In this paper, we derive the outage probability for Mary phase shifting keying (MPSK) and Mary quadrature amplitude modulation (MQAM) burst transmission (BT) of adaptive decodeandforward (ADF) cooperative relay systems over quasistatic Rayleigh fading channels. Within a burst, there are pilot symbols and data symbols. Pilot symbols are used for channel estimation schemes and each relay node’s transmission mode selection schemes. At first, we focus on ADF relay systems in which the probability density function (PDF) is derived on the basis of error events at relay nodes corresponding to channel estimation errors. Next, the average outage probability is derived as an approximate expression for an arbitrary link signaltonoise ratio (SNR) for different modulation orders. Its accuracy is demonstrated by comparison with simulation results. Further, it is confirmed that BTADF relay systems with pilot symbol based channel estimation schemes enables to select correctly decoded relay nodes without additional signaling between relay nodes and the destination node, and it is verified that the ideal performance is achieved with small SNR loss.
1. INTRODUCTION
Many researches have widely discussed cooperative relay schemes. In general, there are two main relay protocols for cooperative diversity schemes: amplifyandforward (AF) and decodeandforward (DF). AF amplifies the received signal and retransmits it to the destination, whereas DF detects the received signal and then retransmits a regenerated signal
[1]

[3]
. A third option is adaptive DF (ADF) scheme, in which relays forward only correctly decoded messages
[4]

[6]
. At ADF relay nodes, errors are assumed to be correctly detected by using a cyclic redundancy check (CRC) code from a higher layer (e.g., data link layer)
[3]
,
[7]

[9]
. At the destination node, the receiver can enhance performance by employing one of various diversity combining techniques based on the multiple signal replicas from the relays and the source. The advantages of general cooperative diversity schemes come at the expense of the spectral efficiency since the source and all the relays must transmit on orthogonal channels (i.e., different time slots or frequency bands) in order to avoid interfering with each other as well
[3]
. Recent studies have examined relayselection schemes in which only two channels are necessary (one for the direct link and the other one for the best relay link)
[10]

[14]
. However, they need additional process or feedback information for channel states.
In
[5]
, the authors have derived an exact bit error rate (BER) applicable for both DF and ADF relaying as wellknown tractable forms. It shows how an erroneous detection at each relay affects both the received signaltonoise ratio (SNR) and the average BER. Even if it can give exact results
[5]
,
[15]
, it is noted that previous researches including relayselection schemes have assumed that each relay can detect symbolbysymbol error
[5]
,
[11]
,
[15]
. It means that at each relay, transmission mode ('Tx. mode') or notransmission ('Sleep mode') can be determined per symbolbysymbol. However, this is not practical and the performance based on this assumption implies only an achievable bound.
In
[16]

[18]
, the authors showed the practical approach based on burst transmission for DF relay systems. Nevertheless, no one has expressed the approximated outage expression as wellknown tractable forms, which can cover both Mary phase shifting keying (MPSK) and Mary quadrature amplitude modulation (MQAM) burst transmission. In
[19]
, the authors provided a framework for analyzing the BER performance of AF relayassisted cooperative transmission in the presence of imperfect channel estimation. However, the framework in
[19]
does not include pilot symbol assistedchannel estimation (PSACE) schemes which can be applied in practical systems, resulting in errorfloor even at high SNR region. We extend the analytical approach in
[19]
to ADF burst transmission systems.
In this paper, we consider burstbyburst error detection for ADF relay systems, instead of symbolbysymbol
[16]

[18]
. At first, we derive the probabilities for all possible errorevents at relay nodes. By considering pilot and data symbol transmission within a burst, we derive error rate expressions over quasistatic independent and nonidentical distributed (INID) Rayleigh fading channels, so that it can be an actual system performance. Furthermore, the average outage probability is approximated to a simplified expression for arbitrary link SNRs related to channel estimation errors and modulation order. In numerical and simulation results where the derived analytical solutions are compared with MonteCarlo simulations, we verify that correctly decoded relay nodes can be selected from transmitted pilot symbols without additional signaling between relay nodes and the destination node. Furthermore, its performance well matches with our analytical results for all SNR regions and different modulation order.
The remainder of this paper is organized as follows: Section 2 describes the system model of BTADF cooperative relay systems. In section 3, the derived outage performance expression is presented. The numerical and simulation results are presented in Section 4 and also concluding remarks are given in Section 5.
2. BTADF COOPERATIVE RELAY SYSTEMS
[
Fig. 1
] shows the block diagram of BTADF relay systems with a source (S), a destination (D), and relays (R) where
L
is the number of relays. We assume that S and
L
relays transmit over orthogonal time slots so that we need
L
+1 time slots for single burst transmission
[3]
. At first, we explain fading channel model used in this paper.
Block diagram of BTADF Cooperative relay systems
 2.1 BTADF COOPERATIVE RELAY CHANNEL MODEL
Let
h_{0}
,
h
_{L}
_{+}
_{r}
, and
h_{r}
(
r
={
1,2,
…,
L
}) be the channel gains of SD, SR, and RD links, respectively, as shown in
Fig. 1
. In this paper, wireless channels between any pair of nodes are assumed to be quasistatic Rayleigh fading
[20]
,
[21]
. It means that channel coefficients are considered to be constant during bursttransmissions and then, the magnitude and the phase of
h_{r}
are Rayleigh distributed and uniformly distributed over [0, 2
π
], respectively. From here,
N_{P}
and
N_{D}
are the number of pilot symbols and the number of modulated data symbols within a
N_{B}
denotes the length of a burst (i.e.,
N_{B}
=
N_{P}
+
N_{D}
). Also, each link channel is corrupted by complex additive white Gaussian noise (AWGN) term of
n_{r}
[
t
]. Without loss of generality, it is assumed that
E
[
n_{r}
[
t
]] = 0,
E
[
n_{r}
[
t
]
^{2}
] =
σ
^{2}
, and {
n_{r}
[
t
]} are mutually independent for different
r
and
t
. The operator
E
[ ⋅] represents statistical expectation.
For simplicity, this paper considers the first burst transmission. Then,
is the pilot symbol known to all nodes and
is MPSK or MQAM data symbol. Also,
are mutually independent for different
t
with
E
[
s
[
t
]] = 0 and
E
[
s
[
t
]
^{2}
] = 1.
 2.2 BTADF COOPERATIVE RELAY SYSTEM MODEL
As shown in [
Fig. 1
], BTADF cooperative relay systems have (
L
+
1
) steps for single burst transmission. The
0
th step is related to the transmission from the source node to all the relays and the destination by using the
0
th time slot. During the
0
th step, the received signals can be presented for the SD link and the
r
th SR link as
where
E_{0}
=
E
_{L}
_{+}
_{r}
=
E_{S}
is the average transmitted symbol energy of the source and
t
=
1,2
,…,
N_{B}
is the time index of the first burst transmission. In (1),
y
_{0}
[
t
] is the received signal at the destination during the
0
th time slot.
For the remaining
L
steps, each relay transmits the regenerated data symbols. Only when all
N_{D}
data symbols are correctly decoded, the
r
th relay transmits the regenerated symbol
in the
r
th transmission step. For the
r
th time slot, the received signal at the destination node is written with
t
=
1,2
,…,
N_{B}
as
where
n_{r}
[
t
]=
n
_{0}
[
t
+
rN_{B}
] and
E_{r}
is the average transmitted symbol energy of the
r
th relay node.
 2.3 PILOTSYMBOL BASED CHANNELESTIMATION
For
in (2) are pilots symbols to estimate RD link channels so that, (1) and (2) can be expressed as single equation of
with
r
∈ {0,1,..., 2
L
}. Then, for pilotsymbol based channelestimation schemes, the channel gains can be obtained as
where
is the channel estimation error with
E
[
e_{r}

^{2}
]=
σ
^{2}
/
N_{P}
and
E
[
e_{r}
]=0. From pilot symbols and the estimated channel gain
the noise variance for pilot symbols can be estimated as
Note that for large
N_{P}
, the estimated noise variance in (4) can be approximated as
The statistical noise variance for data symbol transmission can be expressed as
From (5) and (6), the noise variance for data symbols can be estimated as
Note that in above equations from (3) to (7),
r
=0,
r
=
1
,…,
L
, and
r
=
L
+
1
,…,
2L
mean SD, RD, and SR links, respectively.
