Вы находитесь на странице: 1из 5

On the Power Allocation for Decode-and-Forward Cooperative Transmission

Over Rayleigh-Fading Channels



Lian Zhao and Zaiyi Liao
Electrical and Computer Engineering Department
Ryerson University, ON, Canada, M5B 2K3
Abstract
As an accompany paper to [1], efcient power allocation
strategies for cooperative transmission applying decode-and-
forward (DF) approach are further investigated in this pa-
per. The cooperative ratio, dened as the ratio of the power
used for cooperative-information transmission to the total
power, is investigated in an attempt to minimize bit error
rate (BER) with a constraint of xed total transmit power
for each user. Our results show that efcient power alloca-
tion greatly depends on the cooperative method. The pre-
ferred cooperative ratio changes from low to high by us-
ing DF with no parity check, Amplify-and-forward, DF with
parity check. Simulation results show that with appropri-
ate power allocation, BER performance of all cooperative
schemes can achieve a signicant gain over non-cooperative
systems.
I. INTRODUCTION
Diversity techniques are important strategies to combat
fading occurred in wireless channels. Recently, there are
great research interests on multipath/multiuser diversity,
which can be realized by exploring cooperative transmis-
sion (see for example, the tutorial paper [2] and the ref-
erences cited in). The basic idea of cooperative transmis-
sion is to create multiple channels for each user, while
each user is equipped with only one transmit antenna, to
deliver his/her information to the destination. As a result,
spacial diversity gain can be achieved in the presence of
channel fading. The transmission quality, measured as bit
error rate (BER) at the destination can be improved since
statically speaking, it is less likely that several channels
are fading low simultaneously.
There are mainly two different kind of operations at
the partners (or relays) and the cooperation is then dis-
tinguished as amplify-and-forward (AF) and decode-and-
forward (DF). AF is sometimes also named as nonregener-
ative, where the partners amplify and forward the received
signal without any decoding. DF, on the other hand, also
known as regenerative relaying, the partners decode the

This work has been supported by the Natural Sciences and Engineer-
ing Research Council (NSERC) of Canada under grant number 293237-
04.
signal, recode and send to the next terminal. We will
mainly focus on DF in this paper.
Power control/power allocation has long been playing
an important role in wireless networks to dynamically
combat channel uctuations and control the co-channel
interference [3]. Recently, power allocation problem in
a cooperative system has also attracted lots of research at-
tention. In [4], optimal power and bandwidth allocation
algorithm is solved to maximize the overall system capac-
ity for FDMA-based DF multihop links. In [5], optimal
power allocation is solved to minimize the outage prob-
ability, where iteration is required to obtain the optimal
power allocation for DF system with diversity. While for
AF with diversity, the authors suggest to use the same so-
lution as that of DF due to difculty/complexity in for-
mulation. A more comprehensive survey can be found in
the accompany paper [1], where efcient power allocation
strategy has been discussed for AF cooperations.
In this paper, two methods of DF are considered:
decode-and-forward with parity check (DFWP) and
decode-and-forward without parity check (DFNP). For
DFWP, the partner decodes the symbol received in phase
I, and forwards to the destination if the symbol can be de-
coded correctly assuming using parity check. Otherwise,
the partner does not send and stay in idle; while for DFNP,
the partner decodes what has been received in phase I and
forwards the decoded symbol to the destination without
parity check. Therefore, when the partner decodes erro-
neously, this error propagates.
II. SYSTEM MODEL
A cooperative transmission scheme, consisting of two
phases, is considered. Phase I is referred as self-
information transmission phase, and phase II is
cooperative-information transmission phase. In this
work, CDMA method is applied to create orthogonal
channels as proposed in [6]. Data modulation applies Bi-
nary Phase Shift Keying (BPSK) scheme. The generic
model is shown in Fig.1. The rst and the second sub-
script of the signals distinguish the indexes for the nodes
and the phases respectively (for example, Y
ij
denotes the
received signal for node i at phase j). Let P
ij
(i, j = 1, 2)
1-4244-0264-6/07/$25.00 2007 IEEE
denote the transmit power level of user i at phase j. The
powers, P
11
and P
21
, are the allocated powers to transmit
users self-information directly to the destination; while
P
12
and P
22
are the powers for cooperative transmission
in phase II.
0
1
2
h3
h1
h2
Figure 1: Model for user cooperation.
Similar with [1], let (0 1) be the cooperative
ratio, which implies the ratio of the power used for cooper-
ative transmission over the total power. Thus, P
12
= P
and P
11
= (1)P, where P is the total power of user 1.
For simplicity, it is assumed that user 2 has the same total
power and the same cooperative ratio as user 1 consider-
ing fairness as discussed in [1]. Therefore, P
22
= P and
P
21
= (1 )P. In phase I, the transmitted signals for
mobile user 1 and 2 are respectively given as:
X
11
=
_
P
11
b
1
c
1
(1)
X
21
=
_
P
21
b
2
c
2
(2)
where b
1
denotes user 1s information bit and c
1
denotes
user 1s spreading sequence. The spreading sequence is
selected to be unit energy and be orthogonal to each other,
i.e., c
T
i
c
j
= (i j), where c
T
i
denotes the transpose of
column vector c
i
.
The received signals at user 1, user 2 and the destination
in phase I are given as
Y
11
=
_
h
3
X
21
+n
1
=
_
h
3
P
21
b
2
c
2
+n
1
(3)
Y
21
=
_
h
3
X
11
+n
2
=
_
h
3
P
11
b
1
c
1
+n
2
(4)
Y
01
=
_
h
1
X
11
+
_
h
2
X
21
+n
0
=
_
h
1
P
11
b
1
c
1
+
_
h
2
P
21
b
2
c
2
+n
0
(5)
=
1
b
1
c
1
+
2
b
2
c
2
+n
0
where
1
=

