Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

The 2014 International Conference on Advanced Technologies for Communications (ATC'14)

Optimized Transceivers for Interference Alignment


in MIMO Interference Channels
Ha Hoang Kha
Faculty of Electrical & Electronics Engineering
Ho Chi Minh City University of Technology, Vietnam
Email: hhkha@hcmut.edu.vn

improved as compared to the rate of a single communication


link. Alternatively, by sharing all channel state information
(CSI) and knowledge of data streams to all remote users, the
transmitters can coordinately mitigate interference. This technique can significantly improve the data throughput. However,
the main drawbacks of this method are stringent requirements
on the user coordination and the communication bandwidth
of backhaul links [3]. In contrast to previous results, it has
been shown in [6] that the sum capacity per user of a network
with arbitrary number of users is the order of half the rate
of each user with interference-free. The important results are
obtained by a novel technique called interference alignment.
Due to their potential, interference alignment techniques have
recently become a very active research area [3], [5], [7], [8].
In MIMO wireless networks, the partial dimension offered
by multiple antennas is exploited for interference alignment.
Reference [9] characterized the capacity region of multipleinput single output (MISO) interference channels while [13]
considered for scenarios of the single-input multiple-output
with strong interference. Reference [10] designed the linear
filters at the transmitters and receivers for MIMO X channels
which include two pairs of users equipped with multipleantenna. In [14], the authors analyzed the degree of freedom
region for the X channels with two user pairs equipped
with an arbitrary number of antennas per user. The authors
in [8] extended this transmission technique to the -user
interference channels and showed that an achievable sum-rate
multiplexing gain is /2 per time, frequency and antenna
dimension.
It is shown in [3] that the optimal design of transceivers to
maximize the sum rate of the system is NP-hard. Therefore, the
development of computationally efficient algorithms to obtain
the optimal transceivers is a challenging task but of great
interest. Additionally, the problem of maximizing the DoF for
MIMO interference channels in which each user is equipped
with more than 2 antennas is also NP-hard [3]. Several
papers (see, for example [11], [12], and references therein)
have developed the numerical iterative algorithms to find the
optimal transceiver strategies for MIMO interference channels.
However, the common drawbacks of those algorithms are that
the solutions highly rely on the initial points, step size and are
trapped in a local optimum.
In this paper, we adopt the alternating optimization method
in which the precoding matrices at the transmitters and interference suppression matrices at the receivers are obtained

AbstractThis paper is concerned with the optimal transceiver


design in flat-fading multiple-input multiple-output (MIMO)
interference channels. Interference alignment (IA) is recently
known as an emerging technique to obtain the maximum total
degree of freedom. Finding the transceiver matrices for perfect
IA is mathematically challenging since the interference alignment
problems are highly nonlinear and nonconvex. Interference
leakage minimization and signal to noise plus interference ratio
(SINR) maximization methods were developed to suppress the
interference from the undesired users. However, most of previous
works commonly assume that the equal power is allocated
to each user and each stream. In this paper, we develop an
iterative optimization algorithm to optimize power allocation to
each stream. The simulation results indicate that the proposed
algorithm offers a higher sum-rate than the interference leakage
minimization method.
Index TermsMIMO interference channels, interference alignment, geometric programming, power allocation.

I. I NTRODUCTION
The use of multiple antennas at both transmitters and
receivers can improve the channel capacity, reliable transmission and spectral efficiency. Multiple-input multiple-output
(MIMO) techniques have been deployed in the current and
future wireless networks [1]. In modern wireless systems, the
spectrum scarcity is of the major concerns and, thus, the
communication between multiple user pairs is likely to use
the same radio sources such as time, frequency and space.
Such models are known as interference channels. Interference
channels are appropriate models for cellular networks, wireless
ad-hoc networks, and cognitive radio [2], [3]. It is known
that interference is one of the critical factors that degrade the
system performance. Thus, interference mitigation techniques
are of great interest in wireless network design. Recently,
interference alignment (IA) has been developed to optimize
the degree of freedom (DoF) being an alternative measure
for the sum capacity of networks at high signal to noise
ratio (SNR) [4]. The key idea of interference alignment is
to confine the interference signals in a particular subspace
at each receiver while the desired signals can be transmitted
on interference-free subspace [5]. The optimal transceiver
designs in IA schemes are mathematically challenging and
under investigation, except for several special cases due to the
nonlinear and nonconvex constraints.
In interference channels, as there is no cooperation among
users, each user aims to maximize its own rate. As a result,
the achieved sum of data rates across all user pairs is not

