Research On Massive MIMO Channel Estimation Based On Low-rank Approximation

Massive multiple-input multiple-output(Massive MIMO)technology is one of the hotspots in 5G communication research,which deploys a large number of antennas on the base station side.It has many advantages in cellular communications,including increased system throughput,communication reliability,and spectral efficiency.In order to realize these advantages,it is necessary to obtain accurate channel state information(CSI).However,pilot overhead limits the performance of massive MIMO systems during channel estimation.Based on this,this thesis studies channel estimation algorithms based on low-rank matrix completion and low-rank tensor completion.The purpose is to obtain CSI accurately and efficiently with reduced system training overhead and computational complexity.The main research contents are as follows:Firstly,this thesis studies the channel estimation algorithm based on low-rank matrix completion.Inspired by the small rank of the channel matrix under the physical finite scattering channel model,a channel estimation method based on low rank matrix approximation is proposed.The rank function of the channel matrix is expressed as the non-convex γ-norm to increase the accuracy of rank approximation of channel matrix,and the alternating direction algorithm of multipliers is adopted to solve the channel matrix,which can reduce the training overhead greatly with fewer channel parameters.Moreover,during the iteration,the gradient descent method is adopted to avoid the matrix inverse operation.This effectively reduces the computational complexity.The accuracy of channel estimation is also improved by selecting a smaller iteration step.The simulation results show that the proposed algorithm has better channel estimation performance than other low-rank matrix recovery algorithms.Secondly,this thesis studies the 3D channel estimation problem based on tensors.In order to further improve the channel estimation performance of the massive MIMO system,a uniform planar antenna array is considered on the base station,which can improve the spatial freedom in the vertical direction,and enhance the system performance effectively.Since the antenna structure of the base station has changed,the three-dimensional channel model is first established.Then,the 3D channel tensor is unfolded by using the tensor as a mathematical tool to match the 2D training sequence.Moreover,exploiting the low-rank nature of channel tensors in massive MIMO systems,the Tucker rank of the channel tensor is expressed as the non-convex Geman function to improve the accuracy of rank approximation of channel tensor.Finally,the proximal linearized minimization algorithm is used to solve the channel tensors,and the sub-problems are solved iteratively with the alternating direction algorithm of multipliers.Simulation results show that compared with other algorithms,the proposed algorithm has better channel estimation performance under the condition of low training overhead.