Power Allocation in Cooperative Networks Using

Differential Space-Time Codes

Behrouz Maham and Are Hjrungnes
UniK University Graduate Center, University of Oslo, Oslo, Norway
Email: behrouz@unik.no,arehj@unik.no
AbstractEstimating channel state information (CSI) in the
fast fading conditions is very challenging. In this paper, a simple
structure for cooperative diversity in decode-and-forward mode
is examined, in which the transmitter, the relay and the receiver
do not require to know CSI. In this scheme, transmission
of information is done in a two phase process. In the rst
phase, differential modulated signals are radiated from the
source. After decoding the received signal in the relay side,
the source and the relay collectively send information using
differential space-time codes. The optimum power allocation
in a sense of minimizing pairwise error probability (PEP) in
high SNR scenarios has been obtained using an analytical
method. An interesting property of the optimum point is that it
is independent of the channel statistics and the position of the
relay, which improves its feasibility in the future communication
systems. Numerical simulations verify the analytical results.
Index Terms Cooperation, differential space-time codes,
decode-and-forward, power control.
I. INTRODUCTION
Cooperative diversity allows a collection of radios to relay
signals for each other and effectively create a virtual antenna
array for combating multipath fading in wireless channels. A
majority of the work on cooperative diversity has focused on
scenarios in which the receivers, and perhaps the transmitters,
obtain channel state information (CSI) in the form of accurate
estimates of the fading coefcients (see, e.g., [1]-[5]). The
receivers can exploit available CSI for coherent reception, and
the transmitters can utilize available CSI for power control
and coherent beamforming. Estimating channel coefcients is
achieved by using the training sequences or blind methods,
but when channel conditions are affected by fast fading, the
estimation of CSI becomes inaccurate and impractical. For
the case that CSI is unknown in both the transmitters and
the receivers, some few schemes are presented. Recently,
the non-coherent detection is investigated in a relay-based
network with repetition codes [6]. Another way of dealing
with the issue is to employ differential modulation schemes.
Differential space-time coding has been considered in [7]-
[10] when the CSI is not available at either MIMO source
and destination terminals. The extension of this method to
the cooperative systems is considered in [11]. Furthermore,
[12] and [13] studied on differential modulation schemes in a
distributed manner with repetition-based codes. An introduced
scheme in [13] operates under BPSK. In addition to the
repetition codes, [12] studies the usage of differential unitary
This work was supported by the Research Council of Norway through
the project 176773/S10 entitled "Optimized Heterogeneous Multiuser MIMO
Networks OptiMO".
Fig. 1. Communication between source and destination by employing a relay.
space-time codes in a cooperative manner. But, instead of the
dense orthogonal space-time codes which are studied in this
paper, it uses diagonal matrices as a unitary matrix. Moreover,
decoding at the destination side in [12] is just based on
transmitting signals from relays in phase II, which is not
efcient.
In this paper, we develop a distributed differential space-
time system based on [11], which uses Alamouti scheme as
a unitary matrix. Decoding in the destination is based on
the transmitting M-PSK differentially modulated signals in
both phases and the source node contributes in transmitting
signals in the second phase. Then, we analyze the pairwise
error probability (PEP) of the system. A simple power control
scheme based on minimization of PEP that is independent
of the channel statistics and the position of the relay is
considered.
This paper is organized as follows. In Section II, the system
model is given. Then in Section III, a distributed version
of space-time codes where we employed in our system is
considered. Section IV discusses the PEP analysis in high
SNR scenarios and an optimum power allocation. Finally,
the overall performance comparisons for the system and the
proposed power allocation are presented in Section V.
II. SYSTEM MODEL
Consider a network consisting of a source, a relay, and
a destination node (Fig. 1). It is assumed that each node is
equipped with a single antenna.
Suppose each link has Rayleigh distribution and is inde-
pendent of the others. We denote the source-to-destination,
source-to-relay and relay-to-destination links by SD, SR,
and RD, respectively. Both the destination and the relay
are affected by zero-mean additive white complex Gaussian
noise with the variance of N
0
. In what follows, SD, SR,
and RD links are demonstrated by h
SD
(n), h
SR
, and h
RD
,
respectively. The variances of the related channel coefcients
are denoted by
2
SD
,
2
SR
, and
2
RD
. Path-loss in Rayleigh fading
channel is equal to 1/d

where is the path-loss exponent.

Typically, lies in the range of 2 to 5 [14]. In contrast to
space-time codes in which transmit antennas usually send the
equal power, in distributed space-time codes for improving
the performance of the system, the available power is usually
allocated as a function of the relays position. Note that, the
proposed protocol in this paper exploits all degrees of freedom
for the decoding of the received signal. This implies that in
contrast to the proposed protocol in [4] in which just utilize the
transmitted information through the second phase, we employ
information radiated in both phases for the decoding purpose.
On the other hand, in contrast to the protocols in [3] in
which the source does not contribute in the second phase of
transmission, in our protocol the source contributes to transmit
data as well as relays in the second phase.
In this paper, the transmitted power in the rst phase which
is radiated by the source is denoted by P
1
, and the transmitted
power in the second phase which is radiated by both the source
and relay is represented by P
2
. In the second phase, P
2
and (1 )P
2
are transmitted by the source and the relay,
respectively, where 0 1.
Since DF algorithm is used in this paper for transmitting
the data from the relay, it is necessary for the decoded data
in the relay to be almost without error. For achieving this
goal, transmitted power P
1
is adjusting in such a way that the
average received SNR in the relay, i.e., SNR
R
= P
1
/(
2
n
d

SR
),
exceeds from a suitable threshold.
III. DISTRIBUTED DIFFERENTIAL SPACE-TIME CODE
In this section, we describe the distributed differential space-
time code using the Alamouti structure. In this model, we
extend the differential space-time code proposed in [9] and
investigate its distributed version.
The structure proposed in [9], when using the 2 2
Alamoutis scheme, can be shown as
S(0) = I
2
, S(n) = S(n 1)C(x
1
, x
2
), n = 1, 2, . . . (1)
where the 22 matrix S(0) is an arbitrary unitary matrix that
could be the identity matrix I
2
. The matrix C(x
1
, x
2
) has the
Alamouti structure and consists of symbols x
1
and x
2
chosen
from M-PSK constellation. The time index n is skipped for
simplifying in representation of x
1
and x
2
.
In the following, we explain the distributed form of the
differential structure described in (1). In this model, at the
rst phase, the source S transmits the second row of the matrix
S(n) for R and D. In the second phase, S and R nodes transmit
the scaled rows of the matrix S(n) with appropriate power
ratios, like two antennas belong to a node. We can express the
stated expressions as a following matrix form
S
DDST
(n) =
_
s
21
(n) s
22
(n)
1
s
11
(n)
1
s
12
(n)
0 0
2
s
21
(n)
2
s
22
(n)
_
,
(2)
where s
ij
(n), i, j {1, 2} are the components of S(n). Fur-
thermore, two real coefcients
1
and
2
fulll the equation
P
2
= P
1
(
2
1
+
2
2
). These coefcients are introduced to have a
degree of freedom in distributing the power between S and R,
under a given total transmit power. The parameter introduced
in Section I is related to
1
and
2
by = P
1

2
1
/P
2
.
Assuming r
1
(n) and r
2
(n) are the received signals in the
rst phase, and r
3
(n) and r
4
(n) are the received signals in the
second phase. Stacking the signals received at the destination
terminal to form a 1 4 vector r(n), we obtain the following
input-output relation
r(n) =
_
P
1
h(n)S
DDST
(n) +w(n), (3)
where
r(n) =
_
r
1
(n) r
2
(n) r
3
(n) r
4
(n)

, (4)
h(n) =
_
h
SD
(n) h
RD
(n)

