UrbanFM: Scaling Urban Spatio-Temporal Foundation Models

TL;DR

UrbanFM introduces a scalable approach to modeling urban spatio-temporal data, leveraging large-scale datasets, novel computational units, and a minimalist architecture to improve generalization across cities and tasks.

Contribution

The paper systematically investigates scaling principles for urban spatio-temporal models, proposing new datasets, units, and architecture to enhance generalization and robustness.

Findings

UrbanFM achieves strong zero-shot transfer to unseen cities and tasks.

WorldST standardizes diverse urban signals into a billion-scale dataset.

MiniST enables unified representation of grid and sensor data.

Abstract

Urban systems, as dynamic complex systems, continuously generate spatio-temporal data streams that encode the fundamental laws of human mobility and city evolution. While AI for Science has witnessed the transformative power of foundation models in disciplines like genomics and meteorology, urban computing remains fragmented due to "scenario-specific" models, which are overfitted to specific regions or tasks, hindering their generalizability. To bridge this gap and advance spatio-temporal foundation models for urban systems, we adopt scaling as the central perspective and systematically investigate two key questions: what to scale and how to scale. Grounded in first-principles analysis, we identify three critical dimensions: heterogeneity, correlation, and dynamics, aligning these principles with the fundamental scientific properties of urban spatio-temporal data. Specifically, to address heterogeneity through data scaling, we construct WorldST. This billion-scale corpus standardizes diverse physical signals, such as traffic flow and speed, from over 100 global cities into a unified data format. To enable computation scaling for modeling correlations, we introduce the MiniST unit, a novel split mechanism that discretizes continuous spatio-temporal fields into learnable computational units to unify representations of grid-based and sensor-based observations. Finally, addressing dynamics via architecture scaling, we propose UrbanFM, a minimalist self-attention architecture designed with limited inductive biases to autonomously learn dynamic spatio-temporal dependencies from massive data. Furthermore, we establish EvalST, the largest-scale urban spatio-temporal benchmark to date. Extensive experiments demonstrate that UrbanFM achieves remarkable zero-shot generalization across unseen cities and tasks, marking a pivotal first step toward large-scale urban spatio-temporal foundation models.