Undesired behavior when moving goal closer to start even with smooth_scaling=True

[vhwanger]

Hi there,

I'm trying out your library and was noticing a weird effect that I can't quite fix. I have a simple 1D DMP and I'm changing the goal state to be closer to the start state than the original demonstration. In the attached image, the left plot is the demonstration and the right plot has the goal moved a bit closer. I notice it ends up with a fairly large overshoot. When smooth_scaling is False, the overshoot is even larger.

Do you have any suggestions on how to get rid of that overshoot? Thank you!

[AlexanderFabisch]

Hi @vhwanger ,

does changing the parameters alpha_y and beta_y have any effect on this? (See here). They are usually supposed to be set as $\frac{\alpha}{4} = \beta$ to ensure that there is no overshooting at the end or too slow convergence.

[vhwanger]

Thanks for your response! Yes, it definitely has an effect, though it's fairly finicky - I have many DMPs that are somewhat s-curve-ish of different ranges, and it seems each one needs a different alpha/beta value (assuming the ratio you provided) to make it not overshoot. Perhaps this is unavoidable.

[AlexanderFabisch]

There are several DMP implementations that work slightly different. Would be interesting to see if there is a version that handles this case better.

Would it be possible to provide an example of a demonstration? I don't know if/when I will have time to look at this, but then I will at least have the option to check it.