Prediction of 5G New Radio Wireless Channel Path Gains and Delays Using Machine Learning and CSI Feedback

Abstract

Next generation wireless communication systems use massive Multi Input Multi Output (m-MIMO) antenna arrays for their enhanced beamforming capabilities. Providing accurate Channel State Information (CSI) is vital for optimizing m-MIMO communication systems. The complexity of channel reconstruction grows exponentially with the number of antennas, causing traditional methods to become increasingly complicated. Machine-learning techniques can be a useful alternative for channel reconstruction using partial CSI feedback. This paper presents the results of a simulation study built using the MATLAB 5G Toolbox and a neural network trained using the simulated data. The simulator emulates a 5G channel to generate its path delays and gains, and the realistic CSI feedback. This data was used to train and test a neural network to estimate the dominant path gains and delays. The models showed promising results while operating on limited CSI data.