Hybrid Beamforming for Multi-User Millimeter-Wave Networks

Ali Arshad Nasir, Hoang Duong Tuan, Trung Q. Duong, H. Vincent Poor, Lajos HanzoIEEE Trans. Vehicular Technology 2020 · DOI

Abstract

This paper introduces a high-performance hybrid beamforming strategy designed for millimeter-wave (mmWave) networks. By integrating a novel "normalized" regularized zero-forcing (RZF) baseband precoder with fractional-time scheduling, the proposed framework enables a base station to effectively serve up to twice as many users as it has RF chains. Our approach achieves data rates that closely approach the theoretical limits of fully digital systems while maintaining computational efficiency and robustness against user overload.

Introduction

Millimeter-wave (mmWave) communication is a critical enabler for 5G and future wireless standards due to its significant available bandwidth. To overcome the high pathloss inherent at these frequencies, large antenna arrays are employed to perform directional beamforming. In an ideal scenario, a "fully digital" architecture—where each antenna is driven by its own Radio Frequency (RF) chain—would provide optimal signal control. However, the high cost and power consumption of RF chains make this impractical for large-scale deployments.

Hybrid beamforming offers a solution by using a limited number of RF chains to drive a large number of antennas via analog phase shifters. Despite its promise, hybrid beamforming faces several challenges:

  1. Hardware Constraints: Analog phase shifters are limited to phase adjustments (unit-modulus constraint), making optimization non-convex and NP-hard.
  2. User Overload: Performance typically degrades when the number of users ($N_u$) exceeds the number of RF chains ($N_{RF}$).
  3. Inter-cell Interference: Edge users in multi-cell environments suffer from significant interference from adjacent base stations.

Methodology

The authors propose a tiered beamforming strategy based on the user-to-RF chain ratio.

Analog and Digital Precoding

The analog beamformer is designed using a simple phase-matching approach, aligning the beamformer's phase with the conjugate of the channel phase to maximize power in the virtual channel. For the digital stage, the authors developed a "normalized" Regularized Zero-Forcing (RZF) precoder. Unlike standard ZF, which becomes unstable when $N_u \approx N_{RF}$, this new class of RZF regularizes the normalized version of the effective channel, enhancing stability and interference suppression.

Fractional-Time Scheduling

To handle "overloaded" scenarios where $N_u > N_{RF}$, the system employs fractional-time scheduling rather than complex Non-Orthogonal Multiple Access (NOMA). Users are categorized into "cell-center" and "cell-edge" groups and served during different time fractions. This approach simplifies implementation while effectively managing interference.

Optimization via Path-Following Algorithms

The resulting max-min rate optimization problems are non-convex. The authors developed iterative path-following algorithms that approximate complex rate functions with concave surrogates, ensuring rapid convergence to local optima.

Simulation and Results

The framework was evaluated using a standard Geometric Channel Model at 28 GHz, accounting for scattering clusters and pathloss. The simulations utilized Uniform Linear Array (ULA) configurations with users randomly distributed within a 200-meter radius.

Convergence and Performance

A key result of the proposed path-following algorithms is their computational efficiency. As illustrated in Figure 2, the algorithms typically reach convergence within 5 to 7 iterations, making them suitable for real-time applications.

Figure 2: Convergence of the proposed path-following algorithm across different antenna array sizes (N).

Furthermore, the proposed hybrid beamforming strategy demonstrates performance that is nearly indistinguishable from fully digital systems. Figure 5 shows the max-min rate as a function of the transmit power budget ($P_T$). The hybrid algorithms closely track the performance of the fully digital (FD) benchmarks across all power levels.

Figure 5: Max-min rate performance of the proposed hybrid beamforming algorithms compared to the fully digital (FD) benchmark across various transmit power budgets.

Key Insights

  • Negotiable Bottlenecks: Limited RF chains do not substantially erode channel capacity if the digital precoder is designed to handle the effective channel's condition.
  • Interference Mitigation: In multi-cell scenarios, scheduling cell-edge users in one cell while serving cell-center users in another at lower power creates a "sweet spot" for interference suppression without complex backhaul coordination.

Conclusion

This research demonstrates that high-performance multi-user mmWave communication is achievable with limited hardware resources. By combining normalized RZF with intelligent scheduling and efficient iterative algorithms, the proposed system approaches the performance of fully digital architectures. Future work will explore the impact of imperfect channel state information (CSI) and hardware impairments, as well as extensions into energy efficiency and physical layer security.

Summary updated

BibTeX
@article{nasir2020hybrid,
  author = {Ali Arshad Nasir and Hoang Duong Tuan and Trung Q. Duong and H. Vincent Poor and Lajos Hanzo},
  title = {{Hybrid Beamforming for Multi-User Millimeter-Wave Networks}},
  journal = {IEEE Trans. Vehicular Technology},
  year = {2020},
  doi = {10.1109/tvt.2020.2966122},
}
Figure