City-scale high-resolution traffic datasets with refined networks for hierarchical traffic control

Abstract

Urban-scale traffic control faces significant challenges due to the limited availability of large-scale and high-resolution datasets to support traffic control strategies across various network topologies and scales. To address this limitation, we present comprehensive traffic datasets for benchmarking traffic control strategies, possibly with hierarchical architectures. The datasets include five cities of varying scales with diverse intersection layouts and network topologies. We highlight the Xuancheng dataset, which adopts a refined network representation with hundreds of intersections, and achieves high spatiotemporal resolution by recording complete vehicle paths across the network. The dataset incorporates vehicle type information and provides historical trip data over long-term periods, with demand patterns exhibiting both periodic fluctuations and holiday-induced variations. Together with the refined road network and the high-resolution trip data, the dataset supports the construction of realistic traffic simulation environments. Furthermore, we derive multi-level traffic state data from the processed datasets, which facilitate the traffic control strategies spanning vehicular, lane-intersection, and network levels.