On mMIMO ML-Based Wireless Channel Predictor Using Measurement Data

Channel state information (CSI) acquisition is one of the fundamental problems in wireless communications. Accurate CSI is nearly impossible due to errors caused by CSI estimation, compression, feedback and processing delays. Motivated by these problems and seeing the popularity of machine learning (ML), this article proposes an ML-enabled massive multiple-input multiple-output (mMIMO) channel predictor (CP), which can work on the estimated channel and the quantized (compressed) version of the estimated channel as well. While existing work has evaluated the performance of ML algorithms by only using the artificially generated channel realizations and using only the actual channel values (not estimated or compressed). We train and test the ML algorithm using the realistic channel realizations from a measurement campaign performed at Nokia Bell-Labs, Stuttgart, Germany. The results corroborate the validity of the proposed ML-based CP, which can work both on the estimated and compressed channel.