You are on page 1of 9

Available online at www.sciencedirect.

com

ScienceDirect
ICT Express xxx (xxxx) xxx
www.elsevier.com/locate/icte

A novel low complexity beamforming scheme for coordinated multipoint


networks
Sandeep Gulia ∗, Anwar Ahmad
Department of Electronics and Communication Engineering, Faculty of Engineering and Technology, Jamia Millia Islamia, New Delhi, India
Received 17 January 2021; received in revised form 19 August 2021; accepted 21 August 2021
Available online xxxx

Abstract
This article presents a quick overview of conventional precoding and beamforming techniques in the cellular networks and proposes a
novel coordinated beamforming technique for coordinated multipoint (CoMP) networks, which is based upon signal-to-leakage-plus-noise
ratio (SLNR) metric. The proposed scheme has reduced computational cost and offers improved performance compared to the conventional
multi-cell precoding schemes in the noise-dominated environments. Simulation results reveal that proposed scheme offers a spectrally efficient,
low complexity solution with improved average user throughput and sum rate in the CoMP enabled networks.
⃝c 2021 The Korean Institute of Communications and Information Sciences (KICS). Publishing services by Elsevier B.V. This is an open access
article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Keywords: Coordinated beamforming; SLNR; Inter-cell interference; CoMP; Precoding

1. Introduction The CoMP was introduced as a novel solution that can


mitigate or even exploit inter-cell interference into a construc-
Inter-cell interference is ostensibly a major bottleneck that
tive useful signal. Initially, CoMP was implemented for fourth
adversely affects the performance of cellular mobile users
specifically at the cell-edges. There are several categories of generation (4G) networks, but recently has been incorporated
techniques that have been developed to contain the inter-cell as one of the fifth generation (5G) component technologies [5].
interference. These techniques may be applied at transmitter It is envisaged as a promising technology and key enabler
side, receiver side or both. However, techniques implemented in the foreseeable future to combat interference in dense
at transmitter side are usually preferred to circumvent com- cellular network employing aggressive frequency reuse. Fur-
plexity at the user equipment (UE) [1]. The widely used ther, CoMP is anticipated to play a cardinal role in scenarios
techniques in modern cellular networks to mitigate interfer- where network reliability, coverage imbalance and cell edge
ence implemented at network side are inter-cell interference throughput are prime concerns [6,7]. Fundamentally, CoMP is
coordination (ICIC) [2] with variants such as enhanced ICIC a base station coordination technique that deals with inter-cell
(eICIC) [3], further enhanced ICIC (feICIC) [4] etc. These interference differently than the conventional interference mit-
schemes improve the cell-edge performance by partitioning igation schemes. It is a dynamic approach that enables rapid
and management of resources in power, frequency and time cell cooperation and exploits spatial diversity to diminish the
domains. Nevertheless, since resources are segregated between impact of signal fading. CoMP allows a set of neighbor cells,
center and cell edge users in these techniques, they suffer in termed cluster, to cooperatively transmit data to target UEs
terms of spectral efficiency and cell throughput [2]. Further, with fast information exchange. CoMP is essentially a set of
due to quasi-static nature of these schemes, these schemes techniques broadly categorized into three foremost schemes —
are inefficient to cope with aggressive inter-cell interference joint transmission CoMP (JT-CoMP), dynamic point selection
in challenging radio environments. CoMP (DPS-CoMP) and coordinated scheduling/beamforming
CoMP (CS/CBF-CoMP) [8]. First two techniques jointly pro-
∗ Corresponding author.
cess both user data and control information of co-scheduled
E-mail addresses: sandeepgulia@gmail.com (S. Gulia),
aahmad4@jmi.ac.in (A. Ahmad).
users simultaneously in a cluster whereas latter only processes
Peer review under responsibility of The Korean Institute of Communica- the coordination information jointly with other cluster cells.
tions and Information Sciences (KICS). Fig. 1 illustrates the principle mechanisms of three CoMP
https://doi.org/10.1016/j.icte.2021.08.020
2405-9595/⃝ c 2021 The Korean Institute of Communications and Information Sciences (KICS). Publishing services by Elsevier B.V. This is an open access
article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Please cite this article as: S. Gulia and A. Ahmad, A novel low complexity beamforming scheme for coordinated multipoint networks, ICT Express (2021), https://doi.org/10.1016/j.icte.2021.08.020.
S. Gulia and A. Ahmad ICT Express xxx (xxxx) xxx

Fig. 1. Illustration of coordinated multipoint in three-cell cluster scenario: (a) JT-CoMP (b) DPS-CoMP (c) CS/CBF-CoMP.

