Generate SE3 foot trajectory for a robot using ndcurves

[MaximilienNaveau]

Dear users,

I am trying to generate a smooth C^2 trajectory for my feet robot trajectories.
I am wondering which trajectory are the best one to use for this.
I tried some pieces SE3 by the trajectory are not C^2 during between each piece...

Any help is appreciated.
Thanks

[stonneau]

If you want continuity I guess you need to enforce it yourself ?
Not sure I understand the question here ?

[stonneau]

Also itf it's just for z why do you work in se3 ?

[MaximilienNaveau]

Hi, I trying to generate the full foot SE3 behavior.
For the translation I ended up using the ndcurves::helpers::effector_spline which is ok for my usage.
And for the rotation as a separate spline I used the curve_so3_t.
I am plotting the trajectories yet to see the quality but it is promising.

[MaximilienNaveau]

If you want continuity I guess you need to enforce it yourself ?

Actually it is not the continuity of the curves rather than the continuity of its derivation.
I thought you could parametrize a curve than goes up->forward->down with three pieces of curves and ensure C^2 continuity between them. But I did not figure out how to ensure derivates continuity in the piecewize_curves.

[olivier-stasse]

One of our student as the same problem.
I would suggest to have specific helper classes computing the coefficient for the basic strategies.

[stonneau]

I m sorry but this is still unclear to me. However introducing helper functions to ensure continuity is certainly something that can be considered, although I m a bit perplex on the API here. I think it would be great to better specify the use case, inputs / outputs and then we could discuss how to implement this

[stonneau]

If you want continuity I guess you need to enforce it yourself ?

Actually it is not the continuity of the curves rather than the continuity of its derivation. I thought you could parametrize a curve than goes up->forward->down with three pieces of curves and ensure C^2 continuity between them. But I did not figure out how to ensure derivates continuity in the piecewize_curves.

Again, ensuring the continuity of the derivatives is something you need to do by yourself, either manually or through constrained optimisation. I really don't understand how you would expect the continuity to hold unless you select exactly the right degree for your curves.

[olivier-stasse]

@stonneau Should we exclude specific foot trajectories in ndcurves ?

[stonneau]

Are you referring the the ones defined in the helpers folder ? I would be in favor of removing them if no one argues against. @pFernbach might have a better opinion than me.

Otherwise I m not sure what you are referring to

[olivier-stasse]

Yes, I was thinking of completing them.
I do not see really the point of creating another library/package when everything we do is to instantiate polynomials.

[olivier-stasse]

But you are the one deciding :) and thanks again for maintaining the library.