Beruflich Dokumente
Kultur Dokumente
fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2495163, IEEE Transactions on Wireless Communications
1
School of Electrical and Computer Engineering Purdue University, West Lafayette IN 47906
Email: aelghari@purdue.edu and mikedz@purdue.edu
I. I NTRODUCTION
A large-scale multi-input multi-output (MIMO) (or a socalled Massive MIMO) system in which a large number of
antennas is used at the transmitter and/or receiver is one
of the main components of the future 5G wireless communication systems [2]. The capacity of this MIMO system
can be scaled up by installing more antennas at the transmitter and/or receiver to fulfill the demands for high data
rate applications [3], [4], [5]. The interest in these systems
poses challenges in several design aspects, such as channel
estimation, antenna correlation, hardware implementation, and
detection complexity [6],[5]. In particular, a critical design
challenge in a large-scale MIMO system is to design a reliable
and computationally efficient detector even if the number of
antennas grows very large or the modulation order increases.
There have been many linear detectors and near-Maximum
Likelihood detectors proposed in the literature of conventional
MIMO systems; however, they become noncompetitive when
A preliminary version of this work was presented in IEEE WCNC 2015 [1],
in which only one algorithm is considered. In this paper, further algorithms
are considered with extensive simulation results.
1536-1276 (c) 2015 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2495163, IEEE Transactions on Wireless Communications
2
x k22
x = argmin k
y H
where
Nt is the set of all possible Nt -dimensional complex
candidate vectors of the transmitted vector x. The equivalent
real-valued model of (1) is:
y = Hx + n
"
y=
<{
y}
={
y}
"
,x =
<{
x}
={
x}
(1)
"
,n =
}
<{n
}
={n
(3)
#
"
}
<{H
,H =
}
={H
}
={H
}
<{H
(4)
x t
(2)
x
Nt
argmin
x + n
y = H
where y = [
y1 , . . . , yNr ]T CNr 1 is the received signal
CNr Nt denotes the flat
vector at all Nr antennas, H
fading channel gain matrix whose entries are modeled as
represents the receiver AWGN noise vector
CN (0, 1), and n
whose entries are modeled as i.i.d CN (0, 2 ). A more realistic
MIMO channel will be considered later in section IV. The tilde
symbol in (1) is made to distinguish the complex model from
the real model shown in the next section. We assume ideal
channel estimation and synchronization at the receiver end.
subject to
1 T
z Q z + bT z
2
0 z ( C 1)1
(6)
1536-1276 (c) 2015 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2495163, IEEE Transactions on Wireless Communications
3
i = 1, 2, . . . , 2Nt
(7)
where Q[.] is a quantization function to the appropriate constellation levels of the set . In the next subsections, we propose
improvements to the QP detector in a large MIMO system
through performing further analysis to the problem (6) using
first, two-stage QP detection with interference cancellation,
and second, the concept of the BB search tree [32], [25],
[33]. It is worth noting that in the previous work [22],
[23], a randomization rounding technique is shown to provide
better performance than simple rounding (as in (7)), but with
additional complexity of the order O(Nt2 ). This technique,
however, can still be used with any of our proposed algorithms.
x|J |
subject to
1 T
T z
z Q z + b
2
z ( C 1)
0
1
(9)
= H
TH
, b
= H
T (
where Q
y + ( C 1)H
1)/2, and
1 is a
column vector of length (2Nt |I|) and |I| is the cardinality
and b,
we further
of the set I. In order to avoid recomputing Q
simplify them in terms of Q and b as follows:
= Q(J , J )
Q
= Q(I, J )T z(I) + b(J )
b
(10)
1536-1276 (c) 2015 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2495163, IEEE Transactions on Wireless Communications
4
the subtree below this node. The initial value of the upper
bound can be taken as a very large value, such as , or can
be computed from any available integer solution, such as ZF or
MMSE solutions. And 2) as mentioned above, if the solution
of any node satisfies the integer constraints, then no branching
is needed and the node is pruned.
