For a few years now, collaborative robotics has allowed us to deploy robots quickly in more or less controlled environments. Using kinesthetic teaching techniques, demonstrating a point in space or moving the robot is faster than ever. On the other hand, it is still difficult to demonstrate a complete task using such a method since the robot only knows the points in the space shown this way. In this project, we developed a system to be able to demonstrate an insertion task to the robot and use this demonstration to build a program repeating the executed task.
First, we developed a variable admittance control for the demonstration of insertion tasks. Our technique distinguish itself by its ability to adjust the rigidity of the control when the robotic arm is in the action of inserting an object. Secondly, we studied the use of deep learning for the detection and localization of insertion tasks within a complete demonstration. We used the measured forces at the robot’s end effector to determine if an insertion task occurred during a demonstration. Finally, we developed a program generator capable of building a robotic routine on a Universal Robot. This generator automatically programs the trajectories as well as the insertion techniques to be used to replay the demonstration.
| Date | 27 Jun 2019 |
|---|
| Original language | French |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Vincent Duchaine (Supervisor) |
|---|
Roberge, E. (Author),
Duchaine (Supervisor),
27 Jun 2019Student thesis: Master's thesis › Master in Engineering: Automated Manufacturing Engineering