[Paper Review] The Approximate Capacity Region of the Gaussian Y-Channel

Anas Chaaban, Aydin Sezgin|arXiv (Cornell University)|Oct 7, 2013

Cooperative Communication and Network Coding30 references3 citations

TL;DR

This paper characterizes the approximate capacity region of the Gaussian Y-channel within a constant gap, using a novel three-phase transmission strategy combining bi-directional, cyclic, and uni-directional communication. By leveraging compute-and-forward with lattice codes and network coding, the authors achieve a constant-gap approximation to the capacity region, proving that cut-set bounds are not tight even asymptotically.

ABSTRACT

A full-duplex wireless network with three users that want to establish full message-exchange via a relay is considered. Thus, the network known as the Y-channel has a total of 6 messages, 2 outgoing and 2 incoming at each user. The users are not physically connected, and thus the relay is essential for their communication. The linear-shift deterministic Y-channel is considered first, its capacity region is characterized and shown not to be given by the cut-set bounds. The capacity achieving scheme has three different components (strategies): a bi-directional, a cyclic, and a uni-directional strategy. Network coding is used to realize the bi-directional and the cyclic strategies, and thus to prove the achievability of the capacity region. The result is then extended to the Gaussian Y-channel where the capacity region is characterized within a constant gap independent of the channel parameters.

Motivation & Objective

• To resolve the open problem of characterizing the capacity region of the Gaussian Y-channel, which is not tightly bounded by cut-set bounds.
• To extend prior work on deterministic and MIMO Y-channels to the single-antenna Gaussian case, where cut-set bounds are shown to be loose.
• To develop a transmission strategy that achieves a constant-gap approximation to the capacity region using network coding and lattice codes.
• To demonstrate that the capacity region of the Gaussian Y-channel is not characterized by cut-set bounds, even in the asymptotic degrees-of-freedom sense.

Proposed method

• Analyzes the linear-shift deterministic Y-channel first to gain insights into the structure of the capacity region.
• Proposes a three-component transmission strategy: bi-directional, cyclic, and uni-directional communication, each optimized for different rate regimes.
• Uses network coding to enable efficient bi-directional and cyclic message exchange at the relay.
• Applies compute-and-forward with lattice codes to enable reliable computation of integer linear combinations at the relay, enabling interference alignment in the code domain.
• Derives power allocation schemes for uplink and downlink transmission that satisfy individual power constraints at each source.
• Establishes a constant-gap approximation to the capacity region by showing that the proposed scheme achieves all rate tuples within a fixed gap of the outer bound.

Experimental results

Research questions

• RQ1 Is the cut-set bound tight for the Gaussian Y-channel, even in the asymptotic degrees-of-freedom regime?
• RQ2 Can a constant-gap approximation of the capacity region be achieved for the Gaussian Y-channel using structured signaling?
• RQ3 What are the optimal transmission strategies—bi-directional, cyclic, or uni-directional—for different rate regimes in the Y-channel?
• RQ4 How can network coding and lattice codes be jointly used to enable efficient multi-way relaying with interference management?
• RQ5 Can the deterministic Y-channel model provide insights into the capacity region of the Gaussian Y-channel?

Key findings

• The cut-set bounds are not tight for the Gaussian Y-channel, even in the asymptotic degrees-of-freedom sense, contradicting earlier assumptions.
• The capacity region of the Gaussian Y-channel is characterized within a constant gap independent of channel parameters, using a novel three-phase transmission strategy.
• The proposed scheme achieves the outer bound within a constant gap by combining bi-directional, cyclic, and uni-directional communication phases.
• Lattice codes enable reliable computation at the relay via compute-and-forward, facilitating interference alignment in the code domain.
• Power allocation schemes are derived that satisfy individual power constraints at each source and ensure reliable decoding at the relay and destinations.
• The constant-gap approximation is robust across all rate regimes, including symmetric and asymmetric message exchange patterns.