In this paper, we focus on the Breadth First (BF) search
strategy [36], where the nodes of the tree are explored level
by level as dipicted in Fig. 1. We prefer this strategy because
it suits well our proposed approximation herein. In general,
Input: Q, b
z = quadprog(Q, b) from (6)
Find I that satisfies k z bz e k
z(I) = Q[z (I)]
Find set of indices J
= Q(J , J ), and
Find Q
= Q(I, J )T z(I) + b(J )
b
b)
from (9)
z (J ) = quadprog(Q,
z(J ) = Q[z (J )]
subject to
argmin
z
subject to
1 T
z Q z + bT z
2
j
k
(0)
,
0 z ( C 1)1, and zi zi
(11)
1 T
z Q z + bT z
2
l
m
(0)
0 z ( C 1)1, and zi zi
(12)
where zi is called
variable at index i (0
j
kthel branching
m
(0)
(0)
i 2Nt ), and zi
( zi
) denotes the largest (smallest)
integer smaller (greater) than or equal to zi(0) . There are
various strategies for choosing the branching variable [35],
but in this paper we choose the simplest one, which branches
a node at the first non integr variable. Now solving these
new subproblems again using the interior-point algorithm,
returns (z(1) , f (z(1) )) and (z(2) , f (z(2) )) for nodes 1 and
2, respectively. If the solutions to these subproblems do not
satisfy the integer constraints, each of them will be branched
into two more subproblems and the process of branching will
continue until the optimal integer solution is found, see more
details on [33] and [35]. Two important pruning rules are used
with the BB algorithm: 1) for any node in the tree, whenever its
cost function value is greater than a known upper bound f (up) ,
this node is pruned because no better solution is expected from
(13)
1536-1276 (c) 2015 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2495163, IEEE Transactions on Wireless Communications
5
Q[z(k) ]
z(k)
)
(14)
E. Complexity Analysis
The main ingredient of the computations in the QP detector
is the interior-point algorithm, which finds a point where
the Karush-Kuhn-Tucker (KKT) conditions hold for the optimization problem (6) in an iterative manner. As shown in
[30] and [39], each iteration of the interior-point algorithm
boils down to solving a system of linear equations where it
is required to perform a matrix inversion in every iteration.
Therefore, the complexity of one interior-point iteration is in
the order of O(Nt3 ), and becomes O(nNt3 ) for n iterations. In
practice, the interior-point converges in a number of iterations
which is almost always a constant, independent of the problem
dimension [31]. This is very attractive in high dimensional
optimization problems. From our simulation experiments, we
found that when using the standard interior-point algorithm,
the average number of iterations required for various number
1536-1276 (c) 2015 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2495163, IEEE Transactions on Wireless Communications
6
1536-1276 (c) 2015 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2495163, IEEE Transactions on Wireless Communications
7
10
BER
QP
2QP (=0.1)
2QP (=0.4)
4
10
2QP (=0.2)
2QP (=0.3)
2QP (=0.25)
10
12
14
15
25
=0.1
=0.2
=0.25
=0.3
=0.4
20
15
Blue color : 16
QAM
Black color:
QPSK
10
BER
10
Nt = 32, SNR = 21 dB
Nt = 64, SNR = 21 dB
10
Nt = 64, SNR = 15 dB
Nt = 32, SNR = 15 dB
Nt = 20, SNR = 15 dB
Nt = 64, SNR = 10 dB
Nt =32, SNR = 39 dB
0
0
10
2
10
12
(a)
14 15
Nt = 32, SNR = 49 dB
0.1
0.2
(b)
0.3
0.4
0.5
(c)
Fig. 2 A Two-stage QP detector (a) QPSK BER performance (b) Problem size of the second stage (c) QAM BER performance vs.
