JOINT DESIGN FOR SOURCE AND RELAY PRECODING IN AF-BASED MIMO TWO-WAY
RELAY NETWORKS
Bing Fang, Zuping Qian, Wei Zhong, Wei Shao
College of Communications Engineering
PLA University of Science and Technology, Nanjing, 210007, China
Email: bingfang ch@163.com, qzp811@sina.com, weizhong@ieee.org, weishao ch@163.com
ABSTRACT
Joint precoding for multiple-input multiple-output (MIMO)
two-way relay networks has recently received much attention.
In this paper, we present a coherent way to iteratively opti-
mize the source and relay precoding matrices based on convex
programming. Although the proposed total mean-square error
(MSE) minimization problem is not joint convex for both the
sources and relay, the subproblems for separate optimization
can be both formulated as a standard convex programming
problem. Numerical results further show that the proposed
iterative precoding algorithm converges very fast.
Index Terms— MIMO two-way relaying, convex opti-
mization, minimum mean-square error (MMSE).
1. INTRODUCTION
To overcome spectrum efficiency loss caused by one-way re-
laying, two-way relaying has recently been proposed. Com-
pared to one-way relaying, two-way relaying needs only two
time slots to complete one round of information exchange,
which would otherwise take four time slots. Central idea
of two-way relaying is that two sources can simultaneously
transmit their information to the relay in one phase, and the
relay forwards the combined information to both sources in
another phase. Since each source perfectly knows its own in-
formation, the self-interference caused by itself can be easily
canceled, provided that the required channel state information
(CSI) is available.
Existing two-way relaying protocols include amplify-
and-forward (AF) [1]- [5], decode-and-forward (DF) [6] [7],
compress-and-forward (CF) [8] and physical-layer network
coding (PNC) [9]. Among them, AF is more attractive due to
low-complexity, short processing delay, and low implemen-
tation cost. Let N
s
and N
r
denote the number of antennas
used by the sources and relay. When N
s
≥ N
r
, a gener-
alized singular-value-decomposition (GSVD) scheme [10]
can achieve the asymptotic capacity with high signal-to-noise
This work was supported by the Natural Science Foundation of China
under Grant No. 61201218 and No. 61201241.
ratio (SNR). However, the most practical case N
s
< N
r
(e.g., it is more connivent to implement more antennas at
the relay station than at the user terminals) is still not well
established [11].
To fully realize the benefits of MIMO and two-way relay-
ing protocol, efficient precoding is crucial. The total mean-
square-error (total-MSE) minimization criteria is directly re-
lated to user experience, and thus is very important and adopt-
ed here as the precoding design metric as [2]- [4]. Due to
the nonconvexity of the formulated total-MSE minimization
problem, an iterative precoding algorithm is proposed to al-
ternately optimize the source and relay precoding matrices.
Simulation results further show that the proposed precoding
algorithm converges very fast.
2. SYSTEM MODEL AND PROBLEM
FORMULATION
In this section, we first provide the system model, then for-
mulate the total-MSE minimization problem.
2.1. System Model
Consider a two-way relay network as shown in Fig. 1, where
two sources, denoted as S
1
and S
2
, exchange information by
the aid of a relay, denoted as R. In the system, the two sources
are both equipped with N
s
antennas, and the relay is equipped
with N
r
antennas. Imposed by the two-way relaying protocol,
one round of information exchange between S
1
and S
2
takes
in two time slots.
In the first time slot (also referred to as the MAC phase),
S
1
and S
2
simultaneously transmit signal to R . The vector-
valued signals transmitted by the sources can be denoted as
x
i
= A
i
s
i
, i = 1, 2, (1)
where s
i
is the zero-mean information signal vector with a
covariance matrix E{s
i
s
H
i
} = I
N
s
, A
i
∈ C
N
s
×N
s
represents
the precoding matrix of S
i
. We assume that each source has
an average transmit power constraint and is denoted as
Tr(A
i
A
H
i
) ≤ P
i
, i = 1, 2. (2)