Documente Academic
Documente Profesional
Documente Cultură
RESEARCH ARTICLE
2
School of Computing and Communications, Lancaster University, Lancaster LA1 4WA, UK.
ABSTRACT
In the Fifth generation (5G) wireless communication systems, a majority of the traffic demands is contributed by
various multimedia applications. To support the future 5G multimedia communication systems, the massive multiple-input
multiple-output (MIMO) technique is recognized as a key enabler due to its high spectral efficiency. The massive antennas
and radio frequency (RF) chains not only improve the implementation cost of 5G wireless communication systems but also
result in an intense mutual coupling effect among antennas because of the limited space for deploying antennas. To reduce
the cost, an optimal equivalent precoding matrix with the minimum number of RF chains is proposed for 5G multimedia
massive MIMO communication systems considering the mutual coupling effect. Moreover, an upper bound of the effective
capacity is derived for 5G multimedia massive MIMO communication systems. Two antenna receive diversity gain models
are built and analyzed. The impacts of the antenna spacing, the number of antennas, the quality of service (QoS) statistical
exponent, and the number of independent incident directions on the effective capacity of 5G multimedia massive MIMO
communication systems are analyzed. Comparing with the conventional zero-forcing precoding matrix, simulation results
demonstrate that the proposed optimal equivalent precoding matrix can achieve a higher achievable rate for 5G multimedia
massive MIMO communication systems. Copyright
c 0000 John Wiley & Sons, Ltd.
KEYWORDS
mutual coupling, massive MIMO systems, effective capacity, RF chains, equivalent precoding.
∗
Correspondence
Dr. Qiang Li, School of Electronic Information and Communications, Huazhong University of Science and Technology, Wuhan
430074, Hubei, P. R.China. E-mail: qli patrick@mail.hust.edu.cn
Part of this work appeared in the IEEE IWCMC 2015 [1], which was granted the best paper award.
A lot of studies have achieved great achievements the multi-channel communication systems simultaneously
about mutual coupling among multiple antennas on many according to their simulation results. However, rare
topics such as antenna propagation, signal processing and efforts has been made to study the effective capacity
antenna arrays [9–12]. Utilizing the real measurement of massive MIMO multimedia wireless communication
data, the authors of [9] have made a comparison on systems which consider the QoS constraint and mutual
the antenna array performance between the systems coupling effect.
considering the mutual coupling and the systems not. It Motivated by the above gaps, we propose an optimal
has been proved that mutual coupling has a great influence equivalent precoding matrix to reduce the cost of RF chains
on the performance of antenna arrays for not only small in 5G massive MIMO multimedia communication systems
but also large inter-antenna spacing, because that in order and derive the upper bound of effective capacity with QoS
to contain the changes in all the anticipant vectors, the constraints. The main contributions of this paper are listed
steering vectors of the antenna arrays should be adjusted as follows.
not only in amplitude but also in phase[10]. Clerckx et.
al. studied how the mutual coupling influenced a simple
multi-antenna communication system performance [11]. In 1. We define the receive diversity gain to analyze
order to recover the signals received by separate antennas how the mutual coupling influence the performance
without mutual coupling, the authors of [12] have invented of the rectangular antenna arrays in the massive
a new technique to make a compensation for mutual MIMO wireless communication systems.
coupling in small antenna arrays. 2. An optimal equivalent precoding matrix is proposed
At practical wireless communication transmission to reduce the cost of RF chains and satisfy
terminals, each data stream is first passed through the the multimedia data requirements for 5G massive
baseband precoding to radio frequency (RF) chains and MIMO multimedia communication systems.
then is transmitted to antennas by the RF chains precoding. 3. We refer to the QoS statistical exponent constraint
For MIMO wireless systems, the precoding technologies and the mutual coupling effect, then derive the
are focused on the baseband precoding, i.e., the first upper bound of effective capacity for 5G massive
order precoding, and each RF chain corresponds to an MIMO multimedia communication systems.
antenna. Utilizing the phase matrix between RF chains and 4. Based on numerical results, the proposed optimal
antennas, the joint precoding of baseband and RF chains equivalent precoding matrix is compared with the
was proposed for massive MIMO systems with limited RF conventional zero-forcing (ZF) precoding matrix
chains [13]. However, it is still a great challenge to reduce in 5G massive MIMO multimedia communication
the number of RF chains for saving the cost of massive systems.
