Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Unfolding-Based Channel Estimation for Wideband TeraHertz Near-Field Massive MIMO Systems

Jiabao Gao, Xiaoming Cheng, Geoffrey Ye Li

TL;DR

The paper tackles accurate channel estimation for wideband Terahertz massive MIMO in the near-field by addressing beam-split-induced sparsity degradation. It introduces frequency-dependent near-field dictionaries to preserve sparsity and unfolds the AMP-SBL algorithm into a learnable deep network, where a CNN-based M-step and an attention mechanism exploit channel patterns across frequency and space. A mixed training scheme with a weighted NMSE loss enables a single robust model that adapts to different configurations while maintaining low complexity. Simulation results demonstrate improved NMSE and reduced computational load compared with baselines, highlighting practical benefits for high-rate THz MIMO systems.

Abstract

The combination of Terahertz (THz) and massive multiple-input multiple-output (MIMO) is promising to meet the increasing data rate demand of future wireless communication systems thanks to the huge bandwidth and spatial degrees of freedom. However, unique channel features such as the near-field beam split effect make channel estimation particularly challenging in THz massive MIMO systems. On one hand, adopting the conventional angular domain transformation dictionary designed for low-frequency far-field channels will result in degraded channel sparsity and destroyed sparsity structure in the transformed domain. On the other hand, most existing compressive sensing-based channel estimation algorithms cannot achieve high performance and low complexity simultaneously. To alleviate these issues, in this paper, we first adopt frequency-dependent near-field dictionaries to maintain good channel sparsity and sparsity structure in the transformed domain under the near-field beam split effect. Then, a deep unfolding-based wideband THz massive MIMO channel estimation algorithm is proposed. In each iteration of the unitary approximate message passing-sparse Bayesian learning algorithm, the optimal update rule is learned by a deep neural network (DNN), whose structure is customized to effectively exploit the inherent channel patterns. Furthermore, a mixed training method based on novel designs of the DNN structure and the loss function is developed to effectively train data from different system configurations. Simulation results validate the superiority of the proposed algorithm in terms of performance, complexity, and robustness.

Deep Unfolding-Based Channel Estimation for Wideband TeraHertz Near-Field Massive MIMO Systems

TL;DR

The paper tackles accurate channel estimation for wideband Terahertz massive MIMO in the near-field by addressing beam-split-induced sparsity degradation. It introduces frequency-dependent near-field dictionaries to preserve sparsity and unfolds the AMP-SBL algorithm into a learnable deep network, where a CNN-based M-step and an attention mechanism exploit channel patterns across frequency and space. A mixed training scheme with a weighted NMSE loss enables a single robust model that adapts to different configurations while maintaining low complexity. Simulation results demonstrate improved NMSE and reduced computational load compared with baselines, highlighting practical benefits for high-rate THz MIMO systems.

Abstract

The combination of Terahertz (THz) and massive multiple-input multiple-output (MIMO) is promising to meet the increasing data rate demand of future wireless communication systems thanks to the huge bandwidth and spatial degrees of freedom. However, unique channel features such as the near-field beam split effect make channel estimation particularly challenging in THz massive MIMO systems. On one hand, adopting the conventional angular domain transformation dictionary designed for low-frequency far-field channels will result in degraded channel sparsity and destroyed sparsity structure in the transformed domain. On the other hand, most existing compressive sensing-based channel estimation algorithms cannot achieve high performance and low complexity simultaneously. To alleviate these issues, in this paper, we first adopt frequency-dependent near-field dictionaries to maintain good channel sparsity and sparsity structure in the transformed domain under the near-field beam split effect. Then, a deep unfolding-based wideband THz massive MIMO channel estimation algorithm is proposed. In each iteration of the unitary approximate message passing-sparse Bayesian learning algorithm, the optimal update rule is learned by a deep neural network (DNN), whose structure is customized to effectively exploit the inherent channel patterns. Furthermore, a mixed training method based on novel designs of the DNN structure and the loss function is developed to effectively train data from different system configurations. Simulation results validate the superiority of the proposed algorithm in terms of performance, complexity, and robustness.
Paper Structure (12 sections, 14 equations, 7 figures, 1 table)

This paper contains 12 sections, 14 equations, 7 figures, 1 table.

Table of Contents

  1. Introduction
  2. System model
  3. Massive MIMO system
  4. Wideband THz near-field channel model
  5. Formulation of the MMV-CS problem
  6. Deep unfolding-based channel estimator
  7. AMP-SBL
  8. AMP-SBL unfolding
  9. Training scheme
  10. Gain analysis
  11. Simulation results
  12. Conclusion and future work

Figures (7)

  • Figure 1: Partially-connected hybrid analog-digital massive MIMO system
  • Figure 2: Impact of dictionaries under the default simulation setting
  • Figure 3: The architecture of the $l$-th AMP-SBL layer
  • Figure 4: Impact of optimizer on DNN training. Three AMP-SBL layers are unfolded.
  • Figure 5: NMSE versus $M$ with different dictionaries
  • ...and 2 more figures