 2.4 PILOT SYMBOL BASEDRELAYING MODE SELSECTION
Under pilotsymbol based channel estimation methods, it is assumed that each relaying node can always transmit pilot symbols to the destination node. Then, the destination node can simply detect each relay's data transmission mode and hereafter, it refers to pilot symbol basedrelaying mode selection (PBRMS). During the
r
th time slot, we can examine the average signal powers for pilot symbol part and data symbol part, respectively, as follows:
By comparing
the
r
th relay's data transmission mode
can be estimated as
where
T_{e}
is a threshold. At the destination node, a maximal ratio combing (MRC) scheme can be applied in order to combine signals from SD and RD links. Then, by using
the decision variable can be combined as
3. AVERAGE OUTAGE PERFORMANCE ANALYSIS FOR BTADF COOPERATIVE RELAY SYSTEMS
 3.1 EACH RELAY'S ERROR PROBABILITY
From
the
r
th SR link's decision variable is shown for data symbol transmission as
and then, the received SNR can be obtained as
with
and
Then, the probability density function (PDF) of random variable
γ
_{L}
_{+}
_{r}
can be presented for the Rayleigh fading channel as
where
is the average SNR of
When the number of pilots increases, the channel estimation error decreases and the average SNR merges to the case of ideal channel estimation. Furthermore, above derivations can be also applied to SD and each RD link by replacing
L
+
r
with
r
, so we can obtain
for
r
=
0
,
1
,…,
L
.
Consequently, the
r
th SR link's conditional SER can be approximated for MPSK as
with
and
[20]
[21]
.
 3.2 ERROR EVENET PROBABILITY
For errorevents at relays, the
p
th errorevent vector is defined as
with
p
=
1
,
2
,..,
2^{L}
and the total number of errorevents is
2^{L}
. Generally, we can define that
E^{1}
is allzero vector,
E^{2L}
is allone vector, and so on
[5]
,
[22]
. For the
p
th errorevent,
means that the
r
th relay correctly decodes
N_{D}
symbols (i.e.,
for
N_{P}
<
t
≤
N_{B}
) and its 'Tx. mode' probability is
with
P_{S}
(
γ
_{L}
_{+}
_{r}
) of (15). In addition, the average 'Tx. mode' probability can be written as
with
On the other hand,
indicates that there is at least one symbol error among
N_{D}
data symbols with the 'Sleep mode' probability of
Consequently, the probability of the
p
th errorevent at BTADF relay systems is presented as
with
[5]
,
[22]
. The evaluation of (18) can be carried out by using the 'integral( )' function of MATLAB. When we apply the approximation of the Qfunction, shown in
[23]
, as
with
α
_{1}
/
α
= 0.2 and
b
_{1}
/
b
= 3.2/3 into (18), we can obtain the approximated bound as
and from
the result of (19) can be simplified.
 3.3 COMBINED RECEIVED SNR AND AVERAGE OUTAGE PROBABILITY
For BTADF relay systems, the
r
th relay transmits the regenerated data symbols of
only when
N_{D}
data symbols are correctly decoded. Therefore,
can be two values: one is
with the probability of
and the other is
with the probability of
Under the assumption that the destination node knows correctly decoded relay nodes, the combined decision variable is written by using
as
and then, the received SNR can be written as
with
for all
p
. It is worthwhile to mention that when there is a detectionerror at the
r
th relay node for the
p
th event vector
notransmission gives
The PDF of can be presented as
can be presented as
with
[5]
[20]
. Then, the outage probability can be expressed with respect to
γ_{th}
as
By taking into account for all the possible errorevents, the outage probability is presented as
[5]
,
[22]
4. NUMERICAL AND SIMULATION RESULTS
In this section, we show numerical results of average outage probability and then, evaluate their accuracy by comparing simulation results. For simplicity, we assume that
E_{r}
=
E_{s} / L
for
r
= 1,...,
L
. To capture the effect of pathloss on average outage probability,
α_{r}
=
r
/ (
L
+1) is defined as the relative distance between source and the
r
th relay when the distance between source and destination is 1. Then, we use
with the pathloss factor
μ
. From here, we use
μ
=
3.76
which is the parameter of outdoor hotzone model [Table A.2.1.1.23] in
[24]
and SNR is defined as
'Analysis' indicates the numerical results obtained from (26) with
in (21) and 'Simulation' denotes the imulation results obtained from the assumption that the destination node can perfectly know each relay node's transmission mode (i.e., 'Tx. mode' or 'Sleep mode'). On the contrary, 'Simulation w/ PBRMS' indicates the simulation results which are obtained from each relay's 'Tx. mode' selection based on PBRMS of (9) with
T_{e}
=
1.0
.