schemes. Fig. 1(a) describes the JT-CoMP process where three with all cells concurrently, enabling coordinated beamform-
cell-edge users UE1 , UE2 and UE3 positioned in the adjacent ing among cluster cells to nullify inter-cell interference. The
sectors of eNB1 , eNB2 and eNB 3 respectively are being served coordinating cells generate narrow antenna beams pointing
simultaneously by all eNBs in order to increase their signal towards target UEs while silencing their beams in the direction
strength. The evolved Node B (eNBs) are connected via high of unintended users. For instance, eNB1 , eNB2 and eNB3
capacity backhaul links to a central entity, referred to as CoMP steer their beams towards intended users UE1 , UE2 and UE3
coordination unit (CCU), which is responsible for overall respectively while muting their beams towards victim UEs.
coordination and control. Note that in CoMP frameworks, a Additionally, time–frequency resources are scheduled by each
base station is referred to as a transmission point (TP) and eNB such that no two or more cells transmit in the same
each cell (or sector) of an eNB site may contain one or more frequency resource in vicinity within the cooperation region.
TPs [8]. However, for the sake of simplicity, we consider TP For example, eNB1 employs frequency resource f1 to generate
and eNB equivalent in this article. Under non-coordination beam towards UE1 while eNB2 and eNB3 use resources f2
condition, each UE would experience strong interference from and f3 for UE2 and UE3 respectively. As a result, cross-cell
neighbor eNBs. However, with base station coordination using interference is suppressed and the received signal quality of
JT-CoMP scheme, the interference is harnessed by utilizing it interference prone UEs at cell-edges is improved.
as a beneficial signal. The UEs estimate and feedback local In general, JT and DPS schemes provide higher throughput
channel state information (CSI) to their serving eNBs, which gains than CS/CBF but are delay sensitive requiring tight
is forwarded to CCU by each eNB. The CCU computes a synchronization, high capacity backhaul and accurate CSI [11–
global CSI by using local CSI and shares it among all eNBs 13]. Conversely, CS/CBF has less stringent requirements for
concurrently. The eNBs exploit global CSI to design precoding synchronization and backhaul capacity, however, offers low
matrices for their associated users and thereby schedule radio throughput gains comparatively.
resources in coordination with other cluster cells. Since JT al- There are several conventional single-cell precoding and
lows multiple cells to transmit simultaneously to a mobile user beamforming schemes that have been extended for multi-
in the same frequency resource block, as shown in Fig. 1(a), cell coordination and CoMP schemes. However, conventional
the signal quality and hence signal-to-interference-plus noise beamforming schemes are not efficient for CoMP scenarios
ratio (SINR) of cell-edge users is improved. This eventually and have several downsides. In this article, we first present a
results in the enhanced cell-edge throughput performance in brief overview of existing precoding schemes and then present
JT-CoMP [9,10], providing a consistent and ubiquitous user an efficient low complexity precoding scheme for CS/CBF-
experience. CoMP networks that improves user throughput and sum rate
In the DPS process shown in Fig. 1(b), the serving cell of a capacity.
mobile user is switched dynamically among the coordination
cells to provide strongest signal to the UE. DPS processes 2. Conventional precoding and related work
both data and CSI jointly similar to JT but allows only a
single eNB to transmit to an intended UE at a certain time The optimal design of precoding and beamforming algo-
instant as indicated. Each UE can be scheduled on a subframe rithms in the cellular systems is a formidable task owing
basis to receive intended signal from different serving points to the intricate interference scenarios. The fundamental ap-
depending upon the instantaneous channel conditions. Since a proach in single-cell transmission has been to boost signal
mobile user switches dynamically to best eNB based upon the power at each UE relative to the background noise. This
varying channel parameters, overall signal quality is improved. technique aims to maximize signal-to-noise ratio (SNR) and
In case of CS/CBF, demonstrated in Fig. 1(c), all cluster is referred to as maximum ratio transmission (MRT) [14].
cells coordinate to perform scheduling and beamforming de- Since this scheme does not take into consider the interference
cisions, however, joint processing of data is not performed. from other users, performance jeopardizes severely in harsh
Hence, data bursts of a mobile user are not available at all interference environments. A reverse but appealing subopti-
base stations simultaneously; instead CCU exchanges user data mal approach for single antenna multi-user multiple-input–
separately with each eNB. However, CCU shares global CSI multiple-output (MU-MIMO) systems, known as zero forcing
2
S. Gulia and A. Ahmad ICT Express xxx (xxxx) xxx

(ZF) [15], aims to completely cancel the interference but In multi-cell cooperation scenarios, several works have
does not account noise factor in beamforming design process. extended the single-cell linear precoding approaches. Base
Zero-forcing beamforming (ZFBF) is a channel inversion and station coordination was initially proposed in [28] to enhance
relatively low complexity approach wherein data streams are downlink capacity. Different coordination approaches have
precoded such that each stream is orthogonal to all other been developed to mitigate inter-cell interference in MU-
streams, entirely canceling the interference from other spatial MIMO cellular networks [29,30]. The network coordinated
channels. However, due to exclusion of noise element in the beamforming with ZF as extended approach has been consid-
precoding process, ZF performance gains vanish quickly in ered in [31–33]. Multi-cell block diagonalization is proposed
low SNR regimes [16]. in [34], wherein authors design a multi-cell BD precoder to
Intuitively, considering the pros and cons of above two maximize weighted sum-rate subject to per base station power
approaches, an explicit choice is SINR based approach that constraints. In [35], authors employ uplink–downlink duality
accounts for both the aspects. Since, SINR parameter is more and develop an iterative algorithm to design joint optimal
significant than SNR in cellular environments, goal of this beamforming for minimizing the weighted transmit power
approach should be to design optimal beamforming weights to subject to SINR constraints. More recent related work for
achieve maximized SINR. In [17], authors suggest an optimal 5G massive MIMO networks are presented in [36,37]. CoMP
SINR metric based beamforming scheme that aims to perform consisting of different multi-cell cooperation strategies was
mid-way of MRT and ZF schemes. However, due to the cou- first introduced in Release-11 [8] by third generation part-
pled nature of the optimization problem, SINR metric based nership project (3GPP) for the long term evolution-advanced
approaches require iterative solutions that lead to high com- (LTE-A) networks with several enhancements in subsequent
putational complexity [18]. An extended ZF technique, called releases [5,38]. The CoMP has been applied in several works
regularized zero forcing (R-ZF) [19] or alternatively termed to improve the spectral and energy efficiency in 5G sys-
as minimum mean square error (MMSE) [20], is a heuristic tems [39–41]. Recently, more advanced beamforming tech-
approach designed to cancel the co-channel interference in niques such as hybrid beamforming have gained attention
MU-MIMO systems. While ZF efforts in full cancellation of that leverage benefits of both analog and digital beamforming
techniques. These schemes find applications in 5G millimeter
inter-symbol interference at the expense of noise enhance-
waves (mmWave) massive-MIMO systems due to their low
ment, MMSE precoder balances the multiuser interference
power consumption [42]. The application of enlarged MIMO
and allows a small amount of interference to minimize total
configurations using massive-MIMO approach has been re-
error [21]. Block diagonalization (BD) is another well-known
garded as an alternative to JT-CoMP [6]. The authors in [6]
precoding scheme wherein UEs can be equipped with multi-
propose an efficient and smart combination of small cells, JT-
ple antennas [22,23]. BD based algorithms are essentially an
CoMP and massive-MIMO to improve the spectral efficiency
extension of channel inversion schemes, which abolish inter-
with tolerable complexity levels. It is suggested by authors
user interference and not the inter-stream interference that
that combined framework of JT CoMP and massive MIMO is
originates from the multiple antennas of a single user. In BD
more advantageous since increased number of antennas leads
approach, diagonalization of user channels is implemented in to improvement in robustness of communication links, reduced
blocks of antennas, while allowing interference within the backhaul overhead, and increased localization of interference.
streams of the same user. However, since the elimination A unique precoding solution based on the SLNR criterion
of multi-user interference is achieved at the cost of noise was introduced in [43] for single-cell MU-MIMO systems
enhancement, BD too performs poorly in low SNR regions. where authors transform coupled optimization problem into
Even though the sum rate improves marginally compared to a decoupled closed-form solution. In this leakage-based ap-
ZF in this scheme, it does not grow linearly with increasing proach, the computational intricacy is significantly reduced
SNR [24]. In [25], authors present a regularized block diago- owing to the dissociated nature of the optimization prob-
nalization (RBD) approach that improves maximized sum-rate lem. The authors establish that SLNR metric based multi-user
capacity in noise dominated and high multi-user interference beamforming solutions can achieve substantial performance
scenarios. gain over multi-user zero-forcing beamforming. The advan-
The aforementioned linear precoding techniques decouple tages offered by SLNR based approaches are two-fold. First,
the optimization problem, thereby reducing the overall com- they remove antenna configuration limitations of the channel
putational complexity. However, one major drawback of the inversion schemes. Second, impact of noise is also taken
linear precoding techniques is that they pose restriction on into consideration in the design process of beamforming co-
number of antennas wherein number of base station transmit efficients, thereby improving the performance in noise —
antennas should be larger than the sum of UE receive antennas. dominated environments. The concept of SLNR has been
Further, undesired amplification of background noise in these employed for MU-MIMO systems in several works [44–46].
schemes leads to performance loss, particularly in the low In [47], authors present a hybrid precoding algorithm for
SNR regimes. Several non-linear precoding techniques such multiuser massive-MIMO systems by using R-ZF as the digital
as dirty paper coding (DPC) [26] and Tomlinson–Harashima baseband precoder that maximizes the SLNR. A digital pre-
precoding (THP) [27] provide superior performance than lin- coder is designed in [48] to maximize the conditional average
ear techniques theoretically; their practical implementation is SLNR, which is utilized to construct a two stage beamformer
prohibitively complex. for downlink multiuser massive-MIMO systems.
3
S. Gulia and A. Ahmad ICT Express xxx (xxxx) xxx

