BumpBusters: Pothole Detection with Machine Learning in the Autonomous City of Buenos Aires

Abstract

The presence of potholes in public roads compromises road safety, affects quality of life, and causes economic harm both to the affected citizens and to society. In this context, a system is proposed for urban areas of the Autonomous City of Buenos Aires (CABA) to detect potholes in real time using the accelerometer and geolocation sensors of mobile devices. For this purpose, a machine learning model is developed that combines the Random Forest technique for pothole detection and K-Means to classify them according to their severity. The prototype is tested using a 1:18 scale vehicle, and the visualizations are geolocated and displayed on a dashboard using Google Maps. As a future extension, integration with platforms such as Waze and Cabify is proposed to facilitate adoption and large-scale distribution.