Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Near-Field Spot Beamfocusing: A Correlation-Aware Transfer Learning Approach

Mohammad Amir Fallah, Mehdi Monemi, Mehdi Rasti, Matti Latva-Aho

TL;DR

The paper tackles CSI-free near-field spot beamfocusing with extremely large programmable metasurfaces (ELPMs) by accelerating training via transfer learning. It introduces a correlation-aware similarity metric based on Phase Distribution Images (PDIs) to enable adaptive policy reuse across subarrays and DFPs, augmented by quasi-liquid layers to further speed learning. A policy blending approach leverages previously trained policies to rapidly adapt to moving focal points, achieving up to 8× faster convergence in dynamic scenarios. Numerical results show substantial gains in convergence speed and focal-region quality, with reduced beamfocusing radius and higher steady-state power at the desired focal point. The work enhances practicality of CSI-independent SBF in dynamic near-field applications and provides a framework for TL-enabled intelligent ELPM control.

Abstract

Three-dimensional (3D) spot beamfocusing (SBF), in contrast to conventional angular-domain beamforming, concentrates radiating power within a very small volume in both radial and angular domains in the near-field zone. Recently the implementation of channel-state-information (CSI)-independent machine learning (ML)-based approaches have been developed for effective SBF using extremely large-scale programmable metasurface (ELPMs). These methods involve dividing the ELPMs into subarrays and independently training them with Deep Reinforcement Learning to jointly focus the beam at the desired focal point (DFP). This paper explores near-field SBF using ELPMs, addressing challenges associated with lengthy training times resulting from independent training of subarrays. To achieve a faster CSI-independent solution, inspired by the correlation between the beamfocusing matrices of the subarrays, we leverage transfer learning techniques. First, we introduce a novel similarity criterion based on the phase distribution image (PDI) of subarray apertures. Then we devise a subarray policy propagation scheme that transfers the knowledge from trained to untrained subarrays. We further enhance learning by introducing quasi-liquid layers as a revised version of the adaptive policy reuse technique. We show through simulations that the proposed scheme improves the training speed about 5 times. Furthermore, for dynamic DFP management, we devised a DFP policy blending process, which augments the convergence rate up to 8-fold.

Near-Field Spot Beamfocusing: A Correlation-Aware Transfer Learning Approach

TL;DR

The paper tackles CSI-free near-field spot beamfocusing with extremely large programmable metasurfaces (ELPMs) by accelerating training via transfer learning. It introduces a correlation-aware similarity metric based on Phase Distribution Images (PDIs) to enable adaptive policy reuse across subarrays and DFPs, augmented by quasi-liquid layers to further speed learning. A policy blending approach leverages previously trained policies to rapidly adapt to moving focal points, achieving up to 8× faster convergence in dynamic scenarios. Numerical results show substantial gains in convergence speed and focal-region quality, with reduced beamfocusing radius and higher steady-state power at the desired focal point. The work enhances practicality of CSI-independent SBF in dynamic near-field applications and provides a framework for TL-enabled intelligent ELPM control.

Abstract

Three-dimensional (3D) spot beamfocusing (SBF), in contrast to conventional angular-domain beamforming, concentrates radiating power within a very small volume in both radial and angular domains in the near-field zone. Recently the implementation of channel-state-information (CSI)-independent machine learning (ML)-based approaches have been developed for effective SBF using extremely large-scale programmable metasurface (ELPMs). These methods involve dividing the ELPMs into subarrays and independently training them with Deep Reinforcement Learning to jointly focus the beam at the desired focal point (DFP). This paper explores near-field SBF using ELPMs, addressing challenges associated with lengthy training times resulting from independent training of subarrays. To achieve a faster CSI-independent solution, inspired by the correlation between the beamfocusing matrices of the subarrays, we leverage transfer learning techniques. First, we introduce a novel similarity criterion based on the phase distribution image (PDI) of subarray apertures. Then we devise a subarray policy propagation scheme that transfers the knowledge from trained to untrained subarrays. We further enhance learning by introducing quasi-liquid layers as a revised version of the adaptive policy reuse technique. We show through simulations that the proposed scheme improves the training speed about 5 times. Furthermore, for dynamic DFP management, we devised a DFP policy blending process, which augments the convergence rate up to 8-fold.
Paper Structure (32 sections, 1 theorem, 16 equations, 9 figures, 1 table, 2 algorithms)

This paper contains 32 sections, 1 theorem, 16 equations, 9 figures, 1 table, 2 algorithms.

Table of Contents

  1. Introduction
  2. Motivation and Contributions
  3. Notations:
  4. System Model and Problem Formulation
  5. System Model
  6. Problem Formulation
  7. Subarray Training Propagation
  8. Dynamic DFP Management
  9. Reviewing Existing DRL-Based SBF Solution
  10. A Review on the Distributed DRL-Based SBF Solution Approach monemi20246g
  11. Environment
  12. Actor
  13. Critic
  14. Agent
  15. Limitations and Drawbacks of the Background SBF Scheme
  16. ...and 17 more sections

Key Result

Corollary 1

In the asymptotic case where $N$ is sufficiently large, the BFR minimization problem eq:opt1 is equivalent to maximizing the focused power at the DFP, corresponding to the following optimization problem: where $p$ is formulated in eq:power_def.

Figures (9)

  • Figure 1: Three different types of shaping the beam profile
  • Figure 2: Block diagram of the proposed TL-based SBF system.
  • Figure 3: Phase Distribution Image (PDI) of the ELPM for SBF at different DFPs.
  • Figure 4: Two highly correlated regions in a typical ELPM PDI.
  • Figure 5: Illustration of the PDI and similarity criterion (a) PDI of ELPM with 3-bit phase shifters and a reference typical subarray (the red box). (b) ECC between PDIs of ELPMs subarrays and the reference subarray, for $\theta=0\degree$. (b) ECC between PDIs of ELPMs subarrays and the reference subarray, for $\theta=90\degree$.
  • ...and 4 more figures

Theorems & Definitions (2)

  • Corollary 1
  • proof