Reproduce “A Neural Receiver for 5G NR Multi-user MIMO” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Sebastian Cammerer, Fayçal Aït Aoudia, Jakob Hoydis, Andreas Oeldemann, Andreas Roessler, Timo Mayer, Alexander Keller

A learned receiver for standard-compliant 5G NR PUSCH jointly performs channel estimation, interference separation, symbol detection, and bit demapping across an entire OFDM slot. Combining time-frequency CNN processing with layer-wise GNN message passing, it handles changing numbers of spatial layers and much larger bandwidths than used during training, approaching K-best detection within 1 dB and matching simulation trends in hardware-in-the-loop tests.

Key insights

  • Jointly learning channel estimation and detection can improve interference separation over an explicit LS-channel-estimation plus LMMSE-detection pipeline, while retaining a substantially smaller model than exhaustive-style detectors.
  • Layer-wise message passing provides permutation-compatible handling of variable numbers of users or spatial streams, whereas convolutional processing enables transfer across resource-grid sizes.
  • Generalization is demonstrated within randomized 3GPP-based conditions, but broader modulation, numerology, impairment, and deployment changes would still require additional training or adaptation.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.