A Fast Complex Lattice Reduction Algorithm for
SIC-based MIMO detection
Zhiyong Chen, Xuchu Dai
Key Laboratory of Wireless-Optical Communications,
Chinese Academy of Sciences, School of Information Science and Technology,
University of Science and Technology of China,
No. 96 Jinzhai Road, Hefei, Anhui Province, 230026, P. R. China.
Email: zhiyong@mail.ustc.edu.cn, daixc@ustc.edu.cn.
Abstract—Recently, lattice-reduction-aided detection in
multiple-input multiple-output (MIMO) systems has attracted
significant research efforts for its capability of achieving full
diversity performance with low complexity. However, most
lattice reduction algorithms are not designed directly to enhance
the bit error ratio (BER) performance. In this paper, a fast
lattice reduction (FLR) algorithm for complex-valued matrices,
which aims at maximizing the minimal signal to noise ratios
(SNR) of all layers, is proposed for V-BLAST (Vertical Bell
Laboratories Layered Space-Time) systems, employing pre-
sorting technique and complex Givens rotation to reduce its
computational complexity. The SNRs of all layers are related
to the diagonal elements of the triangular matrix via QR
decomposition of the channel matrix, and can be optimized
by a series of iterations which can be efficiently implemented
by exploiting the complex Givens rotation, while the average
number of iterations is significantly diminished by utilizing the
low complexity pre-sorting technique. Our analysis reveals that
the proposed SIC-FLR decoder can significantly reduce the
computational complexity without sacrificing any performance.
Simulation results show that the proposed algorithm achieves
the same performance as the state-of-art complex LLL (Lenstra-
Lenstra-Lov´asz) algorithm only with a fraction of complexity.
Index Terms—MIMO, complex lattice reduction, V-BLAST,
low complexity, BER, LLL.
I. INTRODUCTION
Communication systems using multiple antenna transceivers
have been extensively investigated due to the potential for
supporting much greater data rate and higher reliability than
the single-input single-output counterpart [1]. It is well known
that the maximal likelihood (ML) receiver provides optimal
performance if all codewords are transmitted with equal prob-
abilities, but it suffers from exponential complexity in terms of
the number of transmit antennas as well as the constellation
size [2]. On the other hand, for the sake of low-complexity
detection, linear receivers such as zero-forcing (ZF) and
minimum-mean-square-error (MMSE) receivers suffer from
inferior performance in comparison with the ML receiver
[3]. Hence, the design of low-complexity high-performance
receiver becomes a challenging and potential topic in both
academia and industry.
In the famous V-BLAST architecture [4], exploiting suc-
cessive interference cancellation (SIC) is a popular method
to achieve better performance than ZF and MMSE linear de-
coders. However, the computational complexity of calculating
optimal detection order of layers for SIC is high and the
performance is still unsatisfactory since the ordering procedure
does not improve the diversity gain [3]. Recently, lattice
reduction techniques [5] have been extensively adopted in
MIMO detection since that they can collect the same diversity
as ML detectors with low complexity [6]. Among all the
lattice reduction algorithms, LLL algorithm has considered
to be the most practical one because of its bounded average
polynomial complexity O(MN
3
log N), where N, M are the
antenna number of transmitter and receiver [8], [10]. The
standard real-valued LLL algorithm is applied by using an
equivalent real-valued substitute MIMO channel model. In
[11], C Ling puts forward the effective LLL algorithm for SIC
decoding with bounded complexity an order lower than that
of the standard LLL reduction (i.e, O((M + N)N
2
log N))
by realizing that full size reduction is unnecessary for SIC
decoding. The authors in [12] suggest that a sorted QR decom-
position can dramatically decrease the computational effort
of LLL lattice reduction algorithm, regardless of the actually
high computational effort of optimal sorted QR decomposition.
However, these algorithms are introduced for reducing real-
valued lattice bases, while the channel matrices are inherently
complex-valued. To tackle this issue, Y.H. Gan proposes the
complex-valued LLL (CLLL) algorithm in [9], reducing the
computational complexity by nearly 50% without sacrificing
any performance.
To further alleviate the complexity of lattice reduction
algorithm and to associate the BER performance with lattice
reduction algorithms, in this paper, from a max-min SNR
optimization point of view, we propose a fast lattice reduction
algorithm for SIC decoding. Compared with the existing LLL
methods, the proposed FLR algorithm focuses on maximizing
the SNRs of all layers which are related to the QR decomposi-
tion. First, a low-complexity pre-sorting QR decomposition is
adopted to obtain a relatively better starting phase. Then, the
SNR-equilibrizing algorithm is proposed for complex-valued
channel matrices. Finally, theoretical analysis and simulations
are conducted to demonstrate that the proposed algorithm
achieves the same performance as the state-of-art complex
IEEE ICC 2015 - Wireless Communications Symposium
978-1-4673-6432-4/15/$31.00 ©2015 IEEE 2283