. (5)
The 1 4 vector w(n) consists of the zero-mean complex
Gaussian noise components with the covariance matrix of
E{w(n)
H
w(n)} = N
0
I
4
, where (.)
H
stands for the complex
conjugate transpose and I
4
denotes a 4 4 identity matrix.
Using (1) and (2) we have
S
DDST
(n) = S
DDST
(n 1)(I
2
C(n)), (6)
where is Kroncker product. Note that, since the matrix
(I
2
C(n)) is a unitary matrix,
1
and
2
in (2) are time-
invariant parameters. If we regard (3) for two successive inter-
vals, and on the other hand, with assumption that the channel
characteristics remains unvarying, i.e., h(n) h(n1), using
(6) we have
r(n) = r(n 1)(I
2
C(n)) + w(n), (7)
where
w(n) = w(n) w(n 1)(I
2
C(n)). (8)
It can be easily checked that w(n) is a zero-mean Gaus-
sian noise with the covariance matrix of E{ w(n)
H
w(n)} =
2N
0
I
4
, where noise variance is doubled in comparison with
the non-differential case. This is in conjunction with the con-
ventional 3dB performance loss in systems with the differential
detection.
IV. PERFORMANCE ANALYSIS AND POWER CONTROL
STRATEGY
For simplicity in expressing equations, we skip the time
index n from the matrix C(n). Thus, the coded matrix C(n)
is shown by C
1
. If an error occurs in detecting C
1
, we
demonstrate it as C
2
.
An interesting property covered up in (7) is that we can
represent it as a non-differential structure with known channel
information where r(n 1) is equivalent to the channel
coefcients vector, and suppose the noise power is doubled.
On the other hand, the matrix (I
2
C(n)) is unitary. Thus, we
can employ pairwise error probability relationships in space-
time codes context [15]
P (C
1
C
2
| r(n 1))
= Q
_
E
s
2
2
b w
_
_
_
C
2
C
1
_
r(n 1)
_
_
F
_
, (9)
where .
F
is the Frobenius norm, and C
1
and C
2
are as
C
1
= (I
2
C
1
), (10)
C
1
= (I
2
C
1
), (11)
and
2
b w
is the noise variance in (7), and it is equal to 2N
0
. For
simplifying (9), we suppose high SNR conditions. Therefore,
by neglecting the noise term in (3), we have
_
_
_
C
2
C
1
_
r(n 1)
_
_
F

P
1
hS
DDST
(n 1)BS
DDST
(n 1)
H
h
H
, (12)
where
B =
_
C
1
C
2
_ _
C
1
C
2
_
H
. (13)
Now, we can normalize channel coefcients, or in other words,
separate the path-loss components from the fading channel.
Hence, (12) can be considered as
_
_
_
C
2
C
1
_
r(n 1)
_
_
F
hAh
H
, (14)
where, h is a 1 2 vector, and has zero-mean complex
Gaussian elements and unit variance. And,
A = P
1
LS
DDST
(n 1)BS
DDST
(n 1)
H
L, (15)
where,
L =
_

SD
0
0
RD
_
. (16)
Thus, PEP is given by [15]
P (C
1
C
2
) =
E[P (C
1
C
2
| r(n 1))]
1

2
i=1
_
1 +
Es
8N0

i
_, (17)
where
i
is the eigenvalues of the matrix A. Under high SNR
scenarios, we have
P (C
1
C
2
)
2

i=1
_
E
s
8N
0

i
_
1
=
_
E
s
8N
0
_
2
det
1
(A) . (18)
where det (A) stands for the trace of the matrix A.Thus, the
minimum of the probability of error is obtained by maximiz-
ing det (A)
max det (A) =
max det (B) det
_
P
1
LS
DDST
(n)S
DDST
(n)
H
L
_
max P
1
det (B) (d
SD
d
RD
)
_
1 +
2
1
_

