Streaming Erasure Codes over the Multiple Access Relay Channel

Gustavo Kasper Facenda, Elad Domanovitz, Ashish Khisti, Wai-Tian Tan, John G. Apostolopoulos

Abstract

This paper investigates the design of streaming erasure codes for the multi-access relay channel (MARC), where two sources transmit data to a destination via a single relay. To meet the stringent low-latency requirements of real-time applications like cloud gaming and VR, the authors propose a Symbol-Wise Decode-and-Forward (SWDF) framework. By introducing a novel delay-spectrum analysis, they develop coding schemes that optimize transmission rates while ensuring all packets are decoded within a fixed delay constraint $T$. The proposed schemes are shown to be optimal in specific network regimes and significantly outperform traditional multiplexing approaches.

Introduction

In the landscape of 5G and beyond, real-time multimedia applications—such as competitive online gaming and immersive virtual reality—demand extremely low latency. In these scenarios, every millisecond is critical. Traditional error correction techniques often rely on Automatic Repeat Request (ARQ), which requires re-transmitting lost packets. However, the round-trip time for such re-transmissions frequently exceeds the application's strict delay budget.

Forward Error Correction (FEC) offers an alternative by sending redundant "parity" packets, allowing the receiver to recover lost data without additional communication rounds. While FEC for point-to-point links is well-understood, modern network topologies often involve intermediate relay nodes. This paper addresses the challenge of designing streaming codes for a Multi-Access Relay Channel (MARC), where multiple users share a single relay to reach a destination under an adversarial erasure model.

System Model

The study focuses on a network with two source nodes ($S_1, S_2$), one relay ($R$), and a single destination ($D$). Each communication link is characterized by a maximum number of erasures ($N_i$) that can occur within any sliding window of time. The goal is to find the maximum achievable transmission rates ($R_1, R_2$) such that every source packet can be decoded at the destination within a fixed delay $T$.

The Multi-Access Relay Channel (MARC) model with two sources, one relay, and a single destination under adversarial erasure constraints.

Methodology: Symbol-Wise Decode-and-Forward

The authors introduce a framework based on Symbol-Wise Decode-and-Forward (SWDF). Unlike traditional block-based relaying, where the relay waits for an entire data block before processing, SWDF allows the relay to decode and re-encode data symbol-by-symbol. This approach is crucial for minimizing the latency introduced at the relay node.

Key Technical Innovations

  1. Delay-Spectrum Analysis: This is the core innovation of the paper. Instead of assuming a uniform delay for all data, the authors analyze the "spectrum" of delays—specifically, how many symbols can be recovered at each possible delay step (e.g., delay 1, delay 2, ..., delay $T$).
  2. Fixed-Bottleneck SWDF (FB-SWDF): A greedy scheme designed for scenarios where the relay-to-destination link is the primary bottleneck. It pushes this link to its theoretical limit and adjusts the source-to-relay codes accordingly.
  3. Optimized-Bottleneck SWDF (OB-SWDF): A more sophisticated numerical approach that uses a bisection-like algorithm to find the optimal balance of rates between the two hops of the network, maximizing the overall achievable rate region.

Key Insights and Results

The primary advantage of the proposed framework lies in its ability to "merge" data streams. By treating the relay-to-destination link as a joint resource, the relay can better utilize the remaining "delay budget" of each symbol. If a symbol arrives at the relay quickly, it has more time available for recovery on the second hop.

Performance Highlights

  • Capacity Achievement: In "strong" bottleneck scenarios—where the link from the relay to the destination is significantly constrained—the authors prove their scheme is mathematically optimal.
  • Rate Gains: In more common "weak" bottleneck scenarios, the OB-SWDF scheme provides a significantly larger achievable rate region compared to naive extensions of single-user codes (such as CSWDF or CMWDF).
  • Practicality: The proposed codes are time-invariant, meaning the relay's strategy does not need to change based on specific erasure patterns. This simplicity makes the schemes highly suitable for high-speed hardware implementation.

Conclusion and Future Work

This research provides a robust theoretical foundation for low-latency streaming over relayed networks. By treating delay as a partitionable resource and utilizing symbol-wise processing, the authors bridge the gap between theoretical limits and practical code design.

Future research directions include scaling the framework to accommodate more than two source nodes and exploring the "downlink" scenario (one source to multiple destinations). Additionally, the authors aim to further tighten the gap between the derived upper and lower bounds in the weak bottleneck regimes to reach the absolute capacity of the MARC.

Summary updated

BibTeX
@article{facenda2023streaming,
  author = {Gustavo Kasper Facenda and Elad Domanovitz and Ashish Khisti and Wai-Tian Tan and John G. Apostolopoulos},
  title = {{Streaming Erasure Codes over the Multiple Access Relay Channel}},
  journal = {IEEE Trans. Information Theory},
  year = {2023},
  doi = {10.1109/tit.2022.3214165},
}
Figure