A Fair and Scalable Power Control Scheme in Multi-cell Massive MIMO

Amin Ghazanfari, Hei Victor Cheng, Emil Björnson, Erik G. LarssonICASSP 2019 · OpenReview

TL;DR

This paper introduces a novel power control scheme for multi-cell massive MIMO systems that addresses the scalability issues of traditional max-min fairness approaches. By maximizing the geometric mean of per-cell max-min spectral efficiencies, the proposed method achieves a better balance between fairness and overall network performance, especially for weaker users, without sacrificing scalability.

What Problem Does This Paper Address?

Massive MIMO, a key technology for 5G and beyond, involves base stations with a large number of antennas to significantly boost spectral and energy efficiency. A desirable feature in cellular networks is "max-min fairness" (MMF), which aims to provide equal quality of service to all users by prioritizing those with the weakest channels. However, when applied to multi-cell massive MIMO systems, network-wide MMF (NW-MMF) faces a critical scalability problem. As the number of cells and users increases, the probability of having users with extremely poor channels (due to shadow fading) rises. This leads to a situation where the performance of the entire network is dragged down by the weakest link, potentially resulting in near-zero spectral efficiency for everyone. This paper tackles this scalability challenge, seeking a power control strategy that maintains fairness for weak users while remaining effective in large-scale deployments.

Methods

The core contribution of this paper is a new power control (PC) scheme that maximizes the geometric mean (GM) of the per-cell max-min spectral efficiencies (SE). This approach aims to overcome the scalability limitations of network-wide MMF. The authors demonstrate that this GM per-cell MMF problem can be reformulated into a convex optimization problem, allowing it to be solved efficiently to global optimality using standard convex optimization tools.

For comparison and evaluation, the paper also formulates and solves two other PC schemes:

  1. Network-wide Max-Min Fairness (NW-MMF): This is the traditional approach that aims to maximize the minimum SE across all users in the network.
  2. Network-wide Proportional Fairness (NW-PF): This scheme focuses on maximizing the sum SE of the network, offering some consideration for weaker users but without strict fairness guarantees.

The paper provides detailed mathematical formulations for these schemes, including the derivation of signal-to-interference-plus-noise ratios (SINRs) and the optimization objectives. They also prove the concavity of key functions involved in their proposed GM per-cell MMF problem, establishing its convex nature.

Datasets

Not discussed in the paper. The evaluation is based on simulated data generated according to the described system model.

Key Insights and Ideas

  • GM per-cell MMF as a Scalable Fairness Solution: The central idea is that by focusing on the geometric mean of per-cell fairness metrics, the negative impact of a single extremely weak user across the entire network is mitigated. This allows for better overall performance while still prioritizing fairness for users in challenging conditions within their respective cells.
  • Convexity is Key: Proving that the proposed GM per-cell MMF problem is convex is crucial. This guarantees that standard optimization algorithms can find the globally optimal solution efficiently, making the scheme practical for real-world deployment.
  • Balancing Act: The proposed scheme effectively balances the strict fairness of NW-MMF with the sum-rate optimization of NW-PF. It offers better performance for the weakest users than NW-PF and avoids the severe scalability issues of NW-MMF.
  • Heuristic vs. Optimal: The paper compares its optimal GM per-cell MMF approach against a previously proposed heuristic method for per-cell MMF, showing that the optimal approach can offer comparable or superior performance.

Key Results

The numerical analysis, based on simulations of a multi-cell massive MIMO setup (9 cells, 100 antennas per BS), highlights the strengths of the proposed GM per-cell MMF scheme:

  • Improved Fairness for Weak Users: In both uplink (UL) and downlink (DL) transmissions, the proposed scheme provides higher spectral efficiency (SE) for the weakest users compared to NW-PF. Specifically, the 10% weakest UL users and 12% weakest DL users benefit significantly.
  • Scalability Advantage: While NW-MMF can lead to very low SE for all users in large networks, the GM per-cell MMF scheme maintains better performance across all cells, not just the weakest one.
  • Comparable to NW-MMF in Lower Tail: The proposed scheme performs comparably to NW-MMF in the lower tail of the SE distribution (i.e., for the weakest users/cells), but significantly outperforms it for other users.
  • Sum SE Performance: The NW-PF scheme generally achieves the highest sum SE, as expected, since it approximates sum SE maximization. However, the proposed GM per-cell MMF offers a more balanced trade-off.
  • User Distribution: Table 1 shows that while NW-MMF can provide better SE for a certain percentage of users (14-18%), these gains are often "tiny." The proposed GM per-cell MMF scheme also shows instances where it provides better SE for a portion of users (17-22%), indicating its effectiveness in improving conditions for many.

Limitations and Future Work

  • Limitations:
    • The paper focuses on UL and DL data transmission and does not explicitly discuss other aspects like control signaling or interference management beyond the scope of power control.
  • The simulations are based on a specific system model and assumptions (e.g., wrap-around for edge effects, specific large-scale fading models).
  • Future Work:
    • The authors suggest that the proposed GM per-cell MMF approach provides more fairness towards weak users compared to NW-PF, implying potential for further exploration in fairness-aware resource allocation.
  • The paper's success in reformulating a complex problem into a convex one opens doors for applying similar techniques to other optimization challenges in massive MIMO.
  • While the paper addresses scalability, further investigation into the performance of the proposed scheme in extremely large-scale networks (e.g., hundreds or thousands of cells) would be valuable.

Summary updated

Abstract

This paper studies the transmit power optimization in a multi-cell massive multiple-input multiple-output (MIMO) system. To overcome the scalability issue of network-wide max-min fairness (NW-MMF), we propose a novel power control (PC) scheme. This scheme maximizes the geometric mean (GM) of the per-cell max-min spectral efficiency (SE). To solve this new optimization problem, we prove that it can be rewritten in a convex form and then solved using standard tools. To provide a fair comparison with the available utility functions in the literature, we solve the network-wide proportional fairness (NW-PF) PC as well. The NW-PF focuses on maximizing the sum SE, thereby ignoring fairness, but gives some extra attention to the weakest users. The simulation results highlight the benefits of our model which is balancing between NW-PF and NW-MMF.
BibTeX
@inproceedings{ghazanfari2019fair,
  author = {Amin Ghazanfari and Hei Victor Cheng and Emil Björnson and Erik G. Larsson},
  title = {{A Fair and Scalable Power Control Scheme in Multi-cell Massive MIMO}},
  booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
  year = {2019},
  url = {https://openreview.net/forum?id=cA4v95cjZ0},
}
Figure