Interference Empowered 5G Networks

Matthieu De Mari, Emilio Calvanese Strinati, Mérouane Debbah, Zdenek Becvar5GU 2014 · DOI

Abstract

In future 5G networks, in-band interference is perceived as one of the most critical performance bottlenecks. While current solutions classically treat interference as an additional source of noise, recent advances in information theory show that interference is not necessarily an opponent, but might be empowered through intelligent management. This paper proposes a paradigm shift for 5G: instead of treating interference as a hurdle to be avoided, networks should "empower" it through intelligent management. By simultaneously optimizing user-to-access-point assignments, pairing up interfering users, and selecting the best interference-handling regime, the authors demonstrate a nearly 30% boost in spectral efficiency compared to traditional methods.

The Challenge of Interference

In traditional wireless networks, interference is the enemy. Radio Resource Management (RRM) typically handles it in two ways: Orthogonalization (giving users different "lanes" in time or frequency so they don't collide) or treating it as Noise (ignoring it and hoping the signal is strong enough to punch through).

System model for a 2-AP, 2-UE interference scenario.

Both approaches are spectrally inefficient. Orthogonalization wastes bandwidth, and treating interference as noise limits data rates. However, information theory suggests a third way: if an interfering signal is strong enough, a receiver can actually decode it, subtract it from the total received signal, and then decode its own intended message with much higher clarity. This is called Successive Interference Cancellation (SIC). The challenge addressed here is how to coordinate a complex network of many users and Access Points (APs) to decide who should interfere with whom and which technique (Noise vs. SIC) they should use to maximize the total data throughput.

Optimization Framework

The authors propose a threefold optimization framework to manage this complexity:

  1. AP-UE Association: Deciding which user (UE) connects to which Access Point. Interestingly, the best AP isn't always the one with the strongest signal; sometimes, a different connection allows for better interference cancellation later.
  2. User Matching: Grouping users who will share the same spectral resources.
  3. Interference Regime Selection: For every pair of users, the system chooses between three regimes: Noisy (weak interference), SIC (strong interference), or Orthogonalization (avoidance).

For a scenario with two APs, the authors map the problem to Graph Theory. They treat each user as a node in a complete graph, where the "weight" of an edge between two users represents the maximum spectral efficiency they can achieve together. They then solve for the Maximum Weight Disjoint Edges Matching using Edmonds’s algorithm, which efficiently finds the best possible pairings for the entire network.

Simulation Environment

The paper relies on Monte-Carlo simulations with 1,000 independent network snapshots. They use standard 3GPP-compliant models for path loss ($L = 131.1 + 42.8\log_{10}(d)$), log-normal shadowing (10 dB), and antenna gain (10 dBi). The environment consists of a 1km distance between APs with 50 unassigned users distributed uniformly.

Performance Gains

The authors compared five Interference Management Strategies (IMS). Using IMS 2 (Random matching/Noisy regime) as the baseline, the results demonstrate significant improvements across the board.

Empirical PDF of average spectral efficiency for various interference management strategies (IMS).
  • Interference Classification Gain: Simply choosing the right regime (Noisy vs. SIC) improved performance by 11.1%.
  • Matching Gain: Using Edmonds's algorithm to pair the "right" interferers together added another 16.8%.
  • Total Optimization Gain: When combining optimal AP assignment, matching, and regime selection (IMS 5), the system achieved a 28.0% average improvement in spectral efficiency over the baseline.
  • Orthogonalization Failure: IMS 1 (standard orthogonalization) was confirmed to be the least efficient, trailing far behind all interference-aware strategies.

Future Directions

While the 2-AP case is solved efficiently with graph theory, the authors acknowledge that as the number of APs ($M$) increases, the problem becomes NP-Hard. Future work aims to explore "Joint Decoding" or "$k$-SIC," where more than two users are matched together. Additionally, the current model assumes perfect channel knowledge and does not yet incorporate multi-antenna beamforming, which are key areas for practical 5G implementation.

Summary updated

BibTeX
@inproceedings{mari2014interference,
  author = {Matthieu De Mari and Emilio Calvanese Strinati and Mérouane Debbah and Zdenek Becvar},
  title = {{Interference Empowered 5G Networks}},
  booktitle = {5GU},
  year = {2014},
  doi = {10.4108/icst.5gu.2014.258227},
}
Figure