Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Machine Learning Decoder for 5G NR PUCCH Format 0

Anil Kumar Yerrapragada, Jeeva Keshav S, Ankit Gautam, Radha Krishna Ganti

TL;DR

Addresses reliable decoding of UCI in 5G NR PUCCH Format 0 by treating cyclic shift detection as a supervised classification task. Uses a $24$-input fully connected classifier with four output classes, trained on simulated data from the MATLAB 5G Toolbox and validated with over-the-air hardware captures. Demonstrates higher accuracy than a DFT-based receiver, with robust performance across SNR and strong conformance to 3GPP requirements, including a notable low-SNR gain. Finds favorable accuracy–complexity trade-offs, showing that smaller networks can approach performance with substantially reduced computational burden, supporting practical deployment potential.

Abstract

5G cellular systems depend on the timely exchange of feedback control information between the user equipment and the base station. Proper decoding of this control information is necessary to set up and sustain high throughput radio links. This paper makes the first attempt at using Machine Learning techniques to improve the decoding performance of the Physical Uplink Control Channel Format 0. We use fully connected neural networks to classify the received samples based on the uplink control information content embedded within them. The trained neural network, tested on real-time wireless captures, shows significant improvement in accuracy over conventional DFT-based decoders, even at low SNR. The obtained accuracy results also demonstrate conformance with 3GPP requirements.

Machine Learning Decoder for 5G NR PUCCH Format 0

TL;DR

Addresses reliable decoding of UCI in 5G NR PUCCH Format 0 by treating cyclic shift detection as a supervised classification task. Uses a -input fully connected classifier with four output classes, trained on simulated data from the MATLAB 5G Toolbox and validated with over-the-air hardware captures. Demonstrates higher accuracy than a DFT-based receiver, with robust performance across SNR and strong conformance to 3GPP requirements, including a notable low-SNR gain. Finds favorable accuracy–complexity trade-offs, showing that smaller networks can approach performance with substantially reduced computational burden, supporting practical deployment potential.

Abstract

5G cellular systems depend on the timely exchange of feedback control information between the user equipment and the base station. Proper decoding of this control information is necessary to set up and sustain high throughput radio links. This paper makes the first attempt at using Machine Learning techniques to improve the decoding performance of the Physical Uplink Control Channel Format 0. We use fully connected neural networks to classify the received samples based on the uplink control information content embedded within them. The trained neural network, tested on real-time wireless captures, shows significant improvement in accuracy over conventional DFT-based decoders, even at low SNR. The obtained accuracy results also demonstrate conformance with 3GPP requirements.
Paper Structure (16 sections, 2 equations, 7 figures)

This paper contains 16 sections, 2 equations, 7 figures.

Table of Contents

  1. Introduction
  2. Background on PUCCH Format 0
  3. Format 0 Allocation and Sequence Generation
  4. Existing PUCCH Format 0 Receiver Methods
  5. ML based receiver for PUCCH Format 0
  6. PUCCH Format 0 detection as a classification problem
  7. Dataset generation
  8. Simulated data for training
  9. Hardware captures for testing
  10. Neural Network Architecture
  11. Structure of the neural network
  12. Regularization
  13. Activations and Backpropagation
  14. Training and Testing
  15. Results and discussion
  16. ...and 1 more sections

Figures (7)

  • Figure 1: The cyclic shifts $m_{cs}$ applied to the PUCCH Format 0 base sequence depend on the specific UCI content.
  • Figure 2: PUCCH Format 0 neural network decoder architecture.
  • Figure 3: IIT-Madras 5G testbed setup with the Remote Radio Head used as a receiver and the VSG used as a transmitter.
  • Figure 4: (a) Training Accuracy and (b) Training Loss for simulated data at an SNR of 10dB
  • Figure 5: Model accuracy on (a) simulated test data, (b) hardware captured test data. In both cases, accuracy is compared with that achieved by the DFT based decoder.
  • ...and 2 more figures