Alternating Minimization for Hybrid Precoding in

Multiuser OFDM mmWave Systems

Xianghao Yu∗ , Jun Zhang∗ , and Khaled B. Letaief∗† , Fellow, IEEE

∗
Dept. of ECE, The Hong Kong University of Science and Technology
†
Hamad Bin Khalifa University, Doha, Qatar
Email: {xyuam, eejzhang, eekhaled}@ust.hk, ‡ kletaief@hbku.edu.qa
∗

Abstract—Hybrid precoding is a cost-effective approach to by phase shifters, which induces a challenging unit modulus
support directional transmissions for millimeter wave (mmWave) constraint and forms the major challenge in designing hybrid
arXiv:1701.01567v1 [cs.IT] 6 Jan 2017

communications. While existing works on hybrid precoding precoders. Given the large dimension of the design space and
mainly focus on single-user single-carrier transmission, in prac- the unit modulus constraint, an important design aspect of
tice multicarrier transmission is needed to combat the much hybrid precoders is the computational complexity.
increased bandwidth, and multiuser MIMO can provide addi-
tional spatial multiplexing gains. In this paper, we propose a There have been many recent works on hybrid precoder de-
new hybrid precoding structure for multiuser OFDM mmWave
sign in mmWave systems [3]–[10]. In [3], [4], efficient hybrid
systems, which greatly simplifies the hybrid precoder design and
is able to approach the performance of the fully digital precoder. precoding algorithms were developed for single-user single-
In particular, two groups of phase shifters are combined to map carrier MIMO systems. The investigation was then extended to
the signals from radio frequency (RF) chains to antennas. Then single-user OFDM [4]–[6] and multiuser single-carrier systems
an effective hybrid precoding algorithm based on alternating min- [7], [8]. The main differences in these existing works are the
imization (AltMin) is proposed, which will alternately optimize approaches to deal with the unit modulus constraints of the
the digital and analog precoders. A major algorithmic innovation analog precoder. Specifically, such constraints were tackled
is a LASSO formulation for the analog precoder, which yields by orthogonal matching pursuit (OMP) in [3], [5], [7], by
computationally efficient algorithms. Simulation results will show manifold optimization in [4], and by channel phase extraction
the performance gain of the proposed algorithm. Moreover, it in [6], [8], respectively. There were also some studies on how
will reveal that canceling the interuser interference is critical in
to achieve the performance of the fully digital precoder with
multiuser OFDM hybrid precoding systems.
the hybrid structure [9], [10], yet requiring a large number
of RF chains. Although various attempts have been made to
I. I NTRODUCTION balance the performance and computational complexity, there
is no systematic approach to design computationally efficient
The spectrum crunch in current wireless systems stimu-
hybrid precoders.
lates extensive interests on exploiting new spectrum bands
for cellular communications, and millimeter wave (mmWave) In this paper, we shall propose a novel hybrid precoder
bands have been demonstrated to be a promising candidate in structure, which relaxes the unit modulus constraints of the
recent experiments [1]. Thanks to the smaller wavelength of analog part and thus significantly simplifies the hybrid pre-
mmWave signals, large-scale antenna arrays can be leveraged coder design. In particular, by adopting two groups of phase
at both the transmitter and receiver sides, which can provide shifters to map the signals out of the RF chains to antennas,
spatial multiplexing gains with the help of multiple-input the constraints for the analog precoder become more tractable.
multiple-output (MIMO) techniques. On the other hand, the We adopt the alternating minimization (AltMin) framework
ten-fold increase of the carrier frequency introduces several [4] for the hybrid precoder design. While the digital part is
challenges to mmWave communication systems, especially the similar to [4], the optimization of the analog RF precoder
high power consumption and cost of hardware components at is formulated as a Least Absolute Shrinkage and Selection
mmWave bands [2]. In addition, the large available bandwidth Operator (LASSO) problem, for which efficient algorithms are
at mmWave frequencies will result in wideband communica- available. In addition, we discover that the hybrid precoder in
tion systems, where multicarrier techniques such as orthogonal the multiuser setting will produce residual interuser interfer-
frequency-division multiplexing (OFDM) will be utilized to ence, as it only approximates the fully digital precoder. Such
overcome the frequency-selective fading. interference will significantly degrade the system performance,
By utilizing a small number of radio frequency (RF) especially at high SNRs. This issue is more prominent in the
chains to combine a low-dimensional digital baseband precoder multicarrier system as the analog precoder is shared by a large
and another high-dimensional analog RF precoder, hybrid number of subcarriers. To this end, we propose to apply an
precoding stands out as a cost-effective transceiver solution additional block diagonalization (BD) precoder at the baseband
[3]. Moreover, to further reduce the power consumption in to cancel the interuser interference, which is shown to be
the RF domain, analog RF precoders are usually implemented effective to further improve the spectral efficiency and mul-
tiplexing gain. Moreover, simulation results demonstrate that
This work was supported by the Hong Kong Research Grants Council under the proposed hybrid precoding algorithm can easily approach
Grant No. 16210216. the performance of the fully digital precoder with a reasonable
amount of RF chains.