h
1
P
11
and
2
=

h
2
P
21
. The terms,
h
1
, h
2
, h
3
, are the fading power gain for the channels of
mobile user 1 to the destination, mobile user 2 to the des-
tination, and the inter-channel of the mobile users, re-
spectively. A reciprocal inter-channel gains have been
assumed. Each h
i
follows exponential distribution with
mean H
i
, (i = 1, 2, 3). Therefore, the amplitude gain

h
i
, (i = 1, 2, 3) follows Rayleigh distribution. We fur-
ther assume a at fading channel that in one cooperation
interval, the fading gain doesnt change and is indepen-
dent with each other and independent for each cooperation
interval. Additive white Gaussian noise terms are repre-
sented in n
i
, i = 0, 1, 2, seen at the destination, user 1
and user 2, respectively. They are modelled as zero-mean,
complex Gaussian random variables with a common vari-
ance
2
N
. The transmitted signals at the partners are
X
12
=
_

P
12

b
2
c
2
(6)
X
22
=
_

P
22

b
1
c
1
(7)
where

P
12
= P
12
for DFNP; and for DFWP,

P
12
= P
12
if
mobile 1 decodes b
2
correctly, otherwise,

P
12
= 0. Simi-
lar notation applies for (7).

b
1
is the decoded bit at mobile
2 from Y
21
, with an inter-user error probability [7]
P
e
1
= Q
_
_
2
3
_
=
_
h
3
P
11

2
N
_
(8)
=
1

_
/2
0
exp
_

h
3
P
11

2
N
sin
2

_
d
where
3
= h
3
P
11
/
2
N
. The function (x), referred as
the desired form in [7], is given as:
(x) =
1

_
/2
0
exp
_

x
sin
2

_
d (9)
where the identity Q(x) =
1

_
/2
0
exp(
x
2
2 sin
2

)d has
been applied. The received signal at the destination is,
Y
02
=
_
h
1
X
12
+
_
h
2
X
22
+n
0
(10)
=
_
h
1