978-1-4799-6956-2/14/$31.00 2014 IEEE

19

The 2014 International Conference on Advanced Technologies for Communications (ATC'14)

where is the channel matrix between


transmitter and receiver . The MIMO channels are assumed
to be frequency-flat Rayleigh-fading. is noise at
receiver and is assumed to be (00, 2 ).
At receiver k, the received signal is linearly processed by
the matrix as follows

from stream signal to interference plus noise ratio (SINR)


maximization. While all most previous works assume that the
power allocated to each data stream is equal, we will further
optimize the power allocation to maximize the sum stream
rate. We employ the geometric programming to efficiently
find the optimal power allocations for given the transceiver
matrices. The numerical results show that the proposed method
can improve the sum-rate as compared to the interference
leakage minimization method.
Notations: Bold-faced lowercase and uppercase characters
denote the vectors and matrices, respectively. The operations
(.) , (.) are respectively the transpose and conjugate transpose. (:, ) presents the -th column of matrix . We denote
a vector x of Gaussian random variables with mean and
, ).
covariance by (

=
= +

+ .

=1,=

or, equivalently,
1/2
=
+

=1,=

1/2

+
.

(3)
The interference alignment is to find the precoding matrices
at the transmitters and the interference suppression matrices at
the receivers such that the interference is completely cancelled
at the high SNR region. For all , the perfect IA
requirements are mathematically posed as [3]

II. S YSTEM M ODEL


Consider the generic MIMO interference channels as depicted in Fig. 1 where K transmitter-receiver pairs communicate with each other. Transmitter and receiver
are respectively equipped with and antennas [13].
Transmitter = {1, ..., } transmits a symbol vector

= 0 ,

rank(
) = .

(4a)
(4b)

From Eq. (2), the sum rate of the system is defined as


=

=1

where

1
log +
,

= 2 +

(5)

(6)

It is challenging to find the transceiver matrices to meet the


perfect alignment (4) such that the sum-rate (5) is maximized
at high SNR. The following section shall present an iterative
algorithm to find the transceiver matrices. For ease of presentation, we assume that all transmitters have the same number
of antennas, = , so do the receivers = . We
also assume that all transmitters transmit the same number of
streams = and noise variance at the receivers is the same,
2 = . The perfect channel state information is assumed to
be available at the transceivers.

Fig. 1. The K user-pair MIMO interference channel.

= [,1 , ,2 , ..., , ] 1 to receiver . Here,



it is assumed that [
] = and is the number
of independent data streams with min{ , }.
The symbol vector is precoded by an orthogonal matrix
yielding
1/2
=
(1)
where and

is a diagonal matrix given by

1 0
0 2
=
...
...
0
...

III. I NTERFERENCE L EAKAGE M INIMIZATION AND S UM


S TREAM R ATE M AXIMIZATION
It is mathematically challenging to finding the set of
transceiver matrices which are satisfying the perfect IA conditions (4). The reference [14] proposed an alternating optimization to minimize the leakage interference. It exploited the
reciprocity of wireless networks. Given the transmit precoding
matrices, the receive matrices are designed to minimize the
leakage interference caused by all undesired transmitters. It is
shown in [14] the receive matrices are given by

= , and +
...
...
...
...

0
0
...

where the diagonal elements present the power allocated to


the streams. The received signal at receiver is given by
= +

[
=


].

(7)

=1,=

This means that the columns of precoding matrices are the


eigenvectors corresponding to smallest eigenvalues of the

(2)

=1,=

20

The 2014 International Conference on Advanced Technologies for Communications (ATC'14)

interference covariance matrix at receiver . Then, the receive


matrices of the original network are used as the transmit
matrices in the reciprocal network. The precoding matrices
in the original network play as receive matrices in the
reciprocal network. Thus, the precoding matrices can be found
by

=
[

(8)
].

and
(:, ) =
with

The SINR expression in (9) is simplified to


SINR =

To optimize power allocation, Eq. (11) can be rewritten as


max

s.t.

log(1 + SINR ).

