Multi-Scale Diffusion Transformer for Jointly Simulating User Mobility and Mobile Traffic Pattern

2025-10-11

AI Summary

Due to the scarcity of real-world user mobility traces and mobile traffic data—stemming from privacy constraints and high acquisition costs—and the prevalent fragmentation in multimodal modeling that fails to capture the coupling between physical movement and network behavior, this paper proposes a multi-scale diffusion Transformer framework. It is the first to jointly integrate wavelet-based multi-resolution decomposition, urban knowledge graph-guided semantic trajectory generation, and cross-modal attention mechanisms for fine-grained co-modeling of trajectories and traffic. Key innovations include: (1) discrete wavelet transform for decoupling multi-scale traffic features; (2) a hybrid denoising network unifying continuous traffic and discrete location sequences; (3) a dynamic transition mechanism leveraging knowledge graph embedding similarity; and (4) multi-scale Transformer architecture capturing cross-modal dependencies. Experiments show significant improvements: Jensen–Shannon divergence decreases by 17.38% (traffic) and 39.53% (trajectories) over SOTA, markedly enhancing generation fidelity and statistical consistency.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

User mobility trajectory and mobile traffic data are essential for a wide spectrum of applications including urban planning, network optimization, and emergency management. However, large-scale and fine-grained mobility data remains difficult to obtain due to privacy concerns and collection costs, making it essential to simulate realistic mobility and traffic patterns. User trajectories and mobile traffic are fundamentally coupled, reflecting both physical mobility and cyber behavior in urban environments. Despite this strong interdependence, existing studies often model them separately, limiting the ability to capture cross-modal dynamics. Therefore, a unified framework is crucial. In this paper, we propose MSTDiff, a Multi-Scale Diffusion Transformer for joint simulation of mobile traffic and user trajectories. First, MSTDiff applies discrete wavelet transforms for multi-resolution traffic decomposition. Second, it uses a hybrid denoising network to process continuous traffic volumes and discrete location sequences. A transition mechanism based on urban knowledge graph embedding similarity is designed to guide semantically informed trajectory generation. Finally, a multi-scale Transformer with cross-attention captures dependencies between trajectories and traffic. Experiments show that MSTDiff surpasses state-of-the-art baselines in traffic and trajectory generation tasks, reducing Jensen-Shannon divergence (JSD) across key statistical metrics by up to 17.38% for traffic generation, and by an average of 39.53% for trajectory generation. The source code is available at: https://github.com/tsinghua-fib-lab/MSTDiff .

Problem

• Simulating coupled user mobility and mobile traffic patterns jointly
• Overcoming data scarcity due to privacy concerns and collection costs
• Capturing cross-modal dynamics between physical movement and cyber behavior

Innovation

• Multi-Scale Diffusion Transformer for joint simulation
• Discrete wavelet transforms for multi-resolution decomposition
• Cross-attention Transformer capturing trajectory-traffic dependencies