Вы находитесь на странице: 1из 3

334 IEEE COMMUNICATIONS LETTERS, VOL. 11, NO.

4, APRIL 2007

Performance Analysis of Cooperative Diversity Wireless Networks over


Nakagami-m Fading Channel
Salama Ikki and Mohamed H. Ahmed, Member, IEEE

Abstract This letter analyzes the performance of cooperative


diversity wireless networks using amplify-and-forward relaying
over independent, non-identical, Nakagami-m fading channels.
The error rate and the outage probability are determined using
the moment generating function (M GF ) of the total signal-to-
noise-ratio (SN R) at the destination. Since it is hard to find
a closed form for the probability density function (P DF ) of
the total SN R, we use an approximate value instead. We first
derive the P DF and the M GF of the approximate value of the
total SN R. Then, the M GF is used to determine the error rate
and the outage probability. We also use simulation to verify the
analytical results. Results show that the derived error rate and
outage probability are tight lower bounds particularly at medium
and high SN R.
Index Terms Average error rate, outage probability, cooper-
ative diversity, amplify-and-forward, Nakagami-m fading.

Fig. 1. Illustration of a cooperative diversity wireless network.


I. I NTRODUCTION
OOPERATIVE diversity networks technology is a
C promising solution for the high data-rate coverage re-
quired in future cellular and ad-hoc wireless communica-
integrations in this case. Furthermore, their lower error bound
is not tight enough at high SN R (see Figs.3and 4 in [5]).
tions systems. The cooperative diversity concept has been In [6], the M GF of the SN R is used to find the bit error
introduced in [1], [2], [3]. The basic idea of cooperative rate. Hence, it is straightforward to extend this work to the
diversity networks is that (in addition to the direct signal from case of cooperative diversity with M relay nodes. However,
the source to the destination) some neighboring nodes can the derived M GF in [6] is limited to the case of identical
relay the signal of the source node to the destination node Nakagami-m fading channels.
as shown in Fig. 1. As a result, the destination node can In [7] the authors studied the performance of cooperative
receive multiple independent copies of the same signal and can diversity networks over independent non-identical Nakagami-
achieve diversity without the need to install multiple antennas m fading channels and they found the M GF of the total SN R
at the source node or the destination node. but their error probability bound is not tight enough at medium
In [4] the authors derived asymptotic average symbol er- and high SNR (see Fig. 2 in [7]).
ror probability for amplify-and-forward cooperative diversity In this letter we derive a tight lower bound for the error rate
networks. The resulting expressions derived in [4] (using and the outage probability of cooperative diversity networks
the bounding approach) are general for any type of fading over independent non-identical Nakagami-m fading channels
distributions provided the probability density function (P DF ) with amplify-and-forward relaying and maximum ratio com-
of zero instantaneous SN R is not zero, which is not applicable bining (M RC) at the destination node. We first find closed-
for the Nakagami-m fading distribution. form expressions for cumulative distribution function (CDF ),
In [5], [6] multi-hop relaying (not cooperative diversity P DF and M GF of the total SN R. Then, the M GF is
) over Nakagami-m fading channels has been studied. The used to determine the lower bounds. We verify the obtained
error rate is determined in [5] using a single integration of analytical results using computer simulations. Results show
the conditional error probability P (e|SN R) times the P DF that the derived error rate and outage probability bounds are
of the SN R. Hence, it is not straightforward to extend the tight bounds particularly at medium and high SN R.
work in [5] to the case of cooperative diversity with M relay
nodes (as shown in Fig. 1) because we have to use M + 1 II. S YSTEM M ODEL
As shown in Fig. 1, a source node (S) and a destination
Manuscript received November 28, 2006. The associate editor coordinating node (D) communicate over a channel with a flat Nakagami-m
the review of this letter and approving it for publication was Dr. Rick Blum.
This work is partially supported by the National Sciences and Engineering fading coefficient (f ). A number of cooperating nodes (Ri , i =
Research Council of Canada (NSERC) through the NSERC Discovery Pro- 1, 2, . . . , M ) relay the signal to provide the destination with
gram. multiple copies of the original signal. The channel coefficients
The authors are with the Faculty of Engineering, Memorial University of
Newfoundland, St. Johns, NL A1B3X5, Canada (e-mail: ikki@engr.mun.ca). between the source S and Ri (hi ) and between Ri and D (gi )
Digital Object Identifier 10.1109/LCOMM.2007.061947. are also flat Nakagami-m fading coefficients. In addition, f ,
1089-7798/07$25.00 
c 2007 IEEE
IKKI and AHMED: PERFORMANCE ANALYSIS OF COOPERATIVE DIVERSITY WIRELESS NETWORKS OVER NAKAGAMI-M FADING CHANNEL 335

