2

On the Effective Capacity of Amplify-and-Forward Multihop Transmission

Over Arbitrary and Correlated Fading Channels
Kostas P. Peppas, Senior Member, IEEE, P. Takis Mathiopoulos, Senior Member, IEEE,
and Jing Yang, Member, IEEE

3
4

AbstractThis letter presents a novel approach for analyzing,

in a unified way, the effective capacity performance of multihop transmission with channel state information (CSI)-assisted
amplify-and-forward (AF) relay systems over arbitrary and correlated fading channels under a maximum delay constraint. Using a
moment generating function (MGF)-based approach, an exact single integral expression for the effective capacity is deduced. This
expression can be easily evaluated by means of standard numerical
integration techniques for a great deal of fading distributions, such
as the Nakagami-m, generalized-K, and generealized Gamma.
Several case studies, including uncorrelated and correlated fading
environments, are employed to demonstrate the versatility of the
proposed analytical approach. Extensive numerically evaluated
results accompanied with complimentary Monte-Carlo simulations are used to substantiate the analytical derivations.

20
21

Index TermsAmplify-and-Forward, delay

effective capacity, fading channels, multisystems.

I. I NTRODUCTION

22

constraints,

NE of the major challenges for future generation wireless

communication systems is to support quality-of-service
(QoS) requirements for different applications. However, many
important emerging applications, including voice over internet
protocol (VoIP), interactive gaming, mobile TV and computing, as well as interactive and multimedia streaming, are largely
delay-sensitive. They further impose stringent QoS constraints,
which typically appear in the form of constraints on queuing
delays or queue lengths. In the past, the concept of effective capacity (EC) has been introduced as a link-layer model
for supporting QoS requirements of communication systems
[1]. EC is the dual of effective bandwidth [2], and can be
interpreted as the maximum constant arrival rate that a wireless channel can support while a given QoS requirement is
guaranteed.
Let us now consider the multi-hop relay technology in which
two users communicate with each other via relay terminals,
This technology has emerged as an inspiring and powerful
approach to improve the reliability, coverage and throughput of wireless communication systems. In the past, the EC
of amplify-and-forward (AF) relaying systems with channel
state information (CSI) has been addressed in several research

EP

23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44

works. In [3], closed form bounds on the EC of dual hop

systems operating in Rayleigh fading environments have been
deduced. In [4], bounds on the EC of two-way dual-hop systems
over Rayleigh fading channels have been presented. It should be
noted that both of these papers consider only the dual-hop case
and their corresponding analysis cant be easily generalized to
multi-hop relaying systems. Furthermore, they consider specific
channel models with emphasis given to simple fading channel
models such as the multi-path Rayleigh fading model. To the
best of the authors knowledge, the more realistic composite
multi-path/shadowing channel model has not been yet considered in connection with EC performance analysis studies. In
addition, the impact of link correlation on the EC performance
of multi-hop systems has not yet been investigated in conjunction with the problem of how to analytically evaluate the EC
performance of such communication systems.
Motivated by the above, in this contribution, a novel
approach for the exact EC performance analysis of CSI-assisted
AF multihop systems over arbitrary and correlated fading channels is presented. A single integral expression for the EC which
uses a moment generating function (MGF) based approach,
is deduced. This expression can be easily evaluated for a
variety of fading distributions by means of standard numerical integration techniques. It is noted that, to the best of the
authors knowledge, a unified analytical approach for the EC
analysis of AF relaying systems is not, to date, available in
the open technical literature. The proposed unified analytical approach is employed to determine the EC performance
of CSI-assisted multi-hop AF relaying systems under uncorrelated Generalized Gamma, uncorrelated Generalized-K and
correlated Nakagami-m fading channels. In fact, as will be
explained later on, the analysis is valid for arbitrary and correlated fading channels as long as the MGF of the reciprocal
end-to-end signal-to-noise ratio (SNR) exists. The validity of
the proposed analysis is substantiated by extensive numerically
evaluated performance results accompanied with equivalent
results obtained by means of Monte-Carlo simulations.
Notations: E denotes expectation, f X () denotes the
Probability Density Function (PDF) of the random variable
(RV) X , M X () denotes the MGF of the random variable
X , Ja () is the Bessel function of the first kind and order
a [5, Eq. (8.402)], Ia () is the modified Bessel function of
the first kind and order a [5, eq. (8.431)], K a () is the modified Bessel function of the second kind and order a [5,
eq. (8.432)],  ()
function [5, Eq. (8.310/1)],
 is the Gamma

