Cross-Layer Combination of Cooperative HARQ with

AMC in Wireless Ad-hoc Networks

Feijin Shi, and Dongfeng Yuan, Senior Member, IEEE
School of Information Science and Engineering
Shandong University
Jinan, Shandong, P. R. China
E-mail: shifeijin@mail.sdu.edu.cn, dfyuan@sdu.edu.cn

Adaptive modulation and coding (AMC) is one of the most

important realizations of link adaptation [6]. The objective of
AMC is to maximize the data rate by adjusting transmission
parameters to channel variation, while meeting the QoS
constraints from the upper layer.

AbstractThis paper proposes a cross-layer design which

combines adaptive modulation and coding (AMC), cooperative
diversity at physical layer (PHY) and hybrid automatic repeat
request (HARQ) at data link layer (LL) for wireless Ad-hoc
networks. First, we provide theoretical analyses of conventional
direct transmission scheme and cooperative transmission scheme
with AMC and HARQ that include analytical expressions for the
packet error rate (PER) and spectral efficiency. Then we propose
a novel hybrid transmission scheme which invokes cooperation
based on the feedback AMC information. Numerical results
demonstrate that the performance of the hybrid transmission
scheme is equivalent to the cooperative transmission scheme
while its cooperation cost is kept much lower compared with that
of the cooperative transmission scheme.
Keywords- Adaptive modulation and coding (AMC); cross-layer
design; cooperative diversity; hybrid automatic repeat request
(HARQ); wireless Ad-hoc network

I.

In this paper, we develop a novel hybrid transmission

scheme combining hybrid ARQ (HARQ) at LL, AMC and
cooperative diversity at PHY. In this scheme, the system
switches from direction transmission to cooperative
transmission based on the feedback AMC information. For the
sake of comparison, the theoretical analyses of direct
transmission scheme and cooperative transmission scheme with
AMC and HARQ are also provided. Our performance metric
for comparison among the three schemes is spectral efficiency.
II.

INTRODUCTION

For the simplicity of illustrating how our scheme works, we

consider a wireless Ad-hoc network composed of a source
node S , a relay node R , and a destination node D , as shown
in Fig. 1. S and R are constrained to employ half-duplex
transmission, i.e., they cannot transmit and receive
simultaneously.

To meet the demands of future wireless communication

systems for large coverage range and high data rates, diversity
techniques and link adaptive transmission techniques must be
considered. Time, frequency, and spatial diversity techniques
can be utilized to mitigate the fading effects. Automatic repeat
request (ARQ) protocol at the data link layer (LL) is commonly
used to exploit temporal diversity. However, temporal diversity
is limited in slow fading (log-normal shadowing) channels.

A. Scheme Descriptions
1) Direct transmission: Only direct transmission between

Cooperative diversity technology is an emerging and

powerful solution that can mitigate fading in slow fading
wireless environment. The initial attempts for developing
cooperative communications focused on physical layer (PHY).
Several cooperative protocols have been proposed such as
amplify-and-forward, decode-and-forward [1].

S and D is allowed which is equivalent to the conventional

transmission.
2) Cooperative transmission: In the cooperative
transmission scheme, we consider a two-phase protocol. In
phase 1, S transmits a packet with a C-bit CRC attached and
both R and D attempt to receive it. If D fails to decode the
packet correctly, it will go into phase 2 and retransmission will
start. To avoid the error propagation, if R was able to receive
the packet from S correctly, it forwards the packet to D .
Otherwise, retransmission will start by S , which is equivalent

Since cooperative communication involves the interaction

between multiple users, it exhibits various forms at different
protocol layers and introduces many opportunities for crosslayer design and optimization [2].
The cooperative ARQ protocol [3-5] can be viewed as the
combination of cooperative diversity at PHY and ARQ at LL
which only allows relay node to forward what it received from
the source when the information is correctly decoded at the
relay and incorrectly decoded at the destination. However, link
adaptation is not considered in these papers.

to the conventional HARQ.

3) Hybrid transmission: Similar to cooperative
transmission scheme, hybrid transmission scheme also has two
phases. In phase 1, S transmits a packet and both R and D
attempt to receive it. If the packet is correctly detected by D
and an ACK is fed back, S will continue to send a new packet.

