Robots in manufacturing need a precise description of the position and orientation of a component to execute the planned tasks. Often, a component’s position is not fixed and cannot be directly taught to the robot. Its position and orientation must then be determined dynamically before starting the process. Solutions addressing this problem are mainly developed using traditional computer vision approaches and cannot be transferred to other use cases or even components in many cases.

Using AI-assisted robotics, an intelligent and versatile pipeline is presented in this work. It can simplify the 6D localization task into smaller modular tasks which can be adapted to multiple use cases for any desired component.