Joint Network-Channel Coding For The Multiple-Access Relay Channel

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Joint Network-Channel Coding for the

Multiple-Access Relay Channel


Christoph Hausl and Philippe Dupraz
Institute for Communications Engineering (LNT), Munich University of Technology (TUM), 80290 Munich, Germany
{christoph.hausl, philippe.dupraz}@mytum.de

Abstract— We propose to use joint network-channel coding has to send one transmission to support the uplink of MS 2.
based on turbo codes for the multiple-access relay channel. Such If the uplink for MS 1 and MS 2 is done cooperatively with
a system can be used for the cooperative uplink for two mobile the help of network coding at the relay, we only have to send
stations to a base station with the help of a relay. We compare
the proposed system with a distributed turbo code for the relay one transmission from the relay to offer the same diversity
channel and with a system which uses separate network-channel gain for both mobile stations. If we fix the allowed number of
coding for the multiple-access relay channel. Simulation results transmissions from the relay to one, it is still possible to gain
confirm that the systems with network coding for the multiple- the full diversity with network coding. However, if we do not
access relay channel gain cooperative diversity compared to the use network coding and treat the two uplinks separately, the
system with the distributed turbo code for the relay channel.
Moreover, the results show that joint network-channel coding two relays have to share the available transmission and it is
outperforms separate network-channel coding. The reason for not possible to obtain the full diversity gain. The diversity
this is that the redundancy which is contained in the transmission gain through network coding increases with the number of
of the relay can be exploited more efficiently with joint network- transmitters. It was shown in [3] how in large-scale wireless
channel coding. networks with many transmitters the network code can be
I. I NTRODUCTION decoded on a graph which can be obtained from the network
graph of the transmitters.
While the concept of network coding [1] was first pro-
posed to increase the throughput in networks with error-free B. Joint Network-Channel Coding
transmission, the application of network coding for wireless One way to gain diversity through network coding for the
networks with disturbed channels is gaining interest. MARC with noisy channels is to treat network and channel
coding separately. Then, channel coding is used in the physical
A. Diversity through Network Coding
layer for each transmission to transform the noisy channels to
In [2]–[7] it was discussed how to gain cooperative diver- erasure-based links. On the network layer, one performs net-
sity through network coding in wireless networks to combat work coding for the erasure-based networks which is provided
channel fading. The concept of cooperative diversity [8] was by the lower layers.
initially applied (without network coding) for the relay channel However, a relay cannot only be used to gain diversity. Its
(RC) which is depicted in Fig. 1 (a). A simple model where transmission can be seen as additional redundancy which
diversity can be gained with network coding is the multiple- improves the performance compared to a point-to-point com-
access relay channel (MARC) [9] which is depicted in Fig. 1 munication if the relay has a better connection to the base
(b). An example for an MARC is the cooperative uplink of station than the mobile station. For this case relays are also
useful for noisy channels without fading where diversity
MS 1 Broadcast MS 1 Broadcast
is not relevant. Distributed channel codes [10]–[13] can be
BS BS applied to efficiently exploit the (direct) redundancy from the
R1 mobile station and the additional redundancy from the relay.
R
R2 Of course, the relay of the MARC delivers also additional
redundancy. To efficiently exploit this redundancy, we have to
(a) MS 2 (b) MS 2 generalize the concept of distributed channel codes to joint
network-channel coding. Distributed channel codes for the
Fig. 1. The uplink for two mobile stations (MS 1 and MS 2) with the help
of relay(s) (R) to a base station (BS) can be done on the relay channel (RC) relay channel can be seen in this context as joint routing-
(a) or on the multiple-access relay channel (MARC) (b). channel coding.
While for the application of network coding to wireline net-
two mobile stations and with the help of a relay to the base works only the network layer is considered and it is assumed
station in a cellular based mobile communication system. If that the lower layers deliver error-free or erasure-based links
the uplink for MS 1 and MS 2 is done separately on two relay with the help of channel coding, the principle of joint network-
channels and we want to gain diversity, relay R 1 has to send channel coding is that the redundancy in the network code
one transmission to support the uplink of MS 1 and relay R 2 should be used to support the channel code for better error
MS 1 BS
Case 1 Case 2
u1 Channel x1 y13 û13 ρMB ρMB ρMB + 10 dB
Channel
encoder 1 ρMB + 21.25 dB ρMB + 21.25 dB
K R N ρMB
Channel
û14 ρMB + 21.25 dB ρMB + 21.25 dB
Channel Joint
y14 ρMB ρMB
decoder 1
Network
x4 y
Channel 43
network/
û24 channel
Channel encoder
y24 NR decoder
decoder 2 Fig. 3. We consider two cases. In case 1 the relay and the mobile stations
MS 2 Channel have the same distance to the base station. In case 2 the SNR between the
u2 Channel y23 û23 relay and the base station is increased by 10 dB.
Channel
encoder 2 x2
K N