32x32 MIMO, QP Detector
10
10
QP Detector
Standard IP algorithm
Avg.IP iter=2
Avg.IP iter=1
Standard IP algorithm
Avg.IP iter=4
Avg.IP iter=3
Twostage QP
with =0.25
QPSK
2
QPSK
10
10
16 QAM
BER
BER
16 QAM
10
Standard IP algorithm
Avg. IP iter =2
Avg. IP iter =1
Standard IP algorithm
Avg. IP iter =4
Avg. IP iter =3
10
10
15
20
25
(a)
10
15
20
25
(b)
Fig. 3 The effect of reducing interior-point iters. on the BER performance in a 32 32 MIMO system (a) QP Detector (b) Two-stage QP
detector. Standard IP is the standard interior-point algorithm
the other hand, from Figs. 4b, c, d where the modulation level
increases, QP, 2QP, and BB(L,M) steadily become superior to
RTS and LAS at all displayed SNRs. For example, in Fig.
4d, the QP detector, which provides an upper bound BER to
2QP and BB(L,M), provides 5 dB improvements over MMSELAS and 3 dB improvements over RTS at 102 BER. The
performance of RTS was improved using a hybrid of RTS and
Belief Propagation (RTS-BP) in [40], but this only achieved
a 1.6 dB improvement at 103 BER with 16QAM (see Fig.
3 in [40]), while our algorithms 2QP and BB(32,4) provide
improvements of 2 dB and 3 dB over RTS, respectively. It is
worth noting that the performance of our proposed algorithms
can be further improved when combined with the LAS or RTS
algorithms, by making the starting initial vector of LAS or
RTS to be the vector results from QP, 2QP, or BB(L,M). The
simulation results for this claim are not extensively shown
here, but two examples for QP with LAS using QPSK, and
BB(32,4) with LAS using 16QAM are shown in Figs. 4a and
5, respectively.
Figs. 6a, b, c, d present a sample of complexity computations in terms of the average number of real operations versus
Nt measured at relatively low SNR and relatively high SNR
for both 16QAM and 256QAM. The important observations
from these figures are as follows: (i) The complexity of QP
and 2QP are almost similar with the advantage of 2QP for its
superior performance. (ii) There is no significant increase in
the computational complexity of the QP and 2QP detectors
1536-1276 (c) 2015 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2495163, IEEE Transactions on Wireless Communications
8
10
10
10
10
10
10
10
10
10
BER
BER
10
MMSE
MMSELAS
3 MIVLAS
QP
BB(8,4)
BB(16,2)
2QP (=0.25)
QPLAS
RTS
SISOAWGN
MMSE
MMSELAS
3 MIVLAS
RTS
QP
BB(4,4)
BB(16,2)
2QP (=0.25)
BB(32,4)
SISOAWGN
10
10
10
12
14
Average received SNR
12
16
14
16
(a)
10
10
10
BER
10
BER
10
10
28
10
10
26
(b)
18
20
22
24
Avgerage received SNR (dB)
20
MMSE
MMSELAS
3 MIVLAS
RTS
QP
2QP ( =0.25)
BB(4,4)
BB(16,2)
BB(32,8)
10
MMSE
MMSELAS
RTS
QP
2QP (=0.25)
BB(16,2)
BB(32,8)
SISOAWGN
10
25
30
Average received SNR, dB
35
40
(c)
30
35
40
Average received SNR (dB)
45
(d)
Fig. 4 Uncoded BER performance of a 32 32 MIMO (a) QPSK (b) 16QAM (c) 64QAM (d) 256QAM
10
MMSE
QP
BB(2,4)
BB(4,4)
BB(8,4)
BB(16,2)
BB(16,4)
BB(16,6)
BB(16,8)
BB(32,4)
BB(32,4)LAS
SISOAWGN
10
BER
10
10
10
10
10
16
18
20
22
24
26
Average received SNR (dB)
28
1536-1276 (c) 2015 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2495163, IEEE Transactions on Wireless Communications
9
10
10
MMSE
MMSELAS
QP
2QP
BB(4,4)
BB(16,2)
RTS
10
10
10
10
10
20
30
40
50
10
10
10
MMSE
MMSELAS
QP
2QP
BB(4,4)
BB(16,2)
RTS
10
10
60
10
20
30
(a)
9
10
MMSE
MMSELAS
QP
2QP
BB(4,4)
BB(16,2)
RTS
10
10
20
30
40
50
60
(c)
60
10
10
10
50
(b)
10
10
40
N
10
10
MMSE
MMSELAS
QP
2QP
BB(4,4)
RTS
BB(16,2)
10
10
10
20
30
40
50
60
(d)
Fig. 6 Avg. Complexity in terms of # of real operations vs. Nt (a) 16QAM at 19dB SNR (b) 16QAM at 26dB SNR (c) 256QAM at 35dB
SNR (d) 256QAM at 45dB SNR
(15)
where Rr and Rt are the correlation matrices for the receive antennas and transmit antennas, respectively, while Aiid
represents an i.i.d. (independent and identically distributed)
Rayleigh fading channel matrix. This model assumes that
the fading statistics of the transmit and receive arrays are
independent. In this paper, the correlation matrices of the
signals at both the transmit and receive sides are computed
based on the distance between antenna elements [44], [45].
