Research On Channel Modeling And Channel Prediction In 3D MIMO System

The construction of a channel model that fits the characteristics of the real channel is the basis for research work such as algorithm testing and performance evaluation of communication systems.Traditional MIMO(Multiple Input Multiple Output)channel models mostly only use the azimuth domain of the antenna,and do not make full use of the spatial freedom.Three-dimensional MIMO(3D MIMO,Three Dimisional MIMO)utilizes the degree of freedom in the vertical direction to improve system performance.However,the increase in the number of antennas brings new features such as complex spatial structure and spatial non-stationarity,so it is necessary to incorporate these new features into the construction of the 3D MIMO channel model.In the MIMO system,the transmitter performs transmission preprocessing based on accurate channel state information(CSI,Channel State Information)is an important means to improve performance.In a massive MIMO system,in order to avoid huge CSI feedback overhead,the Time Division Duplex(TDD,Time Division Duplex)mode is often preferred.In this mode,the reciprocity of the uplink and downlink channels is used to perform precoding based on the detected uplink CSI.However,the alternate transmission of uplink and downlink brings about the problem of CSI expiration,which needs to be solved by appropriate prediction algorithms.Based on this,this paper conducts 3D MIMO channel modeling based on the new characteristics of MIMO channels,and studies channel prediction technology.The main tasks completed are as follows:1.The transmission characteristics of MIMO channels are studied,the influence of coherence characteristics and fading characteristics on wireless transmission is discussed,and several common channel modeling methods and channel modeling examples are introduced.2.A 3D MIMO channel modeling method that describes the statistical characteristics of the vertical angle domain with the von Mises distribution and considers the spatial non-stationary characteristics.Define two coordinate systems to describe the geometric spatial layout of the MIMO system.Aiming at the elevation characteristics of the 3D MIMO system,the von Mises distribution that is more in line with actual data is used to simulate the angular distribution,and the birth and death process of scatterer clusters in space is used to characterize The spatial non-stationary characteristics of large-scale antenna arrays.The channel characteristics were investigated in the time domain,space domain,and angle domain of the improved modeling scheme.The simulation results confirmed that the constructed channel model conforms to the theoretical derivation.3.Propose a real-time channel prediction technology based on recurrent neural network and its variants.Using the time-correlation characteristics of the 3D MIMO channel,based on the Recurrent Neural Network(RNN)structure,a two-step mechanism of offline training and online prediction is proposed to predict CSI in real time.Further use Long Short-Term Memory(LSTM)and Gated Recurrent Unit(GRU)to solve the problems of gradient disappearance and gradient explosion during network training.On this basis,the input sample dimensions are increased,and the multi-input single-output network is expanded to fully explore the spatial correlation characteristics of the 3D MIMO channel.The prediction network is further improved.After simulation,the performance of the traditional prediction algorithm is improved.The accuracy of the prediction proves its validity.