Page 1 of 1
End position measure
Posted: Sun Apr 18, 2021 1:02 am
by alexnoakes
Hi Thomas,
I am working on a project where the end measure is the ability of the model to end in a certain body position, I have tried playing around with a few measures but can't seem to combine them to create this goal. I was wondering if there was a measure than existed or if you had come across a way to achieve this?
Kind regards
Alex
Re: End position measure
Posted: Mon Apr 19, 2021 5:43 am
by tgeijten
The easiest way to do this is via a
MimicMeasure. If you have an sto file with your desired target pose (i.e. an sto with a single frame), you can use:
Code: Select all
MimicMeasure {
file = file_with_target_pose.sto
use_best_match = 1
}
This should set the score based on how close your simulation gets to the target pose.
Please let me know if this works!
Re: End position measure
Posted: Mon May 10, 2021 12:40 pm
by alexnoakes
This has worked great, thank you.
As an addition to this I was wondering if there would be a way to score the simulation based on whether the acceleration of a body was zero at a certain point, so that the body ended in a static position?
Thanks so much
Alex
Re: End position measure
Posted: Tue May 11, 2021 12:29 am
by tgeijten
You could add velocities to the target states in your .sto file. These are the states that end with "_u", and if you set them to zero, the goal will be to reach a position at zero velocity.
Alternatively, you can add a
BodyMeasure to penalize velocity and/or acceleration of a point on a specific body. You can combine it with the MimicMeasure via a
CompositeMeasure; be sure to take care of the individual weights. If you want a velocity or acceleration penalty to start only after a specific time, you can use the "start_time" parameter for that.
Let me know if this helps!
Re: End position measure
Posted: Sun May 16, 2021 1:06 pm
by alexnoakes
Thank you, this has been very helpful.
I was wondering if there are ever any online workshops or drop ins to discuss projects and how we can optimize SCONE simulations to answer the questions we have?
Kind regards
Alex
Re: End position measure
Posted: Sun May 16, 2021 2:27 pm
by tgeijten
Great to hear it was helpful!
I'm not aware of any public online workshops to discuss predictive simulations, but I'm supportive of any such initiative. Other than that, I'm always happy to help out via this forum.