Default orientation Bug #60
Conversation
- If end_orientation is not provided, it is now correctly defaulted to start_orientation, ensuring the object maintains a consistent orientation throughout the interpolated path.
|
@liquidcronos @MaHajo |
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #60 +/- ##
==========================================
+ Coverage 88.14% 88.32% +0.18%
==========================================
Files 17 17
Lines 1130 1131 +1
==========================================
+ Hits 996 999 +3
+ Misses 134 132 -2 ☔ View full report in Codecov by Sentry. |
This change was necessary in order to pass the code coverage test for the update of the interpolation-orientation branch
this change was required in order to pass the test for code coverage
|
Hello everyone @liquidcronos @MaHajo , |
- This forms a test class for the interpolation method - the spline interpolation is left out
- this is the first approach of spherical interpolating between orientations
- combining test cases for each interpolation
Fixed angle step calculation in circular path generation by dividing by step_num - 1 to ensure proper angular range coverage and avoid over/undersampling
interpreting the sample size correctly
- changing sampling to number of points between start and end
- sampling size should always be at least 2, as start and end point always have to be part of the samples - some classes like test_endeffector use the sampling size wrong, but without changing there test methods, this bug can not be fixed
- this branch is ready for merge
tests/test_interpolation.py
Outdated
| rotations = R.from_quat([start_rot.as_quat(), end_rot.as_quat()]) | ||
| t_vals = np.linspace(0, 1, samples) | ||
| slerp_spline = RotationSpline([0, 1], rotations) | ||
| expected_orientations_both = slerp_spline(t_vals).as_quat() |
There was a problem hiding this comment.
Is it possible to have preset expected orientations without using the RotationSpline function? (analog to expected_positions)
tests/test_interpolation.py
Outdated
| rotations = R.from_quat([start_rot.as_quat(), end_rot.as_quat()]) | ||
| t_vals = np.linspace(0, 1, samples) | ||
| slerp_spline = RotationSpline([0, 1], rotations) | ||
| expected_orientations_both = slerp_spline(t_vals).as_quat() |
| result_with_only_start_orientation.orientations, | ||
| expected_toolpath_with_start_only.orientations, decimal=6) | ||
|
|
||
| def test_zero_sample_size(self): |
There was a problem hiding this comment.
missing function description
Update GitHub Actions workflows to specify Ubuntu version to ensure compatibility and avoid issues with the upcoming change of ubuntu-latest to ubuntu-24.04." actions/runner-images#10636
There was a problem hiding this comment.
@MaHajo Update GitHub Actions workflows to specify Ubuntu version to ensure compatibility and avoid issues with the upcoming change of ubuntu-latest to ubuntu-24.04." After January 17th, 2025 we should be able to change back to ubuntu-latest
There was a problem hiding this comment.
@MaHajo with the new ubuntu 24.04 version this change was necessary.
Description
This PR updates both the
linear_interpolationandcircular_interpolationfunctions to handle quaternion interpolation using Spherical Linear Interpolation (SLERP) throughRotationSpline. Additionally, the functions now correctly handle cases whereend_orientationis not provided. Whenend_orientationisNone, the functions default to usingstart_orientationfor the entire path, ensuring consistent behavior.The test suite has been updated to include combined test cases for both linear and circular interpolation. These tests validate scenarios with both
start_orientationandend_orientation, as well as the case where onlystart_orientationis provided.Additionally, an error in circular path generation has been fixed by adjusting the angle step calculation. Dividing by
step_num - 1now ensures accurate angular range coverage. It ensures that the start and end points align precisely with the input angles.The
GCodeProcessorclass was updated due to a misconception in sampling size, as it always corresponds to the total number of samples in a path including start and end point. Two calculate the correct interpolation step precision+ 2has to be added to theinterpolation_steps.Motivation and Context
Previously, quaternion interpolation relied on methods that didn't correctly handle SLERP, which could result in unexpected behavior, particularly in cases involving smooth rotational transitions. By using
RotationSpline, this PR ensures that quaternion interpolation is smooth and handles edge cases like gimbal lock more effectively.In cases where
end_orientationis omitted, the functions now default tostart_orientation, making it easier for users who expect the object to maintain a consistent orientation along the path. This update enhances the robustness and usability of bothlinear_interpolationandcircular_interpolation.How has this been tested?
end_orientation) for bothlinear_interpolationandcircular_interpolation.RotationSplineto generate expected SLERP results, ensuring consistency between the actual and expected behavior.Screenshots (if appropriate):
N/A
Types of changes
Checklist: