Adding Autonomy to Robotic Enabled Sensing

Abstract

The capabilities of most non-destructive testing methods have been combined with some degree of automation in recent years, to enhance data acquisition speed, part coverage and inspection reliability. A plethora of automated or semi-automated inspection systems have been engineered to enable the robotic manipulation of specific types of sensors. Robotic inspection systems are usually operated through off-line programmed tool-paths. This approach works well when an accurate model of the part is available and the robotic inspection takes place in a well-structured environment, where the part position is precisely registered with respect to the robot reference system. However, it makes the inspection setup for each new part very time-consuming and dependent on the skills and experience of the robot programmer. Moreover, the real geometry of a part may significantly deviate from its digital counterpart, resulting in inaccurate tool paths. This work introduces a new approach capable of conferring full autonomy to robotic sensing applications, providing a breakthrough in the stateof-the-art. As a result of this work, fully autonomous single-pass geometric and volumetric inspection of complex parts, using one single robotized sensor, becomes possible. This concept can find wide applicability to the open problems of structural health monitoring of the modern age.