A Tactile-Driven System for Automated Wire-Harness Insertion
Namya Lakkavalli Giridhar
Abstract
This paper presents a system with learned tactile perception for grasp localization during wire insertion, a step of wire-harness assembly that vision alone does not resolve once the wire is held between the fingers. A CNN-based classifier on an off-the-shelf vision-based tactile sensor distinguishes bare wire from the crimped terminal, and a pinch and grip motion advances the gripper along the wire until the crimp is detected. The same sensor and classifier handle both wire types despite their different contact shapes, making the design extensible across wire types. The control acting on the perception is classical, and the setup, built around a standard industrial gripper, uses no force-torque sensor and no force feedback from the arm.
Chat is not available.
Successful Page Load