Diffusion-Based Generative Priors for Efficient Beam Alignment in Directional Networks

TL;DR

This paper introduces a diffusion-based probabilistic model for beam alignment in mmWave and THz systems, improving accuracy and efficiency over traditional methods by capturing uncertainty and guiding beam sweeping.

Contribution

It recasts beam alignment as a generative task using a conditional diffusion model to learn probabilistic beam priors from geometric and multipath features.

Findings

Achieves high ranking performance with Hit@1 ≈ 0.61 and Hit@3 ≈ 0.90

Outperforms deterministic classifiers with a 180% improvement in Hit@1

Reduces beam training overhead while maintaining SNR

Abstract

Beam alignment is a key challenge in directional mmWave and THz systems, where narrow beams require accurate yet low-overhead training. Existing learning-based approaches typically predict a single beam and do not quantify uncertainty, limiting adaptive beam sweeping. We recast beam alignment as a generative task and propose a conditional diffusion model that learns a probabilistic beam prior from compact geometric and multipath features. The learned priors guide top-$k$ sweeps and capture the SNR loss induced by limited probing. Using a ray-traced DeepMIMO scenario with an 8-beam DFT codebook, our best conditional diffusion model achieves strong ranking performance (Hit@1 $\approx 0.61$, Hit@3 $\approx 0.90$, Hit@5 $\approx 0.97$) while preserving SNR at small sweep budgets. Compared with a deterministic classifier baseline, diffusion improves Hit@1 by about 180\%. Results further highlight the importance of informative conditioning and the ability of diffusion sampling to flexibly trade accuracy for computational efficiency. The proposed diffusion framework achieves substantial improvements in small-$k$ Hit rates, translating into reduced beam training overhead and enabling low-latency, energy-efficient beam alignment for mmWave and THz systems while preserving received SNR.