P
12

b
2
c
2
+
_
h
2

P
22

b
1
c
1
+n
0
. (11)
III. BER ANALYSIS
At the receiver, in order to recover the transmitted sym-
bols, despreading, maximal ratio combining and decision
procedures are carried out. Due to symmetry between user
1 and user 2, we mainly treat user 1s bit retrieval.
A. Decode-and-Forward with Parity (DFWP)
DFWP, also known as selective relay, is under the assump-
tion that the partner/relay re-encodes and forwards the re-
ceived signal in phase I only when the decoding is suc-
cessful. Despreading of user 1s data leads to
Y = c
T
1
_
Y
01
Y
02
_
=
_
h
1
P
11
b
1
+n

0
_
h
2

P
22

b
1
+n

0
_
(12)
Similar with the treatment in [8] for DFWP, the power
saving in case of

P
22
= 0 is negligible, since for high SNR
of inter-user channel, the probability of partner decoding
error is small. Applying MRC to Y, based on the law of
total probability, the instantaneous error probability can
be expressed as
P
b
(h
i
) = (1 P
e1
)
_
h
1
P
11
+h
2
P
22

2
N
_
+P
e1

_
h
1
P
11

2
N
_
(13)
The gain of using cooperation is clearly depicted in the
above equation, where when P
e1
is small, the rst term
at the right-hand-side of (13) dominates the BER result,
which achieves the MRC diversity gain. Averaging over
the fading gains, h
i
, i = 1, 2, 3, the average BER can be
obtained as
P
b
=
_
1 G
_
1 +

3
sin
2

__
G
__
1 +

1
sin
2

__
1 +

2
sin
2

__
+ G
_
1 +

3

2
N
sin
2

_
G
_
1 +

1
sin
2

_
(14)
where
1
= h
1
P
11
/
2
N
= h
1
(1 )P/
2
N
,
2
=
h
2
P
22
/
2
N
, and
i
is the mean value of
i
. The function
G(x()) is given as
G(x()) =
1

_
/2
0
1
x()
d. (15)
Since
i
/ sin
2
, i = 1, 2, 3 is often much larger than 1 for
reasonable BER values in fading environment, the above
equation can be upper bounded by ignoring 1 in the de-
nominators of the integrand. Further approximating the
[1 G()] term by 1 since inter-channel BER is normally
low, leading a BER upper bound given as
P
b


2
+ 3
3
16
1

2

3
=
H
2
+ 3H
3
(1 )
16H
1
H
2
H
3
(1 )
2

2
=
1
16H
1
H
2
H
3

2
_
H
2
+ 3H
3
(1 )
(1 )
2
_
(16)
where = P/
2
N
and is referred as SNR (signal-to-noise
power ratio). Taking the rst derivative with respect to
and equating to zero, we can solve the optimal value
minimizing BER upper bound as

WP
=
9H
3

_
9H
2
3
+ 24H
2
H
3
4(3H
3
H
2
)
. (17)
Compared

WP
with

AF
in [1], it is interesting to nd
that the two expressions are nearly identical. Both of them
are only functions of H
2
and H
3
. When replacing H
3
with 3H
3
in

AF
, we can obtain (17). It implies that H
3
plays a more important role in DFWP than that of AF al-
gorithm. We can also conclude that

WP
is between (0,1);
if 3H
3
> H
2
,

WP
< 0.5.
B. Decode-and-Forward withnot Parity (DFNP)
DFNP forwards what has been decoded in Phase I. The
detector can be expressed as

b
1
= sgn
__
_
h
1
P
11
_
h
2
P
22
_
Y
_
(18)
= sgn
_
_
_
h
1
P
11
_
h
2
P
22
_
_
h
1
P
11
b
1
+n

h
2
P
22
b
1
+n

0
__
where sgn() is signumfunction. The indicator parameter,
= 1 when correct decoding at the partner/relay, and =
1 when incorrect decoding. This detector is similar as
-MRC in [6] with = 1. Conditioned on the value,
the instantaneous BER is given as
P
b
(h
i
) = (1 P
e1
)
_
h
1
P
11
+h
2
P
22

