Download as doc, pdf, or txt
Download as doc, pdf, or txt
You are on page 1of 18

Outage Probability Analysis for MIMO Ad-hoc Network with

Quantized Beamforming
G.Ananthi, S.J.Thiruvengadam
TIFAC CORE in Wireless Technologies
Department of Electronics and Communication Engineering
Thiagarajar College of Engineering, Madurai 625015, Tamil Nadu, India
E-mail:gananthi@tce.edu,sjtece@tce.edu
Abstract: In this paper, exact closed form expressions are derived for sum
throughput/capacity and outage probability for Multiple Input -Multiple Output
(MIMO) ad-hoc network with quantized beamforming and finite rate feedback. In
quantized beamforming, each receiver sends the label of the best beamforming vector
obtained from the codebook to the transmitter through a finite rate limited feedback.
The code book is shared between each transmitter-receiver pair to prevent decoding of
other users information in a MIMO ad-hoc network comprising of K simultaneous
communicating transceiver pairs with each receiver employing single-user detection.
In such a practical environment, feedback links can only convey finite number of bits,
from which transmit beamformer designs are investigated using either the outage
probability or average Signal to Interference Noise Ratio (SINR) as the figure of
merit. Hence, by deriving closed form expressions for outage probability and network
sum throughput interesting insights in the high SINR regimes are provided. In lieu of
validating the obtained simulation results, theoretical results have also been
compared.
Index Terms - MIMO, Ad-hoc Networks, Network Sum Throughput, Outage
Probability, Quantized Beamforming, Finite Rate feedback, Code Book Vector.

1. Introduction
Multiple-Input and Multiple-Output (MIMO) wireless systems have attracted much
interest since they significantly increase the capacity of band-limited wireless
channels to meet the requirements of the future high data rate wireless
communications [1],[2]. Research interests on ad-hoc networks have greatly increased
in the past decade, mainly due to the challenging task of designing effective and
distributed protocols that yield a good throughput, fairness and energy saving abilities
[3],[4]. A mathematical model for determining the capacity of fixed multi-hop ad-hoc
networks was introduced in [3].

Prior results on the capacity analysis typically

focused on scaling laws of network throughput with an asymptotically large number


of nodes.
This paper focuses on mobile ad-hoc networks that have no fixed base station
along with multiple pairs of users intending to communicate with each other.
Employing multiple antennas in each transmit- receive pair has shown great promise
in providing higher throughputs on wireless links in comparison to the traditional
communication systems. However, practical limitations due to channel impairments
and distributed access, may decrease the performance of protocols for ad-hoc
networks [3],[4].
In [5], MIMO capacity with interference is studied assuming that single-user
detection is employed at the receiver. Without channel state information at the
transmitter(CSIT), it is shown that the transmitter should either put equal power into
each antenna (weak interference mode) or the transmitter should put all the power on
a single antenna (singular mode), depending on the interference to noise power ratio.
This concept is extended to MIMO ad-hoc networks with simultaneous pair wise
transmissions [6]. In MIMO ad-hoc networks without CSIT, the total capacity