(10)

log(1 + SINR )
(11a)
=1 =1

(:, ) = 1, , (11b)
(:, )

(:, ) = 1, , (11c)
(:, )

, .

(14b)

, .

IV. N UMERICAL S IMULATION R ESULTS


This section will provide numerical simulation results to
illustrate the performance of our proposed method. In the
simulations, the maximum number of iterations for Algorithm
1 is set at 100. The noise variance is normalized, 2 = 1. The
channel elements are generated from i.i.d. Gaussian random
variables with zero mean and unit variance. The numerical
results are averaged over 100 channel realizations.
We consider the MIMO system models with = 3 userpairs, each equipped with = = {3, 4} antennas. Each
user transmits = 2 streams. We compare the average sumrate of our iterative algorithm with that of the interference

(11d)

=1

where = {1, ..., }. It is difficult to jointly design


, , }
{
=1 due to the high nonlinearity and nonconvexity of the objective function. To solve (11) for SINR larger
than 1 , we employ the alternating optimization method. Given
the power allocation matrices, the transceiver matrices are
found by [14]
(:, ) =

Initialization: Choose
= , = .
repeat
repeat
Obtain the receive matrices from Eq. (12) .
Obtain the transmit matrices from Eq. (13) .
until convergence or the maximum allowable number of
iterations is reached.
Solve (14) to obtain , .
until convergence or the maximum allowable number of
iterations is reached.

The sum stream rate maximization (SSRM) can be presented


as the following optimization

s.t.

(14a)

Algorithm 1 Alternating optimization for SSRM

=1 =1

,
,
}
{
=1

log(1 + SINR )

=1 =1

where is the maximum allowable power of each user. As


shown in [16], the power allocation in (14) can be cast as
geometric programming when SINR is much larger than 1. It
is well-known that geometric programming can be transformed
into convex optimization which is efficiently solved. Consequently, our proposed iterative algorithm is given by Algorithm
1.

is the interference plus noise covariance matrix. The sum


stream rate is defined as [15]

max

=1

, (:, )
(:, )

2 +

(, )

.
(, )
(,)=(,)

(,)=(,)

.
,
(:, )
(:, )

Given the transceiver matrices, we define the effective


channel gain from stream of user to stream of user
as

(:, )2 .
(, ) =
(:, )

(:, )

,
(:, )
(:, )
(:, )

(:, )
(:, )
(9)

= 2 +

(13)

(,)=(,)

The iterative IA algorithm is to start with arbitrary orthogonal


matrices
= , . It alternatively finds the
precoding and receive matrices by Eqs. (7) and (8) until the
solution is converged.
It is important to notice that the interference leakage minimization (ILM) aims at suppressing the interference caused
by all undesired transmitters. It ignores the effects of desired
signals and noise. Additionally, it implicitly assumes that the
power levels allocated to all streams are fixed. Since each
stream can suffer different fading, the fixed power allocation is
clearly not optimal. To find the power allocation which adapts
to channel fading, the ILM method is not applicable since it
results in a trivial solution of all zero powers. Therefore, to
optimize power allocation, we adopt the sum stream rate as
a performance metric since the sum stream rate maximization
(SSRM) offers the same sum rate as the ILM at high SNR
[15]. From (3), define the SINR for a stream of user as

where

= 2 +

=1,=

SINR =

1
(:, )
, , ,
1
(:, )

1
(:, )
, , ,
1

(:, )

(12)

21

The 2014 International Conference on Advanced Technologies for Communications (ATC'14)

leakage minimization (ILM) in [14]. Fig. 2 plots the average


sum-rate achieved for = = 4. As shown in Fig. 2, our
iterative algorithm offers an average sum rate improvement of
about 4 bps/Hz as compared the ILM method. Fig. 3 plots the
average sum-rate achieved for the case = = 3. It can
been seen from Fig. 3 the average sum-rate of the ILM method
is significantly reduced as the number of antennas is reduced
from 4 to 3. That is because the system with = = 3
+
) [7] and, therefore, the rate of the
is not proper ( > +1
infeasibility of the perfect interference alignment increases.
However, in this case, our iterative algorithm is still better
than the ILM in terms of the sum rate.