2
N
_
+P
e1
Q
_
h
1
P
11
h
2
P
22
_
(h
1
P
11
+h
2
P
22
)
2
N
/2
_
(19)
For the second term of the above equation, we cannot
write it into the desired () form since the argument of
the corresponding Q function needs to be nonnegative to
write into the preferred () function. It is possible that
the sign of the numerator, (h
1
P
11
h
2
P
22
), be negative.
Furthermore, random variables h
1
and h
2
cannot be aver-
aged out even we write (19) into the preferred () form
conditioned on the sign of the numerator. Simulation has
to be used to solve power allocation problem.
IV. SIMULATION RESULTS
In the simulation, we generate 5,000,000 fading realiza-
tions and capture the average performance. Fig 2 shows
BER changes vs. the cooperative ratio, . The dashed
curve without any marker is the BER for inter-user de-
coding (P
e1
), which monotonically increases with the in-
creasing of . This is expected since P
e1
only depends on
the allocated power at Phase I given average inter-channel
fading power H
3
. With a larger value, the allocated
power at Phase I decreases and leading to an increased
P
e1
. Circle and plus marked solid curves are respectively
the BER results for user 1 and user 2 with DFNP. The
preferable

value is around 0.2. The two lower dashed


curves are the BER results for user 1 and user 2 respec-
tively with DFWP. The markers denote the simulation re-
sults and the dashed curves are obtained from the closed-
form BER expression in (14) using numerical integration.
We can observe an exact match of (14) with the simula-
tion. With DFWP, the preferable value is

= 0.45.
It also shows with the same channel condition and same
SNR value, DFWP performs superior than DFNP. In both
cases, user 2s BER is lower than that of User 1, which
is due to the fact that H
2
> H
1
. Nevertheless, DFWP
reduces greatly the BER differentiation between the users
over DFNP.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0.01
0.02
0.002
0.004
0.006

B
E
R


User1, no parity
User2, no parity
Interuser
User1, with parity
User2, with parity
Figure 2: BER as a function of for DFWP and DFNP
systems (H
1
= 0.5, H
2
= 1, H
3
= 2, P/
2
N
= 14 dB).
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
10
5
10
4
10
3
10
2
10
1

B
E
R


10 dB
14 dB
18 dB
22 dB
Figure 3: BER as a function of for DFWP: a compar-
ison of the upper bound in (16) with simulation (H
1
=
0.5, H
2
= 1, H
3
= 2, P/
2
N
= [10, 14, 18, 22] dB).
Fig. 3 compares the BER upper bound in (16) with the
simulated BER vs. . The four solid curves, from the
top to the bottom, denote the simulated BER when SNR
increases from 10, 14, 18, to 22 dB respectively. The cor-
responding dashed curves denote the upper bound in (16).
It shows that with the increasing of the SNR values, (16)
exhibits incrementally tight bounding of the true BER val-
ues. More importantly, the bound exhibits an identical
changing trend with respect to . Therefore, it is conjec-
tured that the optimal power ration,

WF
, given in (17) is
a desirable solution for DFWP.
Fig. 4 and Fig. 5 show the BER results vs. the coopera-
tive ratio with and without parity check respectively when
SNR value changes from 10, 12, 14 to 16 dB from the top
to the bottom. Solid curves denote the results for user 1
and dashed curves for user 2. For DFWP, the desirable
value isnt sensitive to the changing of the SNR value
and the

is around 0.45. From the analytical result in


(17),

WP
is calculated as 0.4417 for user 1 and 0.4661
for user 2, which is a good reference of

. BER is shown
not sensitive to the little deviation of values around

due to atness of the curves in the vicinity of

. On the
other hand,

is more sensitive for DFNP. With the in-


creasing of the SNR value,

decreases. Overall, DFNP


has a much lower

value compared with DFWP. Once


again, the gures show that DFWP reduces greatly the dif-
ferentiation of the BER between the two users compared
with DFNP.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
10
3
10
2

B
E
R


10 dB
12 dB
14 dB
16 dB
Figure 4: BER vs. for DFWP (H
1
= 0.5, H
2
=
1, H
3
= 2, P/
2
N
= [10, 12, 14, 16] dB).
Fig. 6 compares the BER performance for three coop-
erative systems with the baseline systems when H
1
=
H
2
= 1, and H
3
= 2. The leftmost and the rightmost
solid curves are the BER curves for second-order full di-
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
10
4
10
3
10
2
10
1