MIMO wireless communication systems.
Lots of excellent studies in the field of wireless
The rest of this paper is summarized as follows.
multimedia communication have emerged [14–18]. In
In Section 2, a system model in which there is a
order to evaluate the QoS of wireless multimedia networks,
2D antenna array is described for massive MIMO
the authors in [14, 15] created a constrained model of
wireless communications. In Section 3, the effect of
statistical QoS to study the transmission characteristics
mutual coupling on the massive MIMO wireless systems
of data queues. In [16, 17], the authors referred to the
is evaluated by the receive diversity gain. Moreover,
effective capacity of the block fading channel model and
an optimal equivalent precoding matrix is proposed
proposed a rate and power adaption scheme in which
to reduce the cost of RF chains and satisfy the
the power is driven by QoS. And in [18], the authors
multimedia data requirements for 5G massive MIMO
further combined the effective capacity with information
multimedia communication systems. Furthermore, the
theory and developed some rate adaptation and QoS-
upper bound of effective capacity is derived for 5G massive
driven power schemes which were suitable for the systems
MIMO multimedia communication systems. Numerical
of multiplexing and diversity. Also they concluded that
simulations and analysis are presented in Section 4.
stringent QoS and high throughput can be achieved by
Finally, Section 5 summarizes the paper.
(r)
2. SYSTEM MODEL in which hp is defined as the real part of hp , and
(i)
hp is defined as the imaginary part of hp . Furthermore,
A massive MIMO wireless transmission system is both of them are Gaussian random variables distributed
illustrated in Fig. 1. The wireless down-link between a user independently and identically, whose expectation and
equipment (UE) with multi-antenna and a BS with a 2D variance are 0 and 0.5 respectively.
rectangular antenna array is studied in this paper. Definitely, P will be very large if considerable scatterers
First of all, we define some basic parameters for this exist in the propagation environment. According to [19,
model. We define λ as the wavelength of the carrier, d as 20], we divide the angular domain into P independent
the antenna spacing of this antenna array, aλ (a ≥ 1) as the incident directions with P being large but finite.
length of this antenna array and bλ (b ≥ 1) as the width of Here we assume both of the azimuth angle φq (q =
this antenna array. If we would like to deploy m antennas 1, ..., P ) and elevation angle θ are within the scope
in each row and n antennas in each column for this antenna of [−π/2, π/2]. Each independent incident direction
array, then we will have the relationship as listed in (1), corresponds to one steering vector a (φq , θ) ∈ CM ×1 , so
all the P steering vectors can constitute the steering matrix
aλ bλ A of the rectangular antenna array which is expressed as
d= = , (1)
m−1 n−1
and the total number of antennas in this antenna array M A = [a (φ1 , θ) , ..., a (φq , θ) , ..., a (φP , θ)] . (6)
can be derived easily as
If we define Aq ∈ Cn×m as the steering matrix of the qth
M = mn. (2) incident direction of the rectangular antenna array, we will
get the following relationship,
If we define SN RBS as the signal-to-noise ratio (SNR)
at the BS, H and β stand for the small scale fading matrix vec (Aq ) = a (φq , θ) , (7)
and large scale fading coefficient of the channel in this
in which vec (·) is defined as the matrix vectorization
model respectively, the signal the BS transmits is defined
operation.
