Joint interference alignment and bi-directional scheduling for MIMO two-way multi-link networks

Ali Mohammad Fouladgar, Osvaldo Simeone, Onur Sahin, Petar Popovski, Shlomo ShamaiICC 2015 · DOI

TL;DR

This paper tackles interference in two-way wireless communication networks by jointly optimizing the transmission direction scheduling ("interference spin") and the linear transceiver designs. They propose algorithms that significantly reduce interference leakage, leading to improved network performance, with complexity far lower than exhaustive search.

What Problem Does This Paper Address?

In modern wireless systems, especially cellular and Device-to-Device (D2D) networks, two-way communication (where data flows in both directions) is ubiquitous. A major challenge in these systems is interference, particularly when multiple links operate simultaneously. Traditional approaches often fix the uplink and downlink transmission schedules first and then optimize the physical layer parameters. However, recent work has shown that dynamically adjusting the order of uplink and downlink transmissions within a frame, known as "bi-directional scheduling" or "interference spin," can significantly reduce interference. This paper extends this idea to complex MIMO (Multiple-Input Multiple-Output) systems with multiple interfering two-way links, aiming to jointly optimize both the transmission direction scheduling and the linear precoding/equalization strategies to minimize interference leakage.

Methods

The core of the paper lies in proposing algorithms for optimizing the "interference spin" (a binary choice for each link: transmit left-to-right or right-to-left in a given time slot) and the linear transceiver matrices (precoders and equalizers).

  • Interference Leakage Minimization (ILM): The authors adopt an ILM criterion, aiming to minimize the total interference power. This is a common approach for interference alignment.
  • Separate Optimization: They first propose optimizing the interference spins and then, using the determined spins, applying a standard ILM algorithm (from prior work) to find the optimal linear transceivers. This is done separately for uplink and downlink phases.
  • Joint Optimization: To overcome the limitations of separate optimization, they introduce a novel algorithm that jointly optimizes spins and transceivers. This algorithm uses a local search strategy to explore the vast space of possible spin configurations, guided by the interference leakage. Instead of exhaustively checking all $2^K$ spin combinations, it iteratively flips spins that contribute most to interference, aiming to find a near-optimal solution with much lower complexity.
  • Channel Model: The system uses a realistic channel model that includes path loss, log-normal shadowing, and fast fading, with MIMO antennas at both ends of each link.

Datasets

Not discussed in the paper. The evaluation is performed using simulated network scenarios with randomly located links.

Key Insights and Ideas

  • Interference Spin Matters: The choice of transmission direction order (interference spin) has a profound impact on the effectiveness of MIMO techniques for interference mitigation. The optimal transceiver design is highly dependent on the chosen spins.
  • Joint Optimization is Key: Simply optimizing spins and then transceivers separately is suboptimal. Jointly optimizing both aspects leads to significantly better interference reduction.
  • Complexity vs. Performance Trade-off: Exhaustive search for optimal spins is computationally infeasible for more than a few links ($2^K$ complexity). The proposed joint optimization algorithm offers a practical solution, achieving performance close to exhaustive search with substantially lower complexity.
  • Transmitted vs. Received Interference: The paper explores two criteria for guiding the spin flipping in the joint optimization: minimizing transmitted interference power and minimizing received interference power. The former appears to be more effective in practice.
  • Long-term CSI Sufficiency: For certain aspects (like optimizing spins based on long-term fading), using only long-term CSI (path loss, shadowing) is sufficient and significantly reduces the need for frequent updates, making the system more practical.

Key Results

  • The proposed joint optimization algorithm achieves interference leakage performance very close to that of exhaustive search, but with significantly lower complexity.
  • Compared to random spin selection, the joint optimization strategy can reduce interference leakage by over 10dB.
  • The proposed method significantly outperforms separate optimization schemes in terms of interference leakage.
  • While interference leakage reduction is substantial, the resulting sum-rate gains are more modest, suggesting that separate optimization might be sufficient to capture most of the sum-rate benefits.
  • The performance gains are observed even with a limited number of comparison steps in the joint optimization algorithm (around 10 steps for a system with 5 links).

Limitations and Future Work

  • Complexity of Joint Optimization: While significantly better than exhaustive search, the joint optimization still involves iterative steps and requires full CSI.
  • Sum-Rate vs. Leakage: The paper notes that the significant gains in interference leakage do not always translate directly into proportional sum-rate gains, indicating that other factors might become limiting.
  • Future Work: The authors suggest investigating a distributed version of the proposed scheme, which would be more practical for large-scale networks. They also mention that other performance criteria, like maximizing SINR, could be explored.

Summary updated

BibTeX
@inproceedings{fouladgar2015joint,
  author = {Ali Mohammad Fouladgar and Osvaldo Simeone and Onur Sahin and Petar Popovski and Shlomo Shamai},
  title = {{Joint interference alignment and bi-directional scheduling for MIMO two-way multi-link networks}},
  booktitle = {ICC},
  year = {2015},
  doi = {10.1109/icc.2015.7248970},
}
Figure