Autonomous and Crowdsourced Sensing Technologies for Smart Infrastructures

Abstract

The demands of deteriorating civil infrastructure require effective methods for its integrity monitoring. While various sensor technologies have been developed for an effective monitoring of the infrastructure and enormous sensor data associated with its health conditions has become available, there still exist some barriers that prevent the widespread adoptions of the technologies and data in the real-world practice of civil infrastructure monitoring. This dissertation seeks to fill the possible missing gaps between the available sensing technologies and their real-world facilitation, particularly focusing on the mesoscale structural strain sensing and crowd-sourced smartphone data in the context of infrastructure monitoring. First, to put it in context, the recent advances in functional material science have enabled the soft elastomeric capacitor (SEC) that allows such mesoscale strain sensing thanks to its highly stretchable and large-coverage feature. Despite its validated mesoscale strain sensing performances under a laboratory environment, full-scale and real-world applications of the SEC are still limited due to the complicated and yet wired data acquisition that requires extremely sensitive manual intervention. Second, various embedded sensors in smartphones can collect lots of information from the daily use of vehicles with the smartphones in them, particularly regarding roadway health conditions. However, almost infinite variations of vehicle types, dynamic characteristics, and driving speeds raise a critical calibration issue which hinders the wide-spread practical use of the smartphone method for road condition monitoring. This research aims to address the two specific issues, as examples, by 1) developing a smart sensor hardware and software that can implement the mesoscale strain sensing capability into a wireless sensor network with fully automated features and 2) exploring deep machine leaning approaches that can extract the road roughness information from crowdsourced smartphone sensor data which is measured while driving regardless of vehicle characteristics (i.e., calibration-free). The performances of the developed wireless hardware and software for mesoscale strain sensing and the deep learning architecture for smartphone-based road roughness monitoring have been experimentally validated. This research has resulted in two achievements; 1) an autonomous wireless smart sensor for mesoscale structural strain sensing and 2) a calibration-free road roughness monitoring using anonymous vehicles and crowdsourced smartphone data.