technol. Conf. (IEEE VTC Spring),
is adopted, the energy efficiency of the proposed mmWave Nov. 2017, pp. 1–6.
massive MIMO-OMA systems with SWIPT is higher than all [13] Y. Zhao, W. Xu, and S. Jin, “An minorization-maximization based
hybrid precoding in NOMA-mMIMO,” in Proc. IEEE Int. Conf. Wireless
of other schemes even the number of users is very large. Commun. Signal Process. (IEEE WCSP), Dec. 2017, pp. 1–6.
[14] W. Hao, M. Zeng, Z. Chu, and S. Yang, “Energy-efficient power allo-
VI. C ONCLUSIONS cation in millimeter wave massive MIMO with non-orthogonal multiple
In this paper, we propose to apply SWIPT in HP-based access,” IEEE Wireless Commun. Lett., vol. 6, no. 6, pp. 782–785,
Dec. 2017.
mmWave massive MIMO-NOMA systems to achieve a [15] L. Dai, B. Wang, Y. Yuan, S. Han, C.-L. I, and Z. Wang,
trade-off between spectrum efficiency and energy efficiency. “Non-orthogonal multiple access for 5G: Solutions, challenges, opportu-
nities, and future research trends,” IEEE Commun. Mag., vol. 53, no. 9,
To enable the spectrum- and energy-efficient systems, user pp. 74–81, Sep. 2015.
grouping, hybrid precoding, power allocation, and power split- [16] Z. Ding, M. Peng, and H. V. Poor, “Cooperative non-orthogonal mul-
ting are carefully designed. Specifically, the CHS algorithm is tiple access in 5G systems,” IEEE Commun. Lett., vol. 19, no. 8,
pp. 1462–1465, Aug. 2015.
first proposed to select one user for each beam as the cluster [17] Z. Ding, F. Adachi, and H. V. Poor, “The application of MIMO to non-
head, and then the analog precoding is designed according orthogonal multiple access,” IEEE Trans. Wireless Commun., vol. 15,
to the selected cluster heads for all beams. After that, user no. 1, pp. 537–552, Jan. 2016.
[18] S. Chen, B. Ren, Q. Gao, S. Kang, S. Sun, and K. Niu, “Pattern
grouping is performed based on the correlation of users’ division multiple access—A novel nonorthogonal multiple access for
equivalent channels. Then, the digital precoding is designed fifth-generation radio networks,” IEEE Trans. Veh. Technol., vol. 66,
by selecting users with the strongest equivalent channel gain no. 4, pp. 3185–3196, Apr. 2017.
[19] International Telecommunications Union—Radiocommunication
in each beam. Furthermore, the joint optimization of power (ITU-R), “IMT vision-framework and overall objectives of the future
allocation and power splitting is proposed to maximize the development of IMT for 2020 and beyond,” Sep. 2015.
achievable sum rate, and an iterative optimization algorithm [20] L. Varshney, “Transporting information and energy simultaneously,”
in Proc. IEEE Int. Symp. Inf. Theory (IEEE ISIT), Aug. 2008,
is developed to solve the non-convex optimization problem. pp. 1612–1616.
Simulation results show that the proposed mmWave mas- [21] T. D. P. Perera, D. N. K. Jayakody, S. K. Sharma, S. Chatzinotas, and
sive MIMO-NOMA systems with SWIPT can achieve higher J. Li, “Simultaneous wireless information and power transfer (SWIPT):
Recent advances and future challenges,” IEEE Commun. Surveys Tuts.,
spectrum and energy efficiency compared with mmWave vol. 20, no. 1, pp. 264–302, 1st Quart., 2018.
massive MIMO-OMA systems with SWIPT. In the future, [22] J. Huang, C.-C. Xing, and C. Wang, “Simultaneous wireless information
we will consider more sophisticated hybrid precoding design and power transfer: Technologies, applications, and research challenges,”
IEEE Commun. Mag., vol. 55, no. 11, pp. 26–32, Nov. 2017.
for the proposed mmWave massive MIMO-NOMA systems [23] I. Krikidis, S. Timotheou, S. Nikolaou, G. Zheng, D. W. K. Ng, and
with SWIPT to further improve the performance. R. Schober, “Simultaneous wireless information and power transfer in
modern communication systems,” IEEE Commun. Mag., vol. 52, no. 11,
R EFERENCES pp. 104–110, Nov. 2014.