(, x, b, ) = x r 1 exp r br dr is the extended
incomplete gamma function [6], p Fq () is the generalized
hypergeometric function [5, eq. (9.14/1)] and G m,n
p,q [] is the
Meijers G-function [5, Eq. (9.301)].

ro

5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

of

IEE

Q1

IEEE WIRELESS COMMUNICATIONS LETTERS

Manuscript received September 12, 2015; accepted February 10, 2016. The
associate editor coordinating the review of this paper and approving it for
publication was H. H. Nguyen.
K. P. Peppas is with the Department of Informatics and Telecommunications,
University of Peloponnese, Tripoli 22100, Greece (e-mail: peppas@uop.gr).
P. T. Mathiopoulos is with the Department of Informatics and
Telecommunications, National and Kapodistrian University of Athens,
Athens 15784, Greece (e-mail: mathio@di.uoa.gr).
J. Yang is with the School of Information Engineering, Yangzhou University,
Yangzhou, China (e-mail: jingyang905@163.com).
Digital Object Identifier 10.1109/LWC.2016.2530787

2162-2337 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93

IEEE WIRELESS COMMUNICATIONS LETTERS

95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118

Let us consider a multi-hop transmission system in which

a source node is communicating with a destination node via
N 1 AF relay nodes in series. All nodes are assumed to have
slow mobility, are equipped with a single antenna and work
in half duplex mode. The available bandwidth is allocated to
the source node for data transmission, which is further divided
into orthogonal sub-channels across time using a time-division
scheme. It is also assumed that CSI is only known at the receiving terminals and that only one terminal transmits in each time
slot [3].
As far as its queuing model is concerned, a simple first-input
first-output (FIFO) buffer with constant arrival rate (source data
rate) at the source data link layer is considered. By considering ideal modulation and coding at the source physical layer,
the service rate of the buffer will be equal to the instantaneous
channel capacity which is time varying [3]. No buffering takes
place at the relays and thus their transmissions are limited to
the physical layer. Consequently, the arriving signal/packet is
immediately amplified and sent to the next node without any
buffering requirements or additional delays.
Considering the previously described system model and
assuming that the transmitter sends uncorrelated circularly
symmetric zero-mean complex Gaussian signals, the instantaneous EC can be deduced as [1], [3, Eq. (9)]:
1
ln E(1 + end ) 
Tf B

(1)

IEE

where end is the end-to-end instantaneous SNR of the AF CSIassisted multi-hop transmission over arbitrary fading channels,
= T f B/N ln(2) where T f is the fading block length, B the
system bandwidth and the exponentially decay rate of the QoS
violation probability. Note that a large value for corresponds
to a fast decaying rate thus having more stringent QoS requirements, while a smaller one refers to slower decaying rate and
thus looser QoS requirements.
The end-to-end instantaneous SNR, end , is characterized by
the normalized harmonic mean of the instantaneous SNRs at
each hop as [7]
end

 N
1
 1
=
n

(2)

n=1

130
131
132
133
134
135
136
137
138
139
140
141
142

III. E XACT A NALYTICAL E XPRESSIONS

146

Proposition: The EC for AF CSI-assisted multi-hop trans- 147

mission over arbitrary fading channels can be expressed in 148
terms of a single integral as
149



M end (u)
1
ln
du .
R( ) =
1 F1 (; 1; u)
Tf B
u
0
(3)
Proof: Let us consider the well-known identity [5, 150
Eq. (1.512/4)]
151

exp [(x + 1)s] s 1 ds = ()(1 + x) .
(4)
0

By substituting (4) into the expectation in (1) and after some

simple algebraic manipulations while exploiting the definition

of the MGF, i.e., M end (s)  0 exp(s ) f end ( )d , EC

can be deduced as



1
1
ln
R( ) =
exp(s)s 1 M end (s) ds .
Tf B
() 0
(5)

152
153
154
155

Integrating [8, Eq.(18)] by parts and employing the identity 156

d
1 J (2 su), the MGF of
1
end can be 157
d u J0 (2 su) = su
expressed in terms of the MGF of end as
158

M end (u)
du.
(6)
M end (s) =
J0 (2 su)
u
0

EP

R( ) =
119
120
121
122
123
124
125
126
127
128
129

it will be shown next, by using the MGF of the reciprocal end- 143
to-end SNR, end , a generic expression for the exact evaluation 144
of R( ) can be obtained.
145

of

II. S YSTEM M ODEL

ro

94

where n denotes the instantaneous SNR of the n-th hop.