Fig. 2. Block diagram of the system. for i ∈ {1, 2, 4} where the noise ni is zero mean and Gaussian
with variance σi2 and the elements of the code blocks xi are
either −1 or +1. We assume that the power at the receiver
protection. It is similar to the principle of joint source-channel is normalized. Therefore, the variance of the noise includes
coding [14], where the remaining redundancy after the source the path-loss. The channel factor ai , which is constrained
encoding helps the channel code to combat noise. We know by E[ai 2 ] = 1, is Rayleigh distributed and represents the
from [15] and [16] that in general, capacity can only be fading due to multipath propagation and the motion of the
achieved by treating network and channel coding jointly, if transmitter. The fading factors ai (i ∈ {1, 2, 4}) of the three
we consider the communication in wireless relay networks. It channels are statistically independent and constant over one
was shown in [4], how joint network-channel coding based on block. The channels are divided in time or frequency and thus,
low-density parity-check (LDPC) codes can be used for the have no interference. The signal-to-noise ratio (SNR) is given
MARC. In [17] joint network-channel coding based on turbo by Eb /N0 = 1/(RS · 2 · σ12 ) = ρMB where we assume that both
codes was applied for the two-way relay channel [18], [19]. mobile stations have the same distance to the base station and
thus σ12 = σ22 . The relay is positioned close to the mobile
C. Main Contributions and Organization of the Paper stations. The distance between a mobile station and the relay
In this paper, we explain how to use joint network-channel is one fourth of the distance between a mobile station and
coding based on turbo codes for the MARC. We want to show the base station. According to the Okumura-Hata model and
that even if the full diversity gain can be obtained with separate the parameters from [20] the SNR of the mobile station -
network-channel coding, it is necessary to use joint network- relay links is given by ρMR = ρMB + 35.2 · log10 (4) dB =
channel coding to exploit the additional redundancy which is ρMB + 21.25 dB. We consider two cases for the relay - base
contained in the transmission of the relay. station link. Case 1 describes a scenario where the relay and
the two mobile stations have the same distances to the base
II. S YSTEM M ODEL station. Therefore, the SNR of the relay - base station link is
A. System Setup given by ρRB = ρMB .
Case 2 describes a scenario where the SNR between relay
In a cellular based mobile communication system (Fig. 1 and base station is higher than the SNR between mobile and
(b)) two users MS1 and MS2 want to transmit statistically base station. Possible reasons for the larger SNR are a lower
independent data which is segmented in blocks u1 and u2 of distance between the relay and base station, a line of sight
length K to the base station BS. A block diagram of the system connection or better antenna characteristics. The SNR of the
is depicted in Fig. 2. The information bits u1 and u2 are relay - base station link is given by ρRB = ρMB + 10.0 dB.
protected against transmission errors with channel encoders Fig. 3 depicts the characteristics of both cases.
which output the code bits x1 and x2 with the block length
N . The relay R receives the disturbed versions of the code bits III. J OINT N ETWORK -C HANNEL C ODING FOR THE MARC
x1 and x2 to obtain the estimates û14 and û24 . Regardless We explain how joint network-channel coding can be per-
whether u1 and u2 are decoded without errors at the relay, formed with a turbo network code [17] for the multiple-access
these estimates of the information bits of MS1 and MS2 are relay channel (MARC). Note that we use also the expression
jointly combined to the network code bits x4 and sent to the parallel concatenated convolutional code (PCCC) for a turbo
base station to provide additional error protection. The network code.
code bits x4 have the block length NR . The code rate of the
system is given by RS = 2 · K/(2 · N + NR ). A. Channel Coding
The channel codes at the mobile stations contain a rate
B. Channel Model
1/2 recursive systematic convolutional code with constraint
The channels from the mobile stations and from the relay to length 4 and information bit block length K = 1500. The
the base station BS are assumed to be block Rayleigh fading feedforward generator is 15 and the feedback generator is 13,
channels and thus, the received samples after the matched filter both in octal. As the code is systematic the output of the
are channel code contains 1503 systematic bits (including three
yi3 = ai · xi + ni , (1) tail bits) vi and 1503 parity bits pi where i ∈ {1, 2}. As
û13
y13 Channel
Dec. 1
| L−
e (u1 ) Π
û14 Le(u1)
−1
Π

