Project Details

Abstract

Uncrewed Aerial Systems (UAS) hold promise for revolutionizing the inspection of transportation infrastructure by enabling rapid and safe assessments. However, the application of UAS is predominantly limited to detecting surface-level defects, such as visible cracks, due to the reliance on vision sensors. This approach inherently misses subsurface damage, which, to date, requires direct contact-based methods (e.g., ultrasonic, magnetic, and radiographic techniques) that are currently carried out by manual inspection. This project aims to investigate a transformative approach to infrastructure inspection by developing 1) an integrated UAS platform equipped with a continuum robotic arm to enable contact-based inspection, 2) novel machine learning algorithms to fuse multimodal sensors (e.g., vision, ultrasonic) to predict damage modes more accurately. The continuum robotic arm will be based on lightweight and collapsible tensegrity structures, whereas the machine learning algorithms will be based on the recent transformer architecture. This proposed system aims to establish a foundational approach for future developments in multimodal and autonomous infrastructure inspection, significantly advancing the field by overcoming current limitations in damage assessment capabilities.

Project Word Files