2
2
(19)
since
S
DDST
(n)S
DDST
(n)
H
=
diag
__
|s
21
|
2
+|s
22
|
2
_
+
2
1
_
|s
11
|
2
+|s
12
|
2
_
,
2
2
_
|s
21
|
2
+|s
22
|
2
_
=
_ _
1 +
2
1
_
0
0
2
2
_
, (20)
where diag(.) denotes a diagonal matrix. One can verify (20)
considering (1) and the fact that multiplying two unitary
matrices become a unitary matrix. Therefore, for minimizing
the probability of error in high SNR scenarios, it is just enough
to maximize
_
1 +
2
1
_

2
2
, which leads to a simple power
control strategy that is independent of code and channel states.
As it is stated in Section II, the power P
1
is chosen in
order to achieve a required average SNR
R
at the relay, dened
as SNR
R
= P
1
/(
2
n
d

SR
). As a result, P
2
,
1
, and
2
can
be obtained by the following three equations, given the total
power in the network to be constant.
_
1 +
2
1
_
=
2
2
, (21)
P
2
= P
1
_

2
1
+
2
2
_
, (22)
P
1
+P
2
= P. (23)
Since = P
1

2
1
/P
2
, when P
1
P
2
, we have
=
1
2

P
1
2P
2
. (24)
It is important to note that this optimum point is independent
of the channel statistics and the position of the relay, which
improves its feasibility in the future communication systems.
When P
1
P
2
, = 0 can be used. This means that all
the power in the second phase is transmitted by the relay
node. Although in this case the diversity of the second order is
achieved, but the usage of relay may decrease the performance
of the system. This is because the fact that the required P
1
for the errorless condition in the relay side could be high.
Thus, non-cooperative transmission may be preferable in this
case. Therefore, when P
1
P
2
, in high SNR conditions,
we can transmit all the power in the second phase from the
selected relay, and in low SNR conditions non-cooperative
transmission can be used. However, as we emphasis on high
SNR channels in our analysis, the amount of P
1
which is
obtained by the required SNR
R
is low enough in most of cases
to avoid P
1
P
2
case.
V. SIMULATION RESULTS
In this section, the performances of distributed differential
space-time codes are studied through simulations. The error
event is bit error rate (BER). In all simulations, 500 informa-
tion symbols are used in each frame. The fading is assumed to
be constant over each frame and vary independently from one
frame to another. The fading channels are at with Rayleigh
distribution in which path-loss is also regarded. The path-
loss exponent is set to 2. The results are presented for
QPSK. Moreover, for position of the nodes, we have supposed
d
SD
= 1 and d
RD
= 0.75. Furthermore, it is assumed that
the relay decodes the received information without error. This
assumption is easily achievable by reducing the SR link or
increasing the required SNR
R
.
Fig. 2 conrms the analytical results attained in Section IV
for nding the optimum of . In this gure, is dened as the
ratio of the transmitted power in Phase II to the transmitted
power in Phase I, i.e., = P
2
/P
1
. Then, substituting the
values of equal to 1, 3 and 5 in (24) lead to the optimum
values of as = 0, = 1/3, and = 0.4, respectively. One
observes in Fig. 2 that the optimum values of via analytical
result lead to the optimum points in a sense of minimizing the
Fig. 2. The average BER curves versus of a distributed differential
space-time system with the employment of different detection with when the
destination SNR is 20 and d
RD
= 0.75.
BER by Monte-Carlo numerical simulations. This conrms the
correctness and usefulness of the brief equation of (24).
Fig. 3 illustrates the average bit error rate (BER) curve
as function of the average SNR per symbol for different
power allocations. Two different values of the parameter
are employed. In each of them, the optimum power allocation
based on (24) is compared with the equal power allocation
to the source and the relay during the second phase, i.e.,
= 0.5. Note that the total transmitted power in network
is assumed to remain constant in all cases. Comparing two
