Learning Successive Interference Cancellation for Low-Complexity Soft-Output MIMO Detection

TL;DR

This work tackles the problem of low-complexity MIMO detection with reliable soft information for edge devices. It introduces recurSIC, a physics-informed, SIC-inspired detector that operates on a QRD-transformed signal $\tilde{\bm{y}}$, uses a single forward pass, and employs multi-path hypothesis tracking with a tunable parameter $K$ and SNR embeddings. Key contributions include a shallow, shared-weight NN architecture, LLR clipping, and a multi-path extension that delivers near-ML hard and soft detection across realistic 5G NR channels for $16$QAM and $64$QAM, while maintaining ultra-low memory and compute. The results demonstrate that recurSIC is well suited for IoT/RedCap edge receivers, providing reliable soft information for LDPC decoding with significantly reduced complexity.

Abstract

Low-complexity multiple-input multiple-output (MIMO) detection remains a key challenge in modern wireless systems, particularly for 5G reduced capability (RedCap) and internet-of-things (IoT) devices. In this context, the growing interest in deploying machine learning on edge devices must be balanced against stringent constraints on computational complexity and memory while supporting high-order modulation. Beyond accurate hard detection, reliable soft information is equally critical, as modern receivers rely on soft-input channel decoding, imposing additional requirements on the detector design. In this work, we propose recurSIC, a lightweight learning-based MIMO detection framework that is structurally inspired by successive interference cancellation (SIC) and incorporates learned processing stages. It generates reliable soft information via multi-path hypothesis tracking with a tunable complexity parameter while requiring only a single forward pass and a minimal parameter count. Numerical results in realistic wireless scenarios show that recurSIC achieves strong hard- and soft-detection performance at very low complexity, making it well suited for edge-constrained MIMO receivers.

Figures (6)

• Figure 1: Multi-path recurSIC framework for $L=2$ layers with the recurSIC block architecture detailed in Fig. \ref{['fig:network_architecture']} and the sic module implementing \ref{['eq:sic']}.
• Figure 2: nn architecture of the recurSIC block with snr embedding. The nn parameters are shared across all detection stages and snr.
• Figure 3: Performance over tdl-A30 channel with 16QAM. Top: uncoded ber (hard-decision); bottom: relative throughput $(1-\text{BLER})$ using soft-output.
• Figure 4: Performance over tdl-A30 channel with 64QAM. Top: uncoded ber (hard-decision); bottom: relative throughput $(1-\text{BLER})$ using soft-output.
• Figure 5: Performance over tdl-B100 channel with 16QAM. Top: uncoded ber (hard-decision); bottom: relative throughput $(1-\text{BLER})$ using soft-output.