Approximate Capacity of Gaussian Interference-Relay Networks with Weak Cross Links

Soheil Mohajer, Suhas N. Diggavi, Christina Fragouli, David N. C. TsearXiv 2010 · OpenReview

Abstract

In this paper we study a Gaussian relay-interference network, in which relay (helper) nodes are to facilitate competing information flows over a wireless network. We focus on a two-stage relay-interference network where there are weak cross-links, causing the networks to behave like a chain of Z Gaussian channels. For these Gaussian ZZ and ZS networks, we establish an approximate characterization of the rate region. The outer bounds to the capacity region are established using genie-aided techniques that yield bounds sharper than the traditional cut-set outer bounds. For the inner bound of the ZZ network, we propose a new interference management scheme, termed interference neutralization, which is implemented using structured lattice codes. This technique allows for over-the-air interference removal, without the transmitters having complete access the interfering signals. For both the ZZ and ZS networks, we establish a new network decomposition technique that (approximately) achieves the capacity region. We use insights gained from an exact characterization of the corresponding linear deterministic version of the problems, in order to establish the approximate characterization for Gaussian networks.

Introduction

In traditional wired networks, managing multiple information flows is a well-solved "multi-commodity flow" problem. However, wireless networks are far more complex because the medium is shared; signals broadcast by one transmitter interfere with others due to superposition. While the capacity of a simple one-hop interference channel has been studied for decades, adding relays—nodes that help facilitate these flows—creates a "relay-interference network" that is significantly harder to analyze. This paper addresses the challenge of determining the capacity region for two-stage networks with weak cross-links, specifically the "ZS" and "ZZ" topologies.

The Linear Deterministic Framework

To tackle the complexity of Gaussian noise, the authors employ a multi-step analytical framework starting with a linear deterministic model. This model ignores noise and discretizes signals into bit levels, allowing the authors to find exact capacity regions and gain structural insights. The paper reinforces the idea that the capacity of a noisy Gaussian network is often a "shifted" version of its deterministic counterpart, where bit-level strategies translate into power-allocation strategies.

The linear deterministic model discretizes signals into bit levels to find exact capacity regions and structural insights.

Interference Neutralization and Network Topologies

A core methodological contribution of this work is interference neutralization. Unlike interference alignment, which packs interference into a small signal space, neutralization uses relay nodes to combine signals such that they cancel each other out "over the air" before reaching the destination. To implement this in noisy Gaussian environments, the authors utilize structured lattice codes, which allow relays to process functions of messages without needing to decode individual messages first.

The Gaussian ZS network (a) and its corresponding simplified parameterization (b) used for capacity analysis.

Key Results

The authors establish an approximate characterization of the capacity region for both Gaussian ZS and ZZ networks. Key findings include:

  • 2-Bit Gap: The proposed transmission schemes achieve any rate pair within 2 bits (or less) of the theoretical outer bound, independent of channel gains or SNR.
  • Genie-Aided Outer Bounds: The paper develops new upper bounds by providing "genies" (extra information) to certain nodes, creating bounds tighter than standard information-theoretic cut-set bounds by incorporating relay decodability requirements.
  • Network Decomposition: For the ZS network, the authors show the network can be decomposed into sub-node disjoint partitions, treating the complex network as two simpler, independent unicast sessions.

Conclusion

By bridging the gap between deterministic and Gaussian models, this research provides a robust framework for managing interference in multi-hop wireless networks. While currently limited to specific ZS and ZZ topologies, the "deterministic approach" and "interference neutralization" techniques offer a path toward solving more general layered wireless networks and the K-user interference-relay problem. Future work will investigate the robustness of neutralization to channel uncertainty.

Summary updated

Abstract

In this paper we study a Gaussian relay-interference network, in which relay (helper) nodes are to facilitate competing information flows over a wireless network. We focus on a two-stage relay-interference network where there are weak cross-links, causing the networks to behave like a chain of Z Gaussian channels. For these Gaussian ZZ and ZS networks, we establish an approximate characterization of the rate region. The outer bounds to the capacity region are established using genie-aided techniques that yield bounds sharper than the traditional cut-set outer bounds. For the inner bound of the ZZ network, we propose a new interference management scheme, termed interference neutralization, which is implemented using structured lattice codes. This technique allows for over-the-air interference removal, without the transmitters having complete access the interfering signals. For both the ZZ and ZS networks, we establish a new network decomposition technique that (approximately) achieves the capacity region. We use insights gained from an exact characterization of the corresponding linear deterministic version of the problems, in order to establish the approximate characterization for Gaussian networks.
BibTeX
@misc{mohajer2010approximate,
  author = {Soheil Mohajer and Suhas N. Diggavi and Christina Fragouli and David N. C. Tse},
  title = {{Approximate Capacity of Gaussian Interference-Relay Networks with Weak Cross Links}},
  year = {2010},
  eprint = {1005.0404},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1005.0404},
}
Figure