Summer internship diary: first steps with a collaborative robot

It has already been three full weeks since I started my summer internship in the Smart Systems Research Group. The main goal of my internship is to learn new things and apply them in practice, and the Rokae collaborative robot, or cobot, is helping me do exactly that.

I started my first day by studying the documentation and software, but I could not spend too long on theory. I soon decided to get to know the robot through hands-on experience. My first task was to set up the connection between the robot and the gripper. To be honest, I do not have much experience with electronics, so it took me a few days to connect the gripper and the robot via the Modbus protocol and to make a custom cable. It was also my first experience soldering wires, and it was fun.

In the second week, I started thinking more carefully about the project plan and the tasks the robot could perform. My plans turned out to be quite ambitious: to develop an intelligent controller for grasping objects and to use computer vision to determine the position of objects in space, as well as their dimensions and colour.

After studying the SDK a bit, I wrote a few simple Python functions to move the robot. However, controlling the gripper turned out to be a challenge, as the SDK did not include any ready-made functions for our gripper. I had to control it using low-level Modbus commands, sending values directly to the device registers. Eventually, I got it working.

In the third week, the main task was to implement a simple pick-and-place routine. Overall, the task was quite straightforward, and I completed it quickly. However, the perfectionist in me was not happy that I had to guess the desired robot position. I could not initially work out how to specify the coordinates precisely using the grid on the work surface. Through trial and error, I finally managed to calibrate the tool correctly and create a coordinate system for the workspace. Now I can specify the desired position directly using the grid on the table.

You can see a short demonstration of the cobot performing the pick-and-place task here: [YouTube]

The next step is adaptive grasping: the robot will automatically adjust its grip force to suit the object.

Stay tuned to see how this turns out.