Lattice reduction aided transceiver designs in mobile MIMO communications : toward the deterministic complexity

다국어 초록 (Multilingual Abstract)

Recently, a lattice reduction-aided (LRA) multiple-input multiple-output (MIMO) detection scheme has been proposed in junction with linear (as well as non-linear) detectors. It is well known that these detection schemes provide a full diversity. Also, its complexity is comparable to that of linear detectors, when the channel remains constant for a several symbol time, which referred as block fading channel. Since lattice reduction (LR) should be performed at the beginning of the frame-block, the overall complexity of LR procedure can be negligible for large block length on slowly varying channels. For the fast varying channels, however, the decoding complexity of LRA detection scheme is unreasonably high. Therefore it is desirable that complexity should be further reduced to make LRA detector more attractive for MIMO detection implementation in practical wireless environment. The focus of the LR implementation lies not only on the complexity but also on the running-time of algorithm. For the real-time communication systems, the detection procedure, which includes the LR, should be ended within a certain period of the processing time. Unfortunately, the running-time and complexity of common LR technique, which called Lenstra, Lenstra, Lov´asz (LLL) algorithm are unknown. In this dissertation, low and deterministic complexity lattice reduction aided mobile MIMO systems are investigated. We first introduce the low complexity scheme, named pre-multiplication scheme in temporally and spectrally correlated MIMO channel. This scheme can give the better starting point of the LLL algorithm so that it can offer the same performance as conventional LLL with significantly reduced complexity. We also propose a lattice reduction algorithm with guaranteed running-time and deterministic complexity in conjunction with pre-multiplication scheme. Our conjecture is that there may exist an algorithm that does not require the any iteration, when the better starting point is applied. The proposed algorithm can obtain the roughly LLL-reduced matrix with one-shot size reduction so that it takes the deterministic complexity. Moreover, we investigate the simple but effective lattice reduction scheme for stacked space time block coding (STBC) system, as a special case. By taking advantage of the inherent STBC structure of the transmitted symbols, the proposed scheme provides the same performance as a brute-force lattice reduction while significantly saving the computational complexity. The proposed scheme iteratively performs the deterministic structure, until certain conditions are satisfied. Therefore, it can guarantee the deterministic complexity and running-time of detection algorithm by limiting the iteration. In conjunction with pre-multiplication scheme, the proposed scheme with limiting the iteration to one can provide the almost same BER performance even in the fast fading channel. In addition, we show how proposed schemes can also be applied in the (lattice reduction aided) precoding systems at the transmitter and give simulation results that underline the usefulness of this approach.