Padua Traffic Dataset Combines Flows, Travel Times, and Urban Context

Riccardo Cappi and colleagues from multiple institutions released a comprehensive city-scale traffic dataset derived from Automatic Vehicle Identification (AVI) records in Padua, Italy, covering February to April 2026. The dataset includes traffic volume time series aggregated at 10-minute intervals, along with time-varying trajectory-based flow statistics such as transition probability matrices, average travel times, and flow residuals. To enrich the traffic measurements, the authors integrated urban contextual data: Points of Interest (POIs), demographic information, meteorological variables, and road infrastructure details. All components are encapsulated in a Python class that simplifies access to temporal and contextual data via a spatio-temporal graph representation.

Validation analyses confirmed the dataset captures expected traffic patterns, including morning and evening rush hours, as well as differences between weekdays and weekends. This resource is particularly valuable for researchers and practitioners working on traffic prediction, urban planning, and AI-driven transportation optimization. The paper is available on arXiv under a Physics and Society subject, with links to code and data. The dataset's multi-modal nature and rich contextual features make it a strong foundation for developing and benchmarking spatio-temporal machine learning models in real-world urban environments.