RadioDiff-v2: Generative Angular Radio Maps for Multi-Beam Selection and Localization

2026-07-08

AI Summary

This work addresses the ill-posedness and poor generalization in predicting angular power spectra (APS) from geometric information under non-line-of-sight (NLOS) conditions in 6G networks, where conventional regression methods often oversmooth and lose multipath structure. The authors formulate APS generation as a perception-distortion trade-off problem and propose a dual-branch generative model based on a one-dimensional diffusion Transformer. Trained via flow matching, the model jointly outputs distributional samples and point estimates, enabling both beam selection and receiver localization. A key innovation is the incorporation of deterministic transport as an inductive bias, which guarantees exact single-step integration of probability flows under concentrated conditional distributions. The architecture further integrates periodic angular encoding, a Fourier-based angular mixer, and a joint velocity-signal head to support multi-task deployment within a single model. In zero-shot evaluation across 99 environments and millions of links, the method achieves state-of-the-art performance: a Wasserstein-1 distance of 0.39 dB, per-bin errors surpassing regression baselines, only 2.43 dB scan loss with 8 beams under NLOS, and a localization error as low as 20.6 pixels using four base stations.

Abstract

Angular radio maps describe the received-power distribution over the angle of arrival and underpin beam selection and receiver localization in sixth-generation (6G) networks. Predicting the angular power spectrum (APS) from geometry is difficult, because the mapping is ill-posed in non-line-of-sight (NLOS) conditions and must generalize to unseen environments. Distortion-minimizing regressors return the conditional mean, which over-smooths the spectrum and erases the multipath structure that downstream tasks need. We cast the task as a perception-distortion problem and propose RadioDiff-v2, a dual-branch one-dimensional diffusion transformer trained with flow matching. It couples periodic angular encoding, adaptive layer-normalization conditioning, a Fourier angular mixer, and joint velocity and clean-signal heads. A per-metric estimator portfolio reads every deployment quantity from this single model, so that samples carry the distribution, the clean-signal head supplies a regression-grade point estimate, Bayes-optimal rules select beams, and the conditional likelihood localizes the receiver. We prove that a concentrated conditional yields a straight probability-flow trajectory that one step integrates exactly, identifying deterministic transport as the correct inductive bias. On a zero-shot test of 99 environments and one million links, RadioDiff-v2 leads every baseline on every metric, with a 0.39 dB Wasserstein-1 distance, per-bin error below the regression baseline, a 2.43 dB eight-beam NLOS sweep loss, and a 20.6-pixel localization error with four base stations. Code is available at https://github.com/UNIC-Lab/RadioDiff-v2.

Problem

• angular radio maps
• beam selection
• receiver localization
• non-line-of-sight
• angular power spectrum

Innovation

• diffusion transformer
• angular radio maps
• flow matching
• perception-distortion tradeoff
• 6G beam selection