B
E
R


10 dB
12 dB
14 dB
16 dB
Figure 5: BER vs. for DFNP (H
1
= 0.5, H
2
= 1, H
3
=
2, P/
2
N
= [10, 12, 14, 16] dB).
versity and no diversity respectively. It is expected that the
BER curves for three cooperative systems fall in between
these two curves. Circle-marked, right-triangle-marked,
and star-marked dashed curves represent for DFNP, AF,
DFWP, applying recommended

values, 0.2, 0.3 [1],


and 0.45 respectively. It shows that DFWP always per-
forms the best among three cooperative schemes and it
is very close to second-order full diversity curve. When
SNR value is less than 10 dB, DFNP and AF perform
nearly identical; while when SNR is larger than 10 dB, AF
outperforms DFNP. With appropriate cooperative trans-
mission, BER performance can be improved signicantly
over that of at Rayleigh fading channel without diversity.
The curve without any marker is obtained from the closed-
form upper bound given in (16). The bound becomes very
tight when SNR is larger than 15 dB.
V. CONCLUSIONS
In this paper, power allocation strategies for decode-and-
forward (DF) cooperative transmission systems are inves-
tigated. By exploiting spacial diversity in wireless chan-
nels, BER performance can be signicantly improved by
properly allocating powers for self-information transmis-
sion and cooperative-information transmission under the
constraint of xed total transmit power of each user. Our
results showthat the desirable cooperative ratio, dened as
the cooperative-information transmission power over total
power for decode-and-forward without parity (DFNP) is
lower than that of AF. The ratio for AF is further lower
than that of decode-and-forward with parity (DFWP).
4 6 8 10 12 14 16 18 20 22 24
10
5
10
4
10
3
10
2
10
1
10
0
P/
2
(dB)
B
E
R


No diversity
Secondorder full diversity
AF, = 0.3
DFNP, = 0.2
DFWP, = 0.45
Figure 6: BER comparison for different cooperative sys-
tems (H
1
= H
2
= 1, H
3
= 2).
REFERENCES
[1] L. Zhao and Z. Liao, On the power allocation for cooperative
amplify-and-forward transmission, Sept. 2007, to appear, VTC-
fall.
[2] A. Nosratinia, T.E. Hunter, and A. Hedayat, Cooperative commu-
nication in wireless networks, IEEE Communications Magazine,
pp. 7480, Oct. 2004.
[3] L. Zhao, J. W. Mark, and J. Ding, Power distribution/allocation in
multirate wideband cdma systems, IEEE Trans. Wireless Commu-
nications, vol. 5, no. 9, pp. 24582467, Sept. 2006.
[4] M. Dohler, A. Gkelias, and H. Aghvami, Resource allocation for
FDMA-based regenerative multihop links, IEEE Trans. Wireless
Communications, vol. 3, no. 6, pp. 19891993, Nov. 2004.
[5] M. O. Hasna and M. S. Alouini, Optimal power allocation for
relayed transmissions over Rayleigh-Fading channels, IEEE Trans.
Wireless Communications, vol. 3, no. 6, pp. 19992004, Nov. 2004.
[6] A. Sendonaris, E. Erkip, and B. Aazhang, User cooperation di-
versity - part II: Implementation aspects and performance analysis,
IEEE Trans. Communications, vol. 51, pp. 19391948, Nov. 2003.
[7] M. K. Simon and M. Alouini, A unied approach to the perfor-
mance analysis of digital communication over generalized fading
channels, Proceedings of the IEEE, vol. 86, no. 9, pp. 18601877,
Sept. 1998.
[8] W. Su, A. K. Sadek, and K. J. R. Liu, SER performance analysis
and optimum power allocation for decode-and-forward cooperation
protocol in wireless networks, in Proc. IEEE Wireless Communi-
cations and Networking Conf., pp. 984989, 2005.

Вам также может понравиться