Title:What is a typical signalized intersection in a city? A pipeline for intersection data imputation from OpenStreetMap

Abstract:Signalized intersections, arguably the most complicated type of traffic scenario, are essential to urban mobility systems. With recent advancements in intelligent transportation technologies, signalized intersections have great prospects for making transportation greener, safer, and faster. Several studies have been conducted focusing on intersection-level control and optimization. However, arbitrarily structured signalized intersections that are often used do not represent the ground-truth distribution, and there is no standardized way that exists to extract information about real-world signalized intersections. As the largest open-source map in the world, OpenStreetMap (OSM) has been used by many transportation researchers for a variety of studies, including intersection-level research such as adaptive traffic signal control and eco-driving. However, the quality of OSM data has been a serious concern.
In this paper, we propose a pipeline for effectively extracting information about signalized intersections from OSM and constructing a comprehensive dataset. We thoroughly discuss challenges related to this task and we propose our solution for each challenge. We also use Salt Lake City as an example to demonstrate the performance of our methods. The pipeline has been published as an open-source Python library so everyone can freely download and use it to facilitate their research. Hopefully, this paper can serve as a starting point that inspires more efforts to build a standardized and systematic data pipeline for various types of transportation problems.