Physical Communication 6 (2013) 43–47

Contents lists available at SciVerse ScienceDirect

Physical Communication
journal homepage: www.elsevier.com/locate/phycom

Full length article

Joint physical network coding and LDPC decoding for two

way wireless relaying✩
Kui Xu ∗ , Zhenxing Lv, Youyun Xu, Dongmei Zhang, Xinyi Zhong, Wenwen Liang
Institute of Communications Engineering, PLAUST, Nanjing, 210007, China

article info abstract

Article history: In this paper, we investigate the joint design of channel and network coding in bi-
Received 16 April 2011 directional relaying systems and propose a combined low complexity physical network
Received in revised form 15 February 2012 coding and LDPC decoding scheme. For the same LDPC codes employed at both
Accepted 26 January 2013
source nodes, we show that the relay can decode the network coded codewords from
Available online 9 February 2013
the superimposed signal received from the BPSK-modulated multiple-access channel.
Simulation results show that this novel joint physical network coding and LDPC decoding
Keywords:
Joint network coding and LDPC decoding
method outperforms the existing MMSE network coding and LDPC decoding method over
Two-way wireless relaying channel the AWGN and complex MAC channel.
Wireless cooperative networks © 2013 Elsevier B.V. All rights reserved.

1. Introduction access (MAC) stage, where sources A and B transmit the

LDPC-coded signals to the relay R simultaneously, and a
The network coding scheme was originally considered broadcast (BC) stage, where the relay R broadcasts to both
as a technique of improving network throughput for wired sources A and B. One critical process at R is to decode the
networks [1]. In wireless networks, the broadcast nature superimposed signal from A and B at the MAC stage in
of the wireless channel is usually considered to cause such a way that A and B could decode the information from
enormous interference if more than two nodes transmit each other reliably at the BC stage. Instead of decoding the
simultaneously at the same frequency. On the other hand, individual information belonging to the sources A and B
physical network coding (PNC) [2,3] can employ this separately, the relay node R aims to decode the received
broadcast nature as a capacity-boosting approach for two- superimposed signal to the network-coded combination of
way or multi-way cooperative communication network. the two sources’ information. We refer to this decoding
A simple two-way wireless relaying system with two process as the joint physical network coding and LDPC
sources A and B and one relay R is depicted in Fig. 1. decoding (JNCLD).
Source A and source B desire to exchange information In [4,5] non-coherent physical layer network coding for
between each other and there is no direct link between FSK and CPFSK modulation is proposed. In [6], joint net-
the two source nodes. Thus, all the transmission between work and channel coding was proposed for the simple
sources A and B must flow through the relay R. The real additive multiple-access white Gaussian noise chan-
relay transmission consists of two states: a multiple nel. By noticing the linearity of both network and chan-
nel coding, the soft Log-likelihood Ratios (LLRs) for the
network-coded codeword can be directly estimated from
the received physically superimposed signals. In [7,8], a
✩ This work was supported by the National Natural Science Foundation
joint network and LDPC coding scheme for bi-directional
of China (No. 60972050), the Major Special Project of China under
Grant (2010ZX03003-003-01) and the Jiangsu Province National Science
relaying is presented and the closed-form expressions for
Foundation under Grant (BK2011002, BK2012055). computing the log-likelihood ratios of the network-coded
∗ Corresponding author. Tel.: +86 13611587232. codewords have been derived for both real and complex
E-mail address: [email protected] (K. Xu). multiple access channels. An adaptive PNC called pseudo
1874-4907/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
doi:10.1016/j.phycom.2013.01.002
44 K. Xu et al. / Physical Communication 6 (2013) 43–47

Fig. 1. Two-way relay channel.