For BPSK (
M
=
2
), the average outage probabilities are shown in
Figs. 2
and
3
when
N_{P}
=∞ and
N_{P}
=8, respectively. As a performance reference, we also plot the SD link's outage performance.
Fig. 2
shows the results of
N_{P}
=∞ which means the case of perfect channel estimation. It is shown that numerical results of 'Analysis' are well matched with simulation results for all SNR regions. On the other hand, in
Fig. 3
where
N_{P}
=8 which means the case of practical channel estimation, some mismatches are shown at lower SNR. Note that for BPSK 'Analysis' with
M
=
2
, single approximation of (20) is used for the burst error rate simplification, whereas the approximation of (15) is not used for BPSK.
Average Outage Probability versus SNR (dB) with respect to different N_{D} and L for the ideal channel estimation (L=1,4, N_{P}=∞, N_{D}=1,32, M=2, μ=3.76).
Average Outage Probability versus SNR (dB) with respect to different N_{D} and L for the practical channel estimation (L=1,4, N_{P}=8, N_{D}=1,32, M=2, μ=3.76, T_{e}=1.0).
[
Fig. 4
] shows the performance comparison with respect to
N_{P}
. From three figures, it is verified that
N_{P}
=8 shows less than 0.5dB SNR loss when it is compared with the ideal channel estimation case. In addition, we can find that average outage probability increases in proportion to
N_{D}
(the number of data symbols within a burst). When
N_{D}
increases, each relay's 'Tx. mode' probability decreases. Consequently, it generates the performance degradation shown as SNR loss at high SNR regions. Also, it is worthwhile to mention that
N_{D}
=1 means the symbolbysymbol detection of previous researches
[5]
,
[15]
. The performance for
N_{D}
=1 is confirmed to be an achievable lower bound for ADF relaying schemes. It is noted that as a practical performance reference, we also plot the simulated performance obtained from using PBRMS of (9) with
T_{e}
=
1.0
. When comparing our approximated analytical results with two simulation results, we can find that they are well matched and the accuracy of the derived analytical method is verified. Also, even if performance loss occurs according to the increase of
N_{D}
, we can still find the diversity gain caused by the increase of
L
.
Average Outage Probability versus SNR (dB) with respect to different N_{P} and L (L=1,4, N_{P}=8,∞, N_{D}=32, M=2, μ=3.76, T_{e}=1.0).
[
Fig. 5
], [
Fig. 6
], and [
Fig. 7
] show the average outage probability versus SNR with respect to
M
for
L=1, L=2
, and
L=4
, respectively. We can see that the diversity order linearly increases as the number of relays,
L
. Moreover, it is worthwhile to mention that the approximated analytical bounds are tight enough for all SNR values and for the different modulation order
M
. It is seen from three figures that the performance loss caused by the increase of
M
is similar for the different
L
(the number of relays).
Average Outage Probability versus SNR (dB) with respect to different M (L=1, N_{P}=8, N_{D}=32, M=2,4,8,16,64, μ=3.76).
Average Outage Probability versus SNR (dB) with respect to different M (L=2, N_{P}=8, N_{D}=32, M=2,4,8,16,64, μ=3.76).
Average Outage Probability versus SNR (dB) with respect to different M (L=4, N_{P}=8, N_{D}=32, M=2,4,8,16,64, μ=3.76).
Consequently, we verify that even though our analytical approach is based on the perfect knowledge of correctly decoded relay nodes, it shows the achievable error rate performance of actual ADF relay systems having pilot symbol transmission schemes. In other words, ADF relay systems with PSACE methods can select correctly decoded relay nodes without additional signaling between relay nodes and the destination node and also, the achievable performance is guaranteed at a cost of negligible SNR loss.