Fig. 2. Illustration of leakage to unintended users in SLNR.

Fig. 3. General system model.


3. Proposed scheme
The SLNR of a cell user is expressed as the ratio of The first term in Eq. (1) represents the product of desired
intended signal power of the user to the aggregate of leakage signal quantities where Hk,l c
∈ C Nr ×Nt , Wk,lc
∈ C Nt ×Nl and
power caused by the intended signal to all other users in the c
xk,l ∈ C Nl ×1
denote complex channel matrix, global precoding
cell. The leakage is defined as the interference caused by
matrix and the received signal vector from serving cell l in
the signal intended for a desired UE on all remaining UEs
the cth cluster corresponding to the{ intended user k. With
as illustrated in Fig. 2. The traditional SINR based models
no loss of generality, we assume E |xl |2 = 1 where E{·}
}
consider the interference suffered by the intended UE. In con- c
trast, SLNR based model considers the interference caused by denotes the statistical expectation operator.Hk,l takes the form
T T T T 1×Nt
intended UE. In the SLNR approach, the achievable sum rates H = [h 1 h 2 · · · h K ] where h k ∈ C is the channel from all
can be improved by designing tractable optimal beamforming the base stations to the kth user in the cluster. The second term
coefficients such that SLNR of all users is maximized simul- corresponds to the inter-user interference experienced by user
taneously by minimizing the leakage elements. This allows k within the serving cell l. The third term represents aggregate
decoupling of the joint optimization problem offering closed- of inter-cluster interference (i.e. the inter-cell interference from
form solutions. The optimization problem in this case concerns the clusters other than the cth cluster) intercepted by user k.
with the total leakage power caused by a specific user to all The last term n k ∈ C Nr ×1 denotes additive white Gaussian
other users rather than the total interference power of all other noise (AWGN) vector having elements with distribution CN
users on that specific user. (0, σ 2 ), where N{·} denotes normal distribution.
The SINR for received signal in Eq. (1) can be expressed
3.1. General beamforming equations as
c c 2
∥Hk,l Wk,l ∥
Consider a cellular system as shown in Fig. 3, divided c
S I N Rk,l = ∑K ∑C ∑ L ′ ∑ K ′ ′ ′
(2)
′ W p ′ ,l ′ ∥ + σ
c c 2 c c 2 2
p=1 ∥Hk,l Wk,l ∥ + c′ =1, l ′ =1 p ′ =1 ∥Hk,l
into C disjoint static clusters using CS/CBF-CoMP scheme p̸=k c ′ ̸ =c

