Research On Beamforming In Milimeter Wave MIMO Systems

With wireless frequency resource shortage becoming more and more severe, the large amount of bandwidth available at millimeter wave(MMW) frequencies brings hope for digital communications with gigabit-per-second data rates, which are demanded by the fifth generation(5G) mobile communication systems. The extremely high path loss of MMW signals is the most difficult problem to solve when MMW systems are applied for outdoor wireless transmissions. Fortunately, the tiny wavelength of MMW signals enables massive antenna arrays(massive MIMO) to be installed on a mobile device while keeping its size small. Massive MIMO can achieve large beamforming gain, thus providing sufficient link margin.Beamforming is an array signal processing technology which can be regarded as linear spatial filtering and is widely used in MIMO systems for counteracting path loss, improving receiving signal to noise ratio and increasing the system capacity. The key of beamforming is to obtain the optimal array signal weighting vectors for the transmitter and receiver, which are designed based on some specified optimization criteria. For the capacity-optimal criteria, the weighting vectors are determined by the singular value decomposition(SVD) of channel matrix, which requires the channel state information(CSI) at both the transmitter and receiver.In time divided duplex(TDD) systems, CSI need not to be estimated explicitly thanks to the channel reciprocity. Alternatively, the weighting vectors are obtained in an iterative way, which saves lots of expensive RF chains and avoids the complicated channel estimation for massive antenna arrays. In frequency divided duplex(FDD) systems, CSI is estimated explicitly and then weighting vectors are acquired by the SVD of channel matrix. No matter TDD or FDD, the optimal weighting vectors are achieved by sending training symbols, the amount of which has significant impact on the performance of the system. In the situation of massive MIMO, the training overheads of traditional methods are formidable, because the training time of these methods are generally larger than the channel coherence time which is limited in MMW system. Moreover, the available independent symbols for training is also limited for such a large number of antennas. Thus, new methods must be developed to overcome these problems, and fortunately the spatial sparsity of MMW channel provides a possibility for solving these problems.This thesis studies the characteristics of the channel of MMW MIMO systems with massive antenna arrays and focuses on the beamforming of such systems in TDD and FDD cases. For TDD systems, power iteration is utilized to implement the SVD beamforming without performing the burdensome channel estimation and matrix decomposition of massive MIMO channel matrix. Moreover, compressive sensing is exploited to reduce the training overhead, which makes SVD based beamforming feasible in MMW systems with large antenna arrays and a limited number of RF chains. For FDD systems, channel estimation of multi-user massive MIMO in MMW cellular system is studied. The multiple measurement vectors(MMV) model is introduced and adapted to exploit the structural sparsity of the multi-user massive MIMO system. A joint multi-user channel estimation algorithm is developed, which not only cuts down the per-link training overhead relying on compressive sensing, but also takes advantage of the correlation of the user channels to improve performance.Simulations show that the proposed beamforming algorithms based on compressive sensing can considerably decrease the training overhead of beamforming in MMW massive MIMO systems and so the proposed methods have practical value.