5. CONCLUSIONS
The average outage probability is derived as the approximated closedform for BTADF relay systems over quasistatic INID Rayleigh fading channels. Our proposed analytical approach includes channel estimation errors related to transmitted pilot symbols within a burst. Firstly, for the relay nodes' error event, its probability is approximated as the form which is related to the error probability of a burst MPSK or MQAM transmission. Then, the average outage probability is derived as the closedform to be simply calculated by numerical operations. It is verified to be an outage performance bound by comparing with simulation results. Therefore, we can conclude that our analytical outage expression is very tractable form, and can be used as a tool to verify outage performance for the different modulation order, the numbers of pilots and data symbols within a burst.
Acknowledgements
This research was supported by the MSIP(Ministry of Science, ICT and Future Planning), Korea, under the CITRC(Convergence Information Technology Research Center) support program (NIPA2014H0401141007) supervised by the NIPA(National IT Industry Promotion Agency) and by Basic Science Research Program through the National Research Foundation of Korea(NRF) funded by the Ministry of Education, Science and Technology (NRF2013R1A1A1A05008009).
BIO
Kyunbyoung Ko
He received the B.S., M.S., and Ph.D. degrees in Electrical and Electronic Engineering at Yonsei University, Seoul, Korea in 1997, 1999, and 2004, respectively. From March 2004 to February 2007, he was a senior engineer in Samsung Electronics, Suwon, Korea where he developed Mobile WiMAX systems for broadband wireless services. Since March 2007, he joined the Department of Control and Instrumentation Engineering at Korea National University of Transportation as an Associate Professor. His current research interests include the field of wireless communications focusing on multicarrier and multiantenna systems, cooperative relaying, and ITS.
Sungmook Lim
He (corresponding author) received the B.S. and Ph.D. degrees in Electrical and Electronic Engineering at Yonsei University, Seoul, Korea in 2005 and 2012, respectively. From September 2012 to March 2014, he was a Postdoctoral Fellow at Yonsei University where his research interents were in 5G wireless communications. Since March 2014, he joined the Department of Electronics Engineering at Korea National University of Transportation as an Assistant Professor. His current research interests include the field of 5G wireless communications focusing on multicarrier and multiantenna systems, cooperative relaying, and ITS.
Hasna M. O.
,
Alouini M. S.
2003
“EndtoEnd performance of transmission systems with relays over Rayleighfading channels,”
IEEE Trans. Wireless Commun.
2
(6)
1126 
1131
DOI : 10.1109/TWC.2003.819030
Anghel P. A.
,
Kaveh M.
2004
“Exact symbol error probability of a cooperative network in a Rayleighfading environment,”
IEEE Trans. Wireless Commun.
3
(9)
1416 
1421
DOI : 10.1109/TWC.2004.833431
Laneman J. N.
,
Tse D. N. C.
,
Wornell G. W.
2004
“Cooperative diversity in wireless networks: efficient protocols and outage behavior,”
IEEE Trans. Inf. Theory
50
(12)
2062 
3080
DOI : 10.1109/TIT.2004.838089
Lee InHo
,
Kim Dongwoo
2007
“BER Analysis for DecodeandForward Relaying in Dissimilar Rayleigh Fading Channels,”
IEEE Commun. Lett.
11
52 
54
Jang J. Y.
,
Ko K. B.
2011
“Exact & closedform BER expressions based on Error Events at Relay nodes for DF Relay Systems over Rayleigh Fading channels,”
IEICE Trans. on Commun.
E94B
(08)
2419 
2422
DOI : 10.1587/transcom.E94.B.2419
KIMURA K.
,
MIYAZAKI H.
,
OBARA T.
,
ADACHI F.
2014
“SignalCarrier Cooperative DF Relay Using Adaptive Modulation,”
IEICE Trans. on Commun.
E97B
(2)
387 
395
DOI : 10.1587/transcom.E97.B.387
Lee S.
,
Han M.
,
Hong D.
2009
“Average SNR and ergodic capacity analysis for opportunistic DF relaying with outage over rayleigh fading channels,”
IEEE Trans. Wireless Commun.
8
(6)
2807 
2812
DOI : 10.1109/TWC.2009.080919
Sagong S.
,
Lee J.
,
Hong D.
2011
“Capacity of reactive DF scheme in cognitive relay networks,”
IEEE Trans. Wireless Commun.