Denoting the reciprocal
 Nof end as end = 1/ end , (2) can be
expressed as end = n=1
n where n is the reciprocal of the
instantaneous SNR of the n-th hop.
In general, the evaluation of the expectation in (1) requires
the computation of an N -fold integral. The numerical evaluation of such integral is, by all means, computationally
intractable, even for small values of N (e.g. 3 or 4).
Furthermore, such integral cannot be expressed as a product of
single integrals because of the non-linear factor (1 + end ) .
Due to this difficulty, existing works dealing with the evaluation of the EC of AF systems, usually employ bounds on the
EC that become tight at high SNR regions [3], [4]. However, as

By substituting (6) into (5), the integral in (5) can be 159

expressed as
160




(u)
M

end
du ds

exp(s)s 1
J0 (2 su)
u
0
0




M end (u)
1
=
exp(s)s
J0 (2 su)ds du.
u
0
0
The inner integral, with respect to s, can be evaluated in closed
form by employing [5, Eq. (6.643/1)] and [5, Eq. (9.220/2)]
yielding (3), which completes the proof.

It is noted that (3) is valid for arbitrary and correlated fading
channels, as long as the MGF of end exists. For the special
case of uncorrelated n , the EC of the considered system can be
readily evaluated by employing the following corollary.
Corollary: When there is no correlation among all hops, the
EC of AF CSI-assisted multi-hop system is deduced as


1
ln
R( ) =
1 F1 (; 1; u)
Tf B
0


N 
N

Mn (u)

Mk (u)du .
(7)
u

161
162
163
164
165
166
167
168
169

Proof: When uncorrelated hops are considered, M end (u)

can be expressed
 N as the product of the MGFs of n , i.e.
M end (u) = n=1
Mn (u). Then, (7) is readily obtained from
(3) by evaluating the derivative of M end (u) with respect
to u.


170
171
172
173
174

n=1

k=1k=n

IV. P ERFORMANCE A NALYSIS AND D ISCUSSION

176
177
178
179
180
181
182
183
184
185
186
187

In this section, the proposed analytical approach, and in

particular (3) and (7), are employed to determine the EC of
CSI-assisted AF relaying. More specifically, the following case
studies are considered: i) Multi-hop systems over uncorrelated
generalized-gamma fading channels, ii) Dual-hop systems over
uncorrelated generalized-K fading channels and iii) Dual-hop
systems over correlated Nakagami-m fading channels. The
numerical integration of (3) and (7) was carried out by applying
an L-point Gauss-Chebyshev quadrature technique, where L
has been selected to be greater than 100 for sufficient numerical
accuracy. Furthermore, similar to [3] and for the three considered case studies, we have assumed that T f = 2 ms, B =
100 kHz and = 0.01 bits1 . The obtained performance evaluation results are substantiated by employing semi-analytical
Monte-Carlo simulations. It is noted that the simulated performance results are obtained by employing 106 random samples
generated have been based on Eq. (2).
1) Multihop Systems Over Uncorrelated GeneralizedGamma Fading Channels: In this case, the first derivative of
the MGF of n , with respect to u, can be easily deduced by
differentiating [9, Eq. (14)] as


Mn (u)
bn
2
bn
2
. (8)
=
 m n , 0, u,
u
gn (m n )
n
gn n

203
204
205
206
207
208
209
210
211
212
213
214
215
216

In the above equation, m n 1/2, n > 0 and gn > 0 are the

fading figure, the shaping parameter and the average SNR of
the n-th hop, respectively, and bn =  (m n + 2/n ) / (m n ).
Note that (8) is quite versatile as it includes as special cases
several well known fading models, such as the Rayleigh
(m n = 1, n = 2), Nakagami-m (n = 2), Weibull (m n = 1),

log-normal (m n a d , n 0), and AWGN noise (m n

a d , n = 2) [9]. Substituting (8) and [9, Eq. (14)] in (7), the

EC performance has been obtained and is illustrated in Fig. 1
as a function of the average SNR per hop, gn = g, for m n =
m = 1.5 and n = = 2.4, as well as for various numbers of
hops, N . As expected, EC decreases as N increases. It is also
underlined that the numerical results and Monte-Carlo simulations are in excellent agreement, thus validating the correctness
of the proposed analysis.
2) Dual-Hop Systems Over Uncorrelated Generalized-K
Fading Channels: In this case, the first derivative of the MGF
of n , with respect to u, can be deduced by differentiating
[10, Eq. (15)] and by employing [5, Eq. (9.31/3)] and [5,
Eq. (9.31/5)] as



Mn (u)
1
2
u
=
G 3,0
m
,k
,1
n
n n
u
u(kn )(m n ) 0,3

217
218
219
220
221
222
223
224

Fig. 1. EC of multi-hop CSI-assisted AF systems over independent and identically distributed generalized gamma fading channels as a function of the
average SNR per hop for m n = m = 1.5, n = = 2.4, and various numbers
of hops, N .

