NSF-BSF: RI: Small: Provably High-Quality Robot Inspection Planning - Theory and Application

University Of North Carolina At Chapel Hill, Chapel Hill NC

Investigators

Abstract

Inspecting the surfaces of objects is a common task required in a variety of applications, from diagnosing a diseased organ by inspecting its tissue surface to verifying the safety of a bridge by inspecting its structure. Robots with mounted sensors have the potential to efficiently and effectively perform such inspection tasks by automatically moving the sensor to view the region of interest on the object's surface while avoiding obstacles and satisfying motion constraints. For example, an endoscopic needle-based robot has the potential to perform an inspection task inside the human body to help diagnose certain diseases that manifest on the surfaces of internal organs. As another example, drones have the potential to be widely used to efficiently inspect the complex geometry of bridges, which is increasingly important since almost 40% of the nation's bridges exceed their 50-year design life, and regular inspections are critical to ensuring bridge safety. In this project, the research team will investigate the problem of robot inspection planning. Specifically, the team will develop and analyze new, efficient computational methods to plan motions for a robot to enable the robot to autonomously maximize the quality of an inspection while safely avoiding obstacles. The goal of this project is to provide a theoretically-grounded efficient and effective algorithmic framework for robust inspection planning demonstrated on real-world robotics applications. Specifically, the research team plans to provide a rigorous explanation to why and when different variants of the inspection-planning problem are computationally hard. Pinpointing exactly why the inspection problem is computationally hard will then enable the research team to develop an efficient algorithmic framework to solve the fundamental version of the inspection-planning problem, namely, when there is no uncertainty with respect to the robot's kinematic model, sensor model, or environment model. Finally, the research team plans to extend the new algorithmic framework to account for different sources of uncertainty in order to make the results applicable to real-world problems. This will be done by combining tools from diverse domains such as computational geometry, graph theory, optimization, and machine learning. In all stages of the project, the research team will demonstrate the results using several applications in a laboratory setting, including the inspection of patient anatomy using needle-based robots and the inspection of bridges using drones. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.