Developement and implementation of control for a collaborative robot arm based on image processing

OData support
Kundra László János
Department of Automation and Applied Informatics

Vision guided robots and robots that collaborate with human operators are topics that are currently actively researched. This paper explores a solution for finding the position and orientation of work-pieces in a plane perpendicular to the optical axis of a camera, and instructing the robot to perform the necessary operations based on this data.

Creating versatile image processing solutions poses many challenges. Errors caused by changing light conditions and the image noise being a few examples. The approach that we examine in this paper is evaluating multiple images and simplifying the contents into elementary geometric features that contain the data that is interesting for us.

To tackle this problem, a prototype camara system was created using a Raspberry Pi\textsuperscript{\textregistered} microcomputer and a 3D printed mount. Image processing was developed using OpenCV python libraries. The system is mounted on a six-axis industrial collaborative robotic arm and communicates with the robot controller through wireless connection using TCP-IP protocol. The solution is integrated into the MTA--SZTAKI SmartFactory Industry 4.0 research and demonstration environment.

As a result, multiple functional methods for object detection were developed and tested. The robot was programmed to use this data to pick up objects and stack them into a tower. These results imply that this direction of development is promising and can improve the flexibility of robot operation, as well as opening possibilities for more advanced object detection and classification.


Please sign in to download the files of this thesis.