Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep CSI Compression for Dual-Polarized Massive MIMO Channels with Disentangled Representation Learning

Suhang Fan, Wei Xu, Renjie Xie, Shi Jin, Derrick Wing Kwan Ng, Naofal Al-Dhahir

TL;DR

The proposed DiReNet disentangles dual-polarized CSI into three components: polarization-shared information, vertical polarization-specific information, and horizontal polarization-specific information, which enables the minimization of information redundancy caused by the polarization correlation and improves the performance of CSI compression and recovery.

Abstract

Channel state information (CSI) feedback is critical for achieving the promised advantages of enhancing spectral and energy efficiencies in massive multiple-input multiple-output (MIMO) wireless communication systems. Deep learning (DL)-based methods have been proven effective in reducing the required signaling overhead for CSI feedback. In practical dual-polarized MIMO scenarios, channels in the vertical and horizontal polarization directions tend to exhibit high polarization correlation. To fully exploit the inherent propagation similarity within dual-polarized channels, we propose a disentangled representation neural network (NN) for CSI feedback, referred to as DiReNet. The proposed DiReNet disentangles dual-polarized CSI into three components: polarization-shared information, vertical polarization-specific information, and horizontal polarization-specific information. This disentanglement of dual-polarized CSI enables the minimization of information redundancy caused by the polarization correlation and improves the performance of CSI compression and recovery. Additionally, flexible quantization and network extension schemes are designed. Consequently, our method provides a pragmatic solution for CSI feedback to harness the physical MIMO polarization as a priori information. Our experimental results show that the performance of our proposed DiReNet surpasses that of existing DL-based networks, while also effectively reducing the number of network parameters by nearly one third.

Deep CSI Compression for Dual-Polarized Massive MIMO Channels with Disentangled Representation Learning

TL;DR

The proposed DiReNet disentangles dual-polarized CSI into three components: polarization-shared information, vertical polarization-specific information, and horizontal polarization-specific information, which enables the minimization of information redundancy caused by the polarization correlation and improves the performance of CSI compression and recovery.

Abstract

Channel state information (CSI) feedback is critical for achieving the promised advantages of enhancing spectral and energy efficiencies in massive multiple-input multiple-output (MIMO) wireless communication systems. Deep learning (DL)-based methods have been proven effective in reducing the required signaling overhead for CSI feedback. In practical dual-polarized MIMO scenarios, channels in the vertical and horizontal polarization directions tend to exhibit high polarization correlation. To fully exploit the inherent propagation similarity within dual-polarized channels, we propose a disentangled representation neural network (NN) for CSI feedback, referred to as DiReNet. The proposed DiReNet disentangles dual-polarized CSI into three components: polarization-shared information, vertical polarization-specific information, and horizontal polarization-specific information. This disentanglement of dual-polarized CSI enables the minimization of information redundancy caused by the polarization correlation and improves the performance of CSI compression and recovery. Additionally, flexible quantization and network extension schemes are designed. Consequently, our method provides a pragmatic solution for CSI feedback to harness the physical MIMO polarization as a priori information. Our experimental results show that the performance of our proposed DiReNet surpasses that of existing DL-based networks, while also effectively reducing the number of network parameters by nearly one third.
Paper Structure (22 sections, 21 equations, 13 figures, 4 tables, 1 algorithm)

This paper contains 22 sections, 21 equations, 13 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. System Model
  3. Disentangled Representation Learning for Dual-polarized CSI
  4. Correlation between Dual-polarized CSI
  5. Disentangled Representation Learning in Network Design
  6. Proposed Architecture of DiReNet
  7. Design of the Encoder Network
  8. Design of the Decoder Network
  9. The Mutual Information Estimator
  10. Training Procedure of DiReNet
  11. Experimental Results
  12. Simulation Setup
  13. Performance of the Proposed DiReNet
  14. NMSE Performance
  15. Network Complexity
  16. ...and 7 more sections

Figures (13)

  • Figure 1: A framework of the proposed DiReNet for dual-polarized CSI.
  • Figure 2: Distribution of the GCS between dual-polarized CSI. (The first quartile is the value below which 25% of the dataset, The third quartile is the value below which 75% of the dataset.)
  • Figure 3: Network architecture of the proposed DiReNet.
  • Figure 4: The proposed disentangled learning structure of the encoder.
  • Figure 5: The structure of the IR module.
  • ...and 8 more figures