[Paper Review] Interference Alignment for Line-of-Sight Channels

Leonard H. Grokop, David Tse+1 more|ArXiv.org|Sep 18, 2008

Cooperative Communication and Network Coding4 references15 citations

TL;DR

This paper proposes a time-domain interference alignment scheme for line-of-sight (LOS) K-user interference channels, demonstrating that spectral efficiency can scale linearly with K even at fixed SNR by exploiting bandwidth scaling. The approach uses a time-indexed interference graph and dynamic programming to achieve near-optimal rates, proving that bandwidth scaling is necessary for such gains in LOS environments.

ABSTRACT

The fully connected K-user interference channel is studied in a multipath environment with bandwidth W. We show that when each link consists of D physical paths, the total spectral efficiency can grow {\it linearly} with K. This result holds not merely in the limit of large transmit power P, but for any fixed P, and is therefore a stronger characterization than degrees of freedom. It is achieved via a form of interference alignment in the time domain. A caveat of this result is that W must grow with K, a phenomenon we refer to as {\it bandwidth scaling}. Our insight comes from examining channels with single path links (D=1), which we refer to as line-of-sight (LOS) links. For such channels we build a time-indexed interference graph and associate the communication problem with finding its maximal independent set. This graph has a stationarity property that we exploit to solve the problem efficiently via dynamic programming. Additionally, the interference graph enables us to demonstrate the necessity of bandwidth scaling for any scheme operating over LOS interference channels. Bandwidth scaling is then shown to also be a necessary ingredient for interference alignment in the K-user interference channel.

Motivation & Objective

• To address the limitation of degrees of freedom as a metric by characterizing spectral efficiency scaling at fixed SNR, not just in the high-SNR limit.
• To demonstrate that linear spectral efficiency scaling with K is achievable in LOS interference channels, even at fixed transmit power.
• To establish that bandwidth scaling is a necessary condition for interference alignment to achieve linear spectral efficiency growth in K-user interference channels.
• To develop a graph-theoretic framework using time-indexed interference graphs to model and solve interference alignment problems efficiently via dynamic programming.
• To prove the necessity of bandwidth scaling through a stationarity property in the interference graph, showing that without it, linear scaling is impossible.

Proposed method

• Models the K-user LOS interference channel using a time-indexed interference graph where vertices represent interfering symbols and edges represent interference constraints.
• Represents the communication problem as finding a maximal independent set (MIS) in the interference graph, which corresponds to a set of non-interfering transmissions.
• Applies dynamic programming to efficiently compute the MIS on the interference graph, leveraging its stationarity property under uniform delay distributions.
• Uses cycle graphs and chain graphs to model periodic interference patterns, enabling analysis of long-term interference alignment feasibility.
• Introduces a phase assignment scheme to cycle graphs to determine feasible transmission patterns, with conditions derived from GCD-based inequalities on path delays.
• Derives necessary and sufficient conditions for achieving a 3/2 independence rate in 3-user channels using GCD comparisons between delay parameters.

Experimental results

Research questions

• RQ1 Can spectral efficiency scale linearly with K in a K-user interference channel at fixed SNR, rather than vanishing as in orthogonal schemes?
• RQ2 Is bandwidth scaling a necessary condition for achieving linear spectral efficiency scaling in LOS interference channels?
• RQ3 Can interference alignment in the time domain achieve high spectral efficiency without relying on frequency-selective fading or frequency-domain sub-channels?
• RQ4 How can the interference alignment problem be efficiently solved in LOS channels with arbitrary delay profiles?
• RQ5 What are the precise delay conditions under which interference alignment achieves a 3/2 independence rate in a 3-user interference channel?

Key findings

• The total spectral efficiency of the K-user LOS interference channel can scale linearly with K at any fixed SNR, not just in the high-SNR limit.
• Bandwidth scaling is a necessary condition for achieving linear spectral efficiency scaling in LOS interference channels, as proven via the interference graph's stationarity and GCD-based constraints.
• For 3-user LOS channels, interference alignment achieves a 3/2 independence rate in at least 1/16 of all delay configurations, under specific GCD conditions on the path delays.
• The interference graph model enables efficient solution via dynamic programming, with the maximal independent set corresponding to a feasible transmission schedule.
• The necessary and sufficient condition for achieving a 3/2 independence rate in the 3-user case is that γ₁ < γ₂, γ₁ < γ₃, and γ₁ < γ, where γᵢ represents the 2-adic valuation of delay differences.
• The result generalizes to K-user channels: when each link has D paths, the total spectral efficiency grows linearly with K, provided that bandwidth scales with K.