exclusive-or (PXOR) for LDPC coded two-way relay block respectively. The encoded vectors are BPSK-modulated to
fading channels is proposed in [9]. Based on the pairwise xA ∈ {−1, 1}N and xB ∈ {−1, 1}N according to the mapping
check decoding (PCD) [10], the check relationship table rule 1 → 1 and 0 → −1.
generated by PXOR mapping obtains the same Hamming
distances as that of conventional XOR mapping. In order to 2.1. Multiple access stage
compensate the amplitude fading and phase deviation of
the TWR block fading channels, the PXOR mapping opti- During the MAC stage, the two sources transmit the
mizes the Euclidean distances by adjusting the symbol dis- modulated signals xA and xB to the relay R simultaneously.
tances dynamically. Under a multiple access white Gaussian noise channel and
An optimal time and rate allocation scheme for the assumption of perfect synchronization, the received
a network-coded bi-directional wireless communication signal at the relay R is
system is proposed in [11]. The closed form expressions yR = xA + xB + nR (1)
for the optimal allocation of the transmission time and
of the data rate in both ways for given channel SNRs to where nR are identically distributed (i.i.d.) zero-mean
maximize the sum-rate are derived. In [12,13], a joint Gaussian random variables with variance σn2 . According
PNC and minimum mean square error (MMSE) based LDPC to the physical network coding scheme introduced in
decoding method is proposed. In this paper, we pro- [6–8,12–14], the XOR of the source information denoted
vide some novel insights into the decode-and-forward by bA⊕B = bA ⊕ bB ∈ {0, 1}K can be estimated at the relay
approach for two-way wireless relaying with BPSK signal- from the received signal yR , i.e., bR = b̂A⊕B ∈ {0, 1}K . Then,
ing. In particular, for the same LDPC codes employed at bR is encoded by the same LDPC code, and the code vector
both source nodes, a novel iterative LDPC decoding algo- cR BPSK-modulated to xR .
rithm is proposed for the physical network coding scheme
at the relay, which outperforms a recently proposed MMSE 2.2. Broadcasting stage
based network coding scheme on bit error rate (BER)
performance. In the BC stage, the relay R broadcasts xR to both A and
The remainder of this paper is organized as follows. B. Thus, the received signals at A and B are given by
In Section 2 we present the considered system model. yA = xR + nA
Section 3 presents the joint PNC and MMSE based LDPC (2)
decoding scheme [12,13] and introduces the proposed yB = xR + nB .
novel joint physical network coding and LDPC decoding At both A and B, the information b̂R , which contains
approach. Finally, in Sections 4 and 5 simulation results are the information of bA⊕B , is estimated from yA and
presented and the conclusion is given. yB , respectively. Since A and B know what has been
transmitted at the MAC stage, A and B can obtain the
2. System model information from each other simply by means of the binary
XOR, i.e., b̂B = b̂R ⊕ b̂A and b̂A = b̂R ⊕ b̂B .
We denote bA ∈ {0, 1}K and bB ∈ {0, 1}K as the in- A critical process at the relay R is to decode the
formation vector of the two source nodes A and B, re- superimposed signal from A and B in such a way that A and
spectively. The information is encoded by the same LDPC B could decode the information from each other reliably
code with a code rate of K /N into the codeword vectors at the BC stage. In this paper, we will focus on deriving a
cA ∈ {0, 1}N and cB ∈ {0, 1}N at the sources A and B, decoding algorithm for yR → bR .
 K. Xu et al. / Physical Communication 6 (2013) 43–47 45

3. Joint physical network coding and LDPC decoding where

 
In [12,13], a joint PNC and MMSE based LDPC decoding 1 (yR − hA ρA + hB ρB )2
P (yR |E1 ) =  exp − (7)
method is proposed. The idea of this algorithm is that un-
4 2π σR2 2σR2
der the assumption of the same LDPC code applied at both
source nodes, i.e., parity check matrix HA = HB = H and  
1 (yR + hA ρA − hB ρB )2
HA cA = 0, P (yR |E2 ) =  exp − (8)
4 2π σR2 2σR2
HB cB = 0, (3)
H (cA ⊕ cB ) = 0  
1 (yR − hA ρA − hB ρB )2
the XOR of the encoded symbols cA⊕B ∈ {0, 1}N is also a P (yR |E3 ) =  exp − (9)
4 2π σR2 2σR2
valid codeword of the LDPC code. The relay R first maps
each pair of the received superimposed signal yR to an esti-  
mation of the joint network and LDPC coded symbol ŷMMSE
R 1 (yR + hA ρA + hB ρB )2
corresponding to cA ⊕ cB by using minimum mean square P (yR |E4 ) =  exp − . (10)
4 2π σR2 2σR2
error (MMSE) estimation; then performs LDPC decoding on
the interim symbol cA ⊕ cB to obtain the network coded
symbol bA ⊕ bB . The LLR of the network coded bit cA⊕B over the AWGN
Different from the above mentioned approach, we will MAC channel with power allocation can be expressed as
compute the LLR Λ(cA⊕B ) of the XOR of the two source 2hA hB ρA ρB
information cA ⊕ cB direct from the received signal yR and ΛAM (cA⊕B ) =
then perform LDPC decoding to obtain the network coded σR2
symbol bA ⊕ bB .   
yR (hA ρA − hB ρB )
+ log cosh
σR2
3.1. AWGN MAC channel with power allocation
  
