Interference Channels With Half-Duplex Source Cooperation

Rui Wu, Vinod M. Prabhakaran, Pramod Viswanath, Yi WangIEEE Trans. Information Theory 2014 · DOI

TL;DR

This paper investigates how two sources can cooperate to improve communication in a wireless interference channel, even when they can only transmit or receive at any given time (half-duplex). The authors develop new coding schemes and characterize the achievable communication rates, showing that cooperation can significantly boost performance, especially in certain channel conditions.

What Problem Does This Paper Address?

In wireless communication, multiple users sharing the same medium inevitably cause interference. While traditional approaches focus on mitigating this interference, this paper explores how cooperation between the interfering sources themselves can be leveraged to improve overall system performance. The challenge lies in the "half-duplex" constraint, meaning each source can either transmit or listen, but not both simultaneously. This limits how they can exchange information to coordinate their transmissions. The authors aim to understand the fundamental limits of communication in such a scenario, particularly for the sum capacity (total data rate) and for scenarios resembling cognitive radio, where a secondary user tries to communicate without significantly impacting a primary user.

Methods

The paper employs a combination of theoretical tools from information theory. A key technique is the use of a "linear deterministic model" (LDM) as a simplified, yet insightful, representation of the wireless channel. This LDM allows the authors to gain intuition about coding schemes before applying them to the more complex Gaussian channel.

The core coding scheme is a generalization of superposition coding, a standard technique for interference channels. They introduce "cooperative private messages" and "pre-shared public messages" to facilitate cooperation over the limited half-duplex links. The strategy involves operating the channel in different "modes" (both transmit, one transmits/one receives) and carefully scheduling these modes over time to effectively create rate-limited "bit-pipes" that mimic noiseless communication links between sources and destinations. This transforms the original problem into a "virtual channel" problem, which is then solved.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Cooperation is beneficial, even with half-duplex constraints: The paper demonstrates that even with the limitation of only transmitting or receiving at a time, sources can still cooperate effectively to improve communication rates.
  • Scheduling is crucial: The performance gains are highly dependent on how the sources schedule their transmissions and receptions across different modes.
  • LDM provides valuable intuition: The linear deterministic model simplifies the problem, allowing for a clearer understanding of the coding strategies before tackling the more complex Gaussian channel.
  • Cognitive radio insights: The study of unidirectional cooperation provides a framework for understanding how secondary users can operate in the presence of primary users, with implications for spectrum sharing.
  • Channel alignment matters: The degree of cooperation benefit is shown to be sensitive to the alignment of channel gains, with certain configurations yielding more significant improvements.

Key Results

The paper provides characterizations of the sum capacity for the symmetric interference channel and the $R_0$-capacity for the cognitive radio scenario, both within a constant gap. They show that half-duplex cooperation can offer substantial gains, particularly when the cooperation link is strong relative to interference links. For instance, in certain scenarios, the sum capacity with half-duplex cooperation can be strictly better than that of the usual interference channel. They also present plots illustrating the generalized degrees of freedom, visually demonstrating the gains from cooperation under various channel conditions.

Limitations and Future Work

The authors acknowledge that their results for the symmetric case are characterizations "within a constant," meaning they provide bounds that are close to the true capacity but may differ by a constant value. They also note that for certain channel conditions (e.g., when the channel matrix is ill-conditioned), the LDM might not perfectly capture the behavior of the Gaussian channel, and further analysis considering channel gain angles is needed. The paper suggests that the developed coding schemes and analysis techniques can be extended to more general interference channel settings.

Summary updated

BibTeX
@article{wu2014interference,
  author = {Rui Wu and Vinod M. Prabhakaran and Pramod Viswanath and Yi Wang},
  title = {{Interference Channels With Half-Duplex Source Cooperation}},
  journal = {IEEE Trans. Information Theory},
  year = {2014},
  doi = {10.1109/tit.2014.2300158},
}
Figure