Design of a Standard-Compliant Real-Time Neural Receiver for 5G NR

TL;DR

This work tackles the deployment of a real-time neural MU-MIMO receiver for 5G NR that can operate under dynamic MCS configurations without retraining. It introduces a TensorRT-optimized NRX architecture with joint channel estimation and demapping, and training schemes that enable mixed-MCS operation via Double-readout and Multi-loss, achieving sub-millisecond latency on an NVIDIA A100. It also investigates site-specific adaptation using ray-tracing channel models to quantify data and fine-tuning requirements, showing that modest environment-specific data and iterations yield substantial gains with careful management of overfitting. Together, the study provides a practical pathway toward real-time, standards-compliant neural receivers for 5G NR and releases code to support replication.

Abstract

We detail the steps required to deploy a multi-user multiple-input multiple-output (MU-MIMO) neural receiver (NRX) in an actual cellular communication system. This raises several exciting research challenges, including the need for real-time inference and compatibility with the 5G NR standard. As the network configuration in a practical setup can change dynamically within milliseconds, we propose an adaptive NRX architecture capable of supporting dynamic modulation and coding scheme (MCS) configurations without the need for any re-training and without additional inference cost. We optimize the latency of the neural network (NN) architecture to achieve inference times of less than 1ms on an NVIDIA A100 GPU using the TensorRT inference library. These latency constraints effectively limit the size of the NN and we quantify the resulting signal-to-noise ratio (SNR) degradation as less than 0.7 dB when compared to a preliminary non-real-time NRX architecture. Finally, we explore the potential for site-specific adaptation of the receiver by investigating the required size of the training dataset and the number of fine-tuning iterations to optimize the NRX for specific radio environments using a ray tracing-based channel model. The resulting NRX is ready for deployment in a real-time 5G NR system and the source code including the TensorRT experiments is available online.

Figures (7)

• Figure 1: NRX for 5G NR .
• Figure 2: architecture with Var-IO layers.
• Figure 3: vs. for a Double-TDL channel with $U \times B \equiv 2\times4$. The MCS indices are $i=9$ (left), $i=14$ (middle) and $i=19$ (right). Dotted curves denote single-MCS NRXs and the Var-MCS NRXs are dashed.
• Figure 4: performance vs. number of NRX iterations for $U\times B \equiv 2\times4$ , Double-TDL channels and 16-QAM. NRX inference latency measured on NVIDIA A100.
• Figure 5: Ray tracing environment using Sionna's Munich map, augmented with coverage map, visualizing as red point, and evaluation trajectories as salmon-colored lines.