[24] D. W. K. Ng, E. S. Lo, and R. Schober, “Wireless information and power
[1] S. Mumtaz, J. Rodriquez, and L. Dai, MmWave Massive MIMO: transfer: Energy efficiency optimization in OFDMA systems,” IEEE
A Paradigm for 5G. New York, NY, USA: Academic, 2016. Trans. Wireless Commun., vol. 12, no. 12, pp. 6352–6370, Dec. 2013.
[2] A. L. Swindlehurst, E. Ayanoglu, P. Heydari, and F. Capolino, [25] Z. Ding et al., “Application of smart antenna technologies in simulta-
“Millimeter-wave massive MIMO: The next wireless revolution?” IEEE neous wireless information and power transfer,” IEEE Commun. Mag.,
Commun. Mag., vol. 52, no. 9, pp. 56–62, Sep. 2014. vol. 53, no. 4, pp. 86–93, Apr. 2015.
[3] F. Rusek et al., “Scaling up MIMO: Opportunities and challenges with [26] X. Zhou, R. Zhang, and C. K. Ho, “Wireless information and power
very large arrays,” IEEE Signal Process. Mag., vol. 30, no. 1, pp. 40–60, transfer: Architecture design and rate-energy tradeoff,” IEEE Trans.
Jan. 2013. Commun., vol. 61, no. 11, pp. 4754–4767, Nov. 2013.

Authorized licensed use limited to: Cornell University Library. Downloaded on September 02,2020 at 03:11:02 UTC from IEEE Xplore. Restrictions apply.
DAI et al.: HYBRID PRECODING-BASED mmWAVE MASSIVE MIMO-NOMA WITH SWIPT 141

[27] E. Boshkovska, N. Zlatanov, L. Dai, D. W. K. Ng, and R. Schober, Bichai Wang (S’15) received the B.S. degree in
“Secure SWIPT networks based on a non-linear energy harvesting electronic engineering from Tsinghua University,
model,” in Proc. IEEE Wireless Commun. Netw. Conf. Workshops (IEEE Beijing, China, in 2015, where she is currently
WCNCW), May 2017, pp. 1–6. pursuing the Ph.D. degree with the Department of
[28] Q. Shi, L. Liu, W. Xu, and R. Zhang, “Joint transmit beamforming Electronic Engineering. Her research interests are
and receive power splitting for MISO SWIPT systems,” IEEE Trans. in wireless communications, with an emphasis on
Wireless Commun., vol. 13, no. 6, pp. 3269–3280, Jun. 2014. non-orthogonal multiple access, mmWave massive
[29] H. Zhang, A. Dong, S. Jin, and D. Yuan, “Joint transceiver and power MIMO, and deep learning-based wireless communi-
splitting optimization for multiuser MIMO SWIPT under MSE QoS cations. She received the Freshman Scholarship of
constraints,” IEEE Trans. Veh. Technol., vol. 66, no. 8, pp. 7123–7135, Tsinghua University in 2011, the Academic Merit
Aug. 2017. Scholarships of Tsinghua University in 2012, 2013,
[30] S. He, Y. Huang, W. Chen, S. Jin, H. Wang, and L. Yang, “Energy and 2014, respectively, the Excellent Thesis Award of Tsinghua University
efficient coordinated precoding design for a multicell system with RF in 2015, the National Scholarship in 2016, the IEEE VTC’17 Fall Best Student
energy harvesting,” EURASIP J. Wireless Commun. Netw., vol. 2015, Paper Award in 2017, the IEEE T RANSACTIONS ON C OMMUNICATIONS
no. 1, p. 67, Mar. 2015. Exemplary Reviewer Award in 2017, the First Prize of the 13th China
[31] Y. Liu, Z. Ding, M. Elkashlan, and H. V. Poor, “Cooperative non- Graduate Electronics Design Contest in 2018, and the IEEE ComSoc Asia-
orthogonal multiple access with simultaneous wireless information Pacific Outstanding Paper Award in 2018.
and power transfer,” IEEE J. Sel. Areas Commun., vol. 34, no. 4,
pp. 938–953, Apr. 2016.
