Fix cost functions for pose graph optimization #2601
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
This only affects a covariance that is non-diagonal so might be overlooked for ages. |
|
OK an update to this. While the approach from SVD is more robust to ill-conditioned covariance and gives correct residuals for least squares problems, the resulting submatrix is not necessarily lower-triangular, resulting in potentially different ordering of the residuals (which will not pass the unit test since we are testing each dimension of residuals there). I have reverted the SVD to inverse + llt for now and will leave it to the users to make sure that the covariance is well-conditioned. |
ahojnnes
approved these changes
Jun 19, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Just found a long-standing bug from #2378 in both cost functions in pose graph optimization:
After LLT it should be
L_matrix.transpose()rather thanL_matrixthat is left multiplied on the residuals.Moreover, SVD is faster and more stable compared to
covariance.inverse().llt(). I have updated this as well.