wherein radio resources of all cluster cells are assumed to be As evident from Eqs. (1) and (2), multi-user interference
scheduled in coordination as described in Fig. 1(c). Assume L from other cell users within the same cluster (cth cluster)
number of cells per cluster with K number of active users per is absent due to the coordinated scheduling/beamforming of
cell. The lth cell transmitter, l ∈ {1, 2, . . . , L}, is considered resources. Therefore, multi-user interference for a user in
to transmit Nl layers over Nt antennas to the kth UE receiver CS/CBF scheme relies merely on other users within the same
with Nr antennas. Consider user k being served in cell l of cell rather than other cell users of the cluster. The intra-cell
cluster c, where k ∈ {1, 2, . . . , K} and c ∈ {1, 2, . . . , C}. As multi-user interference can be completely suppressed by ap-
discussed in Section 1, each cell in CS/CBF scheme serves
propriate precoding scheme and hence second term in (1) can
its associated users only and cooperatively schedules radio
be reduced to zero. Furthermore, neglecting the inter-cluster
resources to alleviate inter-cell interference within the cluster.
interference, Eqs. (1) and (2) respectively reduce to
Consequently, multi-user interference from other cells within
the cluster is non-existent. However, multi-user interference c c c c
yk,l = Hk,l Wk,l xk,l + n k (3)
within a single cell still may exist. Accordingly, the received c c 2
c ∥Hk,l Wk,l ∥
signal vector, yk,l ∈ C Nr ×1 can be expressed as c
S I N Rk,l = (4)
∑K σ2
c c c c c c
yk,l = Hk,l Wk,l xk,l + Hk,l W p,l x cp,l It can be observed from Eqs. (3) and (4) that CS/CBF-CoMP
p=1
p̸=k
technique can significantly improve the signal quality of cell
C ∑
L ′ K ′ (1) edge-users. The achievable user rate for kth user in CS/CBF
∑ ′ ∑ ′ ′
+ c
Hk,l ′ W pc′ ,l ′ x cp′ ,l ′ + n k is given by
c′ =1, l ′ =1 p ′ =1
( )
c ′ ̸ =c Rk,l = log2 1 + Υk,l (5)
4
S. Gulia and A. Ahmad ICT Express xxx (xxxx) xxx

where Υk,l is designated as the SINR of the kth user located where L is cluster size and H p,i,l denotes channel matrix
in lth cell as defined by Eqs. (2) and (4). Further, since each between lth cell and the pth user located in ith cell. Wk,l is
cell serves its own users in the CS/CBF scheme, the sum rate the precoding matrix for kth user corresponding to the cell l.
of lth cell in the cluster can be defined as Therefore, the beamforming weights Wk , l , k ∈ {1, 2, . . . , K}
K
∑ { } for each user in Eqs. (9) and (10) should be designed optimally
R S,l = log2 (1 + Υk,l ) (6) to maximize SLNR. The sum rate in coordinated beamforming
k=1 schemes can be maximized by finding the global beamforming
The average sum rate of a cluster with size L and K active weights such that cross-cell interference within the cluster is
users per cell is given as nullified [35]. Hence, we formulate optimization problem as
opt
the evaluation of optimal precoding weights (Wl ), l ∈ {1, 2,
L ∑
K
∑ { } . . . , L} to achieve maximized sum rate
RS = E log2 (1 + Υk,l ) (7)
l=1 k=1 opt opt opt
(W1 , W2 , . . . .., W L ) = arg max (R S ) (11)
The SINR defined by Eq. (2) can be utilized as an optimization opt opt
(W1 ,W2 ,...,W L )
opt

criterion for designing appropriate beamforming weight (Wl )


for each kth user (k = 1, 2,. . . , K) to maximize the SINR where R S is sum rate over L cells in the cluster given
of each user. However, this leads to intractable non-convex by Eq. (6). Also, without loss of generality we assume,
optimization problem with K number of coupled variables. In opt
∥Wl ∥2 = 1. For simplicity, considering only two cells in
case of zero-forcing scheme, the coupled problem is avoided, the cluster (i.e. L = 2), Eq. (10) can be redefined as
wherein beamforming weight of each user is selected such that
inter-user interference to unintended user i is enforced to zero opt
(Waopt , Wb ) = arg max (R S,a + R S,b ) (12)
by applying the condition opt
(Wa ,Wb )
opt

Hi,l Wk,l = 0 ∀i = {i = 1, . . . , K }, i ̸= k (8) opt opt


where Wa and Wb are the optimized precoding weights
The ZF scheme is able to cancel the inter-user interference and R S , a and R S , b are the sum rates for cell a and cell b
entirely, however, above condition puts restrictions on antenna respectively. To further simply the situation, let us assume each
configurations. Additional detriment associated with ZF is cell in the cluster has only one user. Accordingly, we denote
exclusion of noise term in finding the beamforming weights. users k1 and k2 assumed to be located at the boundaries of
On the other hand, the antenna condition is no more required cell a and cell b respectively. By inserting sum rate expression
in SLNR based approach and it accounts for noise contribution from (6), Eq. (12) modifies as
as well while selecting beamforming weights.
opt opt
(Wk1 ,a , Wk2 ,b ) = arg max
opt opt
3.2. SLNR based coordinated beamforming (Wk ,a ,Wk ,b ) (13)
1 2
( )
× log2 (1 + S I N Rk1 ,a ) + log2 (1 + S I N Rk2 ,b )
In our beamforming CS/CBF-CoMP scheme, we utilize
the notion of SLNR and strive to exploit the benefits of where S I N Rk1 ,a and S I N Rk2 ,b are the SINR for users k1 and
this approach. First, we discuss the transformation of SINR k2 positioned in the cell a cell b respectively. Assuming SINR
expression for CS/CBF into corresponding SLNR expression ≫ 1, we can approximate Eq. (13) as
for kth user. For this, we reconsider the intended signal power
∥Hk,l Wk,l ∥2 defined in Eqs. (2) and (4), where Hk,l and Wk,l opt opt
(Wk1 ,a , Wk2 ,b ) ≈ arg max
denote channel and precoding matrices respectively corre- opt opt
(Wk ,a ,Wk ,b ) (14)
1 2
sponding to the user k from the serving cell l. Further, we ( )
× log2 (S I N Rk1 ,a ) + log2 (S I N Rk2 ,b )
define ∥Hi,l Wk,l ∥2 as the leakage power radiated towards ith
user by the signal intended for user k from the cell l. Consid-
Equivalently,
ering K active users, the SLNR for kth user in a single cell
environment can be expressed as opt opt
(Wk1 ,a , Wk2 ,b ) ≈ arg max
( )
2
log2 (S I N Rk1 ,a · S I N Rk2 ,b )
∥Hk,l Wk,l ∥ opt
(Wk ,a ,Wk ,b )
opt
S L N Rk,l = ∑ K (9) 1 2
i=1,i̸=k ∥Hi,l Wk,l ∥2 + σ 2
(15)
For each user, the intended signal power denoted by numerator
term in Eq. (9) is preferred to be as large as possible as
compared to the aggregate of leakage power represented by ≈ arg max
opt opt
the first term of the denominator. For the clustered multi-cell (Wk ,a ,Wk ,b )
1 2
scenarios, (9) can be modified as ( (
∥Hk1 ,a Wa ∥2
)(
∥Hk2 ,b Wb ∥2
))
× log2
∥Hk,l Wk,l ∥2 (∥Hk1 ,b Wb ∥2 ) + σ 2 (∥Hk2 ,a Wa ∥2 ) + σ 2
S L N Rk,l = ∑ L ∑ K (10)
i=1 p=1, p̸=k ∥H p,i,l Wk,l ∥2 + σ 2 (16)
5
S. Gulia and A. Ahmad ICT Express xxx (xxxx) xxx

