On Achieving Local View Capacity Via Maximal Independent Graph Scheduling

TL;DR

This paper investigates how limited network state information affects sum-rate capacity in interference networks, proposing scheduling schemes that achieve near-optimal capacity with partial knowledge.

Contribution

It formalizes the impact of local view knowledge on sum-capacity and introduces maximal independent graph scheduling to achieve normalized sum-capacity.

Findings

Maximal independent graph scheduling achieves normalized sum-capacity in many cases.

Coded set scheduling extends the benefits of local view in certain scenarios.

Characterization of normalized sum-capacity for 1- and 2-local views in various network classes.

Abstract

"If we know more, we can achieve more." This adage also applies to communication networks, where more information about the network state translates into higher sumrates. In this paper, we formalize this increase of sum-rate with increased knowledge of the network state. The knowledge of network state is measured in terms of the number of hops, h, of information available to each transmitter and is labeled as h-local view. To understand how much capacity is lost due to limited information, we propose to use the metric of normalized sum-capacity, which is the h-local view sum-capacity divided by global-view sum capacity. For the cases of one and two-local view, we characterize the normalized sum-capacity for many classes of deterministic and Gaussian interference networks. In many cases, a scheduling scheme called maximal independent graph scheduling is shown to achieve normalized sum-capacity. We also show that its generalization for 1-local view, labeled coded set scheduling, achieves normalized sum-capacity in some cases where its uncoded counterpart fails to do so.