Precoding Design And Optimization For Massive MU-MIMO System

Recently,with the use of smart phones and tablet computers,the demand for high data rates in wireless networks has increased dramatically,which makes the current forth-generation(4G)communication network's capacity to reach its peak.Therefore,we need to develop fifth-generation(5G)communication network with higher carrying capacity.As one of the key technologies in the 5G communication network system,massive multi-user multiple-input multiple-output(MMU-MIMO)technology has combined many advantages of simultaneously serving multiple users,providing higher spectrum utilization and system capacity.Although the use of large-scale antenna arrays achieves array gains,it also brings an increase in multi-user interference(MUI).In order to balance the contradiction between the array gain and MUI,the base station must precode the signal.Using precoding technology,the system not only increases the antenna array gain but also reduces the MUI.But the use of precoding also brings an increase in hardware costs and power consumption.Although many precoding algorithms have been proposed,they have the disadvantages of high computational complexity and poor accuracy.Therefore,this article mainly studies low-complexity and high-efficiency precoding algorithms.(1)First,this thesis introduces the related classic precoding algorithms,and then mainly designs a constant envelope(CE)precoding with 1-bit DAC based on minimum mean square error(MMSE).For the 1-bit digital-to-analog-converter(DAC)precoding method,we first designed a 1-bit DAC precoding algorithm based on the squared-infinity norm relaxation.The algorithm is mainly based on accelerate proximal gradient method(APGM),which solves the squared-infinity norm relaxation problem in a computationally efficient way.And the algorithm only needs simple matrix and vector operations in each iteration,as well as evaluation of the proximal operator of squared-infinity norm.It can be seen from the simulation analysis that APGM precoding scheme can achieve rapid convergence and effectively reduce the bit error rate(BER).(2)In order to obtain faster and more efficient CE precoding,we then designed another 1-bit DAC precoding method based on MMSE,which is based on the linear alternating direction multiplier method(LADMM).Its main advantage lies in using the linearized technology to replace of the quadratic term in the objective function of the alternating direction method of Multiplier's(ADMM's)subproblem,which avoids complex matrix inversion.In order to improve the convergence speed and stability of the algorithm,smoothing parameters is used to improve they.Compared with existing algorithms,it has better and similar BER performance and lower complexity.(3)Since the application of 1-bit DAC precoding algorithm to high-order modulation will cause serious BER performance loss,a multi-bit DAC precoding method is proposed for this reason.Different from 1-bit DAC precoding,multi-bit DAC precoding can make DAC generate more complex signals,which also makes the corresponding CE precoding problem more difficult to deal with.This method is still based on the LADMM algorithm.But due to its higher quantization requirements,the difficulty of solving this problem has been increased a lot.In order to solve the quantization precoding more accurately,the auxiliary matrix is used to directly add the quantization process to the algorithm.Simulation results show that multi-bit LADMM precoding scheme is obviously better than existing algorithms in terms of BER performance and complexity.