Grip Points Generation on Motorcycle Wire Harnesses Main Branch Based on Computer Vision and Clustering

Abstract

The motorcycle industry has come a long way in automating the assembly processes, yet the automation of electrical wiring still presents a considerable challenge. Due to their complex and non-linear nature, wires prove difficult to automate their manipulation. This paper proposes a solution that utilizes a cable recognition system powered by computer vision and unsupervised machine learning (clustering). By treating each pixel of the cable as a point in Cartesian space and grouping them, we can obtain a trajectory and build a precise map of the cable. With this knowledge, handlers can identify and manage their main branch from any point on the wire harness, which is critical when manipulating and routing wire harnesses. Our proposed solution enables us to leverage industrial robots' high precision and repeatability, making it possible to automate even the most challenging aspect of motorcycle assembly.