On the Implementation and Placement of Hybrid Beamforming for Single and Multiple Users in the Massive-MIMO MmWave Systems

Abstract

This paper discusses the issue of implementing the hybrid beamforming functionality using deep learning techniques for both single users and multiple users massive MIMO mmWave (SU-mMIMO and MU-mMIMO) systems. First, the DeepMIMO dataset is used to collect the location data of a grid of user locations in a street environment. Then, the collected dataset is used to build and optimize the direct precoding (and combining) architectures for both the transmitter (the base station (BS)) and the receiver (the user equipment (UE)), respectively. Different assumptions about the previous systems are discussed and the placement and implementation of the hybrid beamforming in the 5G and beyond systems is discussed in details with some realistic calculations and recommendations. The second part of the paper goes a step further in deriving the expected delay of beamforming when it is done on premise, in a dedicated core network, and in a cloud core networks and suggest the best place to implement the beamforming functionality for both static and mobile users using realistic parameters and calculations.