where H̃k1 ,a = [HkT2 ,a ]T and H̃k2 ,b = [HkT1 ,b ]T are the extended


channel matrices that exclude Hk1 ,a and Hk2 ,b respectively. In
general,
H̃ki ,l = [HkT1 ,l · · · HkTi−1 ,l , HkTi+1 ,l · · · HkTK ,l ]T (22)
where l ∈ {1, 2, . . . , L} and H̃ki , l ∈ C Nr (L−1)×N t . The optimal
precoding vectors can be found by solving Eqs. (20) and (21)
given by [43]
(( )−1 H )
opt
wi ∝ max .eigenvector σ 2 I + H̃iH H̃i Hi Hi (23)
in terms of eigenvector corresponding
)−1 H to the largest eigenvalue
of the matrix σ 2 I + H̃iH H̃i
(
Hi Hi . Note that in the SLNR
metric based CBF scheme discussed above, cluster size of
2 has been assumed. However, the concept can be extended
Fig. 4. Geometric layout of system model: Green circles, red triangles and to desired number of cells in a cluster by considering all
red dots represent eNBs, directional sector antennas and UEs respectively. interfering cells while designing the generalized interference
channel matrix given by Eq. (22). By referring to Eqs. (10),
(18) and (19), we can write
The S I N Rk1 ,a and S I N Rk2 ,b indicated as part (a) and (b) ⎛ ⎞
respectively in Eq. (16) can be rearranged as opt ∥H k,l W l ∥ 2
Wk,l = arg max ⎝ ∑ L ∑ K ⎠ (24)
opt opt Wk,l ∥H W ∥ 2 + σ2
(Wk1 ,a , Wk2 ,b ) ≈ arg max i=1 p=1 p,i,l
p̸=k
l
opt opt
(Wk ,a ,Wk ,b )
1 2 Therefore, for the multi-cell scenario we can directly apply
∥Hk1 ,a Wa ∥2 ∥Hk2 ,b Wb ∥2
( ( )( ))
Eq. (24) for calculating the optimal precoding weights.
× log2
(∥Hk2 ,a Wa ∥2 ) + σ 2 (∥Hk1 ,b Wb ∥2 ) + σ 2
4. Results and discussion
(17)
We consider a downlink statically clustered CoMP system
It can be observed from Eq. (17) that S I N Rk1 ,a and S I N Rk2 ,b with wraparound configuration and cluster size, L = 3.
have taken the form of S L N Rk1 and S L N Rk2 respectively, The geometric layout of system model is depicted in Fig. 4.
resembling SLNR equations (9) and (10). From Eq. (17), we The CoMP framework contains 7 hexagonal grid eNB sites
can write expressions for optimized precoding weights for cell (designated as A, B, C, D, E, F, and G), each consisting of 3
a and cell b respectively as sectors (or cells). Each cell associated with a physical site is
equipped with directional antennas. The three cells associated
∥Hk1 ,a Wa ∥2 with site A are labeled as A1, A2, A3. Likewise, B1, B2, B3
( )
opt
Wk1 ,a = arg max (18) are the cells associated with site B, and so on. The layout
W
opt (∥Hk2 ,a Wa ∥2 ) + σ 2
k1 ,a represents an inter-site CoMP operation where cells associated
with multiple sites are cooperating as indicated by shaded area
opt ∥Hk2 ,b Wb ∥2
( ) (cooperation area). Whereas all three cells associated with site
Wk2 ,b = arg max (19) A are involved in CoMP operation, cells associated with rest
W
opt (∥Hk1 ,b Wb ∥2 ) + σ 2
k2 ,b of the sites located in only inner circle (first tier) of site A are
opt opt coordinating. Further, each cell has K = 3 scheduled users
where ∥Wki ,l ∥2 = 1. Note that Wki ,l , l ∈ {1, 2, . . . , L}, are with uniform random distribution where each base station and
the global precoding weights that are computed by a single UE is configured with Nt = Nr = 2 antennas. We consider
entity (CCU) based on the local CSI shared by cluster cells. distance-dependent path loss propagation model with a decay
Hence, optimized precoding weights for users k1 and k2 can factor of 3.76 as defined in [8]. The SNR threshold for a user to
be calculated using Eqs. (18) and (19) such that SLNR for be grouped as cell-edge user is set at 15 dB. The shadowing ef-
each user is maximized. Accordingly, cluster sum rate can be fect considered follows log normal distribution with a standard
maximized by designing optimal precoding weights for each deviation of 8 dB. Major simulation parameters assumptions
user in the cluster in a coordinated manner. Eqs. (18) and (19) are listed in Table 1.
can further be rewritten as We investigate the impact of noise on the performance of
∥Hk1 ,a Wa ∥2 average sum rate of various precoding schemes as presented
( )
opt
Wk1 ,a = arg max (20) in Fig. 5. For the purpose of comparison of evaluation re-
opt
Wk ,a ∥ H̃k1 ,a Wa ∥2 + σ 2
1 sults of the proposed SLNR based coordinated beamforming
∥Hk2 ,b Wb ∥2
( )
opt scheme, referred to as SLNR-CBF, we consider multi-cell
Wk2 ,b = arg max (21) zero forcing beamforming (MC-ZFBF) scheme [33], multi-cell
W
opt ∥ H̃k2 ,b Wb ∥2 + σ 2
k2 ,b block diagonalization beamforming (MC-BDZF) scheme [34]
6
S. Gulia and A. Ahmad ICT Express xxx (xxxx) xxx