Supported by National Scientific Foundation of China (No. 60672036) and

Key Project of Provincial Scientific Foundation of Shandong (Z2006G04)

1-4244-2424-5/08/$20.00 2008 IEEE

SYSTEM DESCRIPTIONS

896

ICCS 2008

nsr

TABLE I.

TRANSMISSION MODES

n
s r

Mode0
Modulation BPSK
Code Rate
1/2
Rate(bits/sym.) 0.50

S D

hsr

hrd
S

hsd

S D

nsd

nrd
D

Figure 1. System Model

gn

6. 0354

6. 0354

2. 6532 0. 8483

0. 2430

0. 0553

pn

1.2146

1.2146

2.2802

20.1332 74.2056

3.9038

(5)

(6)

1,
an exp ( g n ) ,

if 0 < < pn
if pn

(7)

where n is the transmission mode (TM) index of AMC and

is the instantaneous SNR. The parameters an , g n and pn in
(7) depend on the AMC mode. This expression is used to
approximate the PER of convolutional codes. All possible TMs
are listed in Table I [8].
Chase Combining HARQ based on receiver combines the
soft decision of the original and the retransmitted packet for a
higher probability of successful decoding. Hence, the SNR of
the packet after MRC is 0 +...+ Nr . In order to satisfy that the

PER is no greater than the packet loss probability Plink after at

most N r + 1 transmissions at LL, we have:

PERn ( 0 ) PERn ( 0 + 1 )" PERn 0 +...+ N r Plink . (8)

At PHY, when [ n , n +1 ) , the mode n AMC will be
chosen. So applying (8) to the expression below, we will
choose mode n with probability:

xs is the signal transmitted by S .

In phase 2, if retransmission is started by R , the signal

received by D is given by

where

60.556

PERn ( )

xs is the signal transmitted by S .

yrd = hrd xr + nrd ,

133. 27

C. Modeling HARQ with AMC

To simplify the design, we use the following approximate
packet error rate (PER) expression as [7]:

Second, we describe the signal model for the cooperative

transmission and hybrid transmission schemes. In phase 1, the
signals received by D and R are given by

where

424. 06 27. 429

where = E { } is the average SNR.

(1)

(3)

1525. 9

1

p ( )= exp ,

First, we describe the received signal model for the direct

transmission scheme. The signal received at D from S can be
modeled as

ysr = hsr xs +nsr ,

1525. 9

For Rayleigh fading channels, the instantaneous received

SNR has an exponential distribution with a probability
distribution function (pdf):

all the additive white Gaussian noise (AWGN) terms

nsd , nsr and nrd have equal variance N 0 .

(2)

an

where Pi is the transmitted power at node i .

hrd , respectively. Without loss of generality, we assume that

ysd = hsd xs + nsd ,

Mode3 Mode4 Mode5

QPSK 16-QAM 64-QAM
3/4
3/4
3/4
1.50
3.00
4.50

ij = hij Pi N 0

B. Channel Model
Each link in the network is subject to quasi-static Rayleigh
fading, i.e., the channels characteristic is changed packet by
packet, but stays the same for each packet. Define channels
S D , S R and R D with flat-fading coefficient hsd , hsr and

where

Mode2
QPSK
1/2
1.00

Otherwise, if a packet is detected to be error, a NACK is fed

back. In this case, it will go into phase 2 and retransmission
will start. In phase 2, only when the S D channel is deep
fading which could be determined by feedback AMC
information, the system went into cooperative transmission
scheme. Otherwise, it is equivalent to the direct transmission
scheme. In hybrid transmission scheme, since cooperation is
activated only when the instantaneous S D channel is deep
fading, the overhead of cooperation is much lower than
cooperative transmission scheme.

ysd = hsd xs + nsd ,

Mode1
BPSK
1/2
0.50

Pr ( n ) =

(4)

n+1

xr is the signal transmitted by R .

III.

The instantaneous SNR for the channel between node i

( S or R ) and node j ( R or D ) can be written as:

n+1
n
exp .

p ( )d = exp

(9)

PERFORMANCE ANALYSIS

In this section, the PER and spectral efficiency of hybrid

transmission scheme are analyzed. For the sake of comparison,

897

both of which from the source node.

the PER and spectral efficiency expressions of the direct

transmission and the cooperative transmission schemes are also
provided.