as x, w means the additive white Gaussian noise (AWGN)
Without loss of generality, we assume the antenna which
over wireless channels, and the mutual coupling matrix is
locates at the first place for both the row and the column
configured as K, equivalent precoding matrix is configured
of the rectangular antenna array as the reference point
as Feq , A is defined as steering matrix, then the down-link
of which the phase response is zero. And we normalize
signal vector received at a UE equipped with N antennas
amplitude responses of all the antennas of the antenna
can be expressed as
array as 1. We define Aqce , (1 ≤ c ≤ m, 1 ≤ e ≤ n) as the
√ element in the steering matrix Aq which locates at the cth
y= SN RBS HAKFeq β 1/2 x + w, (3)
row and eth column, and it is expressed as Similar properties can be derived for Zst as
2π (c − 1)dcosφq sinθ Zst = Z(s+1),(t+1) , (13)
Aqce = exp j . (8)
λ +(e − 1)dsinφq sinθ
Zst = Zts . (14)
For a rectangular antenna array with M elements, we
Together with (10)-(14), the mutual impedance matrix
define K ∈ CM ×M as the corresponding mutual coupling
ZM can be readily obtained. It bears noting that with (10)-
matrix, which is expressed as [11]
(14), the computational complexity can be significantly
reduced compared to the direct calculation of the M × M
K = ZL (ZL I + ZM )−1 , (9)
entries of ZM , especially with a large M .
The equivalent precoding matrix Feq = FRF FBB con-
in which ZL denotes the antenna load impedance that
sists of baseband precoding matrix FBB and the RF
is constant for each antenna, ZM denotes the M × M
precoding matrixFRF .
mutual impedance matrix, and I denotes an M × M
t
Ns data streams are transmitted by NRF radio
unit matrix. From Fig. 1, ZM can be constructed by
frequency (RF) chains and M antennas at the BS. All
n × n sub-matrices, i.e.,ZM = [Zst ]n×n , where Zst , as
r
wireless data is received by NRF RF chains and N
an m × m mutual impedance sub-matrix, denotes the
antennas at the UE. In this case, the detected wireless
mutual impedances between the m antennas located at
signals at the UE is expressed by
the sth (s = 1, ..., n) row and the m antennas located
at the tth (t = 1, ..., n) row in the rectangular antenna † † †
ỹ = Weq y = WBB WRF y, (15)
array. For ease of exposition, we define Antsu as the
antenna located at the sth row and uth (s = 1, ..., m; u =
in which † is a conjugate transpose operation, Weq is a
1, ..., m) column of the rectangular antenna array, and
N × Ns equivalent signal detection matrix which consists
define Anttv as the antenna located at the tth row and the
of baseband detection matrix WBB and the RF detection
vth (t = 1, ..., m; v = 1, ..., m) column of the rectangular
matrix WRF , y is the received signal vector at antennas
antenna array, the corresponding distance between which
q of the UE. Essentially, FRF and WRF are phase shift
is given as duv = d (t − s) + (v − u)2 . Thus Zst can
st 2
matrices used for the signal precoding and detection at the
be written as
RF chains. Hence, the absolute value of the RF detection
st
z11 st
z12 ... st
z1m matrix FRF and the RF precoding matrix FRF is equal to
st
z21 st
z22 ... st
z2m
1.
Zst = .. .. .. .. , (10)
. . . .
st st st
zm1 zm2 ... zmm
3. MUTUAL COUPLING EFFECT
Consider a special case where all M antenna elements
MODELING
in the rectangular antenna array are dipole antennas with
st
the same parameters. Then the mutual impedance zuv only
3.1. Receive Diversity Gain Models
depends on the antenna spacing and can be obtained with
the EMF method in [27]. With a fixed antenna spacing d, Deployed in a constrained space at the BS, the number
we have the following properties: of antenna elements is inversely proportional to the
antenna spacing, i.e., a larger number of antennas lead to
st st
zuv = z(u+1),(v+1) , (11) a smaller antenna spacing. As concluded in [21], more
antennas lead to a higher receive diversity gain of the
st st
zuv = zvu . (12) massive MIMO system, whereas the diversity gain can be
compromised by the mutual coupling effect that is caused
by decreasing the antenna spacing. Thus, when a number
of antennas are deployed in a fixed constrained area, there
SN RBS † † † † †
−1
Rmax = log I +
Weq HAKFeq Feq K A H Weq R , (21)
Ns
with
†
R = Weq Weq . (21.1)
exists a tradeoff between M and d, and it is important to In addition, the average SNR seen at the UE side can be
analyze the the effect of mutual coupling on the achievable written as
receive diversity gain of the massive MIMO systems.
