Precoding Techniques For MIMO Wireless Communication Systems

Recently, multiple-input multiple-output (MIMO) technique has achieved continu-ous research by the academic and industry. Related work has been partially appliedto the3rd generation (3G) wireless communication systems. For the long term evolu-tion of3G, MIMO technique is necessary for further applications. If the transmitter canutilize the channel state information (CSIT) for precoding design, higher capacity andbetter performance can be achieved by MIMO systems.For precoding, the first goal is to solve the antenna interference problem introducedby MIMO systems. Precoding can obtain the interference-free spatial parallel subchan-nels via the mathematical processing at the transmitter and/or receiver to support thetransmission of multiple streams. The second goal of precoding is to solve the signal-to-noise ratio (SNR) or signal-to-interference-plus-noise ratio (SINR) imbalance problemamong the different streams. If each stream uses same channel coding and modulationconstellation configuration, power allocation is often used to compensate the SNR/SINRimbalance. However, the performance improvement by conventional schemes is not sat-isfactorily significant.In this dissertation, we consider the precoding design by the new emerging mathe-matical methods and ideas to solve the above two problems. Many precoding schemeswith different structures are proposed for various scenarios. These schemes often out-perform the conventional ones with full diversity gain. Our main contents include:1. Precoding design based on lattice reduction method. Based on the latticetheory, the lattice reduction (LR) method can significantly reduce the interferencebetween the streams by improving the orthogonality among the basis vectors anddecrease the deterioration of noise amplification. Thus better performance can be achieved. At the same time, the famous LLL algorithm shows polynomial complex-ity. We deeply analyze the LR method and discuss its application to MIMO systems.Then we propose the LR-aided (LRA) MMSE precoding design, which outperformsthe LRA-ZF precoding and conventional precoding schemes. Moreover, we discussthe LRA precoding design with joint transceiver design and propose the so-calledunitary precoding and integer precoding schemes, which are applicable to LTEfrequency division duplex (FDD) systems.2. Precoding design based on vector perturbation technique. The conventionalnonlinear precoding uses simple one-dimension modulo operation to reduce thepower amplification, which equivalently add a special perturbation vector to thesignal vectors. Based on vector perturbation technique, we can further optimizethe perturbation vectors. Such scheme is also called vector precoding (VP). Wediscuss the existing VP schemes and propose the generalized model of the pertur-bation vector problem. Then we propose two methods to reduce the complexity ofthe generalized problem. One is to perform sphere encoder after the LR transform,which can greatly reduce the complexity. The other is to use Babai’s proceduresto approximate the perturbation vector, which can reduce the complexity to thatof the conventional precoding schemes. For the MMSE-VP, we propose three low-complexity schemes by the second method.3. Precoding design based on GMD-type triangularization method. The con-ventional triangularization method (such as QRD) will introduce the SNR/SINRimbalance problem for different streams. The very recently proposed geometricmean decomposition (GMD) method can solve this problem. However, the ex-isting GMD-based schemes still use the conventional precoding processing. Wefirstly propose the GMD-based joint transceiver VP design, including the ZF-GMD-VP, MMSE-GMD-VP and the improved versions. Simulation results show thatthe proposed schemes can significantly outperform the existing transmitter-sideVP schemes with great complexity reduction. Moreover, we generalize the jointtransceiver design by setting the unitary matrix at the receiver to be any one. Performance analysis gives insight for the principle of choosing the unitary ma-trix. We further propose the low complexity LRA-VP schemes. Finally consideringthe optimal power allocation, we use the uniform channel decomposition (UCD)method to propose the ZF-UCD-VP and MMSE-UCD-VP design and achieve betterperformance.4. Precoding design based on BD-GMD-type block-diagonalization method.Block-diagonalization (BD) method makes it possible for space division multipleaccess SDMA multiuser MIMO systems can perform joint transceiver processingand obtain better performance than the only base-station-side processing. Dueto the good property of GMD, we deeply analyze the BD-GMD and its applica-tion to the precoding design for multiuser MIMO systems. Firstly, we propose theLRA precoding/transceiver schemes for the uplink/downlink, whose performanceare significantly better than the existing BD-GMD schemes. Then, we proposethe BD-GMD based joint transceiver VP schemes for the downlink and proposethe corresponding schemes with low complexity. Simulation results show that theproposed schemes can have great performance gain compared with the existingschemes.