II. S YSTEM M ODEL AND P ROBLEM F ORMULATION

RF Chain
A. System Model RF Chain

Consider the downlink transmission of a multiuser OFDM

mmWave MIMO system, where the base station (BS) is (a) Conventional analog RF precoder structure.
equipped with Nt antennas and transmits signals to K Nr -
antenna users over F subcarriers. On each subcarrier, Ns data
streams are transmitted to each user. The limitations of the RF
t r
chains are given by KNs ≤ NRF ≤ Nt and Ns ≤ NRF ≤ Nr ,
t r
where NRF and NRF are the number of RF chains facilitated
for the BS and each user, respectively.
RF Chain
The received signal for the k-th (1 ≤ k ≤ K) user on the RF Chain
f -th subcarrier is given by
K
!
X
H H
yk,f = WBB,k,f WRF,k Hk,f FRF FBBk,f sk,f + nk,f ,
i=1
(1) (b) DPS analog RF precoder structure.
where sk,f ∈ CNs is the transmitted symbol vector for the k-th
1
user on the f -th subcarrier such that E[sk,f sH
k,f ] = KNs F INs . Fig. 1. Comparison of two analog RF precoder structures.
The digital baseband precoders and combiners are symbolized
t r
by FBBk,f ∈ CNRF ×Ns and WBBk,f ∈ CNRF ×Ns , respec-
tively. Because the transmitted signals for all the users are i.e., |(FRF )i,j | = |(WRF )i,j | = 1. This is intrinsically a non-
mixed together via the digital baseband precoder and the ana- convex constraint and difficult to tackle with, which forms the
log RF precoder is a post-IFFT (inverse fast Fourier transform) main design challenge.
operation, the analog RF precoder is shared by all the users
t In this paper, we propose a new analog RF precoder
and subcarriers, denoted as FRF ∈ CNt ×NRF . Similarly, the
structure, as shown in Fig. 1(b), where the phase shifter
analog RF combiner is a subcarrier-independent operation for
r network is divided into two groups, referred as the Double
each user k, denoted as WRFk ∈ CNr ×NRF . Furthermore,
Phase Shifter (DPS) structure. For each route from an RF chain
the additive noise at the users is represented by nk,f ∈ CNr ,
to an antenna element, a unique phase shifter in each group will
whose elements are independent and identically distributed
be selected and summed up together to compose the analog
(i.i.d.) according to CN (0, σ 2 ). The mmWave MIMO channel
precoding gain. Under this special structure, each nonzero
between the BS and the k-th user on the f -th subcarrier,
element in the analog RF precoding matrix corresponds to a
denoted as Hk,f , can be characterized by the Saleh-Valenzuela
sum of two phase shifters. In other words, the new constraints
model as [3]
for the analog RF precoder and combiner are |(FRF )i,j | ≤ 2
Ncl,k −1 Nray,k
X X and |(WRF )i,j | ≤ 2 since the amplitude of a sum of two
−j2πif /F
Hk,f = γk αil,k ar (θil,k )aH
t (φil,k )e . phase shifters should be less than 2. By doubling the number
i=0 l=1 of phase shifters, the new constraint becomes convex and
q (2) therefore makes it more tractable and promising to develop
ρk Nt Nr low-complexity design approaches.
The normalization factor γk is specified by γk = Ncl,k Nray,k ,
where Ncl,k and Nray,k represent the number of clusters and
the number of rays in each cluster, and ρk is the path loss C. Problem Formulation
between the BS to the k-th user. The gain of the l-th ray in
the i-th propagation cluster is denoted as αil,k . In addition, As shown in [3], [4], minimizing the Euclidean distance
ar (θil,k ) and at (φil,k ) stand for the receive and transmit array between the fully digital precoder and the hybrid precoder is
response vectors with the corresponding angle of arrival θil,k an effective way to design the hybrid precoder in mmWave
and angle of departure φil,k . The detailed expressions can be MIMO systems, whose formulation1 is given by
found in [3].
minimize kFopt − FRF FBB kF
FRF ,FBB
(
B. Analog RF Precoder Structure |(FRF )i,j | ≤ 2 (3)
subject to 2
As mentioned before, the analog precoder is practically kFRF FBB kF ≤ KNs F,
implemented by phase shifters. Conventionally, in either the h i
fully- or partially-connected structure [4], each route from a where Fopt = Fopt 1,1 , · · · , Fopt k,f , · · · , Fopt K,F is the
certain RF chain to one connected antenna element is imple- combined fully digital precoder with dimension Nt × KNs F ,
mented by a phase shifter, as shown in Fig. 1(a). This mapping
strategy implies that each nonzero element in the analog 1 Here we focus on the precoder design, and the combiner is designed in
precoding and combining matrices should have unit modulus, the same way without the transmit power constraint.
  
