Adaptive Hybrid Beamforming with Massive Phased Arrays in Macro-Cellular Networks

TL;DR

This paper proposes adaptive hybrid beamforming algorithms for macro-cellular networks with massive phased arrays, demonstrating significant throughput improvements by optimizing beamforming based on long-term channel info.

Contribution

It introduces novel adaptive beamforming algorithms for massive phased arrays in macro-cellular networks, including single and multi-beam scenarios, with low-complexity heuristics and gradient-based solutions.

Findings

5X network throughput improvement with massive arrays

Heuristic algorithm performs close to SDR upper-bound

Adaptive beamforming updates enhance performance over time

Abstract

Hybrid beamforming via large antenna arrays has shown a great potential for increasing data rate in cellular networks by delivering multiple data streams simultaneously. In this paper, several beamforming design algorithms are proposed based on the long-term channel information for macro-cellular environments where the base station is equipped with a massive phased array under per-antenna power constraint. Using an adaptive scheme, beamforming vectors are updated whenever the long-term channel information changes. First, the problem is studied when the base station has a single RF chain (single-beam scenario). Semi-definite relaxation (SDR) with randomization is used to solve the problem. As a second approach, a low-complexity heuristic beam composition algorithm is proposed which performs very close to the upper-bound obtained by SDR. Next, the problem is studied for a generic number of RF chains (multi-beam scenario) where the Gradient Projection method is used to obtain local solutions. Numerical results reveal that using massive antenna arrays with optimized beamforming vectors can lead to 5X network throughput improvement over systems with conventional antennas.