Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SAIL: Unsupervised Spatial-Angular Interpretable Feature Learning for RF Map Synthesis

Sopan Sarkar, Marwan Krunz

Abstract

In wireless networks, radio-frequency (RF) maps are critical for tasks such as capacity planning, coverage estimation, and localization. Traditional approaches for obtaining RF maps, including site surveys and ray-tracing simulations, are labor-intensive or computationally expensive, especially at high frequencies and dense network deployments. Generative AI offers a promising alternative for RF map synthesis. However, supervised methods are often infeasible due to the lack of reliable labeled training data, while purely unsupervised methods typically lack explicit control over the synthesis process. To address these challenges, we propose SAIL (Spatial-Angular Interpretable Feature Learning), a generative adversarial network (GAN)-based framework that learns interpretable and controllable latent variables directly from unlabeled RF maps and enables targeted RF map synthesis at inference time through latent-variable manipulation. SAIL builds on the information-maximizing GAN (InfoGAN) principle to learn a structured representation comprising: (i) a categorical latent variable that captures discrete floor-plan regions associated with Tx location and (ii) a continuous latent variable that captures angular variations corresponding to the Tx boresight angle, without requiring any location or orientation supervision during training. We further adopt a Wasserstein GAN objective with a gradient penalty to improve training stability and synthesis quality. Our results using ray-tracing-based RF maps indicate that SAIL learns physically meaningful spatial-angular factors and enables fast controlled RF map synthesis, achieving an average SSIM of 0.8576 and an average PSNR of 23.33 dB relative to ray-tracing simulations.

SAIL: Unsupervised Spatial-Angular Interpretable Feature Learning for RF Map Synthesis

Abstract

In wireless networks, radio-frequency (RF) maps are critical for tasks such as capacity planning, coverage estimation, and localization. Traditional approaches for obtaining RF maps, including site surveys and ray-tracing simulations, are labor-intensive or computationally expensive, especially at high frequencies and dense network deployments. Generative AI offers a promising alternative for RF map synthesis. However, supervised methods are often infeasible due to the lack of reliable labeled training data, while purely unsupervised methods typically lack explicit control over the synthesis process. To address these challenges, we propose SAIL (Spatial-Angular Interpretable Feature Learning), a generative adversarial network (GAN)-based framework that learns interpretable and controllable latent variables directly from unlabeled RF maps and enables targeted RF map synthesis at inference time through latent-variable manipulation. SAIL builds on the information-maximizing GAN (InfoGAN) principle to learn a structured representation comprising: (i) a categorical latent variable that captures discrete floor-plan regions associated with Tx location and (ii) a continuous latent variable that captures angular variations corresponding to the Tx boresight angle, without requiring any location or orientation supervision during training. We further adopt a Wasserstein GAN objective with a gradient penalty to improve training stability and synthesis quality. Our results using ray-tracing-based RF maps indicate that SAIL learns physically meaningful spatial-angular factors and enables fast controlled RF map synthesis, achieving an average SSIM of 0.8576 and an average PSNR of 23.33 dB relative to ray-tracing simulations.
Paper Structure (11 sections, 11 equations, 6 figures)

This paper contains 11 sections, 11 equations, 6 figures.

Table of Contents

  1. Introduction
  2. Preliminaries: Introduction to InfoGAN
  3. SAIL: Architecture and Design
  4. SAIL Architecture
  5. SAIL Objective Function
  6. Ray-Tracing-Based Simulated Dataset
  7. Performance Evaluation
  8. Learning and Controlling Tx Location via ${\boldsymbol{c}}_s$
  9. Learning and Controlling Boresight Direction via ${\boldsymbol{c}}_b$
  10. RF Map Fidelity Relative to Ray Tracing
  11. Conclusions

Figures (6)

  • Figure 1: InfoGAN architecture, comprising a generator $G$, a discriminator $D$, and a recognition network $Q$.
  • Figure 2: Generator ($G$), discriminator ($D$), and recognition network ($Q$) in SAIL. The notation $(m\times m\times n)$ below each block denotes the tensor shape in the format (height, width, channels).
  • Figure 3: Ray-tracing-based simulated dataset: (a) the indoor floor plan of dimension $20 \times 20~\text{m}^2$ and (b) a corresponding RF map generated using $8\times8$ UPA at 28 GHz, with the Tx boresight direction at an azimuth angle of $36^\circ$.
  • Figure 4: RF maps generated by SAIL with fixed ${\boldsymbol{z}}$ and ${\boldsymbol{c}}_b$ while varying ${\boldsymbol{c}}_s$.
  • Figure 5: Tx locations for each region of the floor plan, shown for both synthetic (solid) and simulated (dashed) RF maps.
  • ...and 1 more figures