Unified Channel Estimation and Scheduling For Massive MIMO Systems

ISSN (Online) 2394-6849
International Journal of Engineering Research in Electronics and Communication

Engineering (IJERECE)
Vol 5, Issue 4, April 2018
Unified Channel Estimation and Scheduling for

Massive MIMO Systems
[1]
Dr G.Indumathi, [2] C.Prabhabathi Devi, [3] V.Priyadharshini
Department of ECE, Mepco Schlenk Engineering College, Sivakasi.
Abstract: - This paper proposes a unified channel estimation scheme for multiuser massive multiple input multiple output (MIMO)
systems in time-varying environment. In this paper, a new discrete Fourier transform (DFT) based spatial-temporal basis
expansion model (ST-BEM) is introduced to mitigate the training overhead and feedback cost by reducing the dimensions of uplink
and downlink channel. This model is suitable for both time division duplex (TDD) and frequency division duplex (FDD) systems. A
new greedy user scheduling algorithm is also introduced to improve the Spectral efficiency. Various simulation results are provided
to demonstrate the effectiveness of the proposed method.
Keywords: - Massive MIMO, Discrete Fourier Transform (DFT), ST-BEM, FDD, TDD.
I. INTRODUCTION Strategy is used. Here uplink pilot contamination,

downlink training and feedback overhead are significantly
One of the most important physical layer techniques in 5G reduced. But acquisition of channel covariance matrix
communication is massive MIMO or large scale multiple becomes a difficult task and it also leads to Eigen value
input and multiple output [8]. It can simultaneously serve decomposition (EVD) problem for high dimensional
tens of terminals in the same frequency-time with the help of covariance matrices. In this paper, using antenna array
hundreds of antennas located in Base Station (BS) side that theory and array signal processing we propose an alternate
provides robustness, high Energy efficiency and Spectral low rank model for massive Uniform linear array (ULA).
efficiency [9]. To utilize the benefits of massive MIMO This model is based on mean direction of arrivals and
effectively, the perfect channel state information (CSI) for angular spread (AS) of incident signals of each user. This
both Uplink and downlink should be acquired by Base model is known as Spatial Basis Expansion Model
station. Usually channel is estimated via pilot sequences. But (SBEM). The proposed unified transmission strategy for
in massive MIMO it leads to pilot overhead in downlink the multiuser TDD/FDD massive MIMO systems includes
channel estimation because the no of training sequences Uplink (UL) channel estimation and user scheduling for
should be equal to the no of transmit antennas and length of data transmission. It is shown that the Uplink channel
the training stream should be greater than the no of transmit estimation of multi users can be carried out with very few
antennas based on orthogonal training strategy. In uplink if training resources, and thus, the overhead of training and
the no of users or no of antennas of each user increases, then feedback can be also reduced significantly. Meanwhile,
pilot overhead problem occurs. If the training sequences are the pilot contamination in UL training can be immediately
Non-orthogonal, then so called pilot contamination problem relieved. To enhance the spectral efficiency during the
occurs. These problems are diminishing the system data transmission, a greedy user scheduling algorithm is
performance. In [10] and [11], the closed-loop training proposed where users with orthogonal spatial information
schemes were applied to sequentially design the optimal pilot are allowed to transmit simultaneously.
beam patterns. The Compressive sensing (CS)-based
feedback reduction in [12] and the distributed compressive II. SYSTEM MODEL AND CHANNEL
channel estimation in [3] extracted the channel sparsity to CHARACTERISTICS
reduce large amount of measurements feeding back to the
Base station. This Method requires knowledge of the sparsity As shown in figure.1 multiuser massive MIMO system is
level in channel matrices but it’s not an easy task to considered where the Base station (BS) is equipped with
accurately acquire such information. After these types of M (>>1) antennas in the form of uniform linear array
attempts low rank channel estimation approaches are (ULA) and K single antenna users are spread over the
introduced. These approaches reduce the effective channel coverage area. The propagation from user k to BS is
dimensions. In [1], Covariance aware pilot Assignment composed of Pr-rays (Pr>>1) due to scattering, reflection
All Rights Reserved © 2018 IJERECE 12