Π y43 Network
Dec. Π
Conv.
Enc. Punct. Π −1
v1 x1 p4 x4 |
Le(u2)
u1 Conv. Π L−
e (u2 )
Punct. y23 Channel û23
Enc.
p1 û24 Dec. 2
(a) (b) (c)

Fig. 4. (a): Channel encoder 1 at mobile station 1. (b): Network encoder at the relay. (c): Joint iterative network and channel decoder at the base station.

we only want to send N = 2000 code bits xi , we puncture channel outputs y13 and y23 . We include a value of zero for the
the parity bits pi according to the following rule. We only punctured bits before the decoding starts. The log-likelihood
transmit every third parity bit except the parity bits at position ratios L− −
e (u1 ) and Le (u2 ) are interleaved and mixed in the
373, 748, 1123 and 1498. These position are chosen such that same way as it was done in the network encoder. The log-
the puncturing occurs regularly. A block diagram of channel likelihood ratios after the mixture are a priori knowledge for
encoder 1 at MS 1 is depicted in Fig. 4 (a). the network decoder which contains also an SISO decoder.
The network decoder obtains additional information about
B. Network Coding the parity bits from the channel output y43 . The network
| |
If the relay has decoded the data of both mobile stations decoder calculates extrinsic information Le (u1 ) and Le (u2 )
correctly, the relay network encodes the estimates û14 and which is passed back to the channel decoders. After several
û24 and sends the network code bits x4 to the base station to iterations (e.g. 15) the channel decoders combine all available
provide additional redundancy for both uplinks. information to obtain the estimates û13 and û23 . Note that the
A block diagram of the network encoder is depicted in Fig. 4 network decoder could deliver the estimates as well.
(b). Both estimates are interleaved according to [21]. Then, the
D. Alternative Joint Network-Channel Codes
interleaved bits appear alternately as the input of a convolu-
tional encoder with the same parameters like the convolutional In the previously described system the transmissions from
code used as channel encoder. However, the output of the the mobile stations to the relay are only protected with a con-
network encoder contains only the 3003 parity bits of the volutional code. In certain scenarios it could be advantageous
convolutional encoder. As we only want to send NR = 2000 to protect these links with a PCCC. Such joint network-channel
bits we puncture every third bit and the bits at position 1000 codes are described in [22].
and 2000. The network code of rate RR = (2 · K)/NR = 1.5 In [23] a joint network-channel coding scheme was proposed
provides NR additional parity bits which support the decoding which outperforms turbo network coding for the two-way
at the base station. relay channel. However, for the MARC turbo network coding
Although the different coding operations are processed spa- outperforms the scheme of [23] because the information of
tially distributed, we will treat them as one network-channel both mobile stations has to be decoded. Therefore, the iterative
code with the system rate RS = 2 · K/(2 · N + NR ) = 0.5. exchange of soft information improves the performance. If
The joint network-channel code consists of three constituent we use joint network-channel coding for the two-way relay
encoders which are illustrated graphically in Fig. 5 (a). Chan- channel, the information of only one of the two stations
nel encoder 1 and 2 at the mobile stations form two of the three has to be decoded. The information of the other station is
constituent encoders. As they process the information bits in perfectly known. Therefore, it is sufficient to include the
its original order, they are depicted in horizontal direction. known information at the beginning of the decoding. It is
The third constituent encoder is the network encoder at the not necessary to include the known information in further
relay. As it processes the interleaved information bits, it is iterations.
depicted in vertical direction. The network encoder combines IV. R EFERENCE S YSTEMS
the information bits of MS 1 and MS 2. Therefore, the encoder We want to describe three reference systems for the pro-
at the mobile stations and at the relay form one spatially posed JNCC scheme. To ensure a fair comparison, in all
distributed code with increased cooperative diversity. reference systems 2 · N + NR = 6000 symbols are transmitted
C. Iterative Network and Channel Decoding totally. Therefore, all systems require the same resources to
transmit K = 1500 information bits both from mobile station
The joint network-channel code can be decoded iteratively 1 and 2.
at the base station. Fig. 4 (c) depicts the decoder. First,
the channel decoders calculates extrinsic information L− A. Distributed Turbo Code for the Relay Channel
e (u1 )
and L−e (u2 ). The channel decoders contain a soft-in/soft-out In the first reference system the uplink of the two mobile
(SISO) decoder for the channel encoders 1 and 2. They use the stations is done separately with the help of a relay. This
u1 p1 u1 p1
u1 p1
p4 MS 1 q1 MS 1
u2 p2
MS 1
R1 BS BS
BS
R u2 p2 R2 u2 p2
p4
MS 2 MS 2
MS 2 p5 q2
(a) (b) (c)

