Sum-Capacity of Ergodic Fading Interference and Compound Multiaccess Channels

Lalitha Sankar, Elza Erkip, H. Vincent Poor

Abstract

The problem of resource allocation is studied for two-sender two-receiver fading Gaussian interference channels (IFCs) and compound multiaccess channels (C-MACs). The senders in an IFC communicate with their own receiver (unicast) while those in a C-MAC communicate with both receivers (multicast). The instantaneous fading state between every transmit-receive pair in this network is assumed to be known at all transmitters and receivers. Under an average power constraint at each source, the sum-capacity of the C-MAC and the power policy that achieves this capacity is developed. The conditions defining the classes of strong and very strong ergodic IFCs are presented and the multicast sum-capacity is shown to be tight for both classes.

Introduction

In wireless communications, the Interference Channel (IFC) serves as a classic model where two senders communicate with their respective receivers over a shared medium. Traditionally, interference is viewed as a nuisance to be mitigated. However, information theory suggests that if interference is sufficiently "strong," a receiver can decode the interfering signal, subtract it, and then recover its own intended message without loss of performance.

This paper addresses the challenge of resource allocation in real-world wireless environments where channels exhibit fading—fluctuations in signal strength over time. By assuming perfect channel state information (CSI) at all nodes, the authors explore how transmitters should dynamically vary their power to maximize the total data rate (sum-capacity).

System Model and Problem Formulation

The study focuses on a two-user network where sources $S_1$ and $S_2$ transmit messages to destinations $D_1$ and $D_2$. In the IFC model, each receiver only needs to decode its own message. In the Compound Multiaccess Channel (C-MAC) model, both receivers are required to decode both messages (multicast).

System model of the Interference Channel (IFC) and Compound Multiaccess Channel (C-MAC).

The C-MAC sum-capacity serves as a fundamental lower bound for the IFC. If a rate is achievable when forcing both receivers to decode everything, it is inherently achievable when they only require their own messages. The core of this research is identifying when this multicast bound is "tight" for the unicast IFC.

Methodology: Optimal Power Allocation

The authors employ ergodic capacity modeling to calculate the long-term average rate achievable over many fading states. By applying Karush-Kuhn-Tucker (KKT) conditions to the sum-rate maximization problem under average power constraints, they derive optimal power allocation policies.

A critical component of the analysis is a case-based approach. The optimal sum-rate policy is shown to fall into one of five disjoint cases, depending on which receiver's "bottleneck" is more restrictive at any given moment.

Rate regions for various cases where the sum-capacity is determined by the intersection of individual receiver MAC regions.

As illustrated in the rate regions above, the sum-capacity $R_1 + R_2$ is determined by the intersection of the capacity regions of the two multiaccess channels formed at $D_1$ and $D_2$. The optimal policy often mirrors standard MAC behavior, switching between opportunistic scheduling (where only the user with the best channel transmits) and simultaneous transmission using a temporal "water-filling" approach.

Key Insights and Results

The paper's most significant contribution is the definition of "strong" and "very strong" interference in the context of ergodic fading. Unlike static channels where interference strength is fixed, in fading channels, "strength" is a dynamic property that depends on both channel statistics and the power allocation policy.

  • Strong Ergodic IFC: For this class, the sum-capacity is the minimum of the sum-rates achievable at the two receivers. The authors prove that the C-MAC sum-capacity is tight for these channels.
  • Very Strong Ergodic IFC: In this scenario, the interference is so robust that the sum-capacity equals the sum of two independent point-to-point links. It is as if the interference does not exist at all.
  • Ergodic Tightness: The research demonstrates that even if interference is not strong in every single instant, the multicast strategy remains optimal for sum-capacity as long as the interference is strong on average (ergodically).

Conclusion and Future Work

This work generalizes previous findings on parallel channels to a broader class of fading environments, providing explicit mathematical conditions for power policies that maximize sum-capacity. While the two-user case is solved, extending these results to $K$-user networks remains a complex challenge. Furthermore, the assumption of perfect CSI is an idealized model; practical implementations must account for the overhead and errors associated with channel estimation. Finally, while the sum-capacity is now well-defined for these classes, the full capacity region for the general interference channel remains one of the most difficult open problems in information theory.

Summary updated

Abstract

The problem of resource allocation is studied for two-sender two-receiver fading Gaussian interference channels (IFCs) and compound multiaccess channels (C-MACs). The senders in an IFC communicate with their own receiver (unicast) while those in a C-MAC communicate with both receivers (multicast). The instantaneous fading state between every transmit-receive pair in this network is assumed to be known at all transmitters and receivers. Under an average power constraint at each source, the sum-capacity of the C-MAC and the power policy that achieves this capacity is developed. The conditions defining the classes of strong and very strong ergodic IFCs are presented and the multicast sum-capacity is shown to be tight for both classes.
BibTeX
@inproceedings{sankar2008sum,
  author = {Lalitha Sankar and Elza Erkip and H. Vincent Poor},
  title = {{Sum-Capacity of Ergodic Fading Interference and Compound Multiaccess Channels}},
  booktitle = {ISIT},
  year = {2008},
  url = {https://openreview.net/forum?id=ELMCMiZUcSV},
}
Figure