RadioFlow: Efficient Radio Map Construction Framework with Flow Matching

2025-10-10

AI Summary

Existing diffusion models for radio map (RM) generation suffer from excessive parameter counts, slow denoising iterations, and high inference latency—hindering their applicability to real-time electromagnetic digital twin requirements in 6G networks. To address this, we propose RadioFlow, a novel framework based on flow matching that models the continuous transformation from noise to data, enabling efficient single-step sampling. RadioFlow integrates sparse-aware priors with a lightweight network architecture to jointly optimize reconstruction fidelity and computational efficiency. Experiments demonstrate that RadioFlow reduces model parameters by 8× and accelerates inference by over 4× compared to the state-of-the-art RadioDiff, while maintaining high-fidelity RM reconstruction. It achieves new state-of-the-art performance in RM generation, offering an efficient and practical paradigm for real-time RM synthesis and 6G digital twin deployment.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Diffusion Models for Vision

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Data management and stream processing for Web, mobile and wireless applications
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Accurate and real-time radio map (RM) generation is crucial for next-generation wireless systems, yet diffusion-based approaches often suffer from large model sizes, slow iterative denoising, and high inference latency, which hinder practical deployment. To overcome these limitations, we propose extbf{RadioFlow}, a novel flow-matching-based generative framework that achieves high-fidelity RM generation through single-step efficient sampling. Unlike conventional diffusion models, RadioFlow learns continuous transport trajectories between noise and data, enabling both training and inference to be significantly accelerated while preserving reconstruction accuracy. Comprehensive experiments demonstrate that RadioFlow achieves state-of-the-art performance with extbf{up to 8$ imes$ fewer parameters} and extbf{over 4$ imes$ faster inference} compared to the leading diffusion-based baseline (RadioDiff). This advancement provides a promising pathway toward scalable, energy-efficient, and real-time electromagnetic digital twins for future 6G networks. We release the code at href{https://github.com/Hxxxz0/RadioFlow}{GitHub}.

Problem

• Efficient radio map construction with flow matching
• Overcoming slow iterative denoising in diffusion models
• Achieving real-time electromagnetic digital twins for 6G

Innovation

• Flow-matching framework for radio map generation
• Single-step efficient sampling for high-fidelity reconstruction
• Continuous transport trajectories accelerate training and inference