Research On Efficient Channel Estimation For Millimeter Wave Systems

With the development of modern communication technology,the existing communication frequency bands are increasingly difficult to meet the communication needs of large capacity,low latency,and massive connections.Millimeter-wave(mm Wave)has gradually entered people’s field of vision due to its rich spectrum resources.The combination of mm Wave and massive Multiple-input multiple-output(MIMO)architecture can not only compensate the severe path loss of mm Wave,but also make the deployment of massive MIMO possible due to the size reduction of mm Wave device.Therefore,this technology has become a key technology for the fifth generation(5G),beyond 5G(B5G)and even the sixth generation(6G).Accurate mm Wave channel estimation is critical for mm Wave beamforming design and reliable data transmission.In the mm Wave system,due to the limited scattering characteristics,the mm Wave channel is sparse in the angle domain.Taking advantage of this spare nature,the mm Wave channel estimation problem is often modeled as a compressive sensing sparse signal recovery problem,effectively reducing the required pilot symbols and computational complexity for channel estimation.In light of this background,this thesis advances the preliminary studies and attempts to tackle two open challenges in mm Wave channel estimation and propose efficient algorithm designs to achieve higher channel estimation accuracy and spectral efficiency with lower training costs.First,this thesis addresses studies the problem of multipath channel estimation under the condition of low signal-to-noise ratio(SNR)of mm Wave environment,and proposes a two-stage adaptive mm Wave multipath channel estimation method to overcome the performance limitation of classical compressive-sensing based estimation algorithms under low SNR conditions.Specifically,in the first stage of the scheme,random observations are made on the entire space,and the posterior probability distribution of possible combinations of sparse channel angles is calculated based on the obtained initial observation signals.In the second stage,according to the obtained candidate angle support set,a small number of narrow beams are used to detect the angle pairs in the support set one by one,and channel estimation is performed accordingly.By adaptively aligning the direction of the sensing beam to the direction where the channel angle is more likely to exist,the resource consumption of ineffective observations is avoided,and the equivalent training SNR is improved,thereby effectively improving the performance of channel estimation in a low-SNR environment and realizing efficient channel estimation.Simulation results show that the proposed scheme requires an average of 5d B lower SNR to achieve the same channel estimation performance compared to the baseline scheme.Next,this thesis studies the problem of joint uplink and downlink channel estimation in asymmetric mm Wave systems.In such systems,the reciprocity of uplink and downlink channels no longer holds.In particular,for asymmetric systems with more transmitting antennas than receiving antennas at serving base station,the beamwidth of the uplink receiving array is wider,which limits the path angle resolution,and the traditional uplink-downlink reciprocity based channel estimation schemes are no longer suitable for asymmetric systems.To cope with this challenge,preliminary study uses the uplink received pilot signal to obtain the uplink channel estimate and then transfers to the downlink channel information.However,this method suffers error propagation issue,i.e.,if the uplink estimated channel error is large,it is also difficult to accurately obtain the downlink channel.This thesis proposes a new joint uplink and downlink channel estimation scheme for such asymmetric systems.In particular,the downlink channel estimation problem at hand is modeled as a compressive sensing signal recovery problem with0-1 mask operation,where the mask is introduced for characterizing the activation/deactivation state of each antenna in the antenna array,and the channel estimation problem is solved by combining the sparse Bayesian framework with the expectation maximization algorithm(EM).At the same time,dynamic dictionary parameters are introduced to effectively improve the resolution of the estimated channel angle.The proposed scheme can produce downlink channel estimation directly through the uplink received signal,which effectively solves the challenge of downlink channel estimation due to the nonreciprocal uplink-downlink peroperty.The simulation results show that the normalized mean squared error(NMSE)performance of the proposed scheme is improved by about 3d B on average compared with the baseline algorithm.