Denoising Refinement Diffusion Models for Simultaneous Generation of Multi-scale Mobile Network Traffic

TL;DR

ZoomDiff introduces a unified diffusion-based framework for simultaneous multi-scale mobile traffic generation by mapping Urban Environmental Context to traffic across BS, cells, and grid scales. It uses Denoising Refinement Diffusion Models (DRDM) with Hierarchical Noise-adding Process and Resolution Refinement Denoising Process, supported by environment-aware conditioning via Spatiotemporal Positional Encoding and Urban Environment Context. A diffusion-prior guidance mechanism and tailored noise scheduling (including Segmented Noise/Adding and Segmented Denoising) enable intermediate states to correspond to different resolutions within a single denoising trajectory, yielding superior multi-scale generation and strong cross-city generalization. Empirical results on real-world data show ZoomDiff outperforms state-of-the-art baselines by at least 18.4% on multi-scale generation tasks, while achieving efficiency gains and robust refinement from coarse to fine resolutions.

Abstract

The planning, management, and resource scheduling of cellular mobile networks require joint estimation of mobile traffic across different layers and nodes. Mobile traffic generation can proactively anticipate user demands and capture the dynamics of network load. However, existing methods mainly focus on generating traffic at a single spatiotemporal resolution, making it difficult to jointly model multi-scale traffic patterns. In this paper, we propose ZoomDiff, a diffusion-based model for multi-scale mobile traffic generation. ZoomDiff maps urban environmental context into mobile traffic with multiple spatial and temporal resolutions through a set of customized Denoising Refinement Diffusion Models (DRDM). DRDM employs a multi-stage noise-adding and denoising mechanism, enabling different stages to generate traffic at distinct spatiotemporal resolutions. This design aligns the progressive denoising process with hierarchical network layers, including base stations, cells, and grids of varying granularities. Experiments on real-world mobile traffic datasets show that ZoomDiff achieves at least an 18.4% improvement over state-of-the-art baselines in multi-scale traffic generation tasks. Moreover, ZoomDiff demonstrates strong efficiency and cross-city generalization, highlighting its potential as a powerful generative framework for modeling multi-scale mobile network dynamics.

Figures (13)

• Figure 1: Typical cellular mobile network, with multi-layer network nodes including BSs, cells, and users. Grids are aggregations of users within regularly defined geographical areas. The curves in the figure illustrate BS, cell, and grid traffic patterns with different temporal and spatial resolutions.
• Figure 2: Align the spatiotemporal refinement relationships of multi-scale mobile traffic with the denoising process of diffusion models.
• Figure 3: Heatmaps of RV-coefficients between different levels of traffic across various AoI types.
• Figure 4: Time series decomposition of BS, Cell, and 50m grid traffic.
• Figure 5: Overall framework of ZoomDiff. It consists of two components: the Environment-aware Conditional Information and the Denoising Refinement Diffusion Models (DRDM).