of an ad-hoc network is fundamentally limited by the receive


antenna size and is independent of transmitting antenna size and
transmit power, as the number of transmitter- receiver pairs
increases [6]. However, with CSIT, throughput depends on both transmitting as
well as receiving antenna size. Further, with perfect CSIT, transmit beamforming
achieves best throughput [5]-[7]. But, in practical wireless systems, CSIT suffers from
imperfections originating from various sources such as quantization effects, feedback
delay and feedback errors. The effects of limited feedback in MIMO communication
systems have been analyzed in [8], [9]. However, this concept cannot be directly
extended to MIMO ad-hoc networks as it involves interference from the other
transmit-receive pairs. Recently, a limited feedback beamforming concept for MIMO
ad-hoc network was introduced in [10]. In this method, transmit beamformers of all
transmit-receive pairs are selected based on a centralized optimization technique to
maximize the sum rate. The performance of MIMO beamforming with quantized
feedback in ad-hoc network is analyzed in [11]. A random vector quantization (RVQ)
method is used to design the codebook and to derive the closed-form expressions for
the transmission capacity. It is proved that a moderate number of feedback bits are
sufficient to obtain significant transmission capacity gains compared to non-feedback
schemes. Although many complex schemes are employed for better performance,
RVQ method is used in ad-hoc network for analytical tractability. In [12], code book
is designed.using Grassmannian beamforming criterion to minimize the average SNR
reduction due to finite-rate feedback [13]. The closed-form expressions have been
derived for capacity loss due to quantized feedback in MIMO ad-hoc networks.
The objective of this paper is to study the impact of quantized beamforming with
limited feedback based on Grassmannian criterion, on the network sum

throughput/capacity and outage probability of MIMO ad-hoc networks, per unit time
and unit bandwidth for a given user. The beamforming vector is quantized at the
receiver using a fixed codebook designed based on Grassmannian beamforming
criterion at both the transmitter and the receiver. The receiver is assumed to convey
the index of the best beamforming vector from the codebook and is able to feed a
finite number of bits back to the transmitter. It is assumed that a low-bandwidth, errorfree, zero-delay feedback channel exists between the transmitter and the receiver. The
closed form expressions are derived for network sum throughput and outage
probability in MIMO ad-hoc networks.
This paper is organized as follows. Section II presents the system model of
MIMO ad-hoc network. Section III analyzes outage probability performance of
MIMO ad-hoc network with quantized beamforming and finite rate feedback.
Simulation results are discussed in section IV and section V concludes the paper.
Notations: Boldface capital letters denote matrices and boldface lower case letters
denote vectors. A H , and A denote the Hermitian (complex conjugate transpose) and
determinant of matrix of A . I k is a k k identity matrix , H kj denotes the channel
matrix from the j th transmitter array to the k th receiver array, and H k represents the
channel matrix from the k th transmitter array to the k th receiver array. In this paper,
throughput and capacity are used interchangeably, as both refer to mutual information
per unit time and unit bandwidth for a given system or user.

2. System Model
Consider a MIMO ad-hoc network with K user pairs. Each user pair is
assumed to have t transmit antennas and

r receive

antennas. The r 1 received

signal vector for k th user, in Rayleigh flat fading channel, is given by


y P k Rk / 2 H k c k sk Pkj Rkj / 2 H kjc j s j n k
k
jk

(1)

where c k is k th user pair beamforming vector of size t 1 determined from H k . H k


is a r t channel matrix, sk is the transmitted binary phase shift keying (BPSK)
symbol of k th user, k denotes the combined path loss/shadow fading for the k th user
pair channel, P is the transmit signal power, Rk is the distance between k th
transmitter and k th receiver and is the path loss exponent which can take value in
the range 2-4 taking values between 2-4. Also, H kj represents the r t channel matrix
between k th user receiver and jth user transmitter. c j is jth user-pair beamforming
vector of size t 1 determined from the r t H j , s j is the transmitted symbol of jth
interferer, kj denotes the combined path loss/shadow fading between k th user
receiver and jth user transmitter, Rk is the distance between the transmit and receive
array of kth user pair, Rkj is the distance between k th user receiver and jth user
transmitter and n k is (r x 1) independent and identically distributed (i.i.d) Gaussian
noise vector with zero mean and covariance 2 I r . In (1), the first term is desired
signal and the second term represents the multibeam interference from other users. It
is further assumed that each receiver implements single user detection.

3.

Outage Probability Performance with Quantized Beamforming


and finite rate feedback
It is assumed that the receiver is able to feed a finite number of B bits back to

the transmitter which is designed using Grassmannian beamforming codebook [9].


