Learning the Channel State Information Across the Frequency Division Gap in Wireless Communications

Massive multiple-input multiple-output (MIMO) is a key technology for the fifth-generation (5G) and beyond to enable high-rate and low latency communications. However, due to the large pilot sequences required for the downlink (DL) channel estimation, this solution turns out to be potentially intractable for frequency division duplex (FDD) systems. In this paper, we propose a machine learning-based approach in order to predict the DL channel-state-information (CSI) from the uplink CSI. The novel architecture consists of two autoencoders and a set of random forests. Simulation results validate the effectiveness of the proposed machine learning algorithm for the downlink channel estimation in FDD massive MIMO systems assuming a particular scenario chosen from the deepMIMO dataset.