A Low-complexity near-ML performance achieving algorithm for large MIMO detection

Abstract

Abstract-In this paper, we present a low-complexity, near maximum-likelihood (ML) performance achieving detector for large MIMO systems having tens of transmit and receive antennas. Such large MIMO systems are of interest because of the high spectral efficiencies possible in such systems. The proposed detection algorithm, termed as multistage likelihood-ascent search (M-LAS) algorithm, is rooted in Hopfield neural networks, and is shown to possess excellent performance as well as complexity attributes. In terms of performance, in a 64 × 64 V-BLAST system with 4-QAM, the proposed algorithm achieves an uncoded BER of 10 −3 at an SNR of just about 1 dB away from AWGN-only SISO performance given by Q( √ SNR). In terms of coded BER, with a rate-3/4 turbo code at a spectral efficiency of 96 bps/Hz the algorithm performs close to within about 4.5 dB from theoretical capacity, which is remarkable in terms of both high spectral efficiency as well as nearness to theoretical capacity. Our simulation results show that the above performance is achieved with a complexity of just O(NtNr) per symbol, where Nt and Nr denote the number of transmit and receive antennas. I. INTRODUCTIONMIMO techniques have become popular in realizing spatial diversity and high data rates through the use of multiple transmit antennas [1]. We consider large MIMO systems with tens of transmit and receive antennas, which are of interest due to the high spectral efficiencies possible in such systems. The key issues in realizing large MIMO systems include lowcomplexity detection, channel estimation, and communication terminal size to accommodate large number of antennas. We address the issue of low-complexity detection in large MIMO systems here. More recent approaches to lowcomplexity multiuser detection and MIMO detection involve application of techniques from belief propagation [2], neural networks [3],[4], Markov chain Monte-Carlo methods [5], probabilistic data association [6], etc. Detectors based on these techniques have been shown to achieve an average perbit complexity linear in number of users, while achieving near-optimal performance in large multiuser CDMA system settings. These powerful techniques are increasingly being adopted in MIMO detection. In [4], we presented a Hopfield neural network based likelihood ascent search (LAS) algorithm for large MIMO detection; we showed that the LAS detector achieves near-AWGN SISO performance in a large MIMO setting with hundreds of antennas, while performing close to within 4.6 dB from theoretical capacity.

Low‐complexity soft‐output MIMO uplink detection for large systems iterative detection and decoding

“…As in the works of Wang and Giannakis, 21,22 we used a method for generating soft information by means of hard estimates. It is based on the use of a low-complexity hard-output inner massive MIMO detector [23][24][25][26][27] followed by a series of parallel interference cancelations (PICs) and filtering to produce inputs to the MAP detector, which are independent of the number of transmitting antennas. Besides, differently from the mentioned previous work, we do not consider the matrix channel coefficients to be known by the system.…”

Achieving Near Capacity of Non-Binary LDPC Coded Large MIMO Systems with a Novel Ultra Low-Complexity Soft-Output Detector

“…To the best of our knowledge, only large MIMO systems concatenated with turbo codes have been studied in [9], [10], [11], [12] which are the limited number of publications. Noting that the main objective of the above cited works is only on the development of the low-complexity large MIMO detections.…”

Detecção De Sinais No Enlace Reverso De Sistemas Mimo De Larga Escala Com Múltiplos Usuários E Células

“…Em [21], o detector Multistage LAS (MLAS) foi proposto. Este detector consiste essencialmente em uma sequência de buscas LAS em série, cada qual empregando conjuntos de candidatos de busca com cardinalidade crescente, sendo a primeira busca LAS (primeiro estágio) com SCS unitários, i.e., a diferença entre o vetor de símbolos candidato em um passo durante a busca e o vetor do próximo passo é de apenas 1 símbolo (|L(n)| = 1).…”

“…Tendo em vista a elevada complexidade computacional do detector ML apresentado na subseção anterior, procurou-se empregar a estratégia LAS para a proposição de estimadores/detectores conjuntos que se aproximem do resultado do detector descrito em (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24). A estratégia LAS foi escolhida uma vez que ela busca se aproximar do resultado da detecção ML com complexidade computacional reduzida.…”

“…Este detector, denominado de detector de busca por verossimilhança ascendente (Likelihood Ascent Search -LAS), efetua uma busca em um subespaço do espaço vetorial de vetores de símbolos transmitidos de forma que, a cada passo da busca, seja escolhido um vetor de símbolos candidato com maior valor de verossimilhança que o vetor do passo anterior até que se encontre um ponto fixo neste espaço vetorial. Posteriormente, o algoritmo LAS foi estendido em [21], dando origem ao chamado LAS multi-estágio (Multi-stage LAS -M-LAS) que apresentou melhoria na BER com aumento controlado na sua complexidade computacional. Outros algoritmos baseados no LAS foram propostos em [22], como o LAS com múltiplos vetores de entrada (Multiple Input Vector LAS -MIV-LAS ) e o LAS com múltiplos conjuntos de busca de candidatos (Multiple Search Candidate Set LAS -MSCS-LAS).…”