Research On Hybrid Precoding Technology Of Mmwave Massive MIMO System

With the rapid development of mobile communication,the remaining low-band spectrum resources can no longer satisfy the growing communication needs,the next generation communication systems need to find available spectrum resources on the millimeter wave(mm Wave)frequency band.The combination of massive multi-input multi-output(MIMO)and mm Wave can bring advantages in terms of spectrum efficiency,system throughput,and coverage.At the same time,the precoding technology can provide sufficient antenna array gain to overcome the path loss through mm Wave signal propagation.However,the traditional digital baseband precoding is no longer suitable for mm Wave massive MIMO systems,because numerous radio frequency(RF)chains corresponding to a large number of antennas cause high energy consumption and hardware complexity.Based on current research,the application of digital and analog precoding called hybrid precoding can reduce the number of energy-intensive RF chains.Therefore,this thesis concentrates the research on hybrid precoding schemes in mm Wave massive MIMO systems of different structures,which solves the problems of existing precoding methods.The main works and contributiaons of this thesis are summarized as follows.Firstly,the hybrid precoding under fully-connected structure of mm Wave massive MIMO systems is investigated.Three hybrid precoding schemes are proposed via fully-connected structure.First,a low complexity HEW hybrid precoding scheme is proposed.The analog precoder is obtained from equal gain transmission(EGT)phased array to eliminate inter-user interference.The digital precoder directly applies water-filling method without singular value decomposition(SVD),which reduces the computational complexity.The HEW achieves good sum rate performance under low signal to noise ratio(SNR).Second,a discrete Fourier transform(DFT)codebook-based hybrid precoding is proposed.All users are divided into two groups,i.e.,conflicting users(CUs)and non-conflicting users(NCUs).A conflicting-aware(CA)beam column selection method is proposed to obtain analog precoding matrix,which maximizes the sum rate.Then zero-forcing(ZF)digital precoder is utlized in digital domain to eliminate interference between CUs.The DFT codebook-based hybrid precoding has low complexity and good sum rate performance.Third,a CGJHP hybrid precoding based on cluster grouping and joint space division multiplexing(JSDM)is proposed.The cluster-group based eigen-prebeamforming makes effective channel matrix sparse.A grouping column selection method is implemented to design analog precoding matrix based on codebook to eliminate interferences within non-neighbor clusters.ZF with per-group processing(PGP)is used to obtain the digital precoding matrix of each group to eliminate interference within neighbor clusters.Moreover,the pilot matrix with small DTL is designed based on the graph theory.The CGJHP achieves satisfactory sum rate performance,and reduces the overhead of downlink training length(DTL)and instantaneous channel state information(CSIT)feedback.Secondly,the hybrid precoding under sub-connected structure of mm Wave massive MIMO systems is investigated.A suboptimal iterative approximation(NOIA)hybrid precoding is proposed,which optimization goal is to minimize the mean squared error(MSE).The original optimization problem is reformulated as two sub-problems to solve the challenging non-convex problem,where one of the sub-problems is convex and another is non-convex.The non-convex sub-problem is converted into a convex problem by relaxing the constant modulus constraint from beampattern with interference control(BIC)method.The block diagonal constraint is eliminated from vector operation with the properties of Kronecker product.Then the two convex sub-problems are solved iteratively.The NOIA scheme achieves a highly approximate optimal solution,which can be demonstrated to converge to a Karush-Kuhn-Tucker(KKT)point of the original problem.Finally,the hybrid precoding scheme under broadband frequency-selective structure of mm Wave massive MIMO systems is investigated.The most important difference between broadband structure and narrowband structure of hybrid precoding is that a common analog precoder for all subcarriers and different digital precoders for different subcarriers are required.In the single-user scenario,the SU hybrid precoding scheme is proposed.The general analog precoder is derived from EGT.Digital precoding matrixes for different subcarriers are directly employed based on water-filling technique,since the equivalent channel matrix after analog precoding is almost orthogonal.In a multi-user scenario,the MU hybrid precoding scheme is proposed.Gram-Schmidt orthogonalization is added in the analog domain and ZF is utilized for the digital precoders in order to nullify inter-user interference.In simulation,the SU and MU algorithms can achieve good performance which is close to full digital precoding.