Abstract
This paper explores the two-user interference channel (IC) where transmitters have non-causal knowledge of a common interference state, such as a known signal from a neighboring cell. We propose novel coding schemes that integrate rate splitting with Gel’fand-Pinsker coding and introduce Active Interference Cancellation (AIC). AIC is a technique where transmitter power is used to physically cancel state interference at the receivers. Our analysis shows that AIC significantly expands achievable communication rates in strong and mixed interference environments, while flexible power splitting remains more effective in weak interference scenarios.
Introduction: The State-Dependent Interference Channel
In wireless communications, the Interference Channel (IC) models multiple transmitter-receiver pairs sharing the same frequency. A "state-dependent" IC assumes an additional source of interference (the "state") is known to the transmitters beforehand but unknown to the receivers.
As shown in Figure 1, imagine two base stations communicating with their respective mobiles while a third, distant base station creates a predictable background hum (the state $S$). If the base stations know this hum exactly, they can coordinate transmissions to maximize data rates. While Costa’s "Dirty Paper Coding" (DPC) solves this for single-user channels, the multi-user case is complex because mitigating state interference for one user might inadvertently increase interference for another.
Methods and System Model
The authors develop coding frameworks for Discrete Memoryless (DM) channels and specialize them for Gaussian channels. The Gaussian model is particularly relevant for practical wireless systems.
As illustrated in Figure 3, the received signals $Y_1$ and $Y_2$ are affected by the transmitted signals $X_1, X_2$, the common state $S$, and additive white Gaussian noise $Z_1, Z_2$. The proposed methods include:
- Rate Splitting: Messages are divided into "public" parts (decodable by both receivers) and "private" parts (intended only for one receiver) to manage inter-user interference.
- Gel’fand-Pinsker (GP) Coding: Used to pre-code signals against the known state $S$.
- Active Interference Cancellation (AIC): This is the key innovation for Gaussian channels. Unlike DPC, which is a passive coding strategy to "ignore" the state, AIC uses a portion of the transmitter's power to send a counter-phase signal that physically cancels $S$ at the receivers.
Key Insights and Results
The effectiveness of these strategies depends heavily on the interference regime:
- The "Precious Power" Trade-off: In weak interference scenarios, transmitter power is too valuable to spend on physical cancellation. It is more efficient to use power for data transmission and rely on flexible power splitting between public and private messages.
- AIC as a Multi-user Enabler: In strong or mixed interference, AIC is highly effective. By canceling the common state, a transmitter makes it easier for the other receiver to decode its "public" message. This "altruistic" cancellation facilitates better overall interference management.
- Generalization of DPC: AIC is shown to be a generalization of Costa’s DPC. While DPC is optimal for single users, AIC provides the flexibility needed for the IC where users must balance state mitigation with inter-user interference.
Numerical results demonstrate that AIC significantly enlarges the achievable rate region in strong and mixed interference compared to schemes that either ignore the state or use standard DPC. Furthermore, the authors proved that even with infinite state power, non-trivial communication rates are achievable if the state is known at the transmitters.
Conclusion
This work provides a robust framework for managing state-dependent interference in multi-user channels. While the current model assumes a common state known perfectly by both transmitters, future research could address asymmetric state knowledge or scenarios where different links experience different state interferences. Additionally, simplifying the high-dimensional searches required to optimize these coding schemes will be crucial for real-time implementation.
Summary updated