Fig. 5. (a): Illustration of the turbo network code used for the multiple-access relay channel. (b): Illustration of the distributed turbo code used for the relay
channel. (c): Illustration of the turbo code used for the point-to-point channel.

scenario represents a relay channel where a distributed turbo In all systems MS1 an MS2 have K = 1500 information bits
code [10], [11] can be applied. Fig. 5 (b) depicts the system which they have to transmit to BS. The comparison is fair
model and an illustration of the distributed coding scheme. because in all systems 2 · N + NR = 6000 code bits are used
Each mobile station uses a separate relay for the uplink. The in the complete system. Thus, the system code rate is always
same situation would occur if one relay is shared for the RS = 0.5.
uplink of the two mobile stations.
C. Separate Network-Channel Coding for the MARC
Fig. 5 (b) depicts also an illustration of the distributed turbo
code. The PCCC which is used for the uplink of MS 1 We use the same definition for separate network-channel
consists of two constituent encoders. The first constituent coding (SNCC) as in [16]. First, on the physical layer one
encoder is a convolutional code at MS 1. It is identical to the performs (local) channel coding for each transmission to ob-
one described in Section III-A. We puncture the parity bits tain a network of erasure-based links. Second, on the network
p1 again and transmit K = 1500 information bits u1 and layer one performs network coding for the resulting network.
500 parity bits. The second constituent is processed at the SNCC was applied in [2], [3], [5] for MARCs. We want to
first relay R 1. The information bits u1 are interleaved and show in the next section with simulation results that even if the
channel encoded. We transmit 1000 of the resulting parity full diversity gain can be obtained with SNCC, it is necessary
bits p4 . The puncturing is done analogous to the puncturing to use JNCC to exploit the additional redundancy which is
described in Section III-B. The uplink for the other mobile contained in the transmission of the relay. Let us first explain
station MS 2 is done in the same way. This time a second the system model for SNCC.
relay R 2 processes the parity bits p5 and selects 1000 to The channel codes at the mobile stations are PCCCs which
transmit them to the base station. As the two relays have to consist of the convolutional encoders and the interleavers
share one transmission and can only send 1000 bits, it is not described in Section III. Fig. 6 (a) depicts channel encoder
possible to achieve the full diversity gain. 1 at MS 1. The output of the channel code 1 contains 1506
systematic bits (including 6 tail bits) v1 , 1503 parity bits of the
first constituent encoder p1 and 1503 parity bits of the second
B. Turbo Code for the Point-to-Point Channel constituent encoder q1 . Like in the system described in Section
The second reference system does not contain a relay and III we want to send N = 2000 code bits x1 . We puncture the
the uplink for both mobile station is done with a point-to- parity bits p1 and q1 according to the following rule. We only
point channel. A PCCC is used at the mobile stations. The transmit every sixth parity bit of p1 and q1 except the parity
