Noise Matters: Diffusion Model-based Urban Mobility Generation with Collaborative Noise Priors

TL;DR

This work tackles the privacy challenges of using real mobility data by introducing CoDiffMob, a diffusion-model framework that uses collaborative noise priors to jointly encode individual movement patterns and population dynamics. By coupling a two-stage collaborative noise-prior construction with a UNet-based diffusion denoiser, it produces synthetic urban mobility trajectories that faithfully reflect both personal and collective patterns while reducing privacy risk. Empirical results on ISP and MME datasets show substantial improvements over baselines in trajectory realism and flow accuracy, along with strong privacy protections (low susceptibility to membership inference) and improved utility for downstream tasks. The approach advances synthetic mobility generation for sustainable, equitable, and privacy-preserving urban research and planning, with release of code and data to enable public scrutiny and reuse.

Abstract

With global urbanization, the focus on sustainable cities has largely grown, driving research into equity, resilience, and urban planning, which often relies on mobility data. The rise of web-based apps and mobile devices has provided valuable user data for mobility-related research. However, real-world mobility data is costly and raises privacy concerns. To protect privacy while retaining key features of real-world movement, the demand for synthetic data has steadily increased. Recent advances in diffusion models have shown great potential for mobility trajectory generation due to their ability to model randomness and uncertainty. However, existing approaches often directly apply identically distributed (i.i.d.) noise sampling from image generation techniques, which fail to account for the spatiotemporal correlations and social interactions that shape urban mobility patterns. In this paper, we propose CoDiffMob, a diffusion model for urban mobility generation with collaborative noise priors, we emphasize the critical role of noise in diffusion models for generating mobility data. By leveraging both individual movement characteristics and population-wide dynamics, we construct novel collaborative noise priors that provide richer and more informative guidance throughout the generation process. Extensive experiments demonstrate the superiority of our method, with generated data accurately capturing both individual preferences and collective patterns, achieving an improvement of over 32%. Furthermore, it can effectively replace web-derived mobility data to better support downstream applications, while safeguarding user privacy and fostering a more secure and ethical web. This highlights its tremendous potential for applications in sustainable city-related research. The code and data are available at https://github.com/tsinghua-fib-lab/CoDiffMob.

Figures (8)

• Figure 1: Correlations between noise and trajectories. (a) shows the t-SNE plot of the noise corresponding to the trajectories. (b) and (c) illustrate the correlations between the noise and trajectories from the spatial and temporal domain.
• Figure 2: The overall workflow of mobility generation with collaborative noise priors.
• Figure 3: Visualization of the observed flows and generated flows on ISP dataset.
• Figure 4: Evaluation results of uniqueness testing.
• Figure 5: Success ratio of membership inference attack using different classification algorithms.