The feedback link is considered as error free and delay free. With B bits, each
transceiver pair has a total of N 2 B beamforming vectors w i , i 1, 2....N . The
beamforming vector for kth user c k , is selected such that the capacity is maximized [9].
It is given by,
c k (H k ) arg max H k w i
1i N

2
2

(2)

The received signal vector r 1 for kth user is,


g P k Rk / 2 H k c k sk P kj Rkj / 2 H kjc j s j n k
k
j k

(3)

The network sum throughput for K transmitter-receiver pairs is derived as

I PR /2H c c H H H
r k k kkk k

K
1

CQuan E log2 det

/ 2 H H
2

k1
Ir kj PRkj Hkjc jc j Hkj

jk

(4)
Singular Value Decomposition (SVD) of H kH H k is written as UU
k
c kH H kH H k c k c kH UU
K

H
k

. Thus,

cH k where U k is a unitary matrix and is a diagonal

matrix of eigen values 1 , 2 t of H kH H k . Let u1 , u 2 , ut be the eigen vectors

corresponding to

1 , 2 t

respectively. The entries of H k are assumed to be i.i.d

H
H
Gaussian with zero mean and unit variance. The statistical average of H k c k c k H k is

given by
(5)

Using [9], (5) can be written as

EHk H k c k c k H H k H EHk k ,max EHk u kH,max c k (H k )


(6)
2
t
H
H
H
EHk H k c k c k H k EH k i ui c k
i 1

where k , max and u k ,max are the maximum eigen value and corresponding eigen vector
2

H
of H k H k respectively. In (6) , first factor EHk k ,max indicates the channel quality

H
on average for K users, while the second factor EHk u k ,max c k (H k )

is an indication

of the beamforming codebook quality. Substituting (6) in (4), the capacity is


expressed as,

CQuan log 2 det


k 1

2
H
PR / 2 E

k
k
H k k ,max EH k u k ,max c H k

2
/ 2
EH j kj ,max
I r kj PRkj
j k

(7)

It can be noticed that the quantized beamforming t 1 vector c k does not coincide
with the eigen vector corresponding to the largest eigen value of the matrix. It leads to
quantization distortion measure which is the difference between the perfect channel
estimate network sum throughput and quantized network sum throughput [13].
Assuming that a maximal ratio combiner (MRC) is employed at the receiver,
the r 1 MRC weight vector is given by
r H k c k

(8)

c k is a t 1 quantized beamforming vector selected from the codebook according to


the Grassmannian beamforming criterion given in (2) . With these weights, the
received signal is given by
k = r H g k =(H k c k ) H g k

(9)

Where g k is the r 1 received signal vector with quantized beamforming in (3) and
substituting this expression in (9) , it is expressed as
k = Pk Rk / 2c k H H k H H k c k sk + Pkj Rkj / 2c k H H k H H kjc j s j +c k H H k Hn k
jk

(10)

H
where c k is the t 1 quantized beamforming vector calculated from H k H k , sk is the

transmitted symbol of k th user, c j is t 1 jth user-pair beamforming vector determined


from H j and s j is a transmitted symbol of jth interferer. The received SINR of the
MRC output is calculated as

P k 2 Rk c k H H k H H k c k

P
j k

Using (7) ,

the

2
kj

Rkj c k H H k H H kjc j + 2 c k H H k H

desired

1
t 1

signal

term

(11)

c k H H k H H k c k

is

approximated

as

1
2 max 1 [8],[13] and [14].Here, c k is the eigen vector corresponding to the

H
largest eigen value of H k H k using t number of transmit antennas and N is the

codebook size designed using Grassmannian Beamforming criterion[9]. Further,

c k H k

H 2

is

approximated

t 1
1

as
max
.

