RadioDiff-Flux: Efficient Radio Map Construction via Generative Denoise Diffusion Model Trajectory Midpoint Reuse

TL;DR

RadioDiff-Flux introduces a two-stage latent diffusion framework that reuses diffusion midpoints to decouple static environmental modeling from dynamic refinement in radio map construction. The approach leverages semantic-consistency of midpoints across similar scenes, enabling substantial inference-time reductions with minimal accuracy loss. Theoretical analysis via KL-divergence bounds and empirical results on RadioMapSeer demonstrate up to tens of times speedups (up to ~58× in dynamic scenarios) while preserving high fidelity, making real-time RM updates practical for mobility-aware 6G systems. The work also proposes a practical caching strategy and a reduced-complexity conditioning pipeline, supporting scalable RM generation on resource-constrained edge devices.

Abstract

Accurate radio map (RM) construction is essential to enabling environment-aware and adaptive wireless communication. However, in future 6G scenarios characterized by high-speed network entities and fast-changing environments, it is very challenging to meet real-time requirements. Although generative diffusion models (DMs) can achieve state-of-the-art accuracy with second-level delay, their iterative nature leads to prohibitive inference latency in delay-sensitive scenarios. In this paper, by uncovering a key structural property of diffusion processes: the latent midpoints remain highly consistent across semantically similar scenes, we propose RadioDiff-Flux, a novel two-stage latent diffusion framework that decouples static environmental modeling from dynamic refinement, enabling the reuse of precomputed midpoints to bypass redundant denoising. In particular, the first stage generates a coarse latent representation using only static scene features, which can be cached and shared across similar scenarios. The second stage adapts this representation to dynamic conditions and transmitter locations using a pre-trained model, thereby avoiding repeated early-stage computation. The proposed RadioDiff-Flux significantly reduces inference time while preserving fidelity. Experiment results show that RadioDiff-Flux can achieve up to 50 acceleration with less than 0.15% accuracy loss, demonstrating its practical utility for fast, scalable RM generation in future 6G networks.

Figures (8)

• Figure 1: The illustration of the similarity of latent variables for RMs.
• Figure 2: The illustration of motivation for latent denoise reuse. (a) Visualizes the convergence of diffusion trajectories for semantically similar RMs. (b) and (c) quantitatively measure the similarity of latent variables over the diffusion process. The x-axis represents the diffusion timestep $t$, and the y-axis represents the NMSE between the latent variables of two different scenarios.
• Figure 3: The illustration of latent midpoint reuse for the RM generation framework.
• Figure 4: Illustration of the RM. Pure black regions denote buildings or vehicles, signifying areas impassable to radio signals. The rest of the map is rendered as a grayscale image, with the grayscale level exhibiting a positive correlation to the pathloss value; brighter areas indicate higher pathloss.
• Figure 5: Visual comparison of RM generation under Scenario 1 (Base Station Position Change). Each row presents a distinct test case. The first column displays Ground Truth RMs. Subsequent columns illustrate generated RMs for varying trajectory reuse ratios ($R_{\text{reuse}}$).

Theorems & Definitions (2)

• Theorem 1
• proof