2
SN RBS
G† G
Firstly, with a fixed antenna spacing, the antenna SN RU E =
kG† Gk
number receive diversity gain GM is defined as
, (19)
†
= SN RBS
G G
dmin dmin
GM = ξM − ξM min
, (16)
Then with M antenna elements with antenna spacing d,
dmin the expectation of the SNR at the UE can be obtained as
in which ξM denotes the expectation of the receive
SNR at the UE with M antennas and minimum antenna
d
spacing being dmin at the antenna array of the BS, and ξM = E {SN RU E }
n
o
dmin
= SN RBS E
G† G
ξM min
denotes the expectation of the receive SNR at the , (20)
UE subject to the minimum antenna spacing dmin and
† † † †
= SN RBS N β
Feq K A H HAKFeq
minimum antenna number Mmin at the antenna array of
the BS.
in which E {·} denotes the expectation operation. We can
Secondly, with a fixed number of antennas, the antenna
further obtain Gd and GM through substituting (20) into
spacing receive diversity gain Gd is defined as
(16) and (17), and replacing M and d with Mmin and dmin .
d dmin
Gd = ξM min
− ξM min
, (17)
3.2. Shannon Capacity with Optimal RF Chains
d
in which ξM min
denotes the expectation of the receive
A phase shift matrix is designed to separate the
SNR at the UE with an antenna spacing of d and with at
dmin
RF chains and the antennas. Assume that the relation-
least Mmin antennas at the antenna array of the BS. ξM ,
min ship among the numbers of data stream, RF chains
which can be considered as the baseline for Gd and GM , t
and antennas is configured as Ns 6 NRF 6 M . Con-
is the same as that in (16).
sidering the equivalent precoding matrix (Feq )M ×Ns =
In order to obtain the expectation of the received SNR in
(FRF )M ×N t (FBB )N t ×Ns , about it’s rank, we have
RF RF
(16) and (17), perfect channel state information is assumed t
rank (Feq ) 6 min Ns , NRF , M . This result implies
to be available at the BS, which uses maximal-ratio
that the up-bound of the degree of freedom at the equiva-
combining (MRC) for signal detection. When there are M
lent precoding matrix is depended on the minimum among
antennas and the neighboring antennas are separated with
the numbers of data stream, RF chains and antennas. When
a spacing of d, the received signal after MRC detection at
the number of RF chains is larger than the number of data
the UE is given as [22] t
stream, a part of number of RF chains, i.e., NRF − Ns , has
√ not been utilized by the equivalent precoding matrix. To
ỹ = G† y = SN RU E G† Gx + G† w, (18a)
save the cost, the number of RF chains can be configured as
Ns to satisfy the requirement of the equivalent precoding
in which G† denotes the conjugate transpose of G with
matrix Feq .