Interference

term

c k H k H kjc j

t 1
1

H
H

is approximated as max 1 H kj c j H kj c j . Substituting all

the above said approximations in equation (11) , the received SINR is written as
P k Rk
2

t 1
1

max 1

(12)

H
H
where = P kj Rk H kj c j H kj c j is a sum of independent exponential random
2

j k

variables and is treated as an interference from other user pairs. For simplicity, the

t 1
1

quantization error parameter is given as


. Conditioning on , (12)

can be written as

th
Pout ( th )=Pr max <
+ 2

2
-
Pk R k

The cumulative distribution function (CDF) of max is given by


Pr x

det (c )x

s k 1 z k 1

(13)

(14)

k 1

where z = min(t,r) and s = max(t,r) , where t refers to the number of transmitting


antennas and r is the number of receiving antennas. . is a gamma function,
c x is a Hankel matrix of size z z given by

x
c

i, j

s z i j 1, x , i, j 1, 2...z

(15)

n 1

Here, n, x n n
k 0

xk x
e which is an incomplete gamma function, when
k!

n is a positive integer. It can be observed that the distribution of max depends only
on z and s .
Now, solving for the determinant in (14) , the distribution max is represented by the
following polynomial
z(s-1) z

Fmax (x)=P(x,z,s)= c lk (z,s)x l e-kx


l=0 k=0

where clk z, s depends only on

z and s .

(16)
The values of clk z, s coefficients are

obtained from (14), (15) and (16) . Applying the polynomial of (16) and using
binomial theorem, (13) can be rewritten as
Pout th

z s 1

l 0


i 0 i
l

k 2

th
th
th
Pk 2 Rk Pk 2 Rk
2
c
z
,
s

e
e

lk
2

k 0
P k Rk

th
th 2

2

2

P k Rk Pk Rk

l i

(17)
Then, the expression for outage probability is obtained by averaging over the
distribution of . It is given by,
z(s-1) z

Pout ( th )= clk z , s e
l=0 k=0

-k th 2

Pk 2 Rk


i=0 i
l

th 2

2
-
Pk R k

(l-i)

-k

th
th
Pk 2 Rk
e
f ( )d
0 Pk 2R k -

(18)
It shows that the outage probability for a MIMO ad-hoc network depends on the
probability density function (PDF) f ( ) of the multi-user interference power.
The characteristic function of interference power in a MIMO ad-hoc network

2 -j/b 2/b
j x
is expressed as x ( )=E e =exp - eff 1- e in [15-17].
b

Where

10

eff K / RO 2 refers to as effective node density and b 4 for ground propagation.


The closed-form expression for PDF of interference power is given by [17]

-3/2
f x (x)= xeff exp
2

-3 eff 2

4x

x0

(19)

Substituting (19) and the integral expression [18]

-
-x
v-1 x

Re 0; Re 0

2
dx=2 K v (2)

in (18) the closed-form expression for outage probability for MIMO

ad-hoc networks is expressed as

Pout ( th )

-k th 2

z(s-1) z

c z, s e

Pk Rk

lk

l=0 k=0

l
th 2

K
2 -a a


2
i=0 i
Pk R k
l

(l-i)

a 2

eff
a1

i-1/2
2
i-1/2

1 2

(20)
where the parameters a1 and a 2 are given by a1 =

k th
3 eff 2
,
a
=
2
Pk 2 R k -
4

t 1
1

1 , K i-1/2 2 a2 a1 is the Bessel function of Imaginary argument.

4. Results and Discussion


In this section, the impact of quantized beamforming with finite rate feedback
on network sum throughput and outage probability is investigated by simulations. We
analyze two different schemes such as transmit beamforming with perfect CSIT and
quantized beamforming with finite rate feedback. Simulation results are obtained,
with the assumption that the channel matrix from the j th user transmitter array to the

k th user receiver array H kj consists of i.i.d. complex Gaussian entries with zero mean
and unit variance. The channel matrices are independent across different transmitreceive pairs. The path loss/shadowing effect is summarized using the coefficient

kj

following lognormal distribution. The simulation results correspond to the results

11

obtained in simulated ad-hoc transmit receive pairs with Grassmannian codebook.


