Digital Twin Aided Massive MIMO CSI Feedback: Exploring the Impact of Twinning Fidelity

TL;DR

This work tackles the high data-collection overhead of training DL-based CSI feedback for massive MIMO by introducing site-specific digital twins that synthesize CSI data from an EM 3D model, ray tracing, and hardware models. It presents a fidelity-decomposition framework across $3D$ geometry, $EM$ material, ray tracing, and hardware modeling, and proposes an online refinement workflow using limited real-world data to bridge residual distribution gaps. A data-selection strategy and a CSINet+-based DL design enable efficient refinement, achieving high CSI reconstruction performance at reduced real-world data cost. Experiments in a Boston downtown scenario demonstrate that site-specific twins outperform generic datasets, and fidelity in geometry, ray tracing, and hardware modeling significantly influence CSI reconstruction and spectral efficiency.

Abstract

Deep learning (DL) techniques have demonstrated strong performance in compressing and reconstructing channel state information (CSI) while reducing feedback overhead in massive MIMO systems. A key challenge, however, is their reliance on extensive site-specific training data, whose real-world collection incurs significant overhead and limits scalability across deployment sites. To address this, we propose leveraging site-specific digital twins to assist the training of DL-based CSI compression models. The digital twin integrates an electromagnetic (EM) 3D model of the environment, a hardware model, and ray tracing to produce site-specific synthetic CSI data, allowing DL models to be trained without the need for extensive real-world measurements. We further develop a fidelity analysis framework that decomposes digital twin quality into four key aspects: 3D geometry, material properties, ray tracing, and hardware modeling. We explore how these factors influence the reliability of the data and model performance. To enhance the adaptability to real-world environments, we propose a refinement strategy that incorporates a limited amount of real-world data to fine-tune the DL model pre-trained on the digital twin dataset. Evaluation results show that models trained on site-specific digital twins outperform those trained on generic datasets, with the proposed refinement method effectively enhancing performance using limited real-world data. The simulations also highlight the importance of digital twin fidelity, especially in 3D geometry, ray tracing, and hardware modeling, for improving CSI reconstruction quality. This analysis framework offers valuable insights into the critical fidelity aspects, and facilitates more efficient digital twin development and deployment strategies for various wireless communication tasks.

Figures (15)

• Figure 1: This figure shows the key idea of utilizing a digital twin to train the DL CSI model. A small amount of real-world data can then be used to compensate for the mismatch between the digital twin and real-world CSI data, and refine the DL model to achieve higher performance.
• Figure 2: This figure illustrates the point cloud generated at different sampling densities, where sampling density refers to the number of points per square meter. A higher sampling density captures more details of the building model, resulting in greater geometric fidelity after mesh reconstruction.
• Figure 3: This figure shows the data selection process for the model refinement. The UE first compresses and recovers the CSI matrix using the DL model. The UE then calculates the reconstruction NMSE and feeds back the CSI matrices for which the NMSE is larger than a threshold $\eta$.
• Figure 4: This figure illustrates the geometric layout of the target scenario, representing a real-world section of Downtown Boston. The BS is positioned along a vertical street, oriented toward the negative y-axis. The service area is highlighted in blue, with a foliage object situated in the middle.
• Figure 5: This figure illustrates the geometric layout of the baseline scenario, which is the O1 scenario in the DeepMIMO dataset deepmimo. We choose 'BS 6' as the base station and select rows 1150 to 1650 as the service area, which is marked in blue.