The received superimposed signal yR over the AWGN yR (hA ρA + hB ρB )
− log cosh . (11)
MAC channel with power allocation can be expressed as σR2
yR = hA ρA xA + hB ρB xB + nR (4) The network coded information bit bA⊕B can be ob-
where hA and hB represent the channel gain from nodes tained by the traditional belief broadcasting (BP) decoding
A and B to the relay node R, respectively. PA = ρA2 and algorithm with LLR ΛAM (cA⊕B ).
PB = ρB2 denote the power allocated to the MAC channel,
respectively. Let n index the bit of a codeword. The a priori 3.2. Complex MAC channel with power allocation
probabilities of {cA⊕B (n) = cA (n) ⊕ cB (n) = 0, 1} are

Pr{cA⊕B (n) = 0} = 1/2 The received superimposed signal ỹR over the complex
(5) MAC channel with power allocation can be expressed as
Pr{cA⊕B (n) = 1} = 1/2.

If cA⊕B (n) = 1 holds, the event E1 = {cA (n) = 0, cB (n) ỹR = hA ejθA ρA xA + hB ejθB ρB xB + nR (12)
= 1} or the event E2 = {cA (n) = 1, cB (n) = 0} should be
where nR denotes the zero-mean complex Gaussian noise
satisfied. On the other hand, if cA⊕B (n) = 0 holds, the event
E3 = {cA (n) = 1, cB (n) = 1} or the event E4 = {cA (n) = for the complex MAC channel with covariance σR2 , hA and
0, cB (n) = 0} should be satisfied. Hence, the probability of hB represent the channel gain from nodes A and B to the re-
{cA⊕B (n) = 1} is the sum probability of Pr{E1 } and Pr{E2 }, lay node R, respectively. PA = ρA2 and PB = ρB2 denote the
and the probability of {cA⊕B (n) = 0} is the sum probability power allocated to the complex MAC channel, respectively.
of Pr{E3 } and Pr{E4 }. θA and θB represent the uniformly distributed phase shift
Assume that the relay has knowledge of the parameter over [0, 2π ). The LLR of each network coded bit cA⊕B with
σR2 and channel state information. Soft-decision decoding the assumption that the relay has knowledge of the param-
requires that the relay node R compute the LLR of each eter σR2 and channel state information can be expressed as
network coded bit cA⊕B according to
P (cA⊕B = 1|ỹR )
P (cA⊕B = 1|yR )
ΛCM (cA⊕B ) = log
P (cA⊕B = 0|ỹR )
ΛAM (cA⊕B ) = log
P (cA⊕B = 0|yR )
P (cA ⊕ cB = 1|ỹR )
P (cA ⊕ cB = 1|yR ) = log
= log P (cA ⊕ cB = 0|ỹR )
P (cA ⊕ cB = 0|yR )
= log[P (yR |E1 ) + P (yR |E2 )] = log[P (ỹR |E1 ) + P (ỹR |E2 )]
− log[P (yR |E3 ) + P (yR |E4 )] (6) − log[P (ỹR |E3 ) + P (ỹR |E4 )] (13)
46 K. Xu et al. / Physical Communication 6 (2013) 43–47

where

P (ỹR |E1 )
 
1 ∥ỹR − hA ejθA ρA + hB ejθB ρB ∥2
=  exp − (14)
4 2π σR2 2σR2

P (ỹR |E2 )
 
1 ∥ỹR + hA ejθA ρA − hB ejθB ρB ∥2
=  exp − (15)
4 2π σR2 2σR2

P (ỹR |E3 )
 