The outage probability performance is simulated for various values of SINR and is
compared with the analytical results. The sum throughput is obtained by averaging
over 50 sets of independently generated channel matrices and
shadowing coefficients for all transmitreceive pairs.
Fig.1, shows the performance of sum throughput as a function of user pairs for
transmit beamforming with perfect CSIT and quantized beamforming with finite rate
feedback. The code book size for the quantized beamforming is assumed to be 4. The
sum throughput for quantized beamforming is always less compared to transmit
beamforming with perfect CSIT. Fig.2, shows the performance of sum throughput as a
function of SINR when 4 user pairs are active. Each user pair is assumed to have 2
transmit antennas and 2 receive antennas. The code book size for the quantized
beamforming is assumed to be 4.It is observed that sum throughput for both quantized
beamforming and transmit beamforming with perfect CSIT gets increased
substantially when SINR increases. Fig.3, shows the outage probability performance
of perfect CSIT and quantized beamforming with finite rate feedback for codebook
sizes of N 4 and N 256 . The data rate of the system is set at 2 bits/s/Hz. As the
size of the codebook increases, the outage probability decreases. The simulation
results are obtained by modelling the system as in (6) and the theoretical
approximation results are obtained using (20) directly. Fig. 4, shows the outage
probability performance of a MIMO ad-hoc networks for Quantized beamforming
with finite rate feedback for transmit-receive antenna configurations such as (2,2),
(2,4),(4,4). The data rate is assumed to be 2 bits/s/Hz and the code book size is 256.
As the antenna size increases, the outage probability performance of the system
improves. This is due to the improvement in the inherent diversity/array gain of the

12

system. Fig. 5, shows the outage probability performance as a function of data rate
for perfect CSIT and quantized beamforming with finite rate feedback of codebook
size N 4 and Transmit Receive pairs K=2 at a fixed SINR of 10dB. Further, it is
observed that, as the distance between each transmitter and each receiver increases by
2 times, outage probability increases from 0.15 to 0.5.

5. Conclusion
The closed form expression for network sum throughput/outage probability of MIMO
ad-hoc network with quantized beamforming and finite rate feedback has been
derived. The outage probability performance of the MIMO ad-hoc network has been
analyzed with respect to the parameters such as the size of codebook, size of transmit
& receive antenna, number of user pairs and the distance between the transmit
receive pairs. The code book is designed using the concept of Grassmannian criterion
and it is assumed that the feedback link from the receiver to the transmitter is error
and delay free. The simulation results presented in this paper confirm the accuracy of
the obtained analytical results. In future, the interference from the other user pairs
may be considered for feedback link to obtain more practical solutions.

References
[1] G. J. Foschini and M. J. Gans.: On limits of Wireless Communications in a
fading environment when using Multiple antennas, Wireless Personal
Communication., Kluwer Academic press, 1998, vol. 6, pp. 311335.
[2] E.Telatar.:

Capacity

of

multi-antenna

Gaussian

channels,

European.

Transactions on Telecommunications., Nov./Dec 1999, vol. 10, pp. 585595.


[3] P.Gupta and P.R.Kumar.: The capacity of wireless networks, IEEE Transactions
on Information Theory, Mar. 2000, vol.46, pp.388-404.