From (6), (7) and (13), we can derive the average PER
corresponding to TM n after retransmission as:

In this paper, we assume the maximum number of

retransmission is one ( N r =1) . Let 1 and 2 be the
instantaneous received SNR of the primary transmission and
retransmission, respectively. The LL PER of Chase combining
HARQ can be written as
PERn = PERn ( 1 ) PERn ( 1 + 2 ) .
r

nth+1
1
PERn ( ) PERn (2 ) p sd ( ) d
th

Pr (n) n
1
a 2 exp(3g n nth+1 ) exp( nth+1 sd )
[1 n

3g n sd +1
Pr (n)

pn
2 g exp( pn )
g n sd exp(
)
n sd

sd
2 sd

], n = 0 .
=
an (2 g n sd +1)(3g n sd +1)
2 g n sd +1

2
1 th
1 th
an [exp((3g n + ) n ) exp((3g n + ) n+1 )]
sd
sd

, n =1,...,5

(3g n sd +1) Pr (n)

(15)
r

PER nDT =

(10)

In slow fading channel, we assume the channel condition

stays constant during retransmission as that of the primary
transmission since retransmission occur soon after the primary
transmission. Therefore, if the retransmitted packet is from the
source node, the LL PER can be written as
PERn = PERn ( sd ) PERn (2 sd ) .
r

(11)

The average PER of AMC for direct transmission scheme

can then be written as:

Otherwise, if the retransmitted packet is from the relay node,

the LL PER of Chase combining HARQ can then be derived as
PER = PERn ( sd ) PERn ( sd + rd ) .
r
n

PER DT =

(12)

We assume that the source node only knows the

instantaneous channel state information (CSI) of the sourcedestination link. So the TM selection only depends on the
direct link quality. Let N denote the total number of TMs
available ( N

= 6)

{ }
th
n

and

transmission mode

denote the thresholds for the

= 0,
= ln an2 PMAC 3g n , n = 1,..., N 1,
= +.

nth+1

Pr (n)

th
n

(16)

n=1

Rn Pr (n) (1 PER
N

(13)

r
nDT

n=1

1+ Rn Pr (n) PER n ( sd )
n=1

(17)

Rn Pr (n)
n=1

B. Cooperative Transmission
For cooperative transmission scheme, the SNR of the
combined packet is sd + rd when the retransmission is
started by the relay and the average PER corresponding to TM
n can be written as (18).

Let PER n ( sd ) denote the average PER for mode n after

primary transmission:
1

Rn Pr (n) .

rate of correctly received packets

number of transmission per packet

S DT =

It should be noted that the TM used in retransmission is the

same as that in the primary transmission for all the three
schemes.

PER n ( sd ) =

Rn Pr (n) PER nDT

The spectral efficiency of direct transmission scheme can

be calculated by

n . From (7) and (11), we obtain:

th
0
th
n
th
N

n=1

N
n=0

number of incorrectly received packets

total number of transmitted packets

PERn ( ) p sd ( )d

Let PER nCT denote the average PER corresponding to TM

n of cooperative transmission scheme as:

pn

1 th
g n sd exp(
) + an exp( ( g n +
) n+1 )

sd
sd
1
, n = 0 (14)
Pr (n)
( g n sd +1) Pr (n)
=
.
a [exp(( g + 1 ) th ) exp( ( g + 1 ) th )]
n
n
n
sd n
sd n+1
, n =1,...,5

( g n sd +1) Pr (n)

r R

PER nCT = 1 PER n ( sr ) PER nCT + PER n ( sr ) PER nDT (19)

where PER n ( sr ) denotes the error probability of the packets
that the relay receives from the source node.
Then, similar to (16), the average PER of AMC for
cooperative transmission scheme can then be derived as:

where sd is the average SNR of the source-destination link.

PERCT = Rn Pr (n) PER nCT

A. Direct Transmission
For direct transmission scheme, the receiver will combine
the soft decision of the original and the retransmitted packet,

n=1

Rn Pr (n) .