Table 1
Major simulation parameters.
Parameter Value
Cellular layout Hexagonal grid
Inter-site distance 500 m
Carrier frequency 2 GHz
Bandwidth 20 MHz
UE speed 5 km/h
Scheduling Proportional fair
Feedback channel delay 3
Receiver algorithm MMSE-IRC
Pathloss model 128.1 + 37.6 · log10 (dkm ) dB
Fast fading model SCM, urban macro
Traffic model Full buffer
Receiver noise figure 9 dB
Penetration loss 20 dB
eNB/UE transmit power 46 dBm/23 dBm
Small-scale fading model Rayleigh
Noise spectral density −174 dBm/Hz Fig. 5. Performance of average sum rate vs. SNR.
Network synchronization Synchronized

and SINR constraints based coordinated beamforming (SINR-


CBF) scheme [35]. Since CoMP is more beneficial to cell-edge
users, we emphasis on the performance results of cell edge
users. The average sum rate is found by solving Eq. (10)
and then taking average over 500 random user placements,
each subjected to 100 i.i.d. realizations of Rayleigh small
scale fading channels. As evident from the sum rate curves,
proposed SLNR based beamforming scheme offers improved
performance compared to its counterparts in high noise re-
gions. As expected, MC-ZFBF scheme has poor performance
in low SNR regime; MC-BDZF offers some improvement
in the low to medium SNR regimes. Typically, BD schemes
outperform ZF and even MMSE at high SNRs when the users
are equipped with multiple antennas as discussed earlier. How-
ever both schemes put restriction on number of the antenna
configurations. The coordinated beamforming provides opti-
Fig. 6. CDF of average user spectral efficiency.
mal solution theoretically but with only slight improvement in
sum rate performance. However, it requires iterative algorithms
for convex optimization problem, leading to high complexity. schemes are presented in Fig. 7 and it can be noted that pro-
Whereas the performance of proposed SLNR based scheme posed SLNR-CBF yields gains of 9.53%, 31.11% and 72.65
is very close to SINR-CBF schemes in almost entire region, in terms of average sum rate and gains of 14.29%, 69.01%
it provides additional benefit in terms of low complexity, and 83.83% in terms of average user spectral efficiency over
relaxes constraints imposed by antenna condition and provides SINR-CBF, MC-BDZF and MC-ZFBF schemes respectively.
superior performance than the conventional channel inversion
schemes in noise dominated environments.
We examine the effect on average user spectral efficiency of 5. Conclusion
various schemes, presented as CDF curves in Fig. 6. Note that
lower spectral efficiency region belongs to cell-edge users and In this article, we presented a robust coordinated beamform-
high spectral efficiency region relates to the cell-center users. ing scheme which is based upon the leakage based criteria. The
It can be observed that SINR scheme curve closely follows proposed scheme follows the SLNR metric to maximize the
proposed SLNR metric based schemes in both the cases. Even average sum rate and the average user throughput. The evalu-
though, proposed scheme has superiority over conventional ation results are analyzed and investigated in a three-eNB site
schemes in terms of both the parameters as mentioned above. scenario and compared with multi-cell beamforming schemes
Predominantly, the improvement in throughput gains are influ- namely—MC-ZFBF, MC-BDZF and SINR-CBF schemes. The
enced by resilience offered by SLNR based scheme in the high performance is validated through simulation results and it is
noise regimes. Additionally, it offers reduced computational observed that presented scheme outclasses its counterparts
cost relatively, as stated earlier. The evaluation results of all the and offers a low complexity beamforming solution for the
CS/CBF-CoMP based cellular networks.
7
S. Gulia and A. Ahmad ICT Express xxx (xxxx) xxx