Q2

EP

197
198
199
200
201
202

IEE

188
189
190
191
192
193
194
195
196

ro

175

of

PEPPAS et al.: ON EFFECTIVE CAPACITY OF AMPLIFY-AND-FORWARD MULTIHOP TRANSMISSION

(9)

where kn and
m n are the shaping parameters of the distribution
and n = kn m n /gn . Similarly to the first case study, substituting (9) and [10, Eq. (15)] in (7), the EC performance has
been obtained and is illustrated in Fig. 2, assuming m 1 = 2,
m 2 = 1.5 and for values for the parameters kn , k1 = k2 = k
that account for two different shadowing scenarios, namely,
frequent heavy shadowing (k = 1.0931) and average shadowing (k = 38.0809) [11]. The accuracy of the analysis is again

Fig. 2. EC of dual-hop CSI-assisted AF systems over independent and nonidentically distributed generalized-K fading channels as a function of the first
hop average SNR, g1 , for m 1 = 2, m 2 = 1.5, and various shadowing scenarios.

verified by the excellent agreement between the numerically

evaluated results and those obtained by means of computer
simulations.
3) Dual-Hop Systems Over Correlated Nakagami-m Fading
Channels: In this case, 1 and 2 are correlated gamma RVs
with the following joint PDF [12]
f 1 ,2 (1 , 2 ) =

225
226
227
228
229
230

exp {1 /[(1 )
1 ] 2 /[(1 )
2 ]}

(m)
1
2 (1 )(
1
2 )m1


2 1 2
Im1
(1 2 )(m1)/2

1
2 (1 )
(10)

where m is the fading parameter, is the correlation coefficient,

and
i = Ei /m, i {1, 2}. By employing an infinite series
representation for the modified Bessel function [5, Eq. (8.445)]
and [5, Eq. (3.471/9)], M end (u) can be deduced as

231
232
233
234

IEEE WIRELESS COMMUNICATIONS LETTERS


k

u m+k
4 

(m)
k!(m + k)(
1
2 )m+k 1
k=0




u
u
K m+k 2
.
K m+k 2
(1 )
1
(1 )
2
(11)

Although the first derivative of M end (u), which is required in

(3) for the evaluation of EC, has been obtained in a straightforward way by employing [5, Eq. (8.486/11)], it will not be
presented here due to lack of space. Note that for the special
case of independent and identically distributed Nakagamim links, i.e. for
1 =
2 =
and = 0, the EC can be
expressed in closed form. In this case, all infinite series terms
that appear in (11) with k = 0, k, vanish and (11) simplifies as
 
u
4u m
2
.
(12)
M end (u) =
K 2

(m)2
m m

243
244

The first derivative of M end (u) can be deduced by employing

[5, Eq. (8.486/10)] and [5, Eq. (8.486/11)] as
 
 
M end (u)
8u m1/2
u
u
=
K m1 2
.
Km 2
2
m+1/2
u

(m)

(13)

245
246
247
248

In order to deduce a closed-form expression for R( ), the

confluent hypergeometric function in (3) and the product of
the Bessel functions in (13) are first expressed in terms of
1
the Meijers

 G-function, i.e. 1 F1 (; 1; u) =[()]


1
1,1
[13, Eq. (8.4.45/1)] and K m 2 u/

G 1,2 u  0,0






0

K m1 2 u/
= 0.5 G 3,0
1,3 4u/
1/2,m1/2,m+1/2
[13, Eq. (8.4.24/29)]. Then, by employing [13, Eq. (2.24.1/1)]
and [13, Eq. (8.2.2/15)], R( ) is deduced in closed form as
 2m+1


2
3,1 4 1/2+m,1
1
ln
.
R( ) =
G

Tf B
()(m)2 2,3
,m,2m

250
251
252

260

This letter, presented a comprehensive and unified analytical

approach for the computation of the EC of CSI-assisted multihop communication systems over arbitrary and correlated fading channels employing the AF protocol. Using the MGF of the
reciprocal end-to-end SNR, a novel generic integral expression
for the EC, was deduced. Analytical performance evaluation
results obtained by numerical techniques have been in excellent
agreement with those obtained via Monte-Carlo simulations,
thus verifying the accuracy of the theoretical analysis.

