Research On Precoding And Multi-objective Optimization Of Massive MIMO System

With the official commercialization of 5G mobile communication systems,5G began to provide more efficient services to the society,and the services developed by 5G gradually increased.In recent years,the rapid development of artificial intelligence technology has gradually shown strong capabilities,and traditional industries are also applying various artificial intelligence technologies to improve efficiency.Among the communication algorithms,there are many high-performance algorithms,such as optimization algorithms based on precoding in massive MIMO,but their complexity is high and it is difficult to deploy in actual systems.Therefore,how to reduce the complexity of existing algorithms has also gradually attracted attention.With its powerful learning ability,artificial intelligence technology can provide support for communication algorithms in 5G,reduce the complexity of the algorithm,and benefit practical applications.In addition,the rapid increase in 5G power consumption and the wide variety of services involved,the resource optimization of the communication system is also getting more and more attention from researchers and operators.Faced with multiple services and high demands on system performance,it is necessary to take into account the multi-objective service quality and resource allocation,and the overall service quality and system energy efficiency.Multi-objective optimization technology has also received extensive attention from researchers in the overall optimization of the system.Among them,solving the multi-objective optimal compromise problem and quickly exploring the multi-objective Pareto boundary have become the research focus.This paper mainly studies the precoding and multi-objective optimization of massive MIMO system.First,the paper introduces the theoretical basis of mobile communication system modeling,such as the characteristics of wireless channels,including statistical models of large-scale fading and small-scale fading,in preparation for channel modeling in this paper.It also introduces massive MIMO technology,which is mainly divided into centralized massive MIMO system and distributed massive MIMO system,as well as commonly used precoding technology in the system.Through simulation,the spectral efficiency and energy efficiency of two types of commonly used precoding under different systems are compared.Next,based on the existing sparse precoding algorithm in the distributed massive MIMO system,the paper proposes to use deep learning to learn the features of the sparse precoding results.The RAU is preclustered by the output of the neural network to exclude the RAU that is impossible to activate,reduce the variables in the sparse precoding,and reduce the algorithm complexity.In this paper,the algorithm structure of RAU pre-clustering based on deep learning is given,and the training results are given through simulation,the feasibility of deep learning pre-clustering is verified,and the results and optimization directions are analyzed.Then the thesis studies multi-objective optimization problems in full-duplex distributed massive MIMO system.Considering that complex communication systems not only serve users with one demand,only optimizing a single target will cause performance degradation of other systems.Using the method based on weighted Chebyshev,it is possible to simultaneously optimize the two conflicting objectives of minimizing RAU transmission power and minimizing user transmission power.The performance differences between the full-duplex distributed massive MIMO system and the full-duplex centralized massive MIMO system and the changes caused by different antenna numbers are analyzed through simulation.Finally,using the latest open source genetic algorithm framework,the multi-objective optimization problem of centralized massive MIMO system is studied.Complex mobile communication systems can no longer consider only a single optimization goal,but should be considered as a whole to consider the simultaneous optimization of multiple goals.As a mature and efficient algorithm,genetic algorithm has been applied in different research fields.And there is a specific algorithm for multi-objective optimization,which can quickly give a multi-objective Pareto boundary.We choose NSGA-? and NSGA-? two multi-objective optimization algorithms as the problem solving algorithm.At the same time,the fuzzy logic inference system is introduced,and the framework for resource allocation in multi-objective optimization combined with fuzzy logic inference is given.Finally,the performance of NSGA-? and NSGA-? are studied through simulation.