(20)

n=1

The spectral efficiency of cooperative transmission scheme

is given by

898

 Rn Pr (n) (
N

SCT =

r
1 PER nCT

n=1

N
1+ Rn Pr (n) PER n ( sd )
n=1

10

Rn Pr (n)

n=1

(21)

-1

10

PER

C. Hybrid Transmission
For hybrid transmission scheme, only when the direct
channel is deep fading ( n = 0 ), system will run cooperation.

-2

10

Direct(sim.)
Direct(anal.)
Cooperative(sim.)
Cooperative(anal.)
Hybrid(sim.)
Hybrid(anal.)

Let PER nHT denote the average PER corresponding to TM n

of hybrid transmission scheme as:

-3

10

r
r
r
1 PER n ( sr ) PER nCT + PER n ( sr ) PER nDT , n=0
PER nHT =
. (22)
r
PER nDT , n=1,...,5

-4

10

r
PER HT

Rn Pr (n)

Rn Pr (n) .

n=1

(23)

S HT =

Rn Pr (n) (1 PER nHT )

N
1+ Rn Pr (n) PER n ( sd )
n=1
IV.

n=1

Rn Pr (n)

RESULTS

nth+1

th
n

10

12

14

16

18

20

Then, we consider another scenario (Scenario 2) where the

relay closes to the source. Fig. 4 shows the spectral efficiency
for this scenario. Assume the average SNR for source-relay
channel is sr = 20 dB which indicates a rather good sourcerelay channel. The source-destination channel and relaydestination channel have the same average SNR varying

Consider a scenario (Scenario 1) where all the average link

SNRs are varied simultaneously. In this case, the source-relay
and relay-destination links have the same average SNR, while
the source-destination link has a lower average SNR, which is
a typical scenario when relay is located around the midpoint of
source and destination. We assume sr = rd = 4 sd = . Fig. 2

Pr (n)

Fig. 3 shows the spectral efficiency of the three schemes for

Scenario 1. We notice that both the cooperative transmission
scheme and the hybrid transmission scheme can achieve
substantial spectral efficiency gain over the direct transmission
scheme when is low, while the spectral efficiency gain is
slight when is high enough. This is because when the SNR
of the source-destination channel is high enough, the
retransmission will rarely happen and the cooperation gain
could be rarely got.

(24)

In this section, we present some numerical results to verify

the analytical results and compare the performance of different
schemes. We assume the packet loss probability at LL
4
Plink = 10 and the maximum number of retransmission is one.

n=1

r R
PER nCT = 1

compare theoretical with simulation results for PER. In this

figure, lines correspond to analytical expressions, while each
point signifies the corresponding simulation results. It verifies
the performance analysis that the simulation results are very
close to the theoretical results.

n=1

The spectral efficiency of hybrid transmission scheme can

be calculated by
N

Figure 2. Comparison between analytical and simulated PER .

r
PER nHT

SNR (dB)

Moreover, similar to (16) and (20), the average PER of

AMC for hybrid transmission scheme can then be derived as:

PERn ( 1 ) PERn ( 1 + 2 ) p sd ( 1 ) p rd ( 2 ) d 1 d 2

pn exp( pn ) exp( pn )an2 exp( (2 g n + 1 ) nth )

rd
sd
sd
1 [1exp( pn ) pn exp( pn )+
n = 0, rd = sd
+
],
Pr (n)
rd rd
sd sd (g n rd +1)
(2 g n sd +1)(g n rd +1)
pn
pn
pn
pn
pn

sd exp(
) rd exp(
) rd [exp(
)exp(
)] exp(
)an2 exp( (2 g n + 1 ) nth )
(18)

sd
rd
sd
rd
sd
sd
= 1
+
+
[1+
], n = 0, rd sd
Pr (n)
rd sd
( g n rd +1)( sd rd )
(2 g n sd +1)(g n rd +1)

th
th
2

an [exp( (2 g n + 1 ) n )exp( (2 g n + 1 ) n+1 )]

sd
sd
,
n =1,...,5

(2 g n sd +1)(g n rd +1) Pr (n)

899

Direct
Cooperative
Hybrid

1.8
1.6

Direct
Cooperative
Hybrid

1.1
1

Spectral Efficiency

Spectral Efficiency

0.9

1.4
1.2
1
0.8

0.7
0.6
0.5

0.6

0.4

0.4

0.3

0.2

0.2

0.1

10

12

14

16

18

SNR (dB)

20

10


Figure 4. Spectral efficiency for different schemes in Scenario 2.
1.17

from 0 to 10dB. It can be seen that the hybrid transmission

scheme and the cooperative transmission scheme could
improve the spectral efficiency even when the average SNR of
the source-destination channel is same as the relay-destination
channel.