It has adopted the alternating optimization of sum stream


rate maximization method and has introduced the geometric
programming to optimize the power allocated to each stream.
The numerical results have shown that the optimized power
allocation to streams can improve the average sum rate as
compared to equal power interference alignment.
ACKNOWLEDGMENT
This research is funded by Vietnam National Foundation for
Science and Technology Development (NAFOSTED) under
grant number 102.04-2013.46.
R EFERENCES

Average sum rate (bps/Hz)

70

[1] Q. Li, G. Li, W. Lee, M. Lee, D. Mazzarese, B. Clerckx, and Z. Li,


MIMO techniques in WiMAX and LTE: a feature overview, IEEE
Communications Magazine, vol. 48, pp. 8692, May 2010.
[2] C. Suh and D. Tse, Interference alignment for cellular networks, in
Annual Allerton Conference on Communication, Control, and Computing, pp. 10371044, Sept. 2008.
[3] M. Razaviyayn, M. Sanjabi, and Z. Q. Luo, Linear transceiver design
for interference alignment: complexity and computation, IEEE Trans.
Information Theory, vol. 58, pp. 28962910, May 2012.
[4] S. A. Jafar, Interference alignment: a new look at signal dimensions in
a communication network, Foundations and Trends in Communications
and Information Theory, vol. 7, no. 1, pp. 1136, 2011.
[5] J. Tang and S. Lambotharan, Interference alignment techniques for
MIMO multi-cell interfering broadcast channels, IEEE Transactions on
Communications, vol. 61, pp. 164175, January 2013.
[6] V. Cadambe and S. Jafar, Interference alignment and degrees of
freedom of the K -user interference channel, IEEE Transactions on
Information Theory, vol. 54, pp. 34253441, Aug 2008.
[7] D. Papailiopoulos and A. Dimakis, Interference alignment as a rank
constrained rank minimization, IEEE Transactions on Signal Processing, vol. 60, pp. 42784288, Aug 2012.
[8] S. Jafar and S. Shamai, Degrees of freedom region of the MIMO X
channel, IEEE Transactions on Information Theory, vol. 54, pp. 151
170, Jan 2008.
[9] S. Vishwanath and S. Jafar, On the capacity of vector gaussian
interference channels, in Proc. IEEE Information Theory Workshop,
pp. 365369, Oct 2004.
[10] M. Maddah-Ali, A. Motahari, and A. Khandani, Communication over
MIMO X channels: Interference alignment, decomposition, and performance analysis, IEEE Transactions on Information Theory, vol. 54,
pp. 34573470, Aug 2008.
[11] S. Peters and R. Heath, Interference alignment via alternating minimization, in Proc. IEEE International Conference on Acoustics, Speech and
Signal Processing, (ICASSP), pp. 24452448, April 2009.
[12] K. Gomadam, V. Cadambe, and S. Jafar, A distributed numerical
approach to interference alignment and applications to wireless interference networks, IEEE Transactions on Information Theory, vol. 57,
pp. 33093322, June 2011.
[13] M. Razaviyayn, M. Sanjabi, and Z.-Q. Luo, Linear tranceiver design
for interference alignment: Complexity and computation, IEEE Trans.
Information Theory, vol. 58, pp. 28962910, May 2012.
[14] K. Gomadam, V. Cadambe, and S. Jafar, Approaching the capacity of
wireless networks through distributed interference alignment, in Proc.
of IEEE Global Telecommun., pp. 16, 2008.
[15] C. Wilson and V. Veeravalli, A convergent version of the Max SINR
algorithm for the MIMO interference channel, IEEE Transactions on
Wireless Communications, vol. 12, pp. 29522961, June 2013.
[16] M. Chiang, C. wei Tan, D. Palomar, D. ONeill, and D. Julian, Power
control by geometric programming, IEEE Transactions on Wireless
Communications, vol. 6, pp. 26402651, July 2007.

60
50
40

SSRM
ILM

30
20
10
0
0

10

15

20

25

30

35

40

P (dB)
Fig. 2. Sum-rate versus P for = 3, = = 4, = 2.

40

Average sum rate (bps/Hz)

35
30
25

SSRM
ILM

20
15
10
5
0
0

10

15

20

25

30

35

40

P (dB)
Fig. 3. Sum-rate versus P for = 3, = = 3, = 2.

V. C ONCLUSION
This paper has presented an optimization method for the
optimal transceiver design in MIMO interference channels.

22

You might also like