Beyond TIN: GDoF of K-User Interference Channel With Successive Interference Cancellation and Power Control

This is the authors' abstract. We don't add key points for this paper.

While the exact capacity of large interference networks remains largely intractable, much progress has come about from alternative approaches. In particular, Generalized Degrees of Freedom (GDoF) region characterizations, combined with extremal network theory have significantly advanced the theoretical understanding of the basic scheme of treating interference as (Gaussian) noise (TIN) with power control, producing closed form solutions of the achievable GDoF region via implicit Fourier-Motzkin elimination of power-control variables, as well as extremal characterizations of the relative performance of TIN against other alternatives. The goal of this work is to develop this approach beyond TIN, especially to include successive interference cancellation (SIC). The first contribution is a closed form (i.e., depending only on the channel parameters) characterization of the achievable GDoF region for an SIC scheme for any specified decoding order at each receiver, and anyKuser network topology. This is followed by extremal network theoretic analysis, showing that the maximal sum-GDoF gain from SIC over TIN across all network topologies is precisely min(K,2L+1), when a total of up toLmessages can be successively decoded at every receiver. Remarkably, even when restricted to only weak interference settings where the benefits of SIC are not obvious, the extremal gain of SIC over TIN is shown to be significant, approachingL+1 for largeK. Networks that achieve these extremal gains are explicitly identified. The GDoF analysis is complemented with numerical simulations to illustrate how the insights translate to finite Signal-to-Noise Ratios.