and refraction. The channel is considered to be time selective The normalized DFT of channel vector exhibits some
flat fading and it will change symbol level. It’s safe to specific properties in massive ULA case. This property
assume that the physical position of users seen by the BS is used to minimize the complexity of transceiver design.
unchanged within a single block N. The corresponding M x 1 The normalized DFT can be represented as
uplink channel can be expressed as,
Pr
1

 kp where, F=>MxM DFT matrix whose (p,q)th element is,
hk  kp e a( kp ) (1)
pr p 1
The propagation from user to BS is assumed to be

composed of p rays. While considering single ray i.e.,
hk   k a( k ) and M is infinite, the normalized DFT
has only one non-zero point and this point reflects the
DOA of the impinging signal, namely

When M is large but not infinite, the power leakage may
happen. For most cases, Non-zero point is not an integer
and the channel power will leak from the ([M(d/λ)sinθ k])th
DFT point to other DFT points. The DFT outputs are
discrete samples of DTFT of a (θk).i.e., a sinc function, at
the points of 2πq/M, q=0, 1 ….M-1. Hence the degree of
leakage in fhk is inversely proportional to the M. When
M is infinite, the large but not power leakage is not a
problem because most power of fhk is still concentrates
Fig.1 System model. Users are surrounded by pr local around this Non-zero point as shown in Fig.2. For the Multi-ray
scatterers and the mean DOA and AS f user-k are case, let Dk is the index set of continuous DFT points that
 k and  k respectively. When users are move around contains  % of the channel power [7].
in a circle, the spatial AS seen by BS is generally
unchanged.
where, fd is the Doppler frequency and Ts is the sampling

frequency,  kp is the angle between the uplink transmitted
signal and the motion direction of user-k. kp signifies the
initial phase, which is uniformly distributed in [0,2 ]
and a( kp ) is the array manifold vector and it varies based
on antenna structure.
For ULA case, array manifold vector can be represented as
2d 2d
j sin  kp j ( M 1) sin  kp
a(kp )  [1, e 
,.........e 
] (2)
where, d is the antenna spacing, λ denotes the signal
carrier wavelength, θkp represents DOA of the p-th ray. Fig.2 fhk of single incident ray with =90, M=128
The incident angular spread of user k with mean DOA It can be represented as,
(  k ) is assumed to be limited in a narrow region, i.e.,
[  k   k , k   k ] and this angular spread (AS) of
each user is normally unaltered when the user moves Where, θk ,Δθk are DOA and AS respectively. Cmax is the
within the circular region as shown in Fig.1. maximum no of leakage points. These indices can be
viewed as spatial signature of each user. This model is
known as Spatial Basis Expansion Model (SBEM). To


deduce this, a perfect DOA is to be acquired. For this The preceding optimization can be achieved by sliding a
purpose, a Spatial Rotation operation is introduced. So the window of size τ over the elements in
new channel vector can be formulated as,
together by a 1-D search over
III. CHANNEL ESTIMATION WITH SBEM

where,
A unified transmission strategy for a TDD/FDD massive
MIMO system is introduced that utilizes the spatial
The Spatial Rotation operation, further concentrate the
signatures to realize orthogonal training and data
channel power within the fewer entries fhkro of for a certain
transmission among different users. As shown in Fig.4,
value of ϕ. For one-ray case, When DOA of the incident
this framework always starts with preamble period. The
signal is not arcsin(qM/λd) for some integer q i.e.,
preamble period is used to obtain the spatial signature of
mismatched with DFT points, then power leakage will
each user. After tracking spatial information of each user
happen. Formulate a new channel vector as, fh ro  F ( )h .
k k the users are grouped based on their spatial signature for
If then has only one Non-zero further process.
element and so at q, with the power leakage is