Also, this model does not take into account the structure of
the scattering environment between transmitter and receiver.
The BER performance of the 2QP detector is only considered here for illustration. In this simulation, we consider a 16
1536-1276 (c) 2015 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2495163, IEEE Transactions on Wireless Communications
10
MF
MMSE
QRDM, M=4
MMSEOSIC
MMSELAS
QP
BB(4,4)
2QP (=0.25)
BB(16,2)
BB(2Nt ,2)
RTS
BER
10
10
MMSE
MMSELAS
QP
2QP
RTS
10
BER
10
10
10
16QAM
256QAM
10
3.25
dB
9.75
dB
10
10
5
10
10
10
20
30
40
# of Antennas
50
10
10
BER
10
10
60
10
10
10
10
10
30
40
# of Antennas
50
BER
10
10
MF
MMSE
MMSEOSIC
MMSELAS
2QP ( =0.25)
QP
BB(4,4)
BB(16,2)
RTS
10
10
10
60
BER
20
40
10
35
MF
MMSE
MMSELAS
MMSEOSIC
QRDM, M=16
QP
BB(4,4)
BB(16,2)
2QP (=0.25)
RTS
10
10
15
20
25
30
Avergae received SNR, dB
10
15
20
25
30
Average received SNR, dB
35
40
10
10
10
20
40
60
80
100
# of Antennas
In this section, the aim is to develop a turbo equalizationtype receiver using a QP detector. In the previous sections,
the performance of QP-based detectors were studied based on
uncoded/coded BER. In order to improve the performance of
such detectors in a low SNR regime, a turbo equalizationtype receiver can be used, in which a detector and a decoder
exchange soft information between each others in an iterative
manner (called iterative detection and decoding (IDD)) until
a stopping criteria is reached [46]. There are two challenges
in using QP in an IDD setting. First, how to incorporate a
priori information provided by the channel decoder, in the
form of Log-Likelihood Ratio (LLR), into the QP optimization
problem (6). Second, how to make the QP detector provide
soft information, in the form of LLR, so that it can be used
as a priori information to the soft-input soft-output channel
decoder. Addressing these challenges with implementation and
performance study will be presented in this section for largescale MIMO in a spatial multiplexing setup. We use the same
technique in [47] to incorporate a priori information into the QP
1536-1276 (c) 2015 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2495163, IEEE Transactions on Wireless Communications
11
x1
xi
1
1
||y Hx||2 xT La
2 2
2
Q = HT H,
1 T
z Q z + bT z
2
i
= min
z0
z
z =
(21)
x+1
, and b =
2
1
2
La . The result of (21) can be applied
HT (y + H1)
2
4
Lpost (xi )
p(xi = +1|y, H)
, i = 1, . . . , 2Nt
p(xi = 1|y, H)
(16)
p(y|x, H)P(x) ln
x +1
x
p(y|x, H)P(x)
x 1
x
(17)
p(xi = +1)
, i = 1, . . . , 2Nt
p(xi = 1)
(18)
If the noise in the system is white Gaussian, the probability density function p(y|x,
H) can be represented by
2
2
1
exp
||y
Hx||
/2
.
