Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AI-based CSI Feedback with Digital Twins: Real-World Validation and Insights

Tzu-Hao Huang, Chao-Kai Wen, Shang-Ho Tsai, Trung Q. Duong

TL;DR

The paper tackles the problem of training DL-based CSI feedback with DT-generated data and validating it in RW environments. It combines a DT-driven single-user MIMO-OFDM framework using an autoencoder (EVCsiNet) for implicit feedback with online learning that fine-tunes only the decoder using limited RW data. Key findings show that DT pretraining reduces domain mismatch and OL further improves RW performance, but a dedicated DT remains essential to achieve satisfactory results, particularly under varying environmental and BS-antenna conditions. The work demonstrates a practical approach to site-specific data generation and deployment, highlighting the trade-offs between DT fidelity, OL effort, and feedback overhead, with measurable gains in throughput and reconstruction accuracy.

Abstract

Deep learning (DL) has shown great potential for enhancing channel state information (CSI) feedback in multiple-input multiple-output (MIMO) communication systems, a subject currently under study by the 3GPP standards body. Digital twins (DTs) have emerged as an effective means to generate site-specific datasets for training DL-based CSI feedback models. However, most existing studies rely solely on simulations, leaving the effectiveness of DTs in reducing DL training costs yet to be validated through realistic experimental setups. This paper addresses this gap by establishing a real-world (RW) environment and corresponding virtual channels using ray tracing with replicated 3D models and accurate antenna properties. We evaluate whether models trained in DT environments can effectively operate in RW scenarios and quantify the benefits of online learning (OL) for performance enhancement. Results show that a dedicated DT remains essential even with OL to achieve satisfactory performance in RW scenarios.

AI-based CSI Feedback with Digital Twins: Real-World Validation and Insights

TL;DR

The paper tackles the problem of training DL-based CSI feedback with DT-generated data and validating it in RW environments. It combines a DT-driven single-user MIMO-OFDM framework using an autoencoder (EVCsiNet) for implicit feedback with online learning that fine-tunes only the decoder using limited RW data. Key findings show that DT pretraining reduces domain mismatch and OL further improves RW performance, but a dedicated DT remains essential to achieve satisfactory results, particularly under varying environmental and BS-antenna conditions. The work demonstrates a practical approach to site-specific data generation and deployment, highlighting the trade-offs between DT fidelity, OL effort, and feedback overhead, with measurable gains in throughput and reconstruction accuracy.

Abstract

Deep learning (DL) has shown great potential for enhancing channel state information (CSI) feedback in multiple-input multiple-output (MIMO) communication systems, a subject currently under study by the 3GPP standards body. Digital twins (DTs) have emerged as an effective means to generate site-specific datasets for training DL-based CSI feedback models. However, most existing studies rely solely on simulations, leaving the effectiveness of DTs in reducing DL training costs yet to be validated through realistic experimental setups. This paper addresses this gap by establishing a real-world (RW) environment and corresponding virtual channels using ray tracing with replicated 3D models and accurate antenna properties. We evaluate whether models trained in DT environments can effectively operate in RW scenarios and quantify the benefits of online learning (OL) for performance enhancement. Results show that a dedicated DT remains essential even with OL to achieve satisfactory performance in RW scenarios.
Paper Structure (13 sections, 8 equations, 4 figures, 3 tables)

This paper contains 13 sections, 8 equations, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. System Model and Design with DT
  3. Communication System Model
  4. DT Framework
  5. CSI Feedback with DTs
  6. Experimental Setup
  7. RW Scenario
  8. Virtual Channel in DT
  9. Fidelity of Virtual Channel
  10. Evaluations and Discussions
  11. Performance of the DT-Trained Model in RW Scenarios
  12. Robustness of the DL Model
  13. Conclusions

Figures (4)

  • Figure 1: Neural network architecture and an overview of the efficient AI-driven CSI feedback training approach integrating DTs and OL.
  • Figure 2: (a) Simulated indoor corridor at NSYSU using Wireless InSite®. (b) BS and UE RW measurement point locations. (c) Experimental measurement scenario. (d) Satellite image of the NSYSU campus. (e) Simulated outdoor scenario at the NSYSU campus using Wireless InSite®. (f) 15 cm x 15 cm patch antenna at the BS for the CSI feedback experiment. (g) 7 cm x 15 cm patch antenna at the UE for the CSI feedback experiment. (h) 11.4 cm dipole antenna at the BS for the channel fidelity experiment. (i) 15.5 cm x 17 cm patch antenna at the UE for the channel fidelity experiment.
  • Figure 3: Heatmaps of precoder cosine similarity in (a) DT and (b) RW datasets.
  • Figure 4: RW throughput comparison across different feedback methods.