Research Of Key Technologies In Massive MIMO System With Two-stage Precoding

The development of the data communications and multimedia,the rapid changes in the ecology of mobile communications-related industries,and the growing demand for the mobile data,mobile computing and media applications as well as the the rapid development of the mobile Internet and the Internet of Things(IoT)push an exploded increasing for the data traffic and mobile devices.In order to achieve such a high transmission rate and such a large number of connections,it is necessary to make breakthroughs for the transmission technologies in the physical layer.Massive multiple-input multiple-output(Massive MIMO)has been widely considered as one promising technology for the 5G cellular system,which can effectively solve the problems above.Massive MIMO is also useful in military communications,such as battlefield real-time video transmission and complex battlefield situation information transmission.The high-precision spatial resolution and precise beam steering characteristics of massive MIMO also provide new ideas for anti-interference and anti-monitoring in military communications.Due to the large number of antennas equipped by the base station,massive MIMO brings huge gains to the communication system.At the same time,compared to the traditional MIMO technology,it also faces great challenges in channel state information(CSI)acquisition.In order to overcome this difficulty,researchers have creatively proposed a twostage precoding scheme to make full use of the low-rank characteristics of massive MIMO channels.The precoding for the downlink transmission is divided into two stages in detail,namely outer precoding and inner precoding.During the former stage,the outer precoding,which only depends on the channel covariance matrices,is utilized to eliminate the intercluster interference,and partitions the high-dimensional massive MIMO links into several independent equivalent channels with small sizes.Moreover,users with the same channel covariance matrices can be gathered into the identical cluster while different clusters exhibit orthogonality in the space;during the latter one,each cluster separately performs the inner precoding to eliminate the intra-cluster interferences.The covariance matrix changes very slowly compared to the real-time channel,thus the outer precoding only takes the longer time to update.Each user cluster only needs to obtain the low-dimensional equivalent channel to compute its inner precoding matrix after carrying out the outer precoding.Because the corresponding low-dimensional equivalent channels of each cluster are independent,one pilot sequence can be reused by different clusters.In addition,the number of pilots and the amount of feedback for low-dimensional equivalent channel acquisition is much lower than that of the original massive MIMO channel.Therefore,the two-stage precoding method can overcome the difficulty of obtaining the CSI in massive MIMO system.Based on the two-stage precoding,many researchers have devoted themselves to the research of channel modeling,limited-beam resource scheduling,and efficient beam-field coordinated transmission for massive MIMO.And many academic achievements have been achieved.However,there are still many problems in terms of user scheduling,cooperative transmission and tracking and prediction for the time-vary channel.Inspired by these problems,this dissertation will take full use of the low-rank characteristics of space channel of massive MIMO to explore the effective schemes with respect to these problems,where the advantages of the two-stage precoding framework can be carefully taken into account and efficient mathematical tools such as sparse Bayesian learning,graph theory and combinatorial optimization will be adopted.As a result,the ability of serving users with the same time-frequency resources in the massive MIMO system will be dramatically improved.The specific research content and achievements of this dissertation are as follows:(1)Two effective user scheduling methods with low complexity are proposed to mitigate the impact of users' angular domain overlap on the performance of a two-stage precodingbased massive MIMO network.The first method mitigates the angular overlap of the user clusters by the optimal base station associations.First,a low complexity base station selection algorithm based on SLNR maximization is proposed.The performance of this algorithm is close to optimal,but the base station needs to obtain low-dimensional equivalent channel information from each user cluster to each base station,which will cause huge signaling overhead and waste channel resources.In order to avoid real-time channel acquisition,we seek the expectation of SLNR and derive its lower bound,so that the objective function does not include real-time channel state information.Then,a low-complexity and low-signaling overhead base station selection algorithm is designed based on the lower bound of SLNR.Numerical simulation shows that the SLNR lower bound algorithm can significantly reduce the user cluster angle overlap,improve the system and rate,and its performance is basically the same as the SLNR-based algorithm.The second method jointly gather user clusters with orthogonal subchannels allocaiton,which divides the user cluster into multiple groups,ensuring that the angle overlap between user groups within each group is minimized.Users in the same group then use two-stage precoding to achieve orthogonal transmission,while different groups of users are scheduled on orthogonal time-frequency subchannels.In order to achieve the optimal grouping of users,we will construct an undirected weighted graph to indicate the strength of the angular domain overlapping among clusters.Then,a graph theory based user gruoping method is proposed through improving the traditional Dsatur algorithm.Simulation results show that the performance of the proposed algorithm is near optimal.(2)We propose an interference alignment and soft-space-reuse(IA-SSR)based multi-cell cooperative transmission scheme under the two-stage precoding framework.The IA is introduced into massive MIMO system to remove the inter-cell interference and increase the service rate of cell-edge users.However,the acquisition of the global channel state information(CSI)for IA leads to unacceptable overhead in the massive MIMO systems.To tackle this problem,we employ the two-stage precoding to decompose the high-dimensional channel of massive MIMO into several equivalent MIMO channel with low dimension.Moreover,a low-cost channel estimator is designed for this proposed framework,and then we propose a IA transmission scheme based on the low-dimensional channel.In addition,in order to improve the angle-domain overlapping in the multi-cell scenarios,soft space reuse scheme is designed for the cell-center users.Finally,an optimal power allocation strategy is proposed combining the golden section search algorithm with water-filling algorithm.Finally,plenty of numerical results are presented to show the efficiency of the proposed algorithm.(3)A time-varying channel estimation scheme both for the TDD and FDD massive MIMO networks is proposed for uplink and downlink.Firstly,considering the uplink channel estimation,we formulate the dynamic massive MIMO channel as one sparse signal model where the estimated parameters only related to the channel covariance matrix,which reduces the overheads and complexity of the instantaneous channel tracking.Then,an expectation maximization(EM)based sparse Bayesian learning(SBL)framework is developed to learn the model parameters of the sparse virtual channel.Specifically,the Kalman filter(KF)and the Rauch-Tung-Striebel smoother(RTSS)are applied to track the posterior statistics of the uplink(UL)spatial sparse channel in the expectation step,while a fixed-point theorem based algorithm and a low-complexity searching algorithm are separately developed to recover the temporal varying characteristics and the spatial signatures in the maximization step.For the downlink(DL)channel estimation,utilizing the angle reciprocity,we recover the downlink spatial signature from the UL one.After that,the reduced dimension KF is applied to fully exploit the channel temporal correlations to enhance the DL/UL virtual channel tracking.A monitoring scheme is also designed to detect the model parameters change and trigger the relearning process.Finally,we demonstrate the efficacy of the proposed schemes through simulations...