[32] Y. Xu et al., “Joint beamforming and power-splitting control in downlink
cooperative SWIPT NOMA systems,” IEEE Trans. Signal Process.,
vol. 65, no. 18, pp. 4874–4886, Sep. 2017. Mugen Peng (M’05–SM’11) received the Ph.D.
[33] Z. Yang, Z. Ding, P. Fan, and N. Al-Dhahir, “The impact of power degree in communication and information systems
allocation on cooperative non-orthogonal multiple access networks with from the Beijing University of Posts and Telecom-
SWIPT,” IEEE Trans. Wireless Commun., vol. 16, no. 7, pp. 4332–4343, munications (BUPT), Beijing, China, in 2005.
Jul. 2017. Afterward, he joined BUPT, where he has been
[34] Q. Zhang, Q. Li, and J. Qin, “Robust beamforming for nonorthogonal a Full Professor with the School of Informa-
multiple-access systems in MISO channels,” IEEE Trans. Veh. Technol., tion and Communication Engineering since 2012.
vol. 65, no. 12, pp. 10231–10236, Dec. 2016. He leads a Research Group focusing on wireless
[35] L. Xiao, Y. Li, C. Dai, H. Dai, and H. V. Poor, “Reinforcement learning- transmission and networking technologies with the
based NOMA power allocation in the presence of smart jamming,” IEEE Key Laboratory of Universal Wireless Commu-
Trans. Veh. Technol., vol. 67, no. 4, pp. 3377–3389, Apr. 2018. nications, Ministry of Education, BUPT. He has
[36] Z. Ding, R. Schober, and H. V. Poor, “A general MIMO framework authored/coauthored over 100 refereed IEEE journal papers and over 200 con-
for NOMA downlink and uplink transmission based on signal align- ference proceeding papers. His main research areas include wireless commu-
ment,” IEEE Trans. Wireless Commun., vol. 15, no. 6, pp. 4438–4454, nication theory, radio signal processing, and convex optimizations, with a
Jun. 2016. particular interest in cooperative communication, self-organization network-
[37] J. R. Magnus and H. Neudecher, Matrix Differential Calculus With ing, heterogeneous networking, cloud communication, and Internet of Things.
Application in Statistics and Econometrics. New York, NY, USA: Wiley, Dr. Peng was a recipient of the 2018 Heinrich Hertz Prize Paper Award,
1988. the 2014 IEEE ComSoc AP Outstanding Young Researcher Award, and the
[38] M. Grant and S. Boyd. (Dec. 2017). CVX: MATLAB Software for Best Paper Award in IEEE WCNC 2015, WASA 2015, GameNets 2014,
Disciplined Convex Programming, Version 2.1. [Online]. Available: IEEE CIT 2014, ICCTA 2011, IC-BNMT 2010, and IET CCWMC 2009.
https://cvxr.com/cvx He received the First Grade Award of the Technological Invention Award
[39] Z.-Q. Luo, W.-K. Ma, A. M.-C. So, Y. Ye, and S. Zhang, “Semidefinite in the Ministry of Education of China, and the First Grade Award of the
relaxation of quadratic optimization problems,” IEEE Signal Process. Technological Invention Award from the China Institute of Communications.
Mag., vol. 27, no. 3, pp. 20–34, May 2010. He is on the Editorial/Associate Editorial Board of the IEEE Communi-
cations Magazine, IEEE A CCESS , IEEE I NTERNET OF T HINGS J OURNAL,
Linglong Dai (M’11–SM’14) received the B.S.