261
262
263
264
265
266
267
268
269

R EFERENCES

270

[1] D. Wu and R. Negi, Effective capacity: A wireless link model for support of quality of service, IEEE Trans. Wireless Commun., vol. 2, no. 4,
pp. 630643, Jul. 2003.
[2] C.-S. Chang, Stability, queue length, and delay of deterministic and
stochastic queueing networks, IEEE Trans. Autom. Control, vol. 39,
no. 5, pp. 913931, May 1994.
[3] S. Efazati and P. Azmi, Effective capacity maximization in multirelay networks with a novel cross layer transmission framework and
power allocation scheme, IEEE Trans. Veh. Technol., vol. 63, no. 4,
pp. 16911702, Nov. 2013.
[4] G. G. Ozcan and M. Gursoy, Effective capacity analysis of fixedgain and variable-gain AF two-way relaying, in Proc. IEEE 78th Veh.
Technol. Conf. (VTC Fall), Las Vegas, NV, USA, 2013, pp. 15.
[5] I. Gradshteyn and I. M. Ryzhik, Tables of Integrals, Series, and Products,
6th ed. New York, NY, USA: Academic, 2000.
[6] M. A. Chaudhry and S. M. Zubair, On a Class of Incomplete Gamma
Functions With Applications, 1st ed. London, U.K.: Chapman &
Hall/CRC, 2002.
[7] M. O. Hasna and M.-S. Alouini, Outage probability of multihop transmission over Nakagami fading channels, IEEE Commun. Lett., vol. 7,
no. 5, pp. 216218, May 2003.
[8] M. D. Renzo, F. Graziosi, and F. Santucci, A unified framework for
performance analysis of CSI-assisted cooperative communications over
fading channels, IEEE Trans. Commun., vol. 57, no. 9, pp. 25512557,
Sep. 2009.
[9] F. Yilmaz, O. Kucur, and M.-S. Alouini, Exact capacity analysis of multihop transmission over amplify-and-forward relay fading channels, in
Proc. IEEE Int. Symp. Pers. Indoor Mobile Radio Commun. (PIMRC),
2010, pp. 22932298.
[10] K. Peppas, F. Lazarakis, A. Alexandridis, and K. Dangakis, Momentsbased analysis of dual-hop amplify-and-forward relaying communications systems over generalised fading channels, IET Commun., vol. 6,
no. 13, pp. 20402047, Sep. 2012.
[11] K. Peppas, Accurate closed-form approximations to generalised-k sum
distributions and applications in the performance analysis of equal-gain
combining receivers, IET Commun., vol. 5, no. 7, pp. 982989, May
2011.
[12] M. Nakagami, The m-distributionA general formula of intensity
distribution of rapid fading, in Statistical Methods in Radio Wave
Propagation. New York, NY, USA: Pergamon, 1960, pp. 336.
[13] A. P. Prudnikov, Y. A. Brychkov, and O. I. Marichev, Integrals and Series
Volume 3: More Special Functions, 1st ed. New York, NY, USA: Gordon
and Breach, 1986.

271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313

IEE

EP

235
236
237
238
239
240
241
242

249

V. C ONCLUSIONS

ro

M end (u) =

253
254
255
256
257
258
259

of

Fig. 3. EC of dual-hop CSI-assisted AF systems over correlated Nakagami-m

fading channels as a function of
1 , for m 1 = m 2 = 2.

Substituting the derivative of M end (u) in (3), the EC performance has been obtained and is depicted in Fig. 3, for
the cases where m = 2 and {0, 0.8}. The infinite series in
(11) are truncated to 40 terms for numerical accuracy. Again
here numerically evaluated results and computer simulations
are in excellent agreement. It is also noted that the impact of
correlation on the EC performance is rather small.

QUERIES

IEE

EP

ro

of

Q1: Please be advised that per instructions from the Communications Society this proof was formatted in Times Roman font and
therefore some of the fonts will appear different from the fonts in your originally submitted manuscript. For instance, the
math calligraphy font may appear different due to usage of the usepackage[mathcal]euscript. We are no longer permitted to
use Computer Modern fonts.
Q2: Note that if you require corrections/changes to tables or figures, you must supply the revised files, as these items are not
edited for you.