eliminated, where ϕ is the shift parameter. For example in
Fig.3, Spatial rotation with ϕk= 0.34375 radian can help to
strengthen the channel sparsity of. . For the Multi-ray
cases, a new channel vectors is formulated and define
as the continuous index set such that Fig.4 Communication Framework
contains at least ŋ% of the channel power and search ϕ from Here we consider the users are present within a cell and τ
(< K) orthogonal training sequences with length L (< T)
and select the optimal ϕk that minimizes the are available and the corresponding orthogonal training
The standards are normally regulated by considering channel set will be S=[s1,s2,.....sτ] € CLxτ with SiHSj=Lσp2δ(i-j)
parameters to be estimated as fixed but here a dynamically where σp2 is the signal training power.
changing parameter is considered. Define the set containing
continuous τ integers as where τ is the number of Tracking of Spatial information through preamble Assume
channel parameters that system could handle. Then select the K=Gτ. Since we do not have any prior spatial information
spatial signature set and corresponding shift parameter for about users, we will have to divide the users into G groups
user k by using following optimization, each containing τ users such that τ orthogonal training
sequences are enough for each group and conventional
estimation methods are applied and length of the preamble
is Gτ.
The received signals at the BS is given by,
hk can be estimated using Least Square as,
where, H=[h1,h2,…….hτ] € CMxτ , S=[s1,s2,.....sτ] € CLxτ and

Pkut
D=diag[d1,……dτ] and dk  is used to satisfy the
 p2
uplink training energy constraint for user-k, N is the
Fig.3 Comparison of single ray channels with/without additive white Gaussian noise. Repeating the similar
spatial rotation operations in (7) for all G groups is used to obtain the


estimates of all K users. The next step is to obtain the optimal
shift parameter ϕk and spatial signature set of size τ for each
user, as described in (5). The obtained channel information
from preamble may only last for a short period for example
one coherent time while it needs to be re estimated or
tracked for later transmission. When a user and its
surrounding obstacles does not change the position within the
comparable time then no need to track the spatial information
for few channel coherent times and only the accompanied
[ fhkro ] B ro ,: should be re estimated. The Non-overlapping
k
properties of different spatial signatures are used to

overcome the insufficient problem of orthogonal training
sequences. After obtaining of spatial signature of each user
we may schedule them using Non-overlapping property of
spatial signatures and also throughput is considered to
improve the performance of each group.
IV. DATA TRANSMISSION WITH USER

SECHEDULING
A greedy user scheduling algorithm is considered where the V. SIMULATIONS

strongest channel gain first join the empty group and then
other users with non-overlapping spatial signatures can join
the same group only if the achievable sum rate increases In this section the effectiveness of the proposed strategy is
afterwards. demonstrated using numerical examples. We select
________________________________________ M=128, K=32,d=λ/2. The channel vectors of different
User scheduling Algorithm users are formulated using (1) and we consider Pr=100,
STEP-1:Calculate the Euclidean norm of the estimated fd=200 Hz and Ts=1μs and  kp is randomly taken from
channel vectors, i.e., [ fhkro ]B ro ,: for all users. CN(0,1) for all rays and all users,  kp is distributed
k
dd
STEP-2:Initialize g=1, P=0, Ug = NULL, R(Ug |P)= 0 and the dd
inside [ k   k , k   k ] . And AS = 2 degree for
remaining user set Ur={1,…….K}.
all users. The system coherence interval is set as T=128
STEP-3:For the gth group, select the user with the maximum norm
of channel in user set l '  arg max and τ =16. The length of the pilot symbol should satisfy
lU r
ehl and remove the 16≤L≤128. L may be 16, 32 and 64. The signal to noise
user
from user set and add the user in Ugdd.  p2
STEP-4:Select all users Ur whose spatial signatures are Non- ratio defined as,  . The channel estimation
overlapping with users in Ugdd and denote them by, Ug’. For Eg.,  n2
u' g  {mur Bmro  Blro   , l U gdd } performance metric in terms Mean square error(MSE) is
represented as
Fig.5 shows comparison of MSE performances with

different L values. It is observed that when L increases the
MSE decreases. The total power for uplink training
considered is Pkut= Lρ for all users.


