Perception-driven admittance-based surface alignment for robotic visual inspection

Abstract

Precision visual inspection requires precise alignment between a robot end-effector and local surface geometry despite perception noise, surface irregularities, and real-time human input. This paper presents a closed-loop robotic orientation-control pipeline with an admittance-based framework that unifies operator input and perception-driven surface alignment. The end-effector is modeled as a virtual sphere moving through a viscous medium, producing synchronized, compliant motion from orientation error and operator commands. Experiments on a 6-DOF manipulator demonstrate stable normal tracking and a final mean orientation error of 0.4 degrees.

Antara Banerjee

Ph.D. student in Mechanical Engineering

Colin E. Acton

Ph.D. student in Mechanical Engineering

Xu Chen

McMinn Professor at University of Washington | Director of Boeing Advanced Research Collaboration at UW | Director of UW College of Engineering Robotics Education