G = HAKFeq β 1/2
. (18b) Based on the system model in Fig. 1, the sys-
tem achievable rate, i.e., the maximum Shannon capac-
ity is expressed by (21) and normalized on the unit
bandwidth [23]. Let M IN = min (M, P ), the eigen-
values of wireless channels H are ordered by λ1 >
−1
SN RBS † † † †
Rmax = log I +
HFF H Weq Weq Weq Weq
Ns
SN RBS †
= log I + UH ΣH V H UF ΣF Σ†F U†F VH Σ†H U†H , (25)
Ns
SN RBS †
= log I + VH UF ΣF Σ†F U†F VH Σ†H ΣH
Ns
SN RBS Ns
Rmax = log I + ΣF2 ΣH2
Ns SN RBS
, (32)
SN RBS r Ns
= r log + log Π + fi2 λ2i
Ns i=1 SN RBS
2 2 2
L f1 , f2 · · · fr
Ns Ns Ns
= + λ21 f12 + λ22 f22 · · · + λ2r fr2 . (33)
SN RBS SN RBS SN RBS
+α f12 + f22 + · · · + fr2 − Ns2
∂L f12 , f22 · · · fr2
∂fi2
Ns Ns
= λ2i + λ21 f12 · · · + λ2i−1 fi−1
2
, (34)
SN RBS SN RBS
Ns Ns
× + λ2i+1 fi+1
2
··· + λ2r fr2 + α
SN RBS SN RBS
in which ΣF2 and ΣH2 are diagonal matrices and the in which (i = 1, 2...r), we can further derive the following
values of elements at diagonal line are larger than zero. result
Furthermore, the eigenvalues of ΣF2 and ΣH2 are the Ns Ns
SNRBS SNRBS
elements at the diagonal lines, respectively. U is a unitary + fi2 = + fj2
λ2i λ2j
matrix, i.e., kUk2 = M . When U is configured as a , (36)
Ns 1 1
diagonal matrix, the maximum available rate is achieved ⇒ fi2 − fj2 = − 2
SN RBS λ2j λi
by
SN RBS
Rmax = log I + ΣF2 ΣH2 . (30) in which (i, j = 1, 2...r).
Ns
Based on (31), the square of eigenvalues at the
Assume that the transmission power at the BSs is
equivalent precoding matrix Feq is derived by
independent of the equivalent precoding matrix. This
assumption implies that kFeq kF = Ns , i.e., Ns 1
+ 1
+ ··· + 1
λ2 λ2 λ2
fi2 = 1 2 r
r
X SN RBS r . (37)
fi2 = Ns2 . (31) Ns2 Ns
+ −
i=1
r SN RBS λ2i
The maximum available rate Rmax can be simplified as According to (25) and (30), we know that Weq , VF ,
(32). UH and U are removed in the simplification process of
To achieve the maximum achievable rate, the optimal Rmax , so VF , UH and U can be unit matrices. And
solution of the equivalent precoding is derived by a furthermore, UF and VH can also be unit matrices, and
Lagrange multiplier method in the following. A function UF = VH . Considering the Weq is a N × Ns non-
is first constructed in (33) with Lagrange factor α. And singular matrix, Weq can be decomposed by
then take the derivative of L f12 , f22 · · · fr2 with respect
!
to fi2 , (i = 1, 2...r) in (34). I
Weq = , (38)
Let 0
∂L f12 , f22 · · · fr2
= 0, (35)
∂fi2 In this case, the optimal equivalent precoding matrix is
simplified by
†
Feq = UF ΣF VF
. (39)
= V H ΣF
b1 0 ··· 0 0 ··· 0
b1 0 ··· 0 0 ··· 0
0 b2 ··· 0 0 ··· 0
0 b2 ··· 0 0 ··· 0
.. .. .. ..
ejϑ1,1 ··· ejϑ1,2r 0 ··· 0 .. ..
. . . . . ··· .
.. .. .. ..
Feq = . ··· . . ··· . , (40)
0 0 ··· br 0 ··· 0
ejϑM,1 ··· ejϑM,2r 0 ··· 0 0 0 ··· br 0 ··· 0
| {z }
(FRF )M ×2Ns
0 0 ··· 0 0 ··· 0
.. .. .. .. .. ..
. . . . . ..
. .
0 0 ··· 0 0 ··· 0
| {z }
(FBB )2N ×N
s s
(
r )!
SN RBS
1 −θT B log Ns
+log Π SNN s
RBS
+fi2 λ2
i
CE (θ) = − ln E e i=1
θT
log Π
r !−θT B
Ns 2 2
1 SN RBS
−θT B log +f λ
i i
=− ln e Ns E e i=1 SN RBS . (42)
θT
( − θT B )!
