Development of an intelligent self-learning product assembly system using visual identification

Abstract

Modern automation systems rely on fixed programming to carry out their production routines. These systems are effective for production outputs but do not allow any flexibility within the production routine. Effort is required to change the ongoing production routine through reprogramming, redesign or complete overhaul of the system to cater for new production outputs. These efforts require down time and result in a loss of revenue. If a completely automated flexible system is introduced into such a production line, the complete reprogramming process required to cater for new production needs could be automated without losing production time. Within this study, a real-time KUKA Robotic Control system is introduced. The KUKA Robotic Controller maintains its original programming methods with no reprogramming required when executing a new production assembly. This is achieved through manoeuvring the KUKA Robotic System in real-time to new destinations based on image-processing outputs and feedback. For demonstration purposes and proof of concept, the system learns a design presented to it by an end user and then reproduces this seen design based on the image-processing results in terms of location and orientation. Therefore, instead of reprogramming each new required position, the system takes over real-time control of the KUKA Robotic System and carries out the required steps autonomously. The benefit of such a system would be that the KUKA Robotic System would not require reprogramming to carry out new routines. It is controlled in a real-time environment to carry out new procedures based on external sensors (in this case, image-processing outputs). KUKA Robotic Sensor Interface (RSI) software is used to implement real-time control of the KUKA Robotic System and is explored extensively throughout this study.