[7] S. Schwarz, M. Rupp, Exploring coordinated multipoint beamforming


strategies for 5G cellular, IEEE Access 2 (2014) 930–946, http:
//dx.doi.org/10.1109/ACCESS.2014.2353137.
[8] 3GPP TR 36.819 v11.1.0, Evolved Universal Terrestrial Radio Access
(E-UTRA) and Evolved Universal Terrestrial Radio Access Network
(E-UTRAN); Coordinated multi- point operation for LTE physical
layer aspects (Release 11), 2011, http://www.3gpp.org/ftp/Specs/html-
info/36819.htm.
[9] P. Marsch, G. Fettweis, Coordinated Multi-Point in Mobile Communi-
cations: From Theory to Practice, Cambridge University Press, 2011,
http://dx.doi.org/10.1017/cbo9780511783029.
[10] M. Sawahashi, Y. Kishiyama, A. Morimoto, D. Nishikawa, M. Tanno,
Coordinated multipoint transmission/reception techniques for LTE-
advanced [Coordinated and Distributed MIMO], IEEE Wirel. Commun.
17 (2010) 26–34, http://dx.doi.org/10.1109/MWC.2010.5490976.
[11] C. Yang, S. Han, X. Hou, A. Molisch, How do we design CoMP
to achieve its promised potential? IEEE Wirel. Commun. 20 (2013)
Fig. 7. Performance of various beamforming schemes.
67–74, http://dx.doi.org/10.1109/MWC.2013.6472201.
[12] D. Lee, H. Seo, B. Clerckx, E. Hardouin, D. Mazzarese, S. Nagata,
K. Sayana, Coordinated multipoint transmission and reception in LTE-
CRediT authorship contribution statement advanced: Deployment scenarios and operational challenges, IEEE
Commun. Mag. 50 (2012) 148–155, http://dx.doi.org/10.1109/MCOM.
Sandeep Gulia: Conceptualization, Methodology, Investi- 2012.6146494.
gation, Software, Writing – original draft, Writing – review [13] R. Irmer, H. Droste, P. Marsch, M. Grieger, G. Fettweis, S. Brueck,
& editing, Software, Data curation. Anwar Ahmad: Concep- H.P. Mayer, L. Thiele, V. Jungnickel, Coordinated multipoint: Con-
cepts, performance, and field trial results, IEEE Commun. Mag. 49
tualization, Supervision, Visualization, Validation, Writing – (2011) 102–111, http://dx.doi.org/10.1109/MCOM.2011.5706317.
review & editing. [14] T.K.Y. Lo, Maximum ratio transmission, IEEE Trans. Commun. 47
(1999) 1458–1461, http://dx.doi.org/10.1109/26.795811.
[15] M. Joham, W. Utschick, J.A. Nossek, Linear transmit processing
Declaration of competing interest in MIMO communications systems, IEEE Trans. Signal Process. 53
(2005) 2700–2712, http://dx.doi.org/10.1109/TSP.2005.850331.
The authors declare that they have no known competing
[16] T. Yoo, A. Goldsmith, On the optimality of multiantenna broad-
financial interests or personal relationships that could have cast scheduling using zero-forcing beamforming, IEEE J. Sel. Areas
appeared to influence the work reported in this paper. Commun. 24 (2006) 528–541, http://dx.doi.org/10.1109/JSAC.2005.
862421.
[17] E. Bjornson, M. Bengtsson, B. Ottersten, Optimal multiuser transmit
Acknowledgments beamforming: A difficult problem with a simple solution structure
[Lecture Notes], IEEE Signal Process. Mag. 31 (2014) 142–148,
The work in this article was supported by the Ministry of
http://dx.doi.org/10.1109/MSP.2014.2312183.
Electronics and Information Technology (MeitY), Government [18] M. Schubert, H. Boche, Solution of the multiuser downlink beam-
of India under the Visvesvaraya Ph.D. Scheme. forming problem with individual SINR constraints, IEEE Trans. Veh.
Technol. 53 (2004) 18–28, http://dx.doi.org/10.1109/TVT.2003.819629.
[19] B.M. Hochwald, C.B. Peel, A.L. Swindlehurst, A vector-perturbation
References technique for near-capacity multiantenna multiuser communication -
[1] A.S. Hamza, S.S. Khalifa, H.S. Hamza, K. Elsayed, A survey on Part II: Perturbation, IEEE Trans. Commun. 53 (2005) 537–544,
inter-cell interference coordination techniques in ofdma-based cellular http://dx.doi.org/10.1109/TCOMM.2004.841997.
networks, IEEE Commun. Surv. Tutor. 15 (2013) 1642–1670, http: [20] H. Sampath, P. Stoica, A. Paulraj, Generalized linear precoder and
//dx.doi.org/10.1109/SURV.2013.013013.00028. decoder design for MIMO channels using the weighted MMSE
criterion, IEEE Trans. Commun. 49 (2001) 2198–2206, http://dx.doi.
[2] G. Boudreau, J. Panicker, N. Guo, R. Chang, N. Wang, S. Vrzic,
org/10.1109/26.974266.
Interference coordination and cancellation for 4G networks, IEEE
[21] S. Shi, M. Schubert, H. Boche, Downlink MMSE transceiver op-
Commun. Mag. 47 (2009) 74–81, http://dx.doi.org/10.1109/MCOM.
timization for multiuser MIMO systems: Duality and sum-MSE
2009.4907410.
minimization, IEEE Trans. Signal Process. 55 (2007) 5436–5446,
[3] D. López-Pérez, I. Güvenç, G. De La Roche, M. Kountouris, T.Q.S.
http://dx.doi.org/10.1109/TSP.2007.899283.
Quek, J. Zhang, Enhanced intercell interference coordination chal- [22] Q.H. Spencer, A.L. Swindlehurst, M. Haardt, Zero-forcing methods
lenges in heterogeneous networks, IEEE Wirel. Commun. 18 (2011) for downlink spatial multiplexing in multiuser MIMO channels, IEEE
22–30, http://dx.doi.org/10.1109/MWC.2011.5876497. Trans. Signal Process. 52 (2004) 461–471, http://dx.doi.org/10.1109/
[4] 3GPP R1-113559, Performance evaluation for FeICIC, 2011, TSP.2003.821107.
https://www.3gpp.org/ftp/tsg_ran/WG1_RL1/TSGR1_66b/Docs/R1- [23] L.U. Choi, R.D. Murch, A transmit preprocessing technique for
113559.zip. multiuser MIMO systems using a decomposition approach, IEEE
[5] 3GPP TR 36.741 v 14.0.0, Study on further enhancements to Trans. Wirel. Commun. 3 (2004) 20–24, http://dx.doi.org/10.1109/
Coordinated Multi-Point (CoMP) operation for LTE (Release 14), TWC.2003.821148.
2017. [24] H. Sung, S.R. Lee, I. Lee, Generalized channel inversion methods
[6] V. Jungnickel, K. Manolakis, W. Zirwas, B. Panzner, V. Braun, for multiuser MIMO systems, IEEE Trans. Commun. 57 (2009)
M. Lossow, M. Sternad, R. Apelfröjd, T. Svensson, The role of 3489–3499, http://dx.doi.org/10.1109/TCOMM.2009.11.070404.
small cells, coordinated multipoint, and massive MIMO in 5G, IEEE [25] V. Stankovic, M. Haardt, Generalized design of multi-user MIMO
Commun. Mag. 52 (2014) 44–51, http://dx.doi.org/10.1109/MCOM. precoding matrices, IEEE Trans. Wirel. Commun. 7 (2008) 953–961,
2014.6815892. http://dx.doi.org/10.1109/LCOMM.2008.060709.
8
S. Gulia and A. Ahmad ICT Express xxx (xxxx) xxx

