Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

BeamCKMDiff: Beam-Aware Channel Knowledge Map Construction via Diffusion Transformer

Le Zhao, Yining Wang, Xinyi Wang, Zesong Fei, Yong Zeng

TL;DR

This work tackles environment-aware CKM construction for 6G by enabling high-fidelity, beam-aware maps conditioned on continuous beamvectors. It introduces BeamCKMDiff, a diffusion-transformer framework that uses adaptive layer normalization (adaLN) to inject continuous beam information, decoupling beam steering from environmental geometry. A VAE provides perceptual compression of CKMs, a condition encoder captures topology, and a DiT-based backbone performs diffusion-based reconstruction under beam guidance. On a ray-traced urban dataset, BeamCKMDiff achieves NMSE below $-20$ dB and outperforms discriminative and diffusion baselines, with sub-second inference time, demonstrating strong practical potential for dynamic beam management.

Abstract

Channel knowledge map (CKM) is emerging as a critical enabler for environment-aware 6G networks, offering a site-specific database to significantly reduce pilot overhead. However, existing CKM construction methods typically rely on sparse sampling measurements and are restricted to either omnidirectional maps or discrete codebooks, hindering the exploitation of beamforming gain. To address these limitations, we propose BeamCKMDiff, a generative framework for constructing high-fidelity CKMs conditioned on arbitrary continuous beamforming vectors without site-specific sampling. Specifically, we incorporate a novel adaptive layer normalization (adaLN) mechanism into the noise prediction network of the Diffusion Transformer (DiT). This mechanism injects continuous beam embeddings as {global control parameters}, effectively steering the generative process to capture the complex coupling between beam patterns and environmental geometries. Simulation results demonstrate that BeamCKMDiff significantly outperforms state-of-the-art baselines, achieving superior reconstruction accuracy in capturing main lobes and side lobes.

BeamCKMDiff: Beam-Aware Channel Knowledge Map Construction via Diffusion Transformer

TL;DR

This work tackles environment-aware CKM construction for 6G by enabling high-fidelity, beam-aware maps conditioned on continuous beamvectors. It introduces BeamCKMDiff, a diffusion-transformer framework that uses adaptive layer normalization (adaLN) to inject continuous beam information, decoupling beam steering from environmental geometry. A VAE provides perceptual compression of CKMs, a condition encoder captures topology, and a DiT-based backbone performs diffusion-based reconstruction under beam guidance. On a ray-traced urban dataset, BeamCKMDiff achieves NMSE below dB and outperforms discriminative and diffusion baselines, with sub-second inference time, demonstrating strong practical potential for dynamic beam management.

Abstract

Channel knowledge map (CKM) is emerging as a critical enabler for environment-aware 6G networks, offering a site-specific database to significantly reduce pilot overhead. However, existing CKM construction methods typically rely on sparse sampling measurements and are restricted to either omnidirectional maps or discrete codebooks, hindering the exploitation of beamforming gain. To address these limitations, we propose BeamCKMDiff, a generative framework for constructing high-fidelity CKMs conditioned on arbitrary continuous beamforming vectors without site-specific sampling. Specifically, we incorporate a novel adaptive layer normalization (adaLN) mechanism into the noise prediction network of the Diffusion Transformer (DiT). This mechanism injects continuous beam embeddings as {global control parameters}, effectively steering the generative process to capture the complex coupling between beam patterns and environmental geometries. Simulation results demonstrate that BeamCKMDiff significantly outperforms state-of-the-art baselines, achieving superior reconstruction accuracy in capturing main lobes and side lobes.
Paper Structure (17 sections, 14 equations, 3 figures, 2 tables)

This paper contains 17 sections, 14 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. System Model and Problem Formulation
  3. System Model
  4. Problem Formulation
  5. Proposed BeamCKMDiff
  6. Perceptual Compression via VAE
  7. Condition Encoder
  8. Continuous Beam Conditioning via DiT
  9. Latent Feature Fusion
  10. Beam-Aware adaLN
  11. Diffusion Training and Interference
  12. Training / Diffusion Process
  13. Inference / Reverse Process
  14. Simulation Results
  15. Simulation Settings
  16. ...and 2 more sections

Figures (3)

  • Figure 1: The illustration of the proposed BeamCKMDiff. The VAE is employed to encode the CKM into latent space, thereby reducing the dimension of the diffusion model.
  • Figure 2: Comparisons of constructed CKM via different methods under seen GBS locations and unseen beams.
  • Figure 3: Comparisons of constructed CKM via different methods under unseen GBS locations and beams.