VI. CONCLUSIONS
In this paper we exploited the characteristics of ULA and

proposed a simple DFT based ST-BEM to represent the
channel vectors with reduced parameters which helps to
reduce the pilot overhead and feedback cost. The Uplink
spatial signatures can be used for Downlink also based on
the property angle reciprocity. This method applicable for
both TDD/FDD massive-MIMO systems. Various
numerical results are provided to demonstrate the
effectiveness of the proposed approach.
REFERENCES
Fig.5 MSE performances for different L values
[1] A.Adhikary, J. Nam, J.-Y. Ahn, and G. Caire, ―Joint
spatial division and multiplexing— the large-scale array
Fig.6 explains about MSE performance improvement with
regime,‖ IEEE Trans. Inf.Theory,vol. 59, no. 10, pp.
and without spatial rotation operation.
6441–6463, Oct. 2013.
[2] H. Yin, D. Gesbert, M. Filippou, and Y. Liu, ―A

coordinated approach to channel estimation in large-scale
multiple-antenna systems,‖ IEEE J. Sel.Areas Commun.,
vol. 31, no. 2, pp. 264–273, Feb. 2013.
[3] X. Rao and V. K. Lau, ―Distributed compressive CSIT

estimation and feedback for FDD multi-user massive
MIMO systems,‖ IEEE Trans.Signal Process., vol. 62, no.
12, pp. 3261–3271, Jun. 2014.
[4] Hongxiang Xie, Feifei Gao, Shun Zhang, and Shi Jin,
―A Unified Transmission Strategy for TDD/FDD Massive
MIMO Systems With Spatial Basis Expansion Model,‖
Fig.6 MSE performances for with and without spatial IEEE Transactions on vehicular technology, vol. 66, no. 4,
rotation Apr 2017.
[5] G. B. Giannakis and C. Tepedelenlioglu, ―Basis

expansion models and diversity techniques for blind
identification and equalization of time-varying channels,‖
Proc. IEEE, vol. 86, no. 10, pp. 1969–1986,Oct. 1998.
[6] S. P. Boyd and L. Vandenberghe, Convex

Optimization. Cambridge,MA, USA: Cambridge Univ.
Press, 2004.
[7] Hongxiang Xie, Feifei Gao, Shun Zhang, and Shi Jin,
―A Simple DFT-aided Spatial Basis Expansion Model and
Channel Estimation Strategy for TDD/FDD Massive
Fig.7 AASR of the proposed scheme and conventional MIMO Systems,‖ arXiv:1511.04841v6 [cs.IT] 9 Jan 20.
LS From Fig.7, it is observed that the AASR of proposed
SBEM is providing better performance than conventional [8] F. Rusek et al., ―Scaling up MIMO: Opportunities and
Least square (LS) method. challenges with very large arrays,‖ IEEE Signal Process.
Mag., vol. 30, no. 1, pp. 40–60, Jan. 2013.


[9] E. Larsson, O. Edfors, F. Tufvesson, and T. Marzetta,

―Massive MIMO for next massive MIMO systems: Open-
loop and closed-loop training with memory,‖ IEEE J. Sel.
Topics Signal Process., vol. 8, no. 5, pp. 802–814, Oct. 2014.
[10] J. Choi, D. J. Love, and P. Bidigare, ―Downlink training

techniques for FDD massive MIMO systems: Open-loop and
closed-loop training with memory,‖ IEEE J. Sel. Topics
Signal Process., vol. 8, no. 5, pp. 802–814, Oct. 2014.
[11] A. Duly, T. Kim, D. Love, and J. Krogmeier, ―Closed-

loop beam alignment for massive MIMO channel
estimation,‖ IEEE Commun. Lett., vol. 18, no. 8, pp. 1439–
1442, Aug. 2014.
[12] P. H. Kuo, H. Kung, and P. A. Ting, ―Compressive

sensing based channel feedback protocols for spatially-
correlated massive antenna arrays,‖ in Proc. IEEE WCNC,
Shanghai, China, Apr. 2012, pp. 492–497.

Unified Channel Estimation and Scheduling For Massive MIMO Systems

Uploaded by

Copyright:

Available Formats

Unified Channel Estimation and Scheduling For Massive MIMO Systems

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Unified Channel Estimation and Scheduling For Massive MIMO Systems

Uploaded by

Copyright:

Available Formats

ISSN (Online) 2394-6849

International Journal of Engineering Research in Electronics and Communication