1 ∥ỹR − hA ejθA ρA − hB ejθB ρB ∥2
=  exp − (16)
4 2π σR2 2σR2

P (ỹR |E4 )
  Fig. 2. BER performance of the proposed method and MMSE method for
1 ∥ỹR + hA ejθA ρA + hB ejθB ρB ∥2 various packet lengths and power allocation over AWGN MAC channel.
=  exp − . (17)
4 2π σR2 2σR2

The LLR of the network coded bit cA⊕B over the complex
MAC channel with power allocation can be expressed as
ΛAM (cA⊕B )
∥hA ejθA ρA + hB ejθB ρB ∥2 − ∥hA ejθA ρA − hB ejθB ρB ∥2
=
σR2
ℜ ỹR (hA ejθA ρA − hB ejθB ρB )∗
    
+ log cosh
σR2

ℜ ỹR (hA ejθA ρA + hB ejθB ρB )∗

    
− log cosh . (18)
σR2
The network coded information bit bA⊕B can be
obtained by the traditional BP decoding algorithm with LLR
ΛCM (cA⊕B ).
Fig. 3. BER performance of the proposed method and MMSE method for
various packet lengths with power allocation and uniformly distributed
4. Simulation results phase shift caused by complex MAC channel.

In this section, we demonstrate the simulated perfor- outperforms MMSE about 0.2 dB when the BER is 10−4
mance of the proposed joint physical network coding and for the packet length (PL) of 1010 over the AWGN MAC
LDPC decoding scheme for two-way wireless relay sys- channel. The BER performance of MMSE estimation and
tems with power allocation in each source node. In the the proposed method over the complex MAC channel is
simulation, we assume that the SNR at the relay node R given in Fig. 3, the power ratio of the two source nodes
is defined as (PA h2A + PB h2B )/σR2 for both the AWGN and PA h2A /PB h2B is also set to be 2/3. We can see from Fig. 3
complex MAC channel, where σR2 denotes the noise vari- that the proposed method outperforms the MMSE based
ance received at the relay R and the total transmitting method about 0.1 dB when the BER is 10−4 for the PL of
power PA h2A + PB h2B is set to 2. We check the BER (bit er- 1010.
ror rate) of the decoded packet bA⊕B = bA ⊕ bB at the re-
lay node R. A BPSK modulation scheme is used at both end 5. Conclusion
nodes for all simulations.
For comparison, we also study the performance of In this paper, a novel joint physical network coding and
the MMSE based method proposed in [12,13]. We first LDPC decoding method for a two-way wireless relaying
compare the BER performance of MMSE estimation and system with BPSK signaling is presented. The proposed
the proposed method over the AWGN MAC channel with method employ the same LDPC codes at the source nodes
power allocation for various packet lengths. In Fig. 2, the and the relay decodes the network coded packet from
ratio of PA h2A /PB h2B is set to be 2/3 and the iteration numbers the received superimposed signal by using the proposed
of the two schemes are set to 30. The proposed method method. Simulation results show that the proposed novel
 K. Xu et al. / Physical Communication 6 (2013) 43–47 47

Zhenxing Lv was born in 1986. He received his

method outperforms the MMSE method about 0.2 dB
B.E. degree from Nanjing Institute of Commu-
and 0.1 dB over the AWGN and complex MAC channel, nication Engineering in 2007. He is currently
respectively. a Master’s candidate. His research interests in-
clude channel coding and network coding in
wireless communication.
References