1 SN RBS
Ns
−θT B log r 2 2
ln(2)
=− ln e Ns E Π + f λ
i i
θT i=1 SN RBS
Based on the method in [23], the optimal equivalent 3.3. Effective Capacity with Mutual Coupling
precoding matrix Feq is composed of FRF and FBB , Effect
which are designed in (40) with
From [28], we define the effective capacity under
1 multimedia constraints as
bj = max |fi,j | , (40.1)
2 16i6M
1
CE (θ) = − ln E{e−θT BR } , (41)
|fi,j | θT
ϑi,(2j−1) = ∡fi,j − cos−1 , (40.2)
2bj
in which θ and B denote the QoS statistical exponent and
|fi,j |
ϑi,2j = ∡fi,j + cos−1 , (40.3) bandwidth respectively, E{·} is the expectation operation.
2bj
Without losing generality, we consider independent fading
in which fi,j is the element of Feq located at the ith row
channels that keep static within a frame duration T .
and the jth column∡fi,j is the corresponding angle.
Considering the maximum available rate in (32), (41)
Based on the result in (40), the number of RF chains
can be extended as (42), and it is clear that f (x) =
2Ns can satisfy the requirement of the optimal equivalent
x−a , (a > 0) is a convex function. Then an upper bound
precoding matrix. In general, the number of antennas M is
of the effective capacity can be obtained using Jensen’s
larger than the number of RF chains 2Ns in 5G massive
inequality,
MIMO wireless systems. Hence, our proposed optimal
1 n o
equivalent precoding matrix can save the number of RF CE (θ) = − ln E e−θT BR
chains M − 2Ns . θT
1
6− ln e−θT BE{R} . (43)
θT
= BE {R}
SN RBS r Ns
E {Rmax } = E r log + log Π + fi2 λ2i
Ns i=1 SN RBS
X r , (44)
SN RBS Ns 2 2
6 r log + log E + fi λi
Ns i=1
SN RBS
r
P 1
r
2
SN RBS Ns
X 2 λ
E{Rmax } 6 r log + log E λi Ns + λ2i i=1 i
Ns i=1
r
SN RBS
X r (X r
)!
, (45)
SN RBS Ns r 2
6 r log + log 2
Ns + E λi
Ns i=1
r SN RBS i=1
X r n o
SN RBS Ns r † † †
= r log + log N s + E tr H HAKK A
Ns i=1
r2 SN RBS
n o
SN RBS Ns r † † †
E{Rmax } 6 r log + r log Ns + tr AKK A + E tr H H
Ns r2 SN RBS
, (46)
SN RBS Ns r † †
= r log + r log Ns + tr AKK A + Pr
Ns r2 SN RBS
SN RBS Ns r † †
CE (θ) 6 B r log + r log Ns + tr AKK A + Pr . (47)
Ns r2 SN RBS
SNR_BS=5dB
52 SNR_BS=10dB
the effective capacity in 5G multimedia communication SNR_BS=20dB
50
systems is given by (47). 48
46
44
42
4. NUMERICAL RESULTS AND
40
ANALYSIS 38
36
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
In this section, we demonstrate the performance of Antenna distance d(×λ )
7
x 10
70 2.2
65 2
35 0.8
30 0.6
25 0.4
20
0.2
0 200 400 600 800 1000 1200 1400 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Antenna number M Antenna spacing d(×λ)
Figure 3. Antenna number receive diversity gain with respect to Figure 5. Effective capacity with respect to the QoS statistical
the antenna number considering different SNRs. exponent θ and the antenna spacing.
7
x 10
2.4 2.2
2.2 2
2 1.8
1.6
Shannon capacity (bits/s)
1.8
Effective capacity (bits/s)
0.8 0.6
0.6 0.4
0.4 0.2
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Antenna number M Antenna number M
Figure 4. Effective capacity with respect to the antenna number Figure 6. Shannon Capacity with respect to the antenna
considering different SNRs. number considering different precoding matrices.
the antenna spacing receive diversity gain increases with as 0.01. With a fixed SNR value, it is observed that a higher
the increase of SNR. But if we fix the SNR at the BS, effective capacity is obtained by increasing the antenna
it is shown that there is almost no correlation between number. In addition, with a fixed number of antennas, a
the antenna spacing receive diversity gain and the antenna higher effective capacity is obtained with a higher SNR.
