Randomly-Directional Beamforming in Millimeter-Wave Multi-User MIMO Downlink

Gilwon Lee, Youngchul Sung, Junyeong Seo

Abstract

In millimeter-wave (mm-wave) communication systems, the acquisition of accurate Channel State Information (CSI) is a significant bottleneck due to the massive number of antenna elements and the resulting training overhead. This paper investigates Randomly-Directional Beamforming (RDB) as a low-complexity alternative. By utilizing the unique physical properties of mm-wave channels—specifically their sparsity and directionality—the study demonstrates that RDB can achieve sum-rate scaling that is nearly linear with the number of antennas, provided the number of users also grows linearly. This finding marks a significant departure from conventional microwave systems, where random beamforming requires an exponential number of users to achieve similar gains.

1. Introduction

The transition to millimeter-wave (mm-wave) frequencies (30–300 GHz) is a cornerstone of 5G and future 6G networks, offering vast bandwidth. However, to overcome high path loss, Base Stations (BS) must employ massive MIMO arrays to create highly directive beams. Traditionally, these beams are steered based on precise Channel State Information (CSI). In the mm-wave regime, the number of antennas ($M$) is very large, making the training and feedback required for CSI estimation prohibitively expensive.

This research explores a provocative question: What if the BS does not estimate the channel at all? The authors investigate Randomly-Directional Beamforming (RDB), where the BS selects beamforming vectors randomly. The performance of such a system relies on Multi-User Diversity (MUD)—the probability that at least one user in a large pool is aligned with a random beam.

2. System Model and Methodology

The researchers employ asymptotic analysis, letting the number of antennas ($M$) tend toward infinity to uncover fundamental scaling laws. They move away from the standard i.i.d. Rayleigh fading models typical of microwave analysis, which assume a rich scattering environment, in favor of a model more representative of mm-wave physics.

2.1 The UR-LoS Channel Model

In mm-wave environments, the signal is often dominated by a single direct path. The authors utilize the Uniform Random Line-of-Sight (UR-LoS) channel model, which characterizes the channel as a single path with a random direction and gain. This sparsity is a fundamental physical property that distinguishes mm-wave from lower frequencies.

2.2 Randomly-Directional Beamforming (RDB)

Under the RDB protocol, the BS generates random beamforming vectors. Users measure their received signal strength or Signal-to-Interference-plus-Noise Ratio (SINR) for these beams and feed back only their best "score" and the index of the corresponding beam. The BS then schedules users based on these reports, eliminating the need for explicit downlink channel training.

2.3 Fractional Rate Order (FRO)

To evaluate performance, the authors introduce the Fractional Rate Order (FRO), denoted as $\gamma$. This metric measures the ratio of the log-rate to the log-number of antennas: $$\gamma = \lim_{M \to \infty} \frac{\log R}{\log M}$$ This allows for a clear comparison of how different beamforming strategies scale as the array size grows.

3. Key Insights and Scaling Laws

The analysis reveals several critical properties of RDB in the mm-wave context that challenge traditional assumptions about random beamforming.

3.1 The $\sqrt{M}$ Phase Transition

A pivotal discovery is the existence of a "phase transition" related to the number of users ($K$). For a single random beam, if $K < \sqrt{M}$, the achievable rate converges to zero as $M$ increases. However, if $K > \sqrt{M}$, the system achieves a non-trivial, positive rate. This $\sqrt{M}$ threshold defines the minimum user density required to make random beamforming viable.

3.2 Directionality as a Feature

In traditional microwave frequencies, the "search space" for a good channel is isotropic and massive, making random beams highly inefficient. In mm-wave, the high directionality—mathematically represented by the Fejér kernel—concentrates the beam's energy. This concentration actually makes it easier for a random beam to find a "good enough" user, provided the user distribution is sufficiently dense.

3.3 Linear Sum-Rate Scaling

The most significant result is the scaling law for the sum rate. If the number of users $K$ scales linearly with $M$, the RDB scheme achieves near-linear sum-rate scaling. In contrast, in a Rayleigh fading environment, $K$ would need to grow exponentially with $M$ to achieve the same effect. This suggests that in dense mm-wave deployments, RDB is a highly viable, low-overhead strategy.

4. Results and Validation

The theoretical proofs were validated using Monte Carlo simulations with 5,000 realizations. The results confirmed that as $M$ increases from $10^2$ to $10^4$, the numerical performance curves converge to the predicted FRO lines.

  • Single-Beam Scaling: If $K = M^q$, the RDB scheme achieves a $2q - 1$ fraction of the optimal CSI-based rate.
  • Multi-Beam Gains: By using $S$ multiple random beams, the BS can serve multiple users simultaneously. The analysis shows the sum rate behaves as $\Theta(M^\ell \log M^{2q-1-\ell})$, allowing for massive multiplexing gains without explicit channel training.

5. Discussion and Limitations

While the results are promising, the authors identify several areas for further research:

  • Multipath Effects: The UR-LoS model is a simplification. Real-world mm-wave channels may include a few Non-Line-of-Sight (NLoS) paths, and extending this analysis to multipath models is a necessary next step.
  • Feedback Overhead: Although downlink training is eliminated, the feedback of SINR values from thousands of users could become a bottleneck for the uplink control channel.
  • Hardware Constraints: Practical mm-wave systems often use hybrid analog/digital architectures, which might impose additional constraints on the generation of random beams.

6. Conclusion

This paper demonstrates that the unique physics of mm-wave propagation can be turned into a strategic advantage. By using Randomly-Directional Beamforming, network operators can bypass the "CSI bottleneck" in dense multi-user environments, achieving high spectral efficiency and near-linear scaling with minimal overhead.

Summary updated

BibTeX
@article{lee2016randomly,
  author = {Gilwon Lee and Youngchul Sung and Junyeong Seo},
  title = {{Randomly-Directional Beamforming in Millimeter-Wave Multi-User MIMO Downlink}},
  journal = {IEEE Trans. Wireless Communications},
  year = {2016},
  doi = {10.1109/twc.2015.2483493},
}
Figure