hi and gi are mutually-independent and non-identical. We also the Nakagami-m fading parameters of gi and hi , respectively.
assume here without any loss of generality that all additive Thus, the P DF can be found by taking the derivative of (5)
white Gaussian noise (AWGN) terms have zero mean and with respect to , yielding
equal variance N0 . Assuming MRC at the destination node,
  mh  
mhi mhi mh 1 hi mgi
the total SN R at the destination node can be written as [4], pi () =
hi i e i m g i
,
gi + (6)
[5], [6], [7]   mg  
mgi mgi mg 1 gi mhi
M
hi gi
gi
i e i m h i
,
hi
/ [(mhi )(mgi )]
equ = f + (1)
+ gi + 1
i=1 hi Finally, Mi (s) can be calculated with the help of [10,
2
where hi = hi Es /No is the instantaneous SN R between S eq.(6.455.1)] as Eqn. 7 on the bottom of this page, where
2
and Ri , gi = gi Es /No is the instantaneous SN R between 2 1 F (, , ) is Gauss hypergeometric function defined in [10,
2
Ri and D, f = f Es /No is the instantaneous SN R between eq. 9.100]. If m hi = mgi = mi and hi = gi = i it can be
S and D and Es is the signal transmitted energy. shown that (7) greatly simplifies to the form
 2mi
mi (2mi ) 2
III. P ERFORMANCE A NALYSIS Mi (s) = i mi 2 (mi ) ((2mi /i )+s)2mi
 
A. Error Performance (mi /i )+s
F
2 1 1, 2m i ; m i + 1; (2mi /i )+s (8)
We derive in what follows a tight lower bound on the aver-
By substituting (4) and (7) into (3) we obtain a closed form
age error performance. The total SN R can be approximated
expression of Mb (s). Using Mb (s), the error rate can be
by its upper bound (b ) as follows
evaluated for a wide variety of M-ary modulations (such as
M
equ b = f + i (2) M-ary phase-shift keying (M-PSK) and M-ary quadrature am-
i=1 plitude modulation (M-QAM)) [11]. For instance, the average
where i = min(hi , gi ). The approximate SN R value b symbol error rate (SER) for M-PSK can be written as [11,
is analytically more tractable than the exact value in (1); as p. 271]
a result, this facilitates the derivation of the SN R statistics (M 1)/M

(CDF , P DF , M GF ). This approximation is adopted in 1 gP SK
P (e) = M b d (9)
many recent papers (e.g., [8], [9]) and it is shown to be sin2 ()
0
accurate enough at medium and high SN R values as will
2
be discussed in the results section. where gP SK = sin (M ).
The SER in (9) can be evaluated
Assuming the independence of hi , gi and f the M GF using a single integral and can be done with simple numerical
of b can be written as integration techniques. Furthermore, (9) can be upper bounded
by a simple form: P (e) (1 1/M )Mb (gP SK ) [11, p.275].

M
Mb (s) = Mf (s) Mi (s) (3)
i=1 B. Outage Performance
where Mf (s) and Mi (s) are the M GF of f and i , The outage probability (Pout ) is defined as the probability
respectively. Using the definition of the M GF as MX (s) = that the instantaneous total SN R falls below a given threshold
E(esX ) (where E is the statistical average operator), it can value (0 ). Pout can be calculated using the M GF [11, Ch.
be easily shown that 1] as follows
 mf
Mf = 1 + mff s


(4) Pout = 1 (Mb (s)/s) (10)
0

1
where  () denotes the inverse Laplace transform. The
 mf is the Nakagami-m fading parameter of f and f =
where
E f 2 Es /N0 . In order to find Mi (s), we find the CDF of inverse Laplace transform can be done analytically or using
i as follows simple numerical techniques as in [12], [13].
Fi () = 1 P (hi > ) P (gi > )

mh

mg
 IV. N UMERICAL R ESULTS
mhi , i mgi , g i
= 1
hi i
(5) In this Section, we show numerical results of the analytical
(mhi )(mgi )
bit error rate (BER) and Pout for binary phase shift keying
where (, ) is the incomplete gamma function [10, (BPSK) modulation. We plot the performance curves in terms
eq.(8.350.2)], () is the gamma function [10, eq. (8.310.1)], of average BER and outage probability versus the SN R of
hi = E[h2i ]Es /N0 , gi = E[gi2 ]Es /N0 , and mgi and mhi are the transmitted signal (Es /N0 dB) where Es is the transmit

 mhi  mgi   
mhi mgi (mhi +mgi ) m /
+s
Mi (s) =
hi
gi (mhi )(mgi )
1
1, mhi + mgi ; mhi + 1; mh /hhi+mhgi /g +s
mhi 2 F1 +
  i i i i
mgi /
gi +s
1
mgi 2 F1 1, mhi + mgi ; mgi + 1; mh /h +mg /g +s 1
m +m (7)
i i i i (mgi /gi +mhi /hi +s) gi hi
336 IEEE COMMUNICATIONS LETTERS, VOL. 11, NO. 4, APRIL 2007

0
10
(10). It is clear again that the difference between the exact
10
1 M =0 (from simulation) and analytical results (lower bound) for
2
Pout is small for medium and high values of Es /N0 . It
10 M =1
should be noted that for Figs. 2 and 3 the tightness of the
10
3
error performance improves as Es /N0 increases; however, the
BER

