Research On The Finite-alphabet Constrained Multi-user MIMO Precoding

With the explosive increasing of mobile data traffic,traditional wireless transmission scheme fails to meet the requirement for high-speed and low-latency transmission.Multi-user MIMO technology allows the base station independently to communicate with multiple users at the same frequency and the same time,which significantly improves the system capacity and reduces the transmission delay.Large-scale antenna array not only improves the performance of wireless system,but also brings some problems such as high cost and power explosion,which violates the mainstream trend of green communications.Hybrid precoding technology provides high dimensional space utilization under the condition of a rather small number of RF chains,however,the possible phase candidates of the tunable phase shifters belong to a finite set.How to optimize the spectral efficiency and energy efficiency of the hybrid precoding systems under the constraints of a finite set to obtain a near-optimal solution is one of the topics studied in this thesis.Recently,low resolution digital-to-analog converter with low cost and low power consumption has attracted a wide attention.As an enabling technology of large-scale antenna array systems,the output of digital-to-analog converter is also constrained by finite set.How to make full use of symbol level precoding to design trsnsmitted vector is another topic of this paper.Focusing on these application scenarios of multiuser MIMO,this dissertation aims at two research fields,i.e.,analog digital hybrid precoding design and symbol-level precoding design.Multi-user precoding includes spectral efficiency and energy efficiency optimization with finite-alphabet constraints of analog phase array,and symbol-level precoder design using different modulation types with low resolution analog to digital converters.The main contributions of this dissertation are as follows.1.In order to maximize the spectral efficiency of the analog/digital hybrid precoding with finite-alphabet constrained phase shifters,an iterative algorithm framework based on discrete optimization with global search is proposed,and precoding schemes are designed for singleantenna users and multi-antenna users respectively.The algorithm ensures the convergence of iteration and validity of the results.Firstly,a hybrid precoder with low complexity is proposed based on a cost-efficient and energy saving adaptive switch-fixed phase shifter network,which replaces the variable phase shifter and analog adder in the traditional hybrid precoders.Based on the partially connected structure,the model of partially connected analog/digital hybrid precoding is established,and the problems to be solved in the coding matrix design are also presented.Then,in order to solve the nonconvex combinatorial optimization problem,an iterative algorithm framework of global search is adopted,and the idea of precoding design is given.Finally,the corresponding digital precoders are designed respectively for the single antenna and multiple antennas scenarios.Simulation results verify the effectiveness of the proposed scheme in a variety of hybrid precoding models.Compared with the traditional iterative optimization algorithm,the proposed scheme can achieve higher achievable rates in different SNR intervals and data streams.2.Aiming at the energy efficiency optimization of analog/digital hybrid precoding constrained by finite set of phase shifters,a RF-baseband hierarchical optimization algorithm is proposed considering the hibernation of baseband and RF chain.Different from the traditional algorithm which only considers to reduce the energy consumption of the RF chain,the advantage of the partially-connected system is that it can shut down part of the baseband signal processing circuit without affecting the RF chains that is not connected,which can further reduce the energy consumption of the system by hibernation.Firstly,based on the hybrid precoding structure of partial connection,the energy consumption model is established which fully considers the power of each part.Then,while ensuring that all users are served,taking the overall energy efficiency of the system as the objective function,the analog coding array can be obtained through cross entropy optimization and the digital coding array can be designed to eliminate the interference between data streams.Finally,the energy efficiency of the system is maximized by closing data links that contribute less to the sum rate.Simulation results show that the proposed algorithm has obvious performance improvement compared with traditional fractional optimization,and the RFchain selection algorithm based on matrix decomposition can achieve almost the same performance as the traversal method.3.Consider the bit error rate fairness problem of symbol-level precoding with low-resolution digital to analog converter in multi-user MIMO systems,a low complexity precoding algorithm framework is proposed by taking advantage of PSK modulation’s constructive interference region,which provides a near-optimal closed-form solution while reducing the problem size.Firstly,from the perspective of multi-user fairness,a bit error rate model and combinatorial optimization problem based on decision distance are presented.Then,we relax the problem to obtain a convex problem.By analyzing the KKT condition of the problem,we theoretically prove that the size of the pre-coding variables with finite-alphabet DAC can be greatly reduced.Secondly,smooth and continuous objective function is obtained by approximating the non-differentiable objective function,and the problem is decomposed into three simple sub-problems by using ADMM framework.Finally,the nearly optimal closed-form solution of the subproblem is obtained by using the second-order MM approximation,which makes the real-time operation of the algorithm feasible.Simulation results show that compared with the classical penalty function method and the minimum mean square error algorithm,the performance of the algorithm in this chapter is significantly improved,and the computational complexity is linear because the whole iterative process of the algorithm is closed.4.Bit error rate fairness problem with finite-alphabet constraints is considered in the QAM modulated symbol-level precoding.Through the analysis of the characteristics of the QAM constellation diagram,the constructive interference region is redesigned.Using the symmetry quantitative level,the problem can be converted to binary constraint optimization problem,and the precoding algorithm is obtained by introducing complementary constraint convex relaxation.Firstly,we design corresponding phase-length interference regions according to the nature of internal points in QAM constellation,which not only avoids the performance improvement bottleneck of traditional regions,but also avoids the situation that there may be no possible solution in the phase-length interference regions with distance preservation.Secondly,we show how to transform finite-alphabet constraints into binary constraints equivalently by using the equally-spaced and symmetric property of DAC quantization without causing the variable dimension to increase dramatically.Then,we present the core algorithms adopted in this chapter,including the traditional convex relaxation constraint and the complementary optimization algorithm.Finally,the simulation results verify the feasibility and superiority of the algorithm.Compared with the classical direct quantization algorithm and the minimum mean square error algorithm,the performance of the algorithm in this chapter is significantly improved.