Regularized Zero-Forcing Interference Alignment for the Two-Cell MIMO Interfering Broadcast Channel

Joonwoo Shin, Jaekyun Moon

Abstract

This paper proposes transceiver design strategies for the two-cell multiple-input multiple-output (MIMO) interfering broadcast channel (IBC), where both inter-cell interference (ICI) and inter-user interference (IUI) are present. We introduce a generalized zero-forcing interference alignment (ZF-IA) method and a subsequent regularization scheme based on the minimum weighted-mean-square-error (WMSE) criterion. Unlike existing weighted-sum-rate-maximizing transceivers, our approach avoids iterative weight calculations, leading to significant computational savings and faster convergence. Simulation results demonstrate that the proposed regularized ZF-IA scheme is highly effective, particularly in scenarios with limited iteration budgets.

Introduction: The Challenge of Interference in MIMO Networks

In modern wireless infrastructure, such as 5G macro-pico deployments, base stations (BS) frequently operate in close proximity, leading to significant interference. This is formally modeled as a Two-Cell MIMO Interfering Broadcast Channel (IBC). In this environment, a mobile user must contend with two distinct interference sources:

  1. Inter-User Interference (IUI): Signals intended for other users within the same cell.
  2. Inter-Cell Interference (ICI): Signals leaking from the neighboring base station.

Traditional Zero-Forcing Interference Alignment (ZF-IA) aims to perfectly nullify these interference components. However, ZF-IA is optimized for high signal-to-noise ratio (SNR) regimes and often performs poorly at lower SNRs because it ignores background noise. While more advanced "weighted-sum-rate-maximizing" (max-WSR) schemes offer better performance, they are computationally intensive, requiring numerous iterations and extensive data exchange between base stations, which limits their real-time applicability.

Proposed Solution: Regularized ZF-IA

To address these limitations, the authors propose a two-step transceiver design that balances interference suppression with signal strength and computational efficiency.

1. Generalized ZF-IA (GZF-IA)

The first step extends standard ZF-IA to support multiple data streams per user. By utilizing a "spreading matrix," the transmitter aligns interference into a specific multi-dimensional subspace. This allows the receiver to isolate and ignore the interference while preserving the desired signal.

2. Regularization via WMSE

To improve robustness against noise, the precoders and decoders are "regularized" using a Weighted Mean Square Error (WMSE) criterion. Instead of strictly zeroing out interference, this approach finds a balance that minimizes the overall error at the receiver.

3. The "One-Shot" Weight Innovation

The most significant contribution of this work is the method for calculating MSE weights. Standard WMSE algorithms update weights iteratively, which is slow. The authors discovered that near-optimal weights can be calculated once (non-iteratively) using the results from the initial GZF-IA step. This "one-shot" calculation drastically reduces the computational overhead and communication requirements between base stations.

Key Results and Analysis

The proposed Regularized ZF-IA (Reg. GZF-IA) was evaluated against the state-of-the-art max-WSR method using Monte Carlo simulations.

Rapid Convergence

In mobile networks, the wireless channel can change in milliseconds. An algorithm that requires 100 iterations to converge is often impractical. As shown in the performance analysis, Reg. GZF-IA reaches a stable sum-rate almost immediately, typically within just 1 or 2 iterations. This makes it ideal for real-time applications where the "iteration budget" is strictly limited.

Sum rate vs. number of iterations for different SNR levels.

Computational Efficiency

Beyond convergence speed, the total computational cost is a major factor for hardware implementation. The number of complex multiplications for Reg. GZF-IA grows linearly with iterations and remains consistently lower than that of the max-WSR scheme. This efficiency is primarily due to avoiding the iterative weight update process.

Computational complexity vs. number of iterations.

Conclusion

The Regularized ZF-IA scheme provides a practical and efficient solution for managing interference in two-cell MIMO networks. By combining the alignment principles of ZF-IA with a computationally light regularization step, it achieves fast convergence and high performance at low iteration counts. While the max-WSR method may eventually achieve a higher sum-rate given unlimited time, the proposed scheme's efficiency makes it a superior candidate for real-world, low-latency wireless systems.

Summary updated

Abstract

In this paper, we propose transceiver design strategies for the two-cell multiple-input multiple-output (MIMO) interfering broadcast channel where inter-cell interference (ICI) exists in addition to interuser interference (IUI). We first formulate the generalized zero-forcing interference alignment (ZF-IA) method based on the alignment of IUI and ICI in multi-dimensional subspace. We then devise a minimum weighted-mean-square-error (WMSE) method based on regularizing the precoders and decoders of the generalized ZF-IA scheme. In contrast to the existing weighted-sum-rate-maximizing transceiver, our method does not require an iterative calculation of the optimal weights. Because of this, the proposed scheme, while not designed specifically to maximize the sum rate, is computationally efficient and achieves a faster convergence compared to the known weighted-sum-rate maximizing scheme. Through analysis and simulation, we show the effectiveness of the proposed regularized ZF-IA scheme.
BibTeX
@article{shin2013regularized,
  author = {Joonwoo Shin and Jaekyun Moon},
  title = {{Regularized Zero-Forcing Interference Alignment for the Two-Cell MIMO Interfering Broadcast Channel}},
  journal = {IEEE Communications Letters},
  year = {2013},
  url = {https://openreview.net/forum?id=Twb4JnlSnpG},
}
Figure