This
can be used
2 2
P in (17), and
with the aid of max-log approximation (ln( i exp(i ))
maxi {i }) [48], Eq. (17) can be simplified to:
1
2
||y
Hx||
ln[P(x)]
x 1
2 2
x
i
1
2
min
||y
Hx||
ln[P(x)]
x +1
2 2
x
Lpost (xi )
min
(19)
In order to find the relation between the vector of a priori probability P(x) and La , we follow the same work and assumptions
in [46] and [48]. Thus, (19) can be written as:
1
1 T
2
||y
Hx||
x
L
a
2 2
2
x1
xi
1
1 T
2
min
||y
Hx||
x
L
a
2 2
2
x+1
x
(20)
min
06z61,zi =0
(22)
1 T
z Q z + bT z]
2
(23)
and second, we solve the same problem again 2Nt -times with
bit constraints as follows:
min
z
st
1 T
z Q z + bT z
2
0 6 z 6 1, zi = xor(zi , 1), i = 1, . . . , 2Nt
(24)
(25)
1536-1276 (c) 2015 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2495163, IEEE Transactions on Wireless Communications
12
(
(j)
zi
zi
N (zi )
for i 6= j
for i = j
)
(26)
10
Uncoded 16x16
Uncoded 64x64
Coded, 16x16
Coded, 64x64
IDD , 16x16
IDD , 64x64
10
10
BER
Uncoded
10
Coded
10
iter# 3
iter# 1
10
iter# 2
6
10
8
10
12
14
Average Received SNR, dB
16
18
20
1536-1276 (c) 2015 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2495163, IEEE Transactions on Wireless Communications
13
10
10
LLR using neighbor solutions
LLR using (23) and(24)
10
10
iter# 1
iter# 1
BER
10
BER
10
10
iter# 2
10
iter# 3
iter# 3
iter# 2
10
10
4
5
6
Average received SNR, dB
(a)
4
5
6
Average received SNR, dB
(b)
Fig. 14 IDD BER performance with reduced LLR computation (a) 16 16 (b) 64 64
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
1536-1276 (c) 2015 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2495163, IEEE Transactions on Wireless Communications
14
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
H. Ozcelik,
M. Herdin, W. Weichselberger, J. Wallace, and E. Bonek,
Deficiencies of kroneckermimo radio channel model, Electronics
Letters, vol. 39, no. 16, pp. 12091210, 2003.
J. W. Wallace and M. A. Jensen, Modeling the indoor mimo wireless
channel, IEEE Transactions on Antennas and Propagation, vol. 50,
no. 5, pp. 591599, 2002.
M. A. Jensen and J. W. Wallace, A review of antennas and propagation
for mimo wireless communications, IEEE Transactions on Antennas
and Propagation, vol. 52, no. 11, pp. 28102824, 2004.
M. Tuchler, R. Koetter, and A. C. Singer, Turbo equalization: principles
and new results, IEEE Trans. Communications, vol. 50, no. 5, pp. 754
767, 2002.
B. Steingrimsson, Z.-Q. Luo, and K. M. Wong, Soft quasi-maximumlikelihood detection for multiple-antenna wireless channels, IEEE
Transactions on Signal Processing, vol. 51, no. 11, pp. 27102719, 2003.
B. M. Hochwald and S. Ten Brink, Achieving near-capacity on
a multiple-antenna channel, IEEE Transactions on Communications,
vol. 51, no. 3, pp. 389399, 2003.
Ali Elghariani received both the B.S. and M.S. degrees in Electrical and Electronic Engineering from
University of Tripoli in 1999 and 2008, respectively,
and the Ph.D. in Communications, Networking, and
Signal Processing from the School of Electrical
and Computer Engineering at Purdue University of
West Lafayette in 2014. He joined industry for
several years before he started his PhD. Currently
he is a lecturer at the Department of Electrical and
Electronic Engineering, University of Tripoli, Libya.
During 2013 he was a system engineer intern with
Qualcomm, Inc. at San Diego. He was the recipient of IEEE MILCOM
conference travel grant award in 2012. His current research interests are
signal detection and channel estimation in large-scale MIMO systems, symbol
spreading OFDM systems, turbo equalization, and the application of quadratic
programming optimization techniques in wireless communications.
1536-1276 (c) 2015 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.