Research On Channel Estimation And Beam Alignment In Millimeter Wave Wireless Communications

Millimeter Wave(mmWave)communication is the key technology for the Fifth Gen-eration(5G)mobile communication networks.To meet the demand of the growing data service in wireless communication networks,mmWave communication is able to provide a communication service with high data rates and low latency.Although communication in mmWave frequencies has the advantage of large bandwidth,it suffers from severe path loss and limited coverage due to the large bandwidth and unfavorable weather.To address this issue,large antenna arrays and the beamforming technique should be used to provide sufficient beamforming gain for mmWave communications.Optimal beamforming requires the knowledge of perfect Channel State Information(CSI).The way to obtain the perfect CSI with limited time resource becomes the major problem in mmWave communication systems.Instead of estimating the perfect CSI di-rectly,beam alignment via partial channel estimation is also feasible since the number of paths with prominent path gain is few.Compared with the channel estimation meth-ods,although the beam alignment methods may fail to obtain the maximum beamforming gain,those methods are robust to noise,and are easy to implement.The way to achieve fast beam alignment with limited pilot overhead is also the main problem in mmWave communication systems.To address those issues,this dissertation first studies the problem of channel estima-tion based on the joint sparse and low-rank structure of the channel.Due to the sparse scattering nature and the angular spread within each scatter,mmWave channel has a joint sparse and low-rank structure.In this dissertation,a two-stage compressed sensing based channel estimation algorithm is proposed.It is shown that this joint sparse and low-rank structure can be utilized to reduce the sample complexity,and the channel can be exactly estimated from compressed measurements via the proposed method.A rigorous theoreti-cal analysis is provided to prove that sampling complexity of the proposed method is lower than the that of the compressed sensing method.Simulation results are also provided to show that compared with the compressed sensing method,the proposed two-stage method can save 30% number of measurements when achieving a bit error rate of 0.1.Next,this dissertation studies the problem of channel estimation with the assist of channel covariance matrix.In cellular networks,the base stations are always located at high positions,while the scatters are distributed around the users.In this case,since the Angle of Arrival(Ao A)of paths are bounded to a small interval,the channel covariance matrix has a low-rank structure.With the aid of this low-rank structure,this dissertation shows that the Minimum Mean-Squared Error(MMSE)estimator can be used for channel estimation.A covariance-assisted channel estimation method is then proposed to reduce the training overhead.In the noise-free case,it is shown that the MMSE estimator can achieve exact channel recovery when the number of pilot symbols is no less than the rank of the channel covariance matrix.In the noisy case,the optimal pilot and an asymptotic optimal pilot are both digned for the single-user case and the multi-user scenario,respec-tively.It is also shown that using the optimal pilot design,the number of pilot symbols in time is no less than the rank of the channel covariance matrix.This dissertation develops a simple scheme to estimate the channel covariance matrix.The proposed scheme does not need any additional training overhead.Simulation results are provided to show that compared with the compressed sensing method,the normalized mean-squared error of the proposed method is reduced from 0.15 to 0.03 when signal noise ratio(SNR)is 15 d B.Lastly,this dissertation studies the problem of beam alignment in mmWave commu-nications based on multi-directional beam searching.To reduce the training overhead and avoid redundant interaction between the transmitter and the receiver,a multi-directional beam searching scheme is proposed in this dissertation.By exploiting the sparse scatter-ing nature of mmWave channels,the proposed scheme can reduce the training overhead with multi-directional beams,each beam of which points to multiple directions simultaneously.The optimal beam direction is then obtained by comparing the power of received signals.Theoretical analysis shows that,compared with the exhaustive search method,the proposed algorithm can perfectly recover the support and magnitude of the sparse channel from fewer measurements.The robust multi-directional beam searching design along with the robust beam alignment algorithm are also provided in the noisy case.The proposed algorithm has a simple signal processing procedure that is computationally efficient and noise-robust.Simulation results show that the proposed method provides a beamforming gain of 0.58 when SNR is 0d B,while the beamforming gain of other methods are 0.55,0.25 and 0.2,respectively.