K-User Interference Channel with Backhaul Cooperation: DoF vs. Backhaul Load Trade-Off

Borna Kananian, Mohammad Ali Maddah-Ali, Babak Hossein Khalaj

Introduction

In wireless communications, interference is a fundamental bottleneck. As user density increases, signals overlap, reducing the achievable data rates. Backhaul cooperation—where base stations or devices exchange information over a high-speed side channel—offers a path to mitigate this interference. However, this cooperation incurs a "backhaul load" cost. This paper by Kananian, Maddah-Ali, and Khalaj investigates the fundamental information-theoretic trade-off between the Degrees of Freedom (DoF) in a wireless network and the required communication load on the backhaul.

Problem Formulation

The study focuses on a $K$-user interference channel where $K$ transmitters aim to send messages to $K$ receivers. In an uncoordinated setting, techniques like Interference Alignment typically yield $M/2$ DoF per user. With unlimited backhaul, the system could reach the single-user bound of $M$ DoF. The authors seek to characterize the intermediate regime: how much backhaul capacity ($\alpha$) is needed to achieve a specific DoF?

Methodology

The authors employ a high-SNR approximation, defining backhaul load ($\alpha$) as the normalized entropy of backhaul messages. They analyze two primary cooperation scenarios:

  1. Transmitter Cooperation: Transmitters share message information to perform joint precoding.
  2. Receiver Cooperation: Receivers share quantized signal observations to perform joint decoding.
System models for receiver and transmitter cooperation via a backhaul network.

To establish the fundamental limits, the authors developed new outer bounds by partitioning users and assuming cost-free cooperation within groups. For achievability, they proposed a Centralized Scheme where a single node acts as a central processor, collecting and distributing data to perform joint processing (e.g., Zero-Forcing).

Key Insights

  • Scale Matters: While distributed "Cooperation Alignment" is efficient for 2-user or 3-user systems, these gains do not scale. In large networks, centralized processing becomes the most efficient use of backhaul.
  • Reciprocity: The DoF-backhaul trade-off is identical regardless of whether cooperation occurs at the transmitter or receiver side.
  • Topology Robustness: Using an extended Hall’s condition and Max-Flow Min-Cut algorithms, the authors proved that centralized processing remains optimal even in partially connected networks, provided certain topological conditions are met.

Key Results

The paper characterizes the optimal DoF as a function of the backhaul load $\alpha$. For an even number of users $K$, the optimal trade-off is:

$$DoF^*(\alpha) = \min\left\{M, \frac{1}{2}\left(M + \frac{K}{2(K-1)}\alpha\right)\right\}$$

As $K$ becomes large, this simplifies to $DoF^* \approx \min\{M, \frac{M}{2} + \frac{\alpha}{4}\}$.

Degrees of Freedom (DoF) vs. backhaul load ($\alpha$) trade-off for different network scales.

The results show that the gap between upper and lower bounds for odd $K$ vanishes as the network size increases, confirming the near-optimality of centralized processing in large-scale wireless systems.

Conclusion

This work provides a rigorous framework for understanding the cost of cooperation in wireless networks. By showing that simple centralized schemes are nearly optimal at scale, it simplifies the design space for future high-density wireless architectures like Cloud-RAN. Future work may extend these findings to finite SNR regimes and scenarios with imperfect channel state information.

Summary updated

Abstract

In this paper, we consider K-user interference channels with M antennas per node and with backhaul collaboration in one side (among the transmitters or among the receivers), for M, K ∈ N, and investigate the tradeoff between the rate in the channel versus the communication load in the backhaul. In this investigation, each node is equipped with M antennas and we focus on a first order approximation result, where the rate of the wireless channel is measured by the degrees of freedom (DoF) per user, and the load of the backhaul is measured by the entropy of backhaul messages per user normalized by log of transmit power, at high power regimes. This tradeoff is fully characterized for the case of even values of K and approximately characterized for the case of odd values of K, with vanishing approximation gap as K grows. To achieve DoF of M per user, this result establishes the asymptotic optimality of the most straightforward scheme, called central processing, in which the messages are collected at one of the nodes, centrally processed, and forwarded back to each node. In addition, this result shows that the gain of the schemes, relying on distributed processing, through pairwise communication among the nodes (e.g., cooperative alignment) does not scale with the size of the network. For the converse, we develop a new outer-bound on the tradeoff based on splitting the set of collaborative nodes (transmitters or receivers) into two subsets and assuming full cooperation within each group. We further present a sufficient condition on the wireless channel connectivity, which although more relaxed, guarantees the validity of the above tradeoff. Finally, we show that verifying this condition takes a polynomial time in the network size.
BibTeX
@article{kananian2019k,
  author = {Borna Kananian and Mohammad Ali Maddah-Ali and Babak Hossein Khalaj},
  title = {{K-User Interference Channel with Backhaul Cooperation: DoF vs. Backhaul Load Trade-Off}},
  journal = {IEEE Trans. Information Theory},
  year = {2019},
  doi = {10.1109/tit.2019.2926469},
}
Figure