13

[4] S. Toumpis and A.J. Goldsmith.: Capacity regions for wireless ad hoc networks,
IEEE Transactions on Wireless Communication. Jul. 2003, vol.2, pp. 736-748.
[5] R.S. Blum.: MIMO capacity with interference, IEEE. Journal on Selected Areas
in Communication. June 2003, Vol. 21, pp.793-801.
[6] B. Chen and M.J. Gans.: MIMO Communications in Ad Hoc Networks, IEEE
Transactions on Signal Processing, July 2006, vol.54, no.7 pp.2773-2783.
[7] Paulo Casari, Marco Levorato and Michele Zorzi.: MAC/PHY Cross Layer
Design of MIMO Ad-hoc Networks with layered Multi-user detection, IEEE
Transactions on Wireless Communication, Nov 2008, Vol 7,No 11.pp.4596-4607.
[8] S. Zhou, Z. Wang, and G. B. Giannakis.: Quantifying the power loss when
transmit beamforming relies on finite-rate feedback, IEEE Transactions on
Wireless Communications, Jul. 2005, vol. 4, No.4,pp. 19481957.
[9] D.J. Love, R.W. Heath Jr., T. Strohmer.: Grassmannian beamforming for
multiple-input and multiple-output wireless systems, IEEE Transactions on
Information Theory, Oct.2003, vol.49, no.10, pp.2735-2747.
[10]

Jong Ho Lee, Ye (Geoffrey) Li.: Iterative Limited feedback Beamforming

for MIMO Ad-hoc Networks, IEEE International Conference on Global


Communications (GLOBECOM) 2009.
[11]

Yueping Wu, Raymond H.Y.Louie, Matthew R Mckay,Iain B.Collings.:

MIMO Beamforming with quantized feedback in Ad-hoc networks: Transmission


capacity analysis, IEEE International Conference on Signals, Systems and
Computers,(ASILOMAR) 2010.
[12]

G.Ananthi, S.Annie Shalom, S.J.Thiruvengadam.: Performance Analysis of

MIMO Ad-hoc Networks with Quantized Beamforming and Imperfect Channel

14

State Information, IEEE International conference on Signal Processing and


Communications (SPCOM) 2010.
[13]

Bishwarup Mondal.: Grassmann Quantization for precoded MIMO systems

Ph.D Thesis, The University of Texas, Austin, 2006.


[14]

Juan M. Romero-Jerez, Juan P. Pea-Martn, Gabriel Aguilera and Andrea J.

Goldsmith.:

Performance

of MIMO

MRC Systems

with Co-Channel

Interference, IEEE ICC (International Conference on Communications) 2006


Proceedings.
[15]

E.S Sousa and J.A Silvester.: Optimum Transmission ranges in a direct

sequence spread spectrum multihop packet radio network IEEE Selected areas in
Communications, June 1990, vol.8, No.5, pp-762-771.
[16]

Kaibin

Huang.:

MIMO

Networking

with

Imperfect

channel

state

information, Ph.D Thesis, The University of Texas, Austin, May 2008.


[17]

Kostas Stamatiou, John G. Proakis and James R. Zeidler.:

Information Efficiency of Ad-hoc Networks with FH-MIMO Transceivers,


IEEE ICC(International Conference on Communications) 2007 Proceedings
[18]

I. S. Gradshteyn, and I. M. Ryzhik.: Table of Integrals, Series, and Products,

Orlando, FL: Academic Press, 1994, 5th edition.


[19]

Ming Kang and Mohamed-Slim Alouini.: A Comparative study on the

performance of MIMO MRC Systems with and without channel interference,


IEEE Transactions on Communications, August 2004 ,Vol.52,No.8.pp.1417-1425

15

Fig. 1. Sum Throughput of a MIMO ad-hoc network with t = r =2, P= 2, N=4 and

K 4, t r 2 , N=4 and

Fig. 2. Sum Throughput of a MIMO ad hoc network with

=1

=1

16

Fig. 3. Outage Probability Performance of a MIMO ad hoc network with

K 2, t r 2 and

=1,

Data rate=2bits/s/Hz.

Fig. 4. Outage Probability Performance of a MIMO ad-hoc network for antenna configurations (t,r) :
(2,2),(2,4),(4,4), K=2 and

=1, Data Rate=2bits/s/Hz and codebook size N= 28 .

17

Fig. 5. Outage Probability Performance of a MIMO ad-hoc network with

t r 2 and

=1,

K=2 ,SINR=10dB and codebook size N= 4 for different transmit and receive distances.

18

You might also like