1.165

Spectral Efficiency

1.16

Finally, we study a scenario (Scenario 3) where the average

SNR for source-destination channel is sd =10 dB and the
source-relay channel and relay-destination channel have the
same average SNR . Fig. 5 shows the spectral efficiency of
the three schemes for this scenario. We notice that the hybrid
transmission scheme and the cooperative transmission scheme
could improve the spectral efficiency even when the sourcerelay channel and relay-destination channel are worse than the
source-destination channel.

1.155

Direct
Cooperative
Hybrid

1.15

1.145

1.14

1.135
0

10

12

14

16

18

20

SNR (dB)

From Fig. 3-5, it can be seen that hybrid transmission

scheme has the similar performance as the cooperative
transmission scheme in all three scenarios. It means that the
hybrid transmission scheme benefits to reduce the overhead of
the cooperation without performance degrade.
V.

SNR (dB)

Figure 3. Spectral efficiency for different schemes in Scenario 1.

Figure 5. Spectral efficiency for different schemes in Scenario 3.

[2]

CONCLUSIONS

[3]

In this paper, we developed a novel hybrid transmission

scheme for wireless Ad-hoc networks. We provided theoretical
analyses of conventional direct transmission scheme,
cooperative transmission scheme and hybrid transmission
scheme that included analytical expressions for the packet error
rate and spectral efficiency. Numerical results show that the
performance of the hybrid transmission scheme is as good as
the cooperative transmission scheme with a much lower
cooperation cost. Moreover, the performances for both of them
are obviously better than the direct transmission scheme.

[4]

[5]

[6]

[7]

REFERENCES
[1]

0.8

[8]

J. N. Laneman, D. N. C. Tse, and G. W. Wornell, Cooperative diversity

in wireless networks: Efficient protocols and outage behavior, IEEE
Trans. Info. Theory, vol. 50, no. 12, pp. 3062-3080, Dec. 2004.

900

P. Liu, Z. Tao, Z. Lin, E. Erkip, and S. Panwar, Cooperative wireless

communications: A cross-layer approach, IEEE Wirel. Commun., vol.
13, no. 4, pp. 84-92, Aug. 2006.
F. A. Onat and D. Avidor, Relay-Assisted ARQ in Wireless Ad-Hoc
Networks, IEEE Wireless Commun. and Networking Conference, pp.
500-505, 2008.
G. Yu, Z. Zhang and P. Qiu, Efficient ARQ Protocols for Exploiting
Cooperative Relaying in Wireless Sensor Networks, Elsevier Computer
Commun., vol. 30, no. 14-15, pp. 2765-2773, Oct. 2007.
B. Zhao, M. C. Valenti, Practical relay networks: A generalization of
hybrid-ARQ, IEEE J. Sel. Areas Commun., vol. 23, no.1, pp. 7-18, Jan.
2005.
M. S. Alouini and A. J. Goldsmith, Adaptive modulation over
Nakagami fading channels, Kluwer J. Wireless Commun., vol. 13, no.
12, pp. 119-143, May 2000.
X. Wang, Q. Liu, and G. B. Giannakis, Analyzing and optimizing
adaptive modulation and coding jointly with ARQ for QoS-guaranteed
traffic, IEEE Trans. Veh. Tech., vol. 56, no. 2, pp.710720, Mar. 2007.
F. Shi, D. Yuan, and L. Xu, Performance analysis of Type-III HARQ
scheme in cross-layer design for QoS-guaranteed traffic, IEEE Proc. of
the Int. Workshop on Cross Layer Design, pp. 133-137, Sept. 2007.