Interference and X Networks with Noisy Cooperation and Feedback

TL;DR

This paper explores how delayed channel state information, output feedback, and full-duplex cooperation can improve the degrees of freedom in Gaussian interference and X channels, revealing new potential for increased capacity in noisy, feedback-limited environments.

Contribution

It demonstrates that full-duplex transmitter cooperation and feedback mechanisms can significantly enhance the degrees of freedom in interference and X channels with delayed CSIT.

Findings

Full-duplex cooperation improves DoF with delayed CSIT.

Output feedback yields higher DoF than delayed CSIT alone for large networks.

Combining delayed CSIT and feedback further increases DoF, approaching a maximum of 2.

Abstract

The Gaussian $K$-user interference and $M\times K$ X channels are investigated with no instantaneous channel state information (CSI) at transmitters. First, it is assumed that the CSI is fed back to all nodes after a finite delay (delayed CSIT), and furthermore, the transmitters operate in full-duplex mode, i.e., they can transmit and receive simultaneously. Achievable results are obtained on the degrees of freedom (DoF) of these channels under the above assumption. It is observed that, in contrast with no CSIT and full CSIT models, when CSIT is delayed, the achievable DoFs for both channels with full-duplex transmitter cooperation are greater than the best available achievable results on their DoF without transmitter cooperation. Our results are the first to show that the full-duplex transmitter cooperation can potentially improve the channel DoF with delayed CSIT. Then, $K$-user interference and $K\times K$ X channels are considered with output feedback, wherein the channel output of each receiver is causally fed back to its corresponding transmitter. Our achievable results with output feedback demonstrate strict DoF improvements over those with the full-duplex delayed CSIT when $K>5$ in the $K$-user interference channel and $K>2$ in the $K\times K$ X channel. Next, the combination of delayed CSIT and output feedback, known as Shannon feedback, is studied and strictly higher DoFs compared to the output feedback model are achieved in the $K$-user interference channel when K=5 or $K>6$, and in the $K\times K$ X channel when $K>2$. Although being strictly greater than 1 and increasing with size of the networks, the achievable DoFs in all the models studied in this paper approach limiting values not greater than 2.