Abstract
Image processing and computer vision systems are adequate to enhance the potential functions of industrial robots. However, besides complex implementation, it is required to do time-consuming and complex calibrations between the robots and cameras before running it. This paper first introduces the design of a compact CNN to estimate objects' orientation. Then, the paper presents a desktop-sized pick and place robot while implementing a pixel-based visual feedback (PBVF) controller to move to target objects autonomously. The response of the VF control is quantitatively evaluated by counting control actions until the end-effector reaches the object position. A sliding rail is further considered to enable the robot to work in a broader working space. The PBVF controller is extended to utilise the sliding rail by dealing with the direction of the sliding rail as the robot's y-axis. Finally, the implementation of the PBVF controller and the effectiveness of the robot system with the sliding rail are presented through pick and place experiments of objects randomly placed on a table.
Original language | English |
---|---|
Pages (from-to) | 142-150 |
Number of pages | 9 |
Journal | International Journal of Mechatronics and Automation |
Volume | 9 |
Issue number | 3 |
DOIs | |
Publication status | Published - 2022 |
Keywords
- CNN
- convolutional neural network
- desktop-sized robot
- pick and place
- pixel-based visual feedback control
- sliding rail
ASJC Scopus subject areas
- Control and Systems Engineering
- Computational Mechanics
- Industrial and Manufacturing Engineering
- Computational Mathematics
- Artificial Intelligence
- Electrical and Electronic Engineering