PCCC contains the same convolutional encoders and the same bits at position 295, 595, 895 and 1195. Channel encoder 2 at
interleaver as described in Section III. Fig. 5 (c) depicts an MS 2 works accordingly.
illustration of the PCCCs which are used. In this reference If the relay has decoded the data of both mobile stations
system, the transmission from MS 1 contains the information correctly, the relay network encodes the estimates û14 and
bits u1 , the parity bits of the first constituent encoder p1 and û24 . This is done in the network layer. The network encoder
the parity bits of the second constituent encoder q1 . As we is a modulo-2 addition. The network encoder output u4 is
want to use the same bandwidth and transmission energy like protected against the noisy channel by a local channel encoder
in the case with relay, the channel code at MS 1 and MS 2 which belongs to the physical layer. This channel encoder is
can have a lower rate R = 0.5 instead of R = 0.75 and thus, a punctured PCCC which is identical to the one used at the
we transmit N = 3000 code bits from each mobile station. mobile stations. The output of the punctured PCCC x4 with
Beside the K = 1500 information bits, we choose 750 bits block length NR = 2000 is sent to the base station. A block
from p1 and 750 bits from q1 in a regular way. The uplink diagram of the relay is depicted in Fig. 6 (b).
of MS 2 is done in the same way. The transmission contains Fig. 6 (c) depicts the separate network and channel decoder
K = 1500 information bits u2 , 750 of the parity bits of the at the base station. In the physical layer, each of the three
first constituent encoder p2 and 750 of the parity bits of the channel outputs is decoded with a turbo decoder with 15
second constituent encoder q2 . iterations. We include a value of zero for the punctured bits
Physical Layer Network
y13 ũ13 Layer
PCCC
Dec.
û14 û13
y43 PCCC ũ43 Network
Network Physical Layer
v1 Layer x4 Dec. Dec.
u1 PCCC x1 PCCC û23
Punct. Enc. Punct.
Enc. p1 u4 y23 PCCC ũ23
q1 û24 Dec.
(a) (b) (c)

Fig. 6. Separate Network-Channel Coding: (a): Channel encoder 1 at mobile station 1. (b): Network encoder at the relay. (c): Separate network and channel
decoding at the base station.

before the decoding starts. The three decoder make a hard 0


10 MARC − JNCC
decision and output their estimates ũ13 , ũ43 and ũ23 to the
network decoder. Each channel decoder delivers an additional MARC − SNCC
flag which indicates whether its estimate is error-free. This −1 RC − Distr. Turbo Co.
10
information can be obtained with a cyclic redundancy check PtoPC − Turbo Co.
which has to be included in the data at the channel encoders.
−2
By delivering the hard decisions and the flags, the channel 10

BER
decoders provide erasure channels for the network decoder. If
one of the two estimates ũ13 or ũ23 is not error-free and the −3
other two estimates are error-free, the network decoder makes 10
a modulo-2 addition of the estimates which are error-free to
obtain the estimate which is not error-free. For example, if −4
10
the estimates ũ13 , ũ43 are error-free and ũ23 is not error-
free, the network decoder outputs the estimates û13 = ũ13
and û23 = ũ13 ⊕ ũ43 . If this is not the case, the network 10
−5