and FBB = FBB1,1 , · · · , FBBk,f , · · · , FBBK,F is the con- precoder, we will focus on the analog precoder design in the
t
catenated digital precoder with dimension NRF × KNs F . The following. The optimization of the analog precoder is given by
second constraint is the transmit power constraint at the BS
side. minimize kFopt − FRF FBB kF
FRF (4)
Problem (3) is a matrix decomposition problem, and the subject to |(FRF )i,j | ≤ 2.
goal of this formulation is to find an accurate approximation Note that the power constraint in (3) is temporarily removed.
for an arbitrary fully digital precoder. With this formulation, In fact, a simple normalization operation can be adopted if the
the proposed algorithm can be applied with any fully digital power constraint is not satisfied, which will be discussed later.
precoder. It also has been shown in [3] that minimizing the The optimization problem (4) is a convex problem and can be
objective function in (3) leads to the maximization of spectral solved by solvers such as CVX. Nevertheless, to further reduce
efficiency in the single-user single-carrier case. In this paper, the computational complexity, we will illustrate the inherent
we will adopt the classical BD precoder as Fopt , which is structure of the solution by considering the dual problem. This
asymptotically optimal in the high signal-to-noise ratio (SNR) will lead to a closed-form solution to (4).
regime [11].
Lemma 2. The dual problem of (4) is a LASSO problem, given
by
III. H YBRID P RECODER D ESIGN 1 2
minimize kAx − bk2 + 2kxk1 . (5)
Alternating minimization, which separates the optimization
x 2
of the objective function with respect to different variable The parameters A and b are given by
subsets in each step, has been shown to be effective in hybrid 1
precoding [4] and various other applications, e.g., matrix A = S 2 U, b = ADH fopt , (6)
completion, phase retrieval, and dictionary learning [12]. In −1
= USUH is


this section, we will adopt alternating minimization as the main where D = FTBB ⊗ INt and DH D

−1
approach to design the hybrid precoder under the DPS analog the singular value decomposition (SVD) of DH D . The
RF precoder structure. optimal solution of (4) can be written by
?
fRF = AH (b − Ax? ) , (7)
A. Single-carrier Systems
where fRF = vec(FRF ).
While the main focus of this paper is on multiuser multicar-
rier systems, some advantages of the proposed DPS structure Proof: The proof is omitted due to space limitation.
will be firstly presented in single-carrier systems, as shown in
the following result. Based on Lemma 2, the analog precoder design problem is
transferred to a LASSO problem. This provides the opportunity
Lemma 1. For single-carrier systems, with the DPS precoder to leverage the large body of existing works on efficiently solv-
structure in Fig. 1(b), the fully digital precoder Fopt can be ing the general LASSO problem [13]. Here we are interested
perfectly decomposed into the hybrid precoder FRF and FBB in a special case where we can get a closed-form solution to
t
using the minimum number of RF chains, i.e., NRF = KNs the problem, which will significantly reduce the computational
r
and NRF = Ns . complexity of the hybrid precoding algorithm.
Proof: The proof can be easily obtained by the rank It was shown in [4] that enforcing a semi-orthogonal con-
sufficiency of FRF and FBB in the decomposition when straint to the digital precoder will incur little performance loss
F = 1, and is omitted due to space limitation. in single-user multicarrier systems. Inspired by this work, we
resort to a similar approach, i.e., imposing a semi-orthogonal
Lemma 1 shows that, for single-carrier systems with either constraint to the digital precoder, which is specified as
single-user or multiuser transmissions, the performance of the
fully digital precoder can be easily obtained with a hybrid FBB FH
BB = INRF
t . (8)
precoder via a simple matrix decomposition. Note that, with
the conventional analog precoder structure, the number of RF Under this constraint, the observation matrix A in the LASSO
chains should be at least twice that of the data streams in problem (5) is also semi-orthogonal, i.e.,
t
order to achieve the fully digital precoder, i.e., NRF = 2KNs
−1
r AH A = DH D
and NRF = 2Ns [4], [9]. Considering that the RF chain is  −1
significantly more power hungry than the phase shifter [4], the = (FTBB ⊗ INRF t )
H
(FTBB ⊗ INRFt )
proposed structure is more energy efficient when achieving the  −1 (9)
fully digital precoder. = (F∗BB ⊗ INRF t )(F
T
⊗ I t )
BB N RF
 −1
B. Hybrid Precoder Design in Multicarrier Systems via Alter- = (FBB FH T
BB ) ⊗ INRF t = IN t 2 .
RF
nating Minimization
With the semi-orthogonal observation matrix A, the LASSO
In each step of alternating minimization, one part of the problem (5) has a closed-form solution as [13]
hybrid precoder is fixed while the other part will be optimized.
+
x? = exp{j∠(AH b)} AH b − 2 ,
 