[26] M.H.M. Costa, Writing on dirty paper, IEEE Trans. Inform. Theory [38] 3GPP TR 36.874 v 12.0.0, Coordinated multi-point operation for LTE
29 (1983) 439–441, http://dx.doi.org/10.1109/TIT.1983.1056659. with non-ideal backhaul (Release 12), 2013.
[27] C. Windpassinger, R.F.H. Fischer, T. Vencel, J.B. Huber, Precoding [39] M.S. Ali, E. Hossain, A. Al-Dweik, D.I. Kim, Downlink power
in multiantenna and multiuser communications, IEEE Trans. Wirel. allocation for CoMP-NOMA in multi-cell networks, IEEE Trans.
Commun. 3 (2004) 1305–1316, http://dx.doi.org/10.1109/TWC.2004. Commun. 66 (2018) 3982–3998, http://dx.doi.org/10.1109/TCOMM.
830852. 2018.2831206.
[28] S. Shamai, B.M. Zaidel, Enhancing the cellular downlink capacity via [40] M. Elhattab, M.A. Arfaoui, C. Assi, Comp transmission in downlink
co-processing at the transmitting end, in: IEEE Veh. Technol. Conf., NOMA-based heterogeneous cloud radio access networks, IEEE Trans.
2001, pp. 1745–1749, http://dx.doi.org/10.1109/vetecs.2001.944993. Commun. 68 (2020) 7779–7794, http://dx.doi.org/10.1109/TCOMM.
[29] D. Gesbert, S. Hanly, H. Huang, S. Shamai Shitz, O. Simeone, W. Yu, 2020.3021145.
Multi-cell MIMO cooperative networks: A new look at interference, [41] Y. Al-Eryani, E. Hossain, D.I. Kim, Generalized coordinated mul-
IEEE J. Sel. Areas Commun. 28 (2010) 1380–1408, http://dx.doi.org/ tipoint (GCoMP)-enabled NOMA: Outage, capacity, and power
10.1109/JSAC.2010.101202. allocation, IEEE Trans. Commun. 67 (2019) 7923–7936, http://dx.doi.
[30] J.G. Andrews, W. Choi, R.W. Heath, Overcoming interference in org/10.1109/TCOMM.2019.2931971.
spatial multiplexing mimo cellular networks, IEEE Wirel. Commun. [42] I. Ahmed, H. Khammari, A. Shahid, A. Musa, K.S. Kim, E.
14 (2007) 95–104, http://dx.doi.org/10.1109/MWC.2007.4407232. De Poorter, I. Moerman, A survey on hybrid beamforming techniques
[31] H. Huh, A.M. Tulino, G. Caire, Network MIMO with linear zero- in 5G: Architecture and system model perspectives, IEEE Commun.
forcing beamforming: Large system analysis, impact of channel Surv. Tutor. 20 (2018) 3060–3097, http://dx.doi.org/10.1109/COMST.
2018.2843719.
estimation, and reduced-complexity scheduling, IEEE Trans. In-
[43] M. Sadek, A. Tarighat, A.H. Sayed, A leakage-based precoding
form. Theory 58 (2012) 2911–2934, http://dx.doi.org/10.1109/TIT.
scheme for downlink multi-user MIMO channels, IEEE Trans. Wirel.
2011.2178230. Commun. 6 (2007) 1711–1721, http://dx.doi.org/10.1109/TWC.2007.
[32] M. Kemal Karakayali, G.J. Foschini, R.A. Valenzuela, Network coor- 360373.
dination for spectrally efficient communications in cellular systems, [44] P. Cheng, M. Tao, W. Zhang, A new SLNR-based linear precoding
IEEE Wirel. Commun. 13 (2006) 56–61, http://dx.doi.org/10.1109/ for downlink multi-user multi-stream MIMO systems, IEEE Commun.
MWC.2006.1678166. Lett. 14 (2010) 1008–1010, http://dx.doi.org/10.1109/LCOMM.2010.
[33] O. Somekh, O. Simeone, Y. Bar-Ness, A.M. Haimovich, S. Shamai, 091710.100868.
Cooperative multicell zero-forcing beamforming in cellular downlink [45] H. Du, P.J. Chung, A probabilistic approach for robust leakage-
channels, IEEE Trans. Inform. Theory 55 (2009) 3206–3219, http: based MU-MIMO downlink beamforming with imperfect channel state
//dx.doi.org/10.1109/TIT.2009.2021371. information, IEEE Trans. Wirel. Commun. 11 (2012) 1239–1247,
[34] R. Zhang, Cooperative multi-cell block diagonalization with per-base- http://dx.doi.org/10.1109/TWC.2012.011012.111260.
station power constraints, IEEE J. Sel. Areas Commun. 28 (2010) [46] S. Feng, M.M. Wang, W. Yaxi, F. Haiqiang, L. Jinhui, An efficient
1435–1445, http://dx.doi.org/10.1109/JSAC.2010.101205. power allocation scheme for leakage-based precoding in multi-
[35] H. Dahrouj, W. Yu, Coordinated beamforming for the multicell multi- cell multiuser MIMO downlink, IEEE Commun. Lett. 15 (2011)
antenna wireless system, IEEE Trans. Wirel. Commun. 9 (2010) 1053–1055, http://dx.doi.org/10.1109/LCOMM.2011.081011.110803.
1748–1759, http://dx.doi.org/10.1109/TWC.2010.05.090936. [47] S. Park, J. Park, A. Yazdan, R.W. Heath, Exploiting spatial channel
[36] W. Shin, M. Vaezi, B. Lee, D.J. Love, J. Lee, H.V. Poor, Coordinated covariance for hybrid precoding in massive MIMO systems, IEEE
beamforming for multi-cell MIMO-NOMA, IEEE Commun. Lett. 21 Trans. Signal Process. 65 (2017) 3818–3832, http://dx.doi.org/10.1109/
TSP.2017.2701321.
(2017) 84–87, http://dx.doi.org/10.1109/LCOMM.2016.2615097.
[48] D. Kim, G. Lee, Y. Sung, Two-stage beamformer design for massive
[37] E. Ali, M. Ismail, R. Nordin, N.F. Abdulah, Beamforming techniques
MIMO downlink by trace quotient formulation, IEEE Trans. Com-
for massive MIMO systems in 5G: overview, classification, and trends mun. 63 (2015) 2200–2211, http://dx.doi.org/10.1109/TCOMM.2015.
for future research, Front. Inf. Technol. Electron. Eng. 18 (2017) 2429646.
753–772, http://dx.doi.org/10.1631/FITEE.1601817.

You might also like