4
M =2
proposed lower bound (for BER and Pout ) slightly lose its
10

M =3
tightness at low Es /N0 values particularly when M increases.
10
5
This is due to the fact that the accuracy of total SN R
10
6
approximation (in (2)) improves as Es /N0 increases.
Exact (Simulation)
Lower bound (Analytical eq.9)
Lower bound (Analytical ref.[4]) V. C ONCLUSIONS
7
10
0 5 10 15 20 25
Es /N0 dB Performance bounds for cooperative diversity networks over
independent non-identical Nakagami-m fading channels have
Fig. 2. BER performance for arbitrary E(f 2 )=1, E(h2i )=0.75, been investigated. The total SN R is approximated by its upper
E(gi2 )=0.5, several numbers of cooperative paths (M = 0, 1, 2, 3) bounded. Then, closed-form expressions for the M GF , P DF ,
and mf = mhi = mgi = 0.5.
and CDF of the approximate total SN R have been derived.
10
0
The M GF is used to determine lower bounds for the error
M =0
rate and outage probability. Our numerical results show that
10
1
derived error rate and outage probability are a tight lower
M =1 bound particularly at medium and high SN R.
2
10

R EFERENCES
Pout

3
10 M =3 [1] J. N. Laneman, D. N. C. Tse, and G. W. Wornell, Cooperative diversity
M =2 in wireless networks efficient protocols and outage behaviour, IEEE
4 Trans. Inf. Theory, vol. 50, no. 12, pp. 30623080, 2004.
10
[2] J. N. Laneman and G. Wornell, Energy-efficient antenna sharing and
Exact (Simulation) relaying for wireless networks, in Proc. Wireless Commun. Networking
Lower bound (Analytical eq.10)
10
5 Conf. 2000, vol. 1, pp. 30623080.
0 5 10 15 20 25
Es /N0 dB [3] J. N. Laneman and G. Wornell, Distributed space-time coded protocols
for exploiting cooperative diversity in wireless networks, in Proc. Glob.
Fig. 3. Outage performance for arbitrary E(f 2 )=1, E(h2i )=0.75, Telecomm. Conf. 2002, vol. 1, pp. 7781.
[4] A. Ribeiro, X. Cai, and G. B. Giannakis, Symbol error probabilities
E(gi2 )=0.5, several numbers of cooperative paths (M = 0, 1, 2, 3)
for general cooperative links, IEEE Trans. Wireless Commun., vol. 4,
and mf = mhi = mgi = 0.5. no. 3, 2005.
[5] G. K. Karagiannidis, T. A. Tsiftsis, and R. K. Mallik, Bounds of
multihop relayed communications in Nakagami-m fading, IEEE Trans.
energy signal. We also show computer simulation results for Commun., vol. 54, pp. 1822, 2006.
[6] Hasna and Alouini, Harmonic mean and end-to-end performance of
verification. transmission systems with relays, IEEE Trans. Commun., vol. 52, no.
Fig. 2 shows the BER performance at different values of 1, 2004.
the number of cooperating nodes (M ). We also plotted in Fig. [7] T. A. Tsiftsis, G. K. Karagiannidis, S. A. Kotsopoulos, and F.-N. Pavli-
2 the BER bound derived in [4] for comparison. It is clear dou, BER analysis of collaborative dual-hop wireless transmissions,
IEE Electronic Lett., vol. 40, no. 11, 2004.
that the BER bound derived in this letter is much tighter than [8] A. Bletsas, A. Khisti, D. P. Reed, and A. Lippman, A simple cooper-
that proposed in [4], especially for medium and high Es /N0 ative diversity method based on network path selection, IEEE J. Sel.
values. For instance, for M = 3 and Es /N0 = 15 dB, the Areas Commun., vol. 24, no. 3, 2006.
[9] T. Wang, A. Cano, G. B. Giannakis, and N. Laneman, High-
exact BER (from simulation) is equal to 8 104 , while the performance cooperative demodulation with decode-and-forward re-
BER lower bound from [4] is equal to 2.1 104 and our lays, IEEE Trans. Commun., accepted for publication 2007.
lower bound (from (9)) is equal to 6 104 . This trend (the [10] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series and
Products, Fifth Edition. San Diego: Academic Press, 1994.
tightness of our bound compared with that of [4] at medium [11] M. K. Simon and M.-S. Alouini, Digital Communication over Fading
and high Es /N0 values) is valid at different values of M as Channels. New York: Wiley, 2000.
shown in Fig. 2. From Fig. 2, we can also notice, as expected, [12] J. Abate and W. Whitt, Numerical inversion of Laplace transform of
probability distribution, ORSA J. Comput., vol. 7, no. 1, 1995.
that the number of cooperating relays (M ) has a strong impact [13] Y.-C. Ko, M.-S. Alouini, and M. K. Simon, Outage probability of diver-
on performance enhancement and the achieved diversity order. sity systems over generalized fading channels, IEEE Trans. Commun.,
Fig. 3 illustrates the outage performance determined from vol. 48, no. 11, 2000.

Вам также может понравиться