A 5G Framework for User Distribution Aided Beamforming and Iterative Traffic Sensing

In fifth generation (5G) cellular systems with massive MIMO, coherent hybrid beamforming with phased arrays can be used to achieve high beamforming gain with low system complexity. Typically, hybrid beamforming systems require beam training between the base station (BS) and mobile users to optimize the beamforming coefficients, which causes extra overhead. For avoiding this overhead, long-term beamforming methods are considered to design beams that do not frequently change to serve a specific region over a long period. A framework is hence necessary to acquire the user distribution and apply it for beam design. However, in the absence of beam training, standard localization schemes usually require external assistance or incur extra measurement cost. We propose a framework called Beamforming with Iterative Traffic Sensing (BITS), which alternately infers the active user distribution based on cell-level sum network spectral efficiency and optimizes the beams accordingly. The sum network spectral efficiency can be readily measured during data transmission periods with no extra communication overhead. Simulation results show quick and significant performance improvement when we infer the user distribution with sparsity using Key Performance Indicators (KPIs) that are available to the BS and iteratively improve the beamforming.