Research On Beam Training And Precoding Optimization Design In Millimeter Wave MIMO Communication

Millimeter wave(mmwave)communication,which promises providing high data transmission rate,has been becoming the focus in both academia and industrial community,owing to its abundant spectrum resource.However,the path-loss of mmwave signals is very large due to the high frequencies.Hence,large-scale antenna arrays are required to compensate the path-loss.Nevertheless,the use of large-scale antenna arrays leads to a series of challenging problems in terms of mmwave beam training,beam optimization and precoding designs.This dissertation focuses on beam training and precoding optimization design in mmwave MIMO communication and proposes efficient algorithms in different aspects of mmwave communication,including training codebook design and beam alignment performance analysis,power-efficient digital beam and hybrid analog-digital beam designs and mmwave precoding designs(i.e.,constant envelope precoding and antenna subset selection,wideband hybrid analog-digital precoding,and mmwave directional precoding).The main contributions of this dissertation are listed as follows:1.Beam training is a viable approach to obtain channel state information,which is also adopted by the communication standard.From the perspective of beam pattern/shape approximation,a novel method to design beam training codebook is proposed for the hierarchical beam training algorithm.The problem of beam design is formulated as an optimization problem,where the requirement of small ripple in both mainlobe and sidelobe is incorporated into the constraints of the formulated optimization problem.The training beam of arbitrary beamwidth is optimized to have small ripple in both mainlobe and sidelobe.To address the issue of error propagation in hierarchical BA training algorithm,a novel structure of training codebook is proposed,which helps to achieve a uniform BA performance.For channels of constant and fading path gain,the upper bounds and approximate expressions of BA misalignment probability are derived based on the designed training codebook.Based on the derived expressions,a low-complexity power allocation strategy used in different training stages is proposed to further improve the BA performance.Finally,numerical results are proposed to confirm the effectiveness and superiority of the proposed algorithms.2.For large-scale mmwave antenna array systems,in addition to improving beam properties of designed beams,it is also essential to improve the power efficiency(PE)of the designed beams based on a specific antenna array architecture.To improve the power efficiency of power amplifiers,two power-efficient algorithms are proposed to design fully digital beam and hybrid analog-digital beam.First,a powerefficient algorithm is proposed to design the fully digital beam.Since the formulated optimization problem contains two types of constraints,i.e.,beam property(BP)constraints and PE constraints,and each type contains a large number of constraints,the resulting problem is difficult to tackle and even finding a feasible solution is a non-trivial task.To solve this problem,the idea of outer approximation is introduced,based on which an efficient sequential search algorithm is devised to find the feasible solution.Then a hybrid analog-digital mapping algorithm is proposed to map the designed digital beam into analog part and digital part in order to use the hybrid antenna array.To further improve the BP,a power-efficient algorithm that directly designs the hybrid beam is proposed,where the constraints of finite phases are also incorporated into the procedure of beam optimization.An advantage of the direct method is that there is no need to perform the hybrid analog-digital mapping.Numerical results are provided to confirm the effectiveness and superiority of our proposal.In particular,it is shown that the designed beam greatly reduces the PAPR and thus improves the PE,and can still achieve good BP,even for low-resolution phase shifters.3.With channel state information available,effective precoding designs can further improve system performances,while the issue of power efficiency shall also be considered in precoding designs.Constant envelope precoding(CEP)can effectively reduce PAPR,and thus improves the power efficiency,which,however,has high computational complexity and is difficult to find the globally optimal solution due to its nonlinearity and convexity.The problem of CEP in large-scale multi-antenna system is transformed into the problem of polygon construction in the complex plane,from the geometric viewpoint.The feasibility and equivalence between the CEP and triangle construction is further revealed to minimize the complexity of CEP algorithm.A CEP algorithm with linear complexity is proposed,based on the triangle construction.The joint optimization of CEP and antenna subset selection(ASS)is considered to minimize the total transmit power and improve the PE.First,two ASS design philosophies,namely constellation-oriented and symbol-oriented ASS are analyzed and compared,which shows that the symbol-oriented ASS has great advantages in minimizing the total transmit power and providing robustness to channel uncertainty.A novel ASS algorithm is proposed to minimize the total transmit power while satisfying the Qo S requirement.For both perfect or imperfect channels,the proposed ASS algorithm is guaranteed to find the globally optimal ASS solution with only addition and multiplication operations,whose complexity is at most quadratic in the number of antennas in the worst case.Effectiveness and superiority of the proposed geometric algorithms is verified by comprehensive numerical results.4.An attractive characteristic of mmwave communication is its large bandwidth,which shall be fully utilized to improve the data transmission rate.For OFDM based wideband millimeter wave(mmwave)communication systems,a hybrid analog-digital precoding algorithm is proposed to jointly optimize the digital precoder performed on a per-subcarrier basis and the analog precoder performed in the time domain and used for the entire bandwidth.Since analog-digital precoders are coupled and the resolution of phase shifters is also finite,the resulting optimization problem is difficult to tackle.To address this issue,new optimization variables and equality constraints are introduced to decouple the joint design of digital and analog precoders.The penalty function method is further employed to tackle the introduced equality constraint,which leads to an optimization problem of higher dimension but decomposability.The decomposability of the resulting optimization problem facilitates the use of block coordinate descent methods.Under appropriate constraint qualification,it is shown that the proposed algorithm is converged to the stationary point of the original optimization problem.With some slight modification,the proposed algorithm can also apply to the case of finite resolution phase shifters.Simulation results are provided to further confirm the effectiveness and superiority of our algorithm over the existing algorithms.5.The mmwave channels have sparsity and directionality,based on which directional precoding can be designed to improve the security performance of mmwave communication systems.A unified CEP framework is proposed to enhance security for the sub-connected hybrid analog-digital mmwave systems,by formulating an optimization problem to guarantee the quality of receive constellations for target users,while minimizing the power leaked to the potential eavesdroppers.The proposed framework is able to achieve directional precoding for both single-path and multi-path mmwave channels.However,the resulting optimization problem is very challenging due to the coupling between the analog precoder and digital precoder,the constant modulus constraints,the finite resolution of phase shifters,and the 0-1 constraints caused by subarray selection.To tackle this problem,penalty function methods along with the block coordinate descent methods are leveraged to design hybrid precoders.Then,two typical antenna array architectures,i.e.,multi-subarray architecture and switched-phased-array architecture,are investigated in details.For the multi-subarray mmwave system,the feasibility condition of the hybrid precoding problem is analyzed and the augmented Lagrange penalty method is leveraged to find an efficient solution which is shown to reach a stationary point under some mild conditions.The directional precoding is further investigated for the switched-phased-array mmwave system.Since the resulting problem is a more difficult mixed integer nonlinear problem,an efficient iterative algorithm is proposed to address this problem by using the exact absolute value penalty method as well as augmented Lagrange penalty method.Numerical results are provided to confirm the effectiveness of the proposed algorithms and show their superiority over the existing ones,i.e.,achieve better secure rate performance and high power efficiency.