Robust and real-time teaching of industrial robots for mass customisation manufacturing using stereoscopic vision

Abstract

In this paper a method to control and teach industrial robots in real-time by human demonstration is presented. It aims to decrease down-times for reconfigurations and programming costs, particularly in a mass customisation manufacturing scenario. This system uses high-intensity visual markers that make it sufficiently robust for industrial environments not requiring special lighting. An automated camera calibration method was implemented which enables any non-skilled user a quick and easy configuration of the system. The teaching of the robot is achieved by recording the detected points and replaying them to the manipulator, therefore this is a ¿teaching-by-showing¿ method. The final program that the robot is uploaded with is a smooth trajectory calculated using all the recorded points, which are analysed and grouped into different motion segments according to position and speed.