Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Consider Slerp instead of Squad for rotation interpolation #568

Open
jrutgeer opened this issue Nov 9, 2023 · 5 comments · May be fixed by #577
Open

Consider Slerp instead of Squad for rotation interpolation #568

jrutgeer opened this issue Nov 9, 2023 · 5 comments · May be fixed by #577
Labels
enhancement New feature or request help wanted Extra attention is needed

Comments

@jrutgeer
Copy link

jrutgeer commented Nov 9, 2023

I am using the gz-gui camera tracking plugin to adapt the simulation viewpoint throughout the simulation.

I noticed that motion is sometimes 'shaky' due to the rotation interpolation, e.g. :

$ gz service -s /gui/move_to/pose --reqtype gz.msgs.GUICamera --reptype gz.msgs.Boolean --timeout 500 --req 'pose: {position:{x: 27.2, y:-1.7215,z:2.2255}, orientation:{x:-0.1466788, y:0.0344671, z:0.9623708, w:0.2261413}}'
$ gz service -s /gui/move_to/pose --reqtype gz.msgs.GUICamera --reptype gz.msgs.Boolean --timeout 500 --req 'pose: {position:{x:24.081968307495117,y:1.0910710096359253,z:1.194319486618042},orientation:{x:0.10836052149534225,y:-0.0045970650389790535,z:-0.99320769309997559,w:-0.042135711759328842}}'
squad.mp4

Changing this single line:

gz-math/src/RotationSpline.cc

Line 95 in 6460f7c

return Quaterniond::Squad(_t, p, a, b, q, _useShortestPath);

to use Slerp instead of Squad solves the issue:

  return Quaterniond::Slerp(_t, p, q, _useShortestPath);
slerp.mp4

Is it an option to change this to slerp instead of squad?
Why was squad chosen (I don't find references through 'git blame', seems it was squad from the first commit?)

It seems this is a good reference wrt rotation interpolation:

https://splines.readthedocs.io/en/latest/rotation/index.html
@jrutgeer
Copy link
Author

jrutgeer commented Nov 9, 2023

According to following SE answer, an alternative could be to keep the squad algorithm, but to choose better A and B 'helper quaternions' for that algorithm:

https://math.stackexchange.com/a/2654414

@mjcarroll mjcarroll added the enhancement New feature or request label Nov 20, 2023
@scpeters scpeters added the help wanted Extra attention is needed label Nov 20, 2023
@scpeters
Copy link
Member

I have added the help wanted label to this issue; we will be happy to review pull requests with a proposed change. For example, we could add an overload to RotationSpline::Interpolate with a parameter to specify whether Slerp or Squad should be used.

@sdhar04
Copy link

sdhar04 commented Feb 5, 2024

@scpeters I noticed that the points and orientations given in this issue are too close to each other, and it feels like the 'shaky' movement is due to the radius of the interpolation sphere being too small. Should I add a condition to use Slerp if the final and initial points' and orientations' differences fall within a threshold and Squad if they don't?

@scpeters
Copy link
Member

scpeters commented Feb 6, 2024

that sounds reasonable; go for it!

This was referenced Feb 6, 2024
Closed
Open
@sdhar04
Copy link

sdhar04 commented Feb 7, 2024

that sounds reasonable; go for it!

I played around with the values a little and noted that the ‘shaky’ movement stayed even if the points were far apart. So I focused on the difference between the starting and final orientations and made changes in the code accordingly.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
enhancement New feature or request help wanted Extra attention is needed
Projects
Core development
Status: To do
Milestone
No milestone
4 participants
@mjcarroll @scpeters @jrutgeer @sdhar04