New urdf model for pamy2

[simonguist]

Main changes

• New stl files for pamy2
• Adjusted parameters such that joints can reach closer to 90 degrees

Testing

updated after Jan's comment below

TODO

• More thorough testing of joint limits
• More thorough testing on real system

Future Work

• Replace tendons with direct force transfer to joints (simpler, can reach >90 degrees)

[jan1854]

I think the muscles are pulling in the wrong direction. If I use the master branch and apply a positive command u, the robot moves in positive direction. Here it moves in negative direction. This is the output of the script for the master branch

[simonguist]

@jan1854 Thanks for the comment. When I look at the pressures directly, higher ago pressure increases joint angles, higher antago pressure decreases them. Can you verify this? Isn't this how it was supposed to be?

DoF 0 in your plot is worrying, seems like the joint limits are not working here. Can you send me the script and a list of branches you are working on so I can replicate?

[jan1854]

I think the agonist pulls in negative direction, i.e., high pressures in the agonist decrease the joint angles. This is the way how the model on the master branch behaves. This seems to be also the behavior of the real system. I checked the long-term dataset of the pamy2 paper and there high agonist pressures also seem to lead to negative joint angles.

I generated the plot with the model of the master branch of pam_configuration, so it does not include any of your changes. I just generated the plot to check the pulling directions in the original model.

[simonguist]

@jan1854 Yes, you're right, on the real system the agonist does indeed decrease joint position. I switched the muscle positions (rather than changing the joint axis) to keep orientations in body_template and body_template_no_muscles the same and similar to the real system.
I just tested on the real system, DoFs 1-3 look good, but it seems for DoF 4 the joint angles are inverted (counterclockwise rotation leads to decreasing angles). We should probably fix this on the real system.

[dtrb]

Hi Simon and Jan,Please ask Vincent for the location in the code where switch the direction. On 10. Jan 2025, at 05:07, Simon Guist ***@***.***> wrote: @jan1854 Yes, you're right, on the real system the agonist does indeed decrease joint position. I switched the muscle positions (rather than changing the joint axis) to keep orientations in body_template and body_template_no_muscles the same and similar to the real system. I just tested on the real system, DoFs 1-3 look good, but it seems for DoF 4 the joint angles are inverted (counterclockwise rotation leads to decreasing angles). We should probably fix this on the real system. —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>