Characterizing the urban road network for automated mobility: a scalable typology for evidence-based policy

The successful deployment of Connected and Automated Vehicles (CAVs) is fundamentally linked to the state of physical road infrastructure. Since current assessment methods lack objective and scalable metrics, in this paper we establish a novel and purely data-driven framework to classify urban road environments on a large scale. Applying unsupervised machine learning to over 7 million road segments across 100 cities, we identify four distinct infrastructure profiles, ranging from complex, signal-rich corridors to basic residential streets. A key finding is that most of the urban road network (approximately 87%) falls into a minimal infrastructure category, highlighting a gap between the idealized, well-structured environments assumed in CAV development and the reality of most urban roads. This research provides an empirical alternative to aggregated readiness scores and offers an objective tool for policymakers to understand infrastructure heterogeneity, guide targeted investments, and develop realistic deployment strategies for the transition toward automated mobility.