[论文解读] Foundation Model-Aided Hierarchical Control for Robust RIS-Assisted Near-Field Communications

Mohammad Ghassemi, Han Zhang+3 more|arXiv (Cornell University)|Jan 6, 2026

Advanced Wireless Communication Technologies被引用 0

一句话总结

该论文提出一个双变换器层次强化学习框架（DT-HDRL），联合处理快速CSI估计与缓慢阻塞预测，以在RIS辅助的近场通信中最大化光谱效率。

ABSTRACT

The deployment of extremely large aperture arrays (ELAAs) in sixth-generation (6G) networks could shift communication into the near-field communication (NFC) regime. In this regime, signals exhibit spherical wave propagation, unlike the planar waves in conventional far-field systems. Reconfigurable intelligent surfaces (RISs) can dynamically adjust phase shifts to support NFC beamfocusing, concentrating signal energy at specific spatial coordinates. However, effective RIS utilization depends on both rapid channel state information (CSI) estimation and proactive blockage mitigation, which occur on inherently different timescales. CSI varies at millisecond intervals due to small-scale fading, while blockage events evolve over seconds, posing challenges for conventional single-level control algorithms. To address this issue, we propose a dual-transformer (DT) hierarchical framework that integrates two specialized transformer models within a hierarchical deep reinforcement learning (HDRL) architecture, referred to as the DT-HDRL framework. A fast-timescale transformer processes ray-tracing data for rapid CSI estimation, while a vision transformer (ViT) analyzes visual data to predict impending blockages. In HDRL, the high-level controller selects line-of-sight (LoS) or RIS-assisted non-line-of-sight (NLoS) transmission paths and sets goals, while the low-level controller optimizes base station (BS) beamfocusing and RIS phase shifts using instantaneous CSI. This dual-timescale coordination maximizes spectral efficiency (SE) while ensuring robust performance under dynamic conditions. Simulation results demonstrate that our approach improves SE by approximately 18% compared to single-timescale baselines, while the proposed blockage predictor achieves an F1-score of 0.92, providing a 769 ms advance warning window in dynamic scenarios.

研究动机与目标

• Motivate robust NFC in 6G using ELAA and RIS to overcome spherical-wave propagation and blockage dynamics.
• Develop a dual-timescale control framework to handle millisecond CSI variations and second-scale blockages.
• Leverage a fast-timescale CSI transformer and a vision transformer for blockage prediction within a hierarchical RL setup.
• Achieve near-optimal spectral efficiency while ensuring proactive robustness under dynamic environmental changes.

提出的方法

• Introduce a dual-transformer (DT) architecture: a fast-timescale transformer for CSI estimation from ray-tracing data and a vision transformer (ViT) for blockage prediction from visual data.
• Embed the transformers within a dual-timescale HDRL framework with a high-level meta-controller selecting LoS or RIS-assisted NLoS modes and a low-level controller optimizing BS beamfocusing and RIS phase shifts using instantaneous CSI.
• Formulate a joint beamfocusing and RIS configuration problem to maximize the sum spectral efficiency under power, unit-modulus, and QoS constraints.
• Train the CSI transformer using MSE loss to estimate effective channels and train the ViT using binary cross-entropy to predict blockage probabilities.
• Employ a two-timescale MDP structure where the meta-controller operates on macro-steps using blockage predictions and the sub-controller operates on micro-steps using real-time CSI to realize actions.

实验结果

研究问题

• RQ1 How can near-field NFC channels with RIS be effectively modeled and estimated using transformer-based architectures?
• RQ2 Can a dual-timescale hierarchical reinforcement learning approach improve spectral efficiency under dynamic blockages compared to single-timescale baselines?
• RQ3 How does proactive blockage prediction integrate with fast CSI estimation to optimize RIS and BS transmission parameters?
• RQ4 What is the performance gain in spectral efficiency and blockage prediction accuracy in realistic NFC scenarios?

主要发现

• The DT-HDRL framework yields approximately 18% improvement in spectral efficiency over single-timescale baselines.
• The blockage predictor achieves an F1-score of 0.92, providing a 769 ms advance warning window in dynamic scenarios.
• A high-fidelity ray-tracing dataset (DeepVerse 6G O1) is used to validate robustness across ELAA dimensions and RIS element counts.
• The dual-timescale design demonstrates real-time feasibility under dynamic blockage conditions with analyzed computational complexity.
• The architecture explicitly accounts for nonlinear near-field phase terms via distance-based features enabling accurate energy focusing.