Machine learning with edge and cloud computing for crowdsensing of pavement conditions

PI: Liang Cheng
University: Lehigh University

Modern methods for road pavement condition monitoring take advantage of new technologies in precision instrumentation to realize automatic measurements. However, these current pavement-monitoring procedures are time-consuming and costly. Recently, the U.S. Department of Transportation (DOT) has initiated a Connected Vehicle Program, which promotes applying
vehicle-to-X (V2X) data to pavement monitoring. A study by the Center for Automotive Research and administrated by Michigan DOT reports that using V2X data for pavement monitoring is possible, but it will require novel and proactive techniques of data use and management.

The objective of this seed project is to design, implement, and test a machine-learning based approach that is enabled by edge and cloud computing and that allows for pavement condition monitoring in a low-cost, reliable, and rapid manner. The proposed system collects crowdsensing data from in-vehicle accelerometers in smartphones, dashboard, rear-view or smartphone cameras, and GPS, and generates accelerometer-based roughness indictors (ARI) and photo-based distress indictors (PDI) for existing pavement conditions.