Research On Key Technologies In Massive MIMO Systems

As a key candidate technology in the forthcoming 5G mobile networks,massive MIMO is able to gain higher spatial degree of freedom through de-ploying massive antennas on BSs,to serve multiple users using the same time-frequency resources and to significantly increase the spectral efficiency.How-ever,there still exist some challenges regarding the realistic deployment of mas-sive MIMO,such as pilot contamination in TDD mode acts as the bottleneck for further improvement of the user' performance with the increasing of BS antenna number,the huge feedback overhead caused by the large dimensional channel state information in FDD mode and the multi-beam selection under hybrid precoding architecture in millimeter wave's massive MIMO.This paper studies the three problems listed above.Firstly,a space-time pilot allocation scheme is proposed to mitigate the pilot contamination in TDD massive MIMO systems.This scheme divides the cells into different groups in time dimension,which apply an asynchronous frame structure whose pilot slots for different groups are time-shifted.In spatial dimension,the co-group users with the smallest co-interference are scheduled to use the same pilot sequence with the goal of minimizing the average Bayesian channel estimation MSE.The uplink and downlink asymptotic SINR are given based on random matrix theory.Theoretical analysis and simulation results both illustrate that space-time pilot allocation scheme is able to mitigate pilot contamination from both spatial dimension and time dimension.The grouping strategy greatly reduce the cells and users which proceed the channel estimation simultaneously.Consequently,the MSE computation counts and the feedback overhead of overall second order stochastic channel state information are both reduced significantly.Secondly,a joint space-time smart pilot allocation scheme is proposed to improve the fairness of users' performance.This scheme applies the asyn-chronous frame structure in time dimension.In spatial dimension,an optimiza-tion problem is built with the goal of maximizing the users-minimum uplink SINR in the target cell.The spatial dimension pilot sequences are allocated by sorting the local users' channel quality and the intensity of the interference they suffered.Mathematical analysis proves that this pilot allocation method is the solution of the optimization problem above.The sorting based solution can greatly reduce the searching complexity when compared with exhaustive re-search.Simulations show that the joint space-time smart pilot allocation scheme can not only improve fairness and the minimum uplink SINR performance of the target cell users,but also promote the target cell's overall uplink capacity performance.Thirdly,to solve the huge channel feedback overhead problem in FDD spatially correlated single user massive MIMO systems,wavelet compression method is adopted to explore the spatially correlation redundancy and reduce the feedback overhead.Prior experiment method is adopted in our work to eval-uate the current wavelet basis,and "bior3.1" wavelet basis is found to get the best compression and reconstruction MSE performance for spatially correlated channel.Consequently,"bior3.1" wavelet is adopted at receiver to get the wavelet decomposition of the channel state informaiton.And the wavelet coef-ficients from low frequency to high frequency will be truncated and sent back to transmitter.At transmitter,the missing coefficients can be filled with zero padding and then be reconstructed by wavelet reconstruction algorithm.Sim-ulations show that the compression and reconstruction MSE performance and spectral efficiency performance both outperform that of compressive sensing method which adopts orthogonal matching pursuit reconstruction algorithm.Finally,the beamforming beam selection problem for multi-user millime-ter waves' massive MIMO with hybrid precoding design is considered.A re-alistic deployment is considered when the number of activated BS RF chains is located between user number and BS antenna number(K ? NRF? NBs).A limited feedback based user beam and BS multi-beam selection method is proposed with the goal of maximizing the overall spectral efficiency.Beam evaluation criterions are designed and used for each user to evaluate the trans-mit and receive beam pairs.A high precision feedback mode FB1 and a low precision feedback mode FB2 are proposed.Simulations show that the feed-back scheme proposed in this chapter can exploit the additional RF chains and increase the channel dimension,which get better spectral efficiency than the existed hybrid precoding design under suitable condition.