spacing. Fig. 5 shows the effective capacity with varying values
Fig. 3 illustrates the correlation between GM and of QoS statistical exponent and the antenna spacing. With a
the antenna number considering different SNRs. dmin is fixed antenna spacing, it is observed that a higher effective
configured as 0.1λ. And it is shown that the antenna capacity is reached by decreasing the QoS statistical
number receive diversity gain has a positive correlation exponent. On the other hand, for a fixed QoS statistical
with the antenna number and SNR. exponent and increasing of antenna spacing, the effective
In Fig. 4, the effective capacity is illustrated with capacity almost keeps stationary.
varying values of the antenna number and SNR. For ease of When the number of user and the baseband data
illustration, the antenna spacing is set as 0.5λ and θ is set stream are configured as one, Fig. 6 compares the
7 7
x 10 x 10
2.1 2.3
2.2
2
2.1 SNR_BS=5dB
SNR_BS=10dB
1.9 EC_SNR_BS=5dB
Effective capacity (bits/s)
2 SNR_BS=20dB
1.6 1.6
1.5
1.5
1.4
1.4 1.3
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 0 100 200 300 400 500 600 700 800 900 1000
QoS statistical exponent q Independent incident directions P
Figure 7. Effective capacity with respect to the QoS statistical Figure 8. Effective capacity with respect to the independent
exponent θ considering different SNRs. incident directions P considering different SNRs.
[5] F. Rusek, D. Persson and B. K. Lau, “Scal1ing up [15] T. Taleb, K. Kashibuchi, A. Leonardi, S. Palazzo, K.
MIMO: opportunities and challenges with very large Hashimoto, N. Kato, and Y. Nemoto, “A cross-layer
arrays,” IEEE Signal Proces. Mag., Vol. 30, no. 1, pp. approach for an efficient delivery of TCP/RTP-based
40–60, Jan. 2013. multimedia applications in heterogeneous wireless
networks,” IEEE Trans. on Vehicular Technology
[6] Z. Xu, S. Sfar and R. Blum, “Receive antenna
journal, vol. 57, no. 6, pp. 3801–3814, Nov. 2008.
selection for closely-spaced antennas with mutual
coupling,” IEEE Trans. Wireless Commun., vol. 9, no. [16] J. Tang, X. Zhang, “Quality-of-service driven power
2, pp. 652–661, Feb. 2010. and rate adaptation over wireless links,” IEEE Trans.
Wireless Commun., vol. 6, no. 8, pp. 3058–3068, precoding and downlink pilots,” in Allerton Confer-
Aug. 2007. ence on Communication, Control, and Computing,
pp. 293–298, Oct. 2013.
[17] T. Taleb and A. Ksentini, “QoS/QoE predictions-
based admission control for femto communications,” [27] C. A. Balanis, Antenna Theory: Analysis and Design.
Proc. IEEE ICC 2012, Ottawa, Canda, Jun. 2012. John Wiley and Sons, 2012.
[18] J. Tang, X. Zhang, “Quality-of-service driven power [28] X. Ge, X. Huang, Y. Wang, et. al., “Energy efficiency
and rate adaptation for multichannel communications optimization for MIMO-OFDM mobile multimedia
over wireless links,” IEEE Trans. Wireless Commun., communication systems with QoS constraints,” IEEE
vol. 6, no. 12, pp. 4349–4360, Dec. 2007. Trans. Veh. Technol. , vol. 63, no. 5, pp. 2127–2138,
June 2014.
[19] H. Q. Ngo, E. Larsson and T. Marzetta, “The
[29] Tulino, Antonia M and Verdú, Sergio, “Random
multicell multiuser MIMO uplink with very large
matrix theory and wireless communications,” Now
antenna arrays and a finite-dimensional channel,”
Publishers Inc. , vol. 1, no. 5, 2004.
IEEE Trans. Commun., vol. 61, no. 6, pp. 2350–2361,
June 2013.