IET Communications, and China Communications.
degree from Zhejiang University in 2003, the M.S.
degree (Hons.) from the China Academy of
Telecommunications Technology in 2006, and the
Ph.D. degree (Hons.) from Tsinghua University,
Beijing, China, in 2011. From 2011 to 2013, he was Shanzhi Chen (SM’04) received the bachelor’s
a Post-Doctoral Research Fellow with the Depart- degree from Xidian University in 1991 and the Ph.D.
ment of Electronic Engineering, Tsinghua Univer- degree from the Beijing University of Posts and
sity, where he was an Assistant Professor from Telecommunications, China, in 1997. He joined the
2013 to 2016 and has been an Associate Professor Datang Telecom Technology and Industry Group and
since 2016. He has co-authored the book mmWave the China Academy of Telecommunication Tech-
Massive MIMO: A Paradigm for 5G (Academic Press, Elsevier, 2016). nology (CATT) in 1994, and has been serving as
He has published over 60 IEEE journal papers and over 40 IEEE con- the EVP of Research and Development since 2008.
ference papers. He holds 16 granted patents. His current research interests He is currently the Director of the State Key Labo-
include massive MIMO, millimeter-wave communications, NOMA, sparse ratory of Wireless Mobile Communications, CATT,
signal processing, and machine learning for wireless communications. He has where he conducted research and standardization
received five conference Best Paper Awards at the IEEE ICC 2013, the IEEE on 4G TD-LTE and 5G. He has authored and co-authored four books
ICC 2014, the IEEE ICC 2017, the IEEE VTC 2017-Fall, and the IEEE [among them the textbook Mobility Management: Principle, Technology and
ICC 2018. He has also received the Tsinghua University Outstanding Ph.D. Applications (Springer Press)], 17 book chapters, approximately 100 journal
Graduate Award in 2011, the Beijing Excellent Doctoral Dissertation Award papers, 50 conference papers, and over 50 patents in these areas. He has
in 2012, the China National Excellent Doctoral Dissertation Nomination contributed to the design, standardization, and development of 4G TD-LTE
Award in 2013, the URSI Young Scientist Award in 2014, the IEEE T RANS - and 5G mobile communication systems. His current research interests include
ACTIONS ON B ROADCASTING Best Paper Award in 2015, the Second Prize 5G mobile communications, network architectures, vehicular communication
of Science and Technology Award of the China Institute of Communications networks, and Internet of Things. He served as a member and a TPC Chair of
in 2016, the Electronics Letters Best Paper Award in 2016, the IEEE many international conferences. His achievements have received multiple top
C OMMUNICATIONS L ETTERS Exemplary Editor Award in 2017, the National awards by China central government and honors, especially the Grand Prize of
Natural Science Foundation of China for Outstanding Young Scholars in 2017, the National Award for Scientific and Technological Progress, China, in 2016
the IEEE ComSoc Asia-Pacific Outstanding Young Researcher Award in 2017, (this Grand Prize is the highest category and in some years, it leaves with no
and the IEEE ComSoc Asia-Pacific Outstanding Paper Award in 2018. winners due to its high standard). He is the Area Editor of the IEEE I NTERNET
He currently serves as an Editor of the IEEE T RANSACTIONS ON C OM - OF T HINGS , the Editor of the IEEE N ETWORK , and the Guest Editor of the
MUNICATIONS , the IEEE T RANSACTIONS ON V EHICULAR T ECHNOLOGY, IEEE W IRELESS C OMMUNICATIONS, the IEEE Communications Magazine,
and the IEEE C OMMUNICATIONS L ETTERS . and the IEEE T RANSACTIONS ON V EHICULAR T ECHNOLOGY.

Authorized licensed use limited to: Cornell University Library. Downloaded on September 02,2020 at 03:11:02 UTC from IEEE Xplore. Restrictions apply.