Research On 5G Massive MIMO Channel Features And Downlink Interference Level Prediction Based On Deep Learning

With the application and development of MIMO technology,Massive MIMO technology has become the focus of attention as a key technology for 5G mobile communications.Due to the large number of antennas deployed in the network and high complexity,the performance characteristics estimation and analysis of Massive MIMO systems need to consider the spatial transmission characteristics between the base station and the user and the interference between the users.These characteristics are hidden in the calculation of massive network models.Traditional system simulation methods alone cannot effectively predict and evaluate network performance effectively in actual network planning.Therefore,this paper uses deep learning to research and predict the two aspects of Massive MIMO channel characteristics and downlink interference level.At the same time,this paper collected the ray tracing model data of the macro base station coverage scene in the real scene of the 5G network,and then fused the ray tracing model data with the 3D MIMO channel model to analyze the channel spatial characteristics and user space characteristics.Finally,using deep learning to study channel features and downlink interference prediction.The main research contents of this paper include:1.Based on the existing ray tracing data,which includes the geometric position of the ray between the user and the base station and the large-scale path loss and delay information of each ray,which is fused with the traditional 3D MIMO channel model implementation standard.This fusion channel model is a new model that combines deterministic and statistical models.2.According to the new fusion channel model,the channel matrix is calculated and the downlink interference level is calculated.By analyzing the spatial characteristics between the base station and the user and the characteristics between users,the input and output characteristics in the deep learning network model are modeled and analyzed,and creating your own network training dataset.3.Research and analyze the neural network model methods commonly used in deep learning.Based on the obtained data sets,using deep learning methods to predict and analyze Massive MIMO channel features and downlink interference level.Through the prediction result analysis,the model is continuously trained and optimized to determine the appropriate input,output and deep learning network models.In this paper,the ray tracing data and 3D MIMO channel model are combined into a new model to calculate the channel features and downlink interference level.Compared with the traditional statistical model,the fusion model can better reflect the authenticity and effectiveness of the MIMO network.Based on the analysis results obtained by the fusion model,the specific input vector and output vector of the deep learning model is designed.A method for predicting the channel features and downlink interference level based on deep learning is proposed.Compared with the fusion model and traditional statistical model,the deep learning method proposed in this paper can quickly predict the grid-level network performance.In actual network planning and optimization,this method can reduce the cost of network planning and optimization and reduce the time overhead caused by high network complexity.