curves with = 1 demonstrates a 2 dB gain in SNR at the
BER of 10
3
, when optimum power allocation is utilized.
By increasing the value of to 2, the better performance is
achieved. But, the performance gain due to using the opti-
mum power allocation becomes less signicant, because the
optimum value of becomes close to 0.5. Note that the value
of is assigned by the required SNR
R
and considering the
fact that the total transmitted power is constant. Furthermore,
Fig. 3 demonstrates that our proposed system outperforms the
distributed differential scheme proposed in [12] in DF mode.
Since one relay is utilized in our simulations, two differential
schemes introduced in [12] have the same structure.
VI. CONCLUSION
In this paper, a cooperative diversity scheme using dis-
tributed differential space-time codes in DF mode is proposed,
and examined. An outstanding characteristic of this scheme is
that CSI knowledge is not required in both the transmitters and
the receivers. Performance of the system is analyzed in high
SNR scenarios, and nally a simple and feasible power control
strategy is derived, which is independent of the relay position.
The performance of the system is investigated via numerical
simulations which demonstrate the proposed power allocation
outperforms a system with equal power transmission.
Fig. 3. The average BER performance of a distributed differential space-time
system with the employment of different power allocations.
REFERENCES
[1] A. Sendonaris, E. Erkip and B. Aazhang, "User cooperation diversity.
Part I. System description," IEEE Transactions on Communications, vol.
51, no. 11, pp. 19271938, Nov. 2003.
[2] A. Sendonaris, E. Erkip and B. Aazhang, "User cooperation diversity. Part
II. Implementation aspects and performance analysis," IEEE Transactions
on Communications, vol. 51, no. 11, pp. 19391948, Nov. 2003.
[3] J. N. Laneman and G. W. Wornell, "Distributed space-time-coded pro-
tocols for exploiting cooperative diversity in wireless networks," IEEE
Transactions on Information Theory, vol. 49, no. 10, pp. 24152425,
Oct. 2003.
[4] Y. Jing and B. Hassibi, "Distributed space-time coding in wireless relay
networks," in IEEE Sensor Array and Multichannel Signal Processing
Workshop, Sitges, Spain, July, 2004.
[5] R. U. Nabar, H. Bolcskei, and F. W. Kneubuhler, "Fading relay channels:
Performance limits and space-time signal design," IEEE Journal on
Selected Areas in Communications, pp. 10991109, Aug. 2004.
[6] D. Chen and J.N. Laneman, "Cooperative diversity for wireless fading
channels without channel state information," in Proc. Asilomar Conf.
Signals, Systems, and Computers, Monterey, CA, Nov. 2004.
[7] B. L. Hughes, "Differential space-time modulation," IEEE Trans. Inform.
Theory, vol. 46, pp. 25672578, Nov. 2000.
[8] B. M. Hochwald and T. L. Marzetta, "Unitary spacetime modulation for
multiple-antenna communications in Rayleigh at fading," IEEE Trans.
Inform. Theory, vol. 46, pp. 543-564, March 2000.
[9] B. M. Hochwald and W. Sweldens, "Differential unitary space-time
modulation," IEEE Trans. Commun., vol. 48, pp. 20412052, Dec. 2000.
[10] V. Tarokh and H. Jafarkhani, "A differential detection scheme for
transmit diversity," IEEE J. Select. Areas Commun., vol. 18, pp. 1169
1174, July 2000.
[11] B. Maham and F. Lahouti, "Cooperative communication using dis-
tributed differential space-time codes," in IEEE International Confer-
ence on Wireless Communications, Networking and Mobile Computing
(WiCOM 2006), Wuhan, China, Sep. 2224, 2006.
[12] W. Cho and L. Yang, "Distributed differential schemes for cooperative
wireless networks," IEEE ICASSP, 2006.
[13] Q. Zhao, H. Li, "Decode-based differential modulation for wireless relay
networks," IEEE ICASSP, Philadelphia, PA, March 2005.
[14] M. K. Simon and M.-S. Alouini, Digital Communication over Fading
Channels: A Unied Approach to Performance Analysis. New York, NY:
Wiley, 2000.
[15] H. Jafarkhani, Space-Time Coding Theory and Practice, Cambridge
University Press, 2005.