Question about scene 4

[lingxiaomeng]

Hello, I watched your video. In scene 4, the robotic arm demonstrated compliance when facing external disturbances. Could you please explain how this was achieved and recommend any relevant articles for reference?

[ahaffemayer]

Hello,

Thank you for your interest, and sorry for the delay, I've been wrapping up my PhD and the issue got lost in my mails.

This compliance comes from torque control on the MPC framework, not only the collision avoidance part.

[lingxiaomeng]

Thank you for your response, and congratulations on wrapping up your PhD!
Regarding the compliance behavior of the robotic arm in scene 4, your explanation about torque control in the MPC framework is very helpful. To clarify my understanding, is the compliance primarily achieved by setting smaller position error costs in the MPC formulation?

[ahaffemayer]

Hi!

Thank you!

No, the compliance comes from the fact that instead of controlling the robot by a set of joint positions, you control directly the torques of the motors.

If you are controlling the robot by a set of joint positions in real time, you are basically telling the robot :"go to this position and don't move" and you do this at each millisecond, for instance.
Let's take a step back from this real time and just isolate the motion. Basically you say :go to this position and don't move". But as you do this, you'll have a fixed position of the motors. Which means a very stiff behavior because you are instructing the motors to be at a given position and they'll not move.

However, if you are controlling not the motors positions but the torque they provide, using the known dynamics of the robot, you can still control properly the robot but then the behavior of the robot is not stiff anymore since the motors do not have a fixed position.

Is it clearer?

If you need, send me an email at: arthur.haffemayer@gmail.com, I'll be more responsive. For some reason, I don't recieve the notifications here.