decoder cannot exploit the diversity and outputs the estimates 0 5 10 15 20 25


Eb/N0 on MS−BS link (dB)
û13 = ũ13 and û23 = ũ23 .
V. S IMULATION R ESULTS Fig. 7. Simulation results for case 1: Bit error rate (BER) of system applying
joint network-channel coding (JNCC) and separate network-channel coding
We consider simulation results for the two cases which were (SNCC) for the multiple-access relay channel (MARC), a distributed turbo
explained in Section II and compare the proposed system using code for the relay channel (RC) or a turbo code for the point-to-point channel
(PtoPC).
joint network-channel coding (JNCC) with the three reference
systems. Let us first consider case 1 where the relay and the
mobile stations have the same distance to the base station. Fig. 0
10 MARC − JNCC
7 depicts the bit error rate (BER) and Fig. 8 the frame error rate
(FER) over the signal to noise ratio (SNR) Eb /N0 in dB on the MARC − SNCC
mobile station - base station link for case 1. From the slope of RC − Distr. Turbo Co.
the error rates curves we can recognize the diversity which is −1 PtoPC − Turbo Co.
10
provided by the use of the relay and the use of network coding.
The system using separate network-channel coding (SNCC)
FER

achieves the full diversity gain. However, as the redundancy


−2
which is contained in the transmission of the relay cannot be 10
efficiently exploited, SNCC has of a performance loss of 1.6
dB compared to the JNCC approach for a FER of 10−2 .
Fig. 9 and Fig. 10 depict the BER and FER for case 2 where
−3
the relay is closer to the base station than the mobile stations. 10
As the redundancy which is contained in the transmission of
the relay is received with a higher SNR, it is more important
to exploit it efficiently. Therefore, the performance loss from 0 5 10 15 20 25
E /N on MS−BS link (dB)
SNCC to JNCC is larger than for case 1. The difference at a b 0

FER of 10−2 is 2.0 dB. Fig. 8. Simulation results for case 1: Frame error rate (FER) of system
applying joint network-channel coding (JNCC) and separate network-channel
VI. C ONCLUSION coding (SNCC) for the multiple-access relay channel (MARC), a distributed
turbo code for the relay channel (RC) or a turbo code for the point-to-point
We proposed to use joint network-channel coding (JNCC) channel (PtoPC).
based on turbo codes for the multiple-access relay channel
0 that the frame error rate of JNCC can outperform the one of
10 MARC − JNCC SNCC by up to 2.0 dB.
MARC − SNCC
RC − Distr. Turbo Co. R EFERENCES
−1
10 PtoPC − Turbo Co. [1] R. Ahlswede, N. Cai, S-Y. R. Li, and R. W. Yeung. Network Information
Flow. IEEE Trans. on Information Theory, 46(4):1204–1216, July 2000.
[2] D. Tuninetti and C. Fragouli. Processing Along the Way: Forwarding
vs. Coding. In Proc. International Symposium on Information Theory
BER

−2
10 and its Applications (ISITA), Oct. 2004.
[3] X. Bao and J. Li. Matching code-on-graph with network-on-graph:
Adaptive network coding for wireless relay networks. In Proc. 43rd
Allerton Conf. on Communication, Control, and Computing, Sept. 2005.
−3 [4] C. Hausl, F. Schreckenbach, I. Oikonomidis, and G. Bauch. Iterative
10
network and channel decoding on a tanner graph. In Proc. 43rd Allerton
Conf. on Communication, Control, and Computing, Sept. 2005.
[5] Y. Chen, S. Kishore, and J. Li. Wireless diversity through network
−4 coding. In Proc. IEEE Wireless Communications and Networking
10 Conference (WCNC), May 2006.
0 5 10 15 20 25 [6] X. Bao and J. Li. On the outage properties of adaptive network
E /N on MS−BS link (dB) coded cooperation (ancc) in large wireless networks. In Proc. IEEE
b 0
International Conference on Accoustic, Speech, and Signal Processing
(ICASSP), March 2006.
Fig. 9. Simulation results for case 2: Bit error rate (BER) of system applying [7] L. Xiao, T. E. Fuja, J. Kliewer, and D. J. Costello, Jr. Nested Codes
joint network-channel coding (JNCC) and separate network-channel coding with Multiple Interpretations. In Proc. 40th Annual Conf. on Information
(SNCC) for the multiple-access relay channel (MARC), a distributed turbo Sciences and Systems (CISS), March 2006.
code for the relay channel (RC) or a turbo code for the point-to-point channel [8] A. Sendonaris, E. Erkip, and B. Aazhang. Increasing uplink capacity
(PtoPC). via user cooperation diversity. In Proc. IEEE International Symposium
on Information Theory (ISIT), August 1998.
[9] G. Kramer and A. J. van Wijngaarden. On the White Gaussian Multiple-
0 Access Relay Channel. In Proc. IEEE International Symposium on
10 MARC − JNCC Information Theory (ISIT), June 2000.
[10] B. Zhao and M. Valenti. Distributed turbo coded diversity for the relay
MARC − SNCC
channel. Electronic Letters, 39:786–787, 2003.
RC − Distr. Turbo Co. [11] M. Janani, A. Hedayat, T. E. Hunter, and A. Nosratinia. Coded
−1 PtoPC − Turbo Co. cooperation in wireless communications: Space-time transmission and
10 iterative decoding. IEEE Trans. on Signal Processing, 52(2):362–371,
Feb. 2004.
[12] Y. Cao and B. Vojcic. Cooperative coding using serial concatenated
FER