Since the main difficulty is the constraint on the analog (10)
where (x)+ = max{0, x}, and the corresponding solution to LASSO-AltMin Algorithm: LASSO Based Alternating Min-
FRF in (4) is imization Algorithm

+ Require: Fopt
F?RF = Fopt FH H Fopt FH

 
BB −exp j∠ Fopt FBB BB − 2 . (0)

1: Construct a feasible FRF and set k = 0;
(11)
2: repeat
Note that, to obtain an analog precoder FRF with the fixed
digital precoder FBB , a product between Fopt and FH 3: Fix FRF (k) , solving FBB (k) using the solution to OPP
BB is the
only required step, which is much more computationally effi- (12);
cient than solving the original problem (4) using an algorithm- 4: Fix FBB (k) , and update FRF (k+1) by the LASSO solu-
embedded solver. tion (11);
5: k ← k + 1;
After updating the analog precoder FRF , optimizing the 6: until a stopping criterion triggers.
digital precoder FBB with the semi-orthogonal constraint (8) 7: Compose the effective channels Ĥk,f according to (13).
is a typical semi-orthogonal Procrustes problem (OPP). The 8: Compute BD precoders FBDk,f [11].
solution is similar to [4, Eq. 28], which can be expressed as 9: The final digital baseband precoder FBk,f =
FBBk,f FBDk,f .
FBB = VUH
1 , (12)
10: For the √ digital precoder at the transmit end, normalize
H KNs F
where U1 SV = FH H
opt FRF is the SVD of Fopt FRF , and S is F
bB =
kFRF FB kF FB if the power constraint in (3) is not
t
a diagonal matrix whose nonzero elements are the first NRF satisfied.
nonzero singular values σ1 , · · · , σNRF
t .

C. Interuser Interference Cancellation IV. S IMULATION R ESULTS

While we can perfectly cancel the interuser interference In this section, we will evaluate the performance of the
with the fully digital precoder, there will be residual interuser proposed LASSO-AltMin algorithm and compare it with the
interference for the hybrid precoder, which is an approximation OMP algorithm [3], [7] in multiuser OFDM mmWave systems.
of the fully digital one. Later in Section IV, we will see Assume that Ns = 3 data streams are sent from the BS to
that in multiuser multicarrier systems, interuser interference each user in a 3-user (K = 3) MIMO systems over F =
is a severe problem that will dramatically degrade the hybrid 128 subcarriers, with Nt = 256 and Nr = 16, while both
precoding performance, especially at high SNRs. are equipped with uniform linear arrays (ULAs). The channel
parameters are given by Ncl,k = 3 clusters and Nray,k = 8
In this subsection, after designing the hybrid precoder and rays. The angles of departure and arrival (AoDs and AoAs)
combiner, we propose to cascade another digital baseband follow the Laplacian distribution with uniformly distributed
precoder FBD that is responsible for canceling the residual mean angles in [0, 2π] and angular spread of 10 degrees. The
interuser interference. In particular, with the hybrid precoder antenna elements in ULAs are separated by a half wavelength
and combiner at hand, we define an effective channel for the distance, and all simulation results are averaged over 5000
k-th user on the f -th subcarrier as channel realizations.
H H
Ĥk,f = WBBk,f WRFk Hk,f FRF FBBf , (13) 120
Fully digital BD
  LASSO-AltMin
where FBBf = FBB1,1 , · · · , FBBk,f , · · · , FBBK,f ∈ 110
LASSO-AltMin w/o BD
t
NRF ×KNs OMP
C is the composite digital precoder on the f -th
Spectral Efficiency (bps/Hz)

