• Thomas Pank Roulund
4. term, Electro-Mechanical System Design, Master (Master Programme)
This project covers the design of a visual servoing system for a SCARA type robot. The project is made based on a desire from Aalborg University to develop a visual servoing system. It was decided the system should be implemented on AAU's Smart Lab where its' purpose is to assist placing the lid on the housing of a demo assembly. The system is made to be configurable and efficient, which means the visual servoing system is operating without knowledge about the position, orientation or velocity of the housing. Additionally the lid should be placed on the housing, while the housing is moving past the robot. The vision system is using blob detection to derive the position and orientation of the housing, by converting the RGB image from an USB camera to the HSI colorspace. This is done to make the vision system more robust towards illumination changes. A Kalman filter is used to predict the future position and velocity of the housing, and this is used as an initial reference for the robot. Once this reference was tested on a emulated version of the robot, it showed that the robot controller is not able to track the reference without a steady state error. To improve this PI controllers are implemented to determine a new position and velocity reference for the robot, which reduced the tracking errors such they are now lower than the maximum allowed errors. Because the robot's is not fitted with an suction cup as end-effector, it is not possible to verify physically that the robot places the lid on the housing, but measurement from actual robots shows it is able to track the reference and move in while pretending to place the lid. At the time of the placement, it is verified that the position errors are below the maximum limit.
LanguageEnglish
Publication date2 Jun 2017
Number of pages87
ID: 258871487