convolutional codes. In Proc. IEEE Wireless Communications and


−2 Networking Conference (WCNC), March 2005.
10 [13] A. Chakrabarti, A. de Baynast, A. Sabharwal, and B. Aazhang. Ldpc
code design for half-duplex decode-and-forward relaying. In Proc. 43rd
Allerton Conf. on Communication, Control, and Computing, Sept. 2005.
[14] J. Hagenauer. Source-controlled channel decoding. IEEE Trans. on
−3 Communications, 43(9):2449–2457, Sept. 1995.
10
[15] M. Effros, M. Medard, T. Ho, S. Ray, D. Karger, and R. Koetter. Linear
network codes: A unified framework for source, channel and network
coding. In Proc. DIMACS Works. on Network Information Theory, 2003.
0 5 10 15 20 25 [16] N. Ratnakar and G. Kramer. The multicast capacity of acyclic, deter-
Eb/N0 on MS−BS link (dB) ministic relay networks with no interference. Submitted for publication,
2005.
[17] C. Hausl and J. Hagenauer. Iterative Network and Channel Decoding for
Fig. 10. Simulation results for case 2: Frame error rate (FER) of system the Two-Way Relay Channel. In Proc. IEEE International Conference
applying joint network-channel coding (JNCC) and separate network-channel on Communications (ICC), June 2006.
coding (SNCC) for the multiple-access relay channel (MARC), a distributed [18] Y. Wu, P. A. Chou, and S.-Y. Kung. Information Exchange in Wireless
turbo code for the relay channel (RC) or a turbo code for the point-to-point Networks with Network Coding and Physical-Layer Broadcast. In Proc.
channel (PtoPC). 39th Annual Conf. on Information Sciences and Systems (CISS), March
2005.
[19] B. Rankov and A. Wittneben. Spectral efficient signaling for half-
duplex relay channels. In Asilomar Conference on Signals, Systems,
(MARC). Such as system could be used for the cooperative and Computers 2005, Nov. 2005.
[20] H. Holma and A. Toskala. WCDMA for UMTS. Wiley, Inc., 2001.
uplink for two mobile stations to a base station with the [21] European Telecommunications Standards Institute, Universal mobile
help of a relay. JNCC for the MARC generalizes distributed telecommunications system (UMTS): Multiplexing and channel coding
channel codes for the relay channel. We showed with simu- (FDD). 3GPP TS 125.212 version 3.4.0, 2000.
[22] P. Dupraz. Rate-compatible network code design for joint network
lation results that JNCC for the MARC increases cooperative and channel coding. Diplomarbeit, Munich University of Technology,
diversity compared to distributed channel codes for the relay November 2005.
channel. Although the diversity gain can be also achieved [23] C. Hausl. Improved rate-compatible joint network-channel code for the
two-way relay channel. In Proc. Joint Conference on Communications
with separate network-channel coding (SNCC), JNCC allows and Coding (JCCC), March 2006.
to more efficiently exploit the redundancy which is contained
in the transmission of the relay. Simulation results confirmed

You might also like