[1] R. Ahlswede, N. Cai, S.-Y.R. Li, R.W. Yeung, Network information flow,
IEEE Trans. Inform. Theory 46 (4) (2000) 1204–1216.
[2] C. Hausl, J. Hagenauer, Iterative network and channel decoding for
the two-way relay channel, in: Proc. IEEE International Conference
on Communications, ICC, Istanbul, Turkey, June 2006. Youyun Xu was born in 1966. He graduated
[3] P. Popovski, H. Yomo, Physical network coding in two-way from Shanghai Jiao Tong University with a Ph.D.
wireless relay channels, in: Proc. IEEE International Conference on degree in information and communication engi-
Communications, ICC, Glasgow, Scotland, June 2007. neering in 1999. He is currently a professor with
[4] J.H. Sørensen, R. Krigslund, P. Popovski, et al., Physical layer network the Nanjing Institute of Communication Engi-
coding for FSK systems, IEEE Commun. Lett. 13 (8) (2009) 597–599. neering, China. He is also a part-time professor
[5] M.C. Valenti, D. Torrieri, T. Ferrett, Noncoherent physical-layer with the Institute of Wireless Communication
network coding, in: Proc. IEEE Military Communications Conference, Technology of Shanghai Jiao Tong University
2009. (SJTU), China. He has more than 20 years pro-
[6] A. Zhan, C. He, Joint design of channel coding and physical network fessional experience of teaching and researching
coding for wireless network, in: Proc. IEEE International Conference in communication theory and engineering. Now,
on Neural Networks and Signal Processing, ICNNSP, Zhe Jiang, China, his research interests are focusing on new generation wireless mobile
June 2008. communication systems (IMT-advanced and related), advanced channel
[7] Xiaofu Wu, Weijun Zeng, Chunming Zhao, Xiaohu You, Joint coding and modulation techniques, multi-user information theory and ra-
network and LDPC coding for bi-directional relaying, in: Proc. IEEE dio resource management, wireless sensor networks, cognitive radio net-
International Conference on Information Theory and Information works, etc. He is a senior member of IEEE, and a senior member of the
Security, ICITIS, Hang Zhou, China, November 2011. Chinese Institute of Electronics.
[8] Kui Xu, Youyun Xu, Wenwen Liang, Dongmei Zhang, Zhenxing Lv,
Joint LDPC and physical network coding with power allocation for
two way wireless relaying, in: Proc. IEEE International Conference Dongmei Zhang was born in 1972. She received
on Wireless Communications and Signal Processing, WCSP, Nanjing, her B.E. and M.E. degrees in communication
China, November 2011. engineering in 1993 and 2005, respectively.
[9] J. Liu, M. Tao, Y. Xu, Pseudo exclusive-OR for LDPC coded two-way She is an associate professor with the Nanjing
relay block fading channels, in: Proc. IEEE International Conference Institute of Communication Engineering, China.
on Communications, ICC, Kyoto, Japan, June 2011. Her research interests include new generation
[10] J. Liu, M. Tao, Y. Xu, X. Wang, Pairwise check decoding for LDPC wireless mobile communication system, radio
coded two-way relay fading channels, in: Proc. IEEE International resource management, and network coding in
Conference on Communications, ICC, Cape Town, South Africa, May wireless communication.
2010.
[11] C. Hausl, O. Iscan, F. Rossetto, Optimal time and rate allocation for a
network-coded bidirectional two-hop communication, in: Proc. IEEE
European Wireless Conference, EW, Lucca, Italy, April 2010. Xinyi Zhong was born in 1976. He received
[12] S. Zhang, S. Liew, Channel coding and decoding in a relay system his B.E. and M.E. degrees in communication
operated with physical-layer networking coding, IEEE J. Sel. Areas engineering in 1999 and 2005, respectively, from
Commun. 27 (5) (2009) 788–796. the Institute of Communication Engineering, PLA
[13] S. Zhang, S. Liew, L. Lu, Physical layer networking coding schemes University of Science and Technology, China. He
over finite and infinite fields, in: Proc. IEEE GLOBECOM, 2008. is currently a Doctoral candidate. His research
[14] S. Zhang, S. Liew, P. Lam, Physical layer network coding, in: Proc. interests include channel coding and network
International Conference on Mobile Computing and Networking, coding in wireless communication.
MobiCom, Los Angeles, USA, 2006, pp. 358–365.

Kui Xu was born in China. He received the B.S.

degree, M.S., degree and Ph.D. degree from PLA Wenwen Liang was born in 1986. He received
University of Science and Technology, Nanjing, his B.E. degree from Xidian University in 2009.
China, in 2004, 2006 and 2009. He is currently He is currently a Master’s candidate. His re-
a lecturer in the Wireless Communications De- search interests include channel coding and net-
partment, Institution of Communications Engi- work coding in wireless communication.
neering, PLAUST. His research interests include
multicarrier modulation, synchronization, sig-
nal processing in communications, network cod-
ing and blind source separation. He received
the 2010 ten excellent doctor degree dissertation
award of PLAUST.