Fading Two-Way Relay Channels: Physical-Layer Versus Digital Network Coding

Zhi Chen, Teng Joon Lim, Mehul Motani

Abstract

This paper investigates energy-efficient transmission strategies for fading three-node, two-way relay networks (TWRN). We analyze and compare four strategies: physical-layer network coding (PNC), digital network coding (DNC) with time sharing (DNC-TS), DNC with superposition (DNC-Sup), and codeword superposition (CW-Sup). The objective is to minimize the total average energy required to meet specific average rate pairs under full channel state information (CSI). Theoretical analysis proves that DNC-Sup outperforms DNC-TS and CW-Sup. Furthermore, simulation results reveal a performance trade-off where PNC-Sup is superior for high data rates, while DNC-Sup is more efficient for low data rates. We conclude by proposing an algorithm to select the optimal strategy based on rate requirements.

Introduction

In a Two-Way Relay Network (TWRN), two source nodes ($S_1$ and $S_2$) exchange information via a relay ($R$). Traditional studies often assume static channels and symmetric traffic. However, real-world wireless environments are characterized by fading and asymmetric data demands. This research addresses the problem of minimizing total average energy consumption in such realistic scenarios. A key challenge is managing "leftover" bits when one source has a higher data rate requirement than the other.

Transmission Strategies

The authors evaluate four distinct strategies for managing the uplink and downlink phases in a TWRN. These strategies differ in how they process and combine messages at the relay to handle asymmetric traffic.

Comparison of the four transmission strategies: DNC-TS, DNC-Sup, PNC-Sup, and CW-SuP.
  1. DNC-TS (Digital Network Coding - Time Sharing): The relay decodes both source messages, performs an XOR operation, and then time-shares the downlink between the network-coded message and the remaining bits of the longer message.
  2. DNC-Sup (Digital Network Coding - Superposition): Instead of time-sharing, the relay superimposes the XORed message and the extra bits on the downlink. This leverages the broadcast nature of the channel more effectively.
  3. PNC-Sup (Physical-Layer Network Coding - Superposition): On the uplink, the relay decodes a function of the superimposed signals (e.g., the XOR of the bits) directly. For asymmetric traffic, the extra bits are transmitted separately. The downlink uses superposition coding.
  4. CW-Sup (Codeword Superposition): A baseline strategy where the relay simply superimposes the two original codewords on the downlink without explicit network coding.

The optimization of these strategies involves finding the best power and time-fraction allocations using the Lagrange multiplier method and Karush-Kuhn-Tucker (KKT) conditions, solved via a multi-bisection numerical method.

Key Insights and Results

The Superiority of DNC-Sup

A major theoretical finding is that DNC-Sup consistently outperforms DNC-TS and CW-Sup. By using superposition coding on the downlink, the relay treats the transmission as a broadcast channel where each receiver already possesses its own message as "side information," leading to higher energy efficiency than simple time-sharing.

The PNC vs. DNC Trade-off

The choice between PNC and DNC depends heavily on the required data rates:

  • PNC-Sup is more efficient at high data rates. It handles uplink interference by decoding a combined function of the messages rather than treating them as mutual interference.
  • DNC-Sup is more efficient at low data rates. At lower power levels, multi-access interference is less significant, and DNC's flexibility in resource allocation provides an advantage.

In symmetric Rayleigh fading simulations, the crossover point where PNC-Sup begins to outperform DNC-Sup occurs at approximately 1.2 frames per slot.

Optimal Strategy Selection

To ensure peak efficiency, the authors propose an algorithm ($P_{opt}$) that calculates the energy costs for both PNC-Sup and DNC-Sup for any given rate pair and selects the minimum, ensuring the system adapts to varying traffic demands.

Conclusion

This work provides a rigorous framework for energy minimization in fading TWRNs. While DNC-Sup is the most efficient digital-based method, PNC-Sup offers significant gains in high-throughput scenarios. Future research may extend these findings to multi-antenna (MIMO) systems and explore the impact of imperfect channel state information (CSI) on strategy performance.

Summary updated

Abstract

In this paper, we consider three transmit strategies for the fading three-node, two-way relay network (TWRN) -- physical-layer network coding (PNC), digital network coding (DNC) and codeword superposition (CW-Sup). The aim is to minimize the total average energy needed to deliver a given pair of required average rates. Full channel state information is assumed to be available at all transmitters and receivers. The optimization problems corresponding to the various strategies in fading channels are formulated, solved and compared. For the DNC-based strategies, a simple time sharing of transmission of the network-coded message and the remaining bits of the larger message (DNC-TS) is considered first. We extend this approach to include a superposition strategy (DNC-Sup), in which the network-coded message and the remainder of the longer source message are superimposed before transmission. It is demonstrated theoretically that DNC-Sup outperforms DNC-TS and CW-Sup in terms of total average energy usage. More importantly, it is shown in simulation that DNC-Sup performs better than PNC if the required rate is low and worse otherwise. Finally, an algorithm to select the optimal strategy in terms of energy usage subject to different rate pair requirements is presented.
BibTeX
@article{chen2014fading,
  author = {Zhi Chen and Teng Joon Lim and Mehul Motani},
  title = {{Fading Two-Way Relay Channels: Physical-Layer Versus Digital Network Coding}},
  journal = {IEEE Trans. Wireless Communications},
  year = {2014},
  url = {https://openreview.net/forum?id=b90AGd6_9u0},
}
Figure