Precoding Optimization Algorithm In Downlink Massive MIMO System

With the explosive growth of intelligent devices in mobile communication systems,people put forward requirements for the higher demands of performance in mobile communication systems.Traditional 4G cellular systems are difficult to meet the growing business needs of users,so the 5G communication system emerges at the historic moment.As one of the key technologies of 5G,Massive MIMO precoding technology can effectively reduce the transmission power and system complexity,meet the user's demand for high-rate data transmission and respond to the national call for "green communication".It plays an important role in the field of mobile communications.Therefore,it is significant to improve the performance of 5G systems by optimizing precoding algorithms in Massive MIMO systems,and it has become a research hotspot of 5G communication systems.For solving the problems of high complexity of precoding and complicated in linear matrix inversion in single-cell multi-user Massive MIMO system,this paper proposes a lowcomplexity precoding based on SSOR-PCG with the goal of reducing system complexity.Based on the conjugate gradient algorithm,the symmetric stepwise super-relaxation splitting algorithm is used to preprocess the matrix for reducing its condition number,which can reduce system complexity and accelerate algorithm convergence rate.The simulation results show that the SSOR-PCG precoding algorithm proposed in this paper can effectively avoid matrix inversion,reduce system complexity by about an order of magnitude,has fast convergence speed and approximately optimal system capacity performance.We propose a density-based clustering two-stage precoding scheme to solve the problems of non-channel reciprocity and high complexity of precoding of downlink channels in an FDD Massive MIMO system.The algorithm uses the density clustering algorithm to group users,classifies users with similar channel covariance matrices,and then uses traditional ZF linear precoding algorithm to preprocess users among groups and users within groups,thereby reducing system overhead.The simulation results demonstrate that the two-stage precoding algorithm proposed in this paper is not easy to fall into a local optimal solution,and it can obtain more excellent system performance while reducing system complexity.When the signal-tonoise ratio is 35 d B,compared with the ZF precoding algorithm,the proposed two-stage precoding algorithm increases the system reachable sum rate about 26.30%,and increases the system capacity performance by approximately 73.21%.