100

subcarrier, and Ĥk,f ∈ CNs ×KNs is the effective channel.

90
Our goal is to design the precoders FBDk,f , which satisfy the
conditions 80
Ĥj,f FBDk,f = 0, k 6= j. (14)
70
A simple way to achieve the conditions is the BD precoder, and
60
note that the dimension of the effective channel is sufficient
to perform the BD design. More details can be found in [11]. 50

The overall hybrid precoding algorithm is summarized as 40

the LASSO-AltMin algorithm. Note that Steps 3 and 4 both -10 -8 -6 -4 -2 0 2 4 6 8 10
SNR (dB)
give the globally optimal solution to the digital and analog
precoders, respectively. Hence, the algorithm will converge to Fig. 2. Spectral efficiency achieved by different precoding algorithms when
a stationary point of problem (4) with an additional constraint t
NRF = KNRFr = KNs .
(8), since it is a two block coordinate descent procedure [14].
In the last step, we normalize the digital precoder if the
transmit power constraint is not satisfied. It has been shown Fig. 2 shows the spectral efficiency of different algorithms
t
in [4, Lemma 1] that as long as we can make the Euclidean with the minimum numbers of RF chains, i.e., NRF = KNs
r
distance between the optimal digital precoder and the hybrid and NRF = Ns . First, we see that the proposed algorithm
precoder sufficiently small when ignoring the power constraint, significantly outperforms the existing OMP algorithm imple-
the normalization step will also achieve a small distance to the mented under the conventional analog precoder structure. The
optimal digital precoder. performance gain mainly comes from doubling the number of
phase shifters, which provides more degrees of freedom for • It is beneficial, from both performance and complexity
the analog precoding gain. We observe in the simulation that points of view, to implement twice the number of
the time complexity of the proposed LASSO-AltMin algorithm phase shifters in the analog precoder, as shown in Fig.
is comparable with the OMP algorithm, mainly thanks to the 1(b).
closed-form solutions in the alternating procedures.
• Different from other hybrid precoding systems, in
In [4], [7], it has been pointed out that approximating the multiuser multicarrier systems, interuser interference
fully digital precoder with a hybrid structure will lead to a is a vital problem that we should deal with in addi-
near optimal performance in single-user single-carrier, single- tion to the fully digital precoder approximation. To
user multicarrier, and multiuser single-carrier mmWave MIMO solve this problem, it is effective to cascade a digital
systems. In Fig. 2, we evaluate the performance of the LASSO- baseband precoder that specializes in canceling the
AltMin algorithm without the additional BD operation. We interuser interference.
discover that, without the BD precoder canceling the interuser
interference, there will be residual interuser interference, which It is interesting to extend the proposed DPS analog precoder
results in an obvious performance loss compared to the fully structure to investigate other problems involving hybrid pre-
digital one, especially at high SNRs. This phenomenon il- coder design, e.g., to consider the hybrid precoder design
lustrates that simply approximating the fully digital precoder combined with channel training and feedback.
with the hybrid one is not sufficient in multiuser multicarrier
mmWave systems since the analog precoder is shared by R EFERENCES
a large number of subcarriers. The comparison in Fig. 2 [1] M. R. Akdeniz, Y. Liu, M. K. Samimi, S. Sun, S. Rangan, T. S.
demonstrates the effectiveness and necessity of the BD steps Rappaport, and E. Erkip, “Millimeter wave channel modeling and
in the proposed LASSO-AltMin algorithm. cellular capacity evaluation,” IEEE J. Sel. Areas Commun., vol. 32,
no. 6, pp. 1164–1179, Jun. 2014.
Fig. 3 compares the performance of different precoding [2] T. S. Rappaport, R. W. Heath Jr., R. C. Daniels, and J. N. Murdock,
t
schemes for different RF chain numbers NRF at the BS side Millimeter Wave Wireless Communications. New York, NY, USA:
r Pearson Education, 2014.
while keeping NRF = Ns as the minimum number of RF
chains at each user. It is shown that the proposed LASSO- [3] O. E. Ayach, S. Rajagopal, S. Abu-Surra, Z. Pi, and R. W. Heath Jr.,
“Spatially sparse precoding in millimeter wave MIMO systems,” IEEE
100 Trans. Wireless Commun., vol. 13, no. 3, pp. 1499–1513, Mar. 2014.
[4] X. Yu, J.-C. Shen, J. Zhang, and K. B. Letaief, “Alternating mini-
95 mization algorithms for hybrid precoding in millimeter wave MIMO
systems,” IEEE J. Sel. Topics Signal Process., vol. 10, no. 3, pp. 485–
Spectral Efficiency (bps/Hz)

