Do Vehicles Sense, Detect and Locate Speed Bumps?

Abstract

Pavement surface condition is an essential metric in providing quality and safe road infrastructure to the commuters. One of the keys to roadway condition monitoring is the detection and classification of roadway speed bumps which affect driving comfort and transport safety. The paper presents a data-driven framework and related field studies on the use of supervised machine learning and smartphone sensor technology for the detection and georeferencing of speed bumps. The study proposes a low-cost and automated method to obtain up-to-date information about speed bumps, with the use of smartphones mounted on vehicles. The proposed methodology is based on readily available and accurate technologies, it can be utilized in crowd-sourced applications for pavement management systems (PMS) and geographical information system (GIS) implementations, and it has already been field-tested for the detection and classification of cracks, rutting, ravelling, patches and potholes, exhibiting accuracy levels higher than 90%. The smartphone-based data collection and speed-bump detection algorithms discussed in this paper are complemented with robust regression analysis and Random Under Sampling (RUS) Boosted trees classification models. Ongoing work will further investigate the automated measurement of the geometric properties of the detected bumps and their compliance with regulatory requirements.