Transformer-Based Sparse CSI Estimation for Non-Stationary Channels

2025-11-03

AI Summary

To address the high overhead and poor generalizability of CSI estimation under non-stationary channels, this paper proposes a hybrid model-driven and data-driven framework. Methodologically, we design a physics-informed composite loss function—incorporating phase alignment, time-frequency smoothness, and correlation consistency constraints—and integrate it with a block-wise patch-based self-attention mechanism and Flash-Attention-enabled Transformer architecture, enabling robust CSI reconstruction under dynamic pilot allocation in 3GPP NR. Our key innovation lies in the first integration of block-wise self-attention into a physics-constrained optimization pipeline, jointly leveraging channel prior knowledge and data adaptability. Experiments demonstrate a ~13 dB reduction in normalized mean squared error over LMMSE and LSTM baselines, significant BER improvement, 16× pilot overhead reduction, and substantial gains in spectral efficiency and link reliability.

Technology Category

• Search and Optimization: Learning to Search
• Computer Vision: Learning & Optimization for CV
• Intelligent Robots: State Estimation

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Security and Privacy: Large-scale security measurements

Abstract

Accurate and efficient estimation of Channel State Information (CSI) is critical for next-generation wireless systems operating under non-stationary conditions, where user mobility, Doppler spread, and multipath dynamics rapidly alter channel statistics. Conventional pilot aided estimators incur substantial overhead, while deep learning approaches degrade under dynamic pilot patterns and time varying fading. This paper presents a pilot-aided Flash-Attention Transformer framework that unifies model-driven pilot acquisition with data driven CSI reconstruction through patch-wise self-attention and a physics aware composite loss function enforcing phase alignment, correlation consistency, and time frequency smoothness. Under a standardized 3GPP NR configuration, the proposed framework outperforms LMMSE and LSTM baselines by approximately 13 dB in phase invariant normalized mean-square error (NMSE) with markedly lower bit-error rate (BER), while reducing pilot overhead by 16 times. These results demonstrate that attention based architectures enable reliable CSI recovery and enhanced spectral efficiency without compromising link quality, addressing a fundamental bottleneck in adaptive, low-overhead channel estimation for non-stationary 5G and beyond-5G networks.

Problem

• Estimating Channel State Information accurately under non-stationary wireless conditions
• Reducing substantial pilot overhead in conventional CSI estimation methods
• Addressing performance degradation of deep learning in dynamic channel environments

Innovation

• Flash-Attention Transformer framework for CSI estimation
• Unifies model-driven pilot acquisition with data-driven reconstruction
• Physics-aware composite loss function enforces multiple channel properties