90 500, Apr. 2016.

[5] J. Lee and Y. H. Lee, “AF relaying for millimeter wave communication
85
Fully digital BD
systems with hybrid RF/baseband MIMO processing,” in Proc. IEEE
80
LASSO-AltMin Int. Conf. Commun. (ICC), Sydney, NSW, Australia, Jun. 2014, pp.
LASSO-AltMin w/o BD 5838–5842.
OMP
75 [6] A. Alkhateeb and R. W. Heath Jr., “Frequency selective hybrid pre-
coding for limited feedback millimeter wave systems,” IEEE Trans.
70 Commun., vol. 64, no. 5, pp. 1801–1818, May 2016.
[7] T. E. Bogale and L. B. Le, “Beamforming for multiuser massive MIMO
65
systems: Digital versus hybrid analog-digital,” in Proc. IEEE Global
Commun. Conf. (GLOBECOM), Austin, TX, USA, Dec. 2014, pp.
60
9 9.5 10 10.5 11 11.5 12 12.5 13 13.5 14 4066–4071.
Number of RF chains [8] L. Liang, W. Xu, and X. Dong, “Low-complexity hybrid precoding
in massive multiuser MIMO systems,” IEEE Wireless Commun. Lett.,
Fig. 3. Spectral efficiency achieved by different precoding algorithms for vol. 3, no. 6, pp. 653–656, Dec. 2014.
t , given N r
different transmit RF chain numbers NRF RF = Ns and SNR=5 [9] E. Zhang and C. Huang, “On achieving optimal rate of digital precoder
dB. by RF-baseband codesign for MIMO systems,” in Proc. 80th IEEE Veh.
Technol. Conf. (VTC Fall), Vancouver, BC, Sep. 2014, pp. 1–5.
AltMin algorithm can approach the performance of the fully [10] T. E. Bogale, L. B. Le, A. Haghighat, and L. Vandendorpe, “On the
digital precoder when the number of RF chains is slightly number of RF chains and phase shifters, and scheduling design with
larger than the number of data streams, which cannot be hybrid analog-digital beamforming,” IEEE Trans. Wireless Commun.,
realized by the existing OMP algorithm. Thanks to the newly vol. 15, no. 5, pp. 3311–3326, May 2016.
proposed DPS analog precoder structure and the LASSO- [11] Q. H. Spencer, A. L. Swindlehurst, and M. Haardt, “Zero-forcing meth-
ods for downlink spatial multiplexing in multiuser MIMO channels,”
AltMin algorithm, it turns out that there will not be much IEEE Trans. Signal Process., vol. 52, no. 2, pp. 461–471, Feb. 2004.
performance loss when we adopt hybrid precoding in multiuser [12] P. Netrapalli, P. Jain, and S. Sanghavi, “Phase retrieval using alternating
OFDM mmWave systems. minimization,” IEEE Trans. Signal Process., vol. 63, no. 18, pp. 4814–
4826, Sep. 2015.
V. C ONCLUSIONS [13] T. Hastie, R. Tibshirani, and M. Wainwright, Statistical learning with
sparsity: The lasso and generalizations. Boca Raton, FL, USA: CRC
This paper proposed a new analog precoder structure for Press, 2015.
hybrid precoding, based on which a LASSO based alternating [14] L. Grippo and M. Sciandrone, “On the convergence of the block
nonlinear gauss–seidel method under convex constraints,” Operat. Res.
minimization algorithm was proposed for hybrid precoder Lett., vol. 26, no. 3, pp. 127–136, Mar. 2000.
design in multiuser OFDM systems. The paper helped unravel
some valuable design insights: