Inner and Outer Bounds for the Gaussian Cognitive Interference Channel and New Capacity Results

TL;DR

This paper advances understanding of the Gaussian cognitive interference channel by deriving new bounds, capacity results in specific regimes, and providing performance guarantees and comparisons of achievable schemes.

Contribution

It introduces new outer bounds, extends capacity results to certain regimes, and establishes constant-gap performance guarantees for the Gaussian cognitive interference channel.

Findings

New outer bounds based on broadcast channel ideas

Capacity results for 'primary decodes cognitive' and 'S-channel' regimes

Capacity within one bit/s/Hz or a factor of two for general channels

Abstract

The capacity of the Gaussian cognitive interference channel, a variation of the classical two-user interference channel where one of the transmitters (referred to as cognitive) has knowledge of both messages, is known in several parameter regimes but remains unknown in general. In this paper we provide a comparative overview of this channel model as we proceed through our contributions: we present a new outer bound based on the idea of a broadcast channel with degraded message sets, and another series of outer bounds obtained by transforming the cognitive channel into channels with known capacity. We specialize the largest known inner bound derived for the discrete memoryless channel to the Gaussian noise channel and present several simplified schemes evaluated for Gaussian inputs in closed form which we use to prove a number of results. These include a new set of capacity results for the a) "primary decodes cognitive" regime, a subset of the "strong interference" regime that is not included in the "very strong interference" regime for which capacity was known, and for the b) "S-channel" in which the primary transmitter does not interfere with the cognitive receiver. Next, for a general Gaussian cognitive interference channel, we determine the capacity to within one bit/s/Hz and to within a factor two regardless of channel parameters, thus establishing rate performance guarantees at high and low SNR, respectively. We also show how different simplified transmission schemes achieve a constant gap between inner and outer bound for specific channels. Finally, we numerically evaluate and compare the various simplified achievable rate regions and outer bounds in parameter regimes where capacity is unknown, leading to further insight on the capacity region of the Gaussian cognitive interference channel.