Research On Angle Domain Based CSI Acquisition In Massive MIMO-OFDM Systems

By installing hundreds of antennas at the base station,massive multiple input multiple output(MIMO)can effectively improve the spectral efficiency and energy efficiency of the system and meet the capacity requirements of the future wireless network.Therefore,it becomes one of the key enabling technologies of the future mobile communication network.Massive MIMO combined with orthogonal frequency division multiplexing(OFDM),i.e.,massive MIMO-OFDM,has found its way into 5G new radio(NR)due to the benefits of spatial multiplexing gains from massive MIMO coupled with the robustness of OFDM to multipath fading.Downlink channel state information(CSI)is essential at the base station as the base station needs downlink CSI to carry out channel adaptive techniques(e.g.,downlink precoding and resource allocation)to improve the spectrum efficiency,energy efficiency of massive MIMO-OFDM systems.In frequency division duplex(FDD)mode,because there is no channel reciprocity between the uplink and downlink channels,the downlink pilot overhead and uplink feedback overhead are proportional to the number of base station antennas.Therefore,the acquisition of downlink CSI at the base station has become a bottleneck problem which restricts the applications of FDD massive MIMO technologies.In addition,most of the mobile cellular networks work in FDD mode currently.Considering the smooth evolution of the system,FDD massive MIMO-OFDM system is likely to be widely used in the future.Therefore,it is of great importance to design a downlink channel estimation scheme with low pilot overhead and feedback overhead in FDD massive MIMO-OFDM systems.This dissertation focuses on downlink CSI acquisition in FDD mode for wideband massive MIMO systems,where OFDM is adopted.In massive MIMO systems,the resolution of the channel angle domain increases linearly with increasing number of base station antennas.Due to the limited scatterers at the base station,the signal energy is concentrated in a few angle ranges.The dimension of the channel angle domain that has signal energy is far less than the initial dimension.As a result,the channel presents a sparse feature.With the sparsity of wireless channels in angle domain,the overhead of downlink pilot and uplink feedback can be reduced under the same downlink channel estimation performance.In addition,when the difference between the uplink and downlink carrier frequency is less compared with the uplink and downlink carrier frequencies,the physical paths of uplink and downlink channels have reciprocity,including angle domain infromation.Using the physical path reciprocity of uplink and downlink channels,the pilot overhead for downlink channel estimation can be further reduced to ensure the better performance of downlink channel estimation.By explioting the angle domain properties of the channel in massive MIMO-OFDM systems,this dissertation designs downlink channel estimation schemes based on compressed sensing theory and Bayesian theory,which improve the performance of downlink channel estimation with fixed pilot overhead and reduce the downlink pilot overhead and the uplink feedback overhead in the close loop compressed CSI feedback and estimation mode.The main contributions and research results of this dissertation are summarized as follows:a)In the wireless transmission scenario,channels of different subcarriers within the bandwidth have the same sparse structure in the angular domain because of the common scatterers and angle spreads in the angle domain which leads to a cluster structure in channel angle domain.Based on the common sparsity and clustering of the angular domain of each subcarrier in the broadband system,a downlink channel estimation scheme is proposed to improve the channel estimation performance with fixed pilot overhead.Because the transmission path of each subcarrier in the transmission bandwidth passes through the same scatterer,there is a common sparsity in the angular domain of each subcarrier channel in the bandwidth.Then the downlink channel estimation problem can be modeled as a multipe measurement vector(MMV)model.Since the effective energy of the transmitted signal is concentrated in a few small angular spreads,the sparse support set of the angular domain in the downlink channel presents the property of clustering,and the angular domain channel can be modeled as a block sparse vector.In this dissertation,a binary variable is used to characterize the sparsity,and then a local beta process(LBP)is introduced to characterize the clustering.Finally,a corresponding Bayesian architecture diagram is constructed.And according to the constructed Bayesian architecture diagram,the Bayesian inference based Bayesian channel estimation algorithm is designed.b)In the closed-loop compressed CSI feedback and estimation mode,an adaptive pilot overhead downlink channel estimation scheme is designed by using the closed-loop characteristics,which not only reduces the downlink pilot overhead,but also reduces the uplink feedback overhead synchronously.In order to reduce both the downlink pilot overhead and the uplink feedback overhead with the help of compressed sensing theory,a new closed-loop compressed CSI feedback and estimation mode is proposed in academia.In this mode,the amount of downlink pilot overhead and the uplink feedback overhead is the same.So if the downlink pilot overhead can be reduced,the uplink feedback overhead will also be reduced synchronously.In order to make full use of this closed-loop mode,the base station can design the dedicated downlink pilot sequences according to the estimated channel information,considering that channel estimation is carried out at the base station.Moreover,an adaptive pilot overhead scheme can be achieved by controlling the pilot transmission according to the estimated channel quality.In view of this,we design a relevance vector machine(RVM)based Bayesian channel estimation scheme to obtain the estimation channel as well as the posterior distribution of the estimated channel.Based on the posterior distribution of the estimated channel,a dedicated downlink pilot sequence is designed by minimizing the differential entropy of the estimated channels.Compared with using random pilot sequences,the performance of downlink channel estimation with dedicated downlink pilot sequences is improved.At last,according to the quality of the channel estimation,the downlink pilot transmission is controlled and the adaptive pilot overhead scheme is realized.Under the condition of ensuring certain channel estimation performance,the downlink pilot overhead and uplink feedback overhead are both reduced.c)Under the effect of beam squint,we firstly design an uplink channel extraction scheme of low complexity,then we obtain the number of paths,and angle of arrival(AoA)and delay information of each path of the downlink channel by using the reciprocity of AoA and delay information of each path between uplink and downlink channels.Finally,we design a gain estimation scheme for each path of the downlink channel in the closed-loop mode to improve the performance of downlink channel estimation.More antennas and larger bandwidth will be used in future wireless network,which will inevitably lead to the beam squint effect.Due to the effect of beam squint,the traditional channel model is no longer suitable for this kind of massive MIMO systems,and the existing channel estimation scheme is no longer suitable.In this dissertation,we first analyze the spatial-and frequency-wideband effects of the uplink channel under the influence of beam squint,and then by using the spatial-and frequency-wideband effects of the uplink channel,we design a modified density based spatial clustering applications with noise(M-DBSCAN)algorithm to obtain the potential paths with the true paths contained.Based on the fact that most of the paths in the potential paths are not true paths,an off-grid sparse adapted matching pursuit(OSAMP)algorithm is designed to obtain the number of true paths,the AoA and the delay information of each path of the uplink channel.Then the uplink channel extraction is completed.Based on the physical path reciprocity of the uplink and downlink channels,the number of paths of the downlink channel as well as the AoA and delay of each path are also obtained at the base station.Finally,the downlink path gain estimation scheme is designed by using the characteristics of the closed-loop mode.As a result,the downlink channel can be reconstructed at the BS.The proposed scheme can improve the performance of the downlink channel estimation.