CSI Estimation Using Artificial Neural Network

Abstract

We propose using machine learning to estimate channel state information (CSI) for MIMO communication links. The goal is to use information such as atmospheric conditions, amount of path loss, and Doppler shift to improve the accuracy of CSI estimates. We start by designing an algorithm which estimates the CSI based on previously mentioned factors. Using this algorithm, we simulate a dataset of CSI over varying atmospheric conditions, receiver position, and receiver velocity. We then use this dataset to train an artificial neural network, which is able to estimate the CSI by using the current atmospheric condition, receiver position, and velocity.