10
(10)
3133 
3138
DOI : 10.1109/TWC.2011.081011.101849
Gupta S. D.
2009
Forwarding Strategies and Optimal Power Allocation for Relay Networks: Forwarding Strategies and Position Dependent Optimal Power Allocation for Coherent and Noncoherent Relay Networks
1st ed.
VDM Verlag
Bletsas A.
,
Khisti A.
,
Reed D. P.
,
Lippman A.
2006
“A simple cooperative diversity method based on network path selection,”
IEEE J. of Selected Areas in Commun.
24
(3)
659 
672
DOI : 10.1109/JSAC.2005.862417
Ikki S. S.
,
Ahmed M. H.
2010
“Performance Analysis of Adaptive DecodeandForward Cooperative Diversity Networks with BestRelay Selection,”
IEEE Trans. Commun.
58
(1)
68 
72
DOI : 10.1109/TCOMM.2010.01.080080
Ko K. B.
,
Woo C. W.
2012
“Outage Probability and Channel Capacity for Nth Best Relay Selection AF Relaying over INID Rayleigh Fading Channels,”
International Journal of Communication Systems
25
(11)
1496 
1504
DOI : 10.1002/dac.1357
Nam S. H.
,
Ko K. B.
,
Hong D. S.
2012
“Exact Average SER Performance Analysis for the Nth Best Opportunistic AmplifyandForward Relay Systems,”
IEICE Trans. on Commun.
E95B
(5)
1852 
1855
DOI : 10.1587/transcom.E95.B.1852
Ko K. B.
,
Woo C. W.
2013
“More Accurate ASER Bound for Opportunistic AmplifyandForward Relay Systems,”
Wireless Personal Communications
68
(3)
609 
617
DOI : 10.1007/s1127701104718
Lee Y.
,
Tsai M. H.
,
Sou S. I.
2009
“Performance of DecodeandForward Cooperative Communications with Multiple DualHop Relays over Nakagamim Fading Channels,”
IEEE Trans. Wireless Commun.
8
(6)
2853 
2859
DOI : 10.1109/TWC.2009.080919
Woo C. W.
,
Ko K. B.
2014
“Decodeandforward with partial relay selection,”
International Journal of Control and Automation
7
(10)
99 
108
DOI : 10.14257/ijca.2014.7.10.10
Seo G. W.
,
Baek M.
,
Bong C. H.
,
Ko K. B.
“Performance Evaluation for Cooperative ADF Relaying V2I Communications with Burst transmission and PSACE schemes Over QuasiStatic Rayleigh Fading channels,”
Proc. ITSC’14
Qingdao, China
Oct. 811 2014
2101 
2016
Baek M.
,
Seo G. W.
,
Bong C. H.
,
Ko K. B.
“Performance Analysis of Cooperative HADF Relaying V2I Communication Systems for NDsymbol Burst Transmission using PSACE methods Over QuasiStatic Rayleigh Fading channels,”
Proc. ITSC’14
Qingdao, China
Oct. 811 2014
2081 
2086
Han S. Y.
,
Ahn S. W.
,
Oh E. S.
,
Hong D. S.
2009
“Effect of ChannelEstimation Error on BER Performance in Cooperative Transmission,”
IEEE Trans. Veh. Technol.
58
(4)
2083 
2088
DOI : 10.1109/TVT.2008.2007644
Proakis John G.
1995
Digital Communication
3rd ed.
McGraw Hill
Simon M. K.
,
Alouini M. S.
2000
Digital Communication over Fading Channels
John Wiley & Sons
Ko K. B.
,
Woo C. W.
2013
“More Tractable Average Error Rate Expression for the Nth Best Relay Selection scheme of DF Relaying over Rayleigh fading Channels,”
International Journal of Communication Systems
25
(10)
1356 
1364
Cai X.
,
Giannakis G. B.
“Adaptive Modulation with Adaptive Pilot Symbol Assisted Estimation and Prediction of Rapidly Fading Channels,”
The Johns Hopkins University
2003 Conference on Information Sciences and Systems
Mar. 1214 2003
3GPP TR 36.814 V9.0.0,Evolved Universal Terrestrial Radio Access (EUTRA); further advancements for EUTRA physical layer aspects.