Rich-ARQ: From 1-bit Acknowledgment to Rich Neural Coded Feedback

TL;DR

Rich-ARQ replaces the traditional 1 bit ACK/NACK with a high dimensional neural coded feedback that enables collaborative physical layer coding. The authors develop an asynchronous feedback code AFC that avoids encoder stalls, incorporate SNR conditioned curriculum learning with Langevin perturbations, and implement a lightweight UE encoder alongside a powerful AP decoder. They validate the approach with a full stack SDR prototype compatible with 4G/5G NR and show substantial gains in SNR efficiency, coverage, and latency over conventional HARQ and prior DL based feedback codes. The work demonstrates practical feasibility for intelligent feedback in next generation networks and outlines a path toward semantic and goal oriented feedback in interactive wireless protocols.

Abstract

This paper reimagines the foundational feedback mechanism in wireless communication, transforming the prevailing 1-bit binary ACK/NACK with a high-dimensional, information-rich vector to transform passive acknowledgment into an active collaboration. We present Rich-ARQ, a paradigm that introduces neural-coded feedback for collaborative physical-layer channel coding between transmitter and receiver. To realize this vision in practice, we develop a novel asynchronous feedback code that eliminates stalling from feedback delays, adapts dynamically to channel fluctuations, and features a lightweight encoder suitable for on-device deployment. We materialize this concept into the first full-stack, standard-compliant software-defined radio prototype, which decouples AI inference from strict radio timing. Comprehensive over-the-air experiments demonstrate that Rich-ARQ achieves significant SNR gains over conventional 1-bit hybrid ARQ and remarkable latency reduction over prior learning-based feedback codes, moving the promise of intelligent feedback from theory to a practical, high-performance reality for next-generation networks.

Figures (6)

• Figure 1: The prevailing star-topology architecture in contemporary wireless systems, and a comparison among ARQ, HARQ, and Rich-ARQ.
• Figure 2: The measured SNR variation in an indoor laboratory environment.
• Figure 3: Time line comparison of feedback coding: (a) Synchronous coding timeline, where each transmission round must await feedback from the previous round before proceeding. (b) Asynchronous coding timeline, where the encoder generates the next codeword $\bm{c}^{(t)}$ despite delayed feedback $\widetilde{\bm{c}}^{(t-1)}$, eliminating idle waiting and reducing end-to-end latency.
• Figure 4: Lightweight AFC encoder architecture implemented at the UE to meet hardware constraints.
• Figure 5: Hardware and software architecture of the Rich-ARQ prototype system. The diagram depicts both the AP and a UE, each comprising a host computer running the layered protocol stack and the AFC encoder and decoder, connected via PCIe and Ethernet to a USRP X310 software-defined radio frontend. The software plane illustrates the real-time, multi-threaded execution architecture that decouples neural network inference from time-critical physical layer processing.