Fix Jacobian calculation in `RelativeFrameTask`

stephane-caron / pink

Python inverse kinematics based on Pinocchio

Apache License 2.0

169 stars 11 forks source link

Fix Jacobian calculation in `RelativeFrameTask` #81

Closed stephane-caron closed 3 months ago

stephane-caron commented 3 months ago

See https://github.com/stephane-caron/pink/pull/80#discussion_r1539872284, thanks @ymontmarin!

coveralls commented 3 months ago

Pull Request Test Coverage Report for Build 8481766194

Details

20 of 20 (100.0%) changed or added relevant lines in 3 files are covered.
No unchanged relevant lines lost coverage.
Overall coverage increased (+0.006%) to 99.456%

Totals
Change from base Build 8479606289:	0.006%
Covered Lines:	1280
Relevant Lines:	1287

💛 - Coveralls

stephane-caron commented 3 months ago

I'm concerned by the fact the previous calculation error was not caught by any unit test. I guess this is because it only affected the Jlog6, which slants velocities when the error is large but ultimately does not matter if convergence occurs.

stephane-caron commented 3 months ago

I'm concerned by the fact the previous calculation error was not caught by any unit test. I guess this is because it only affected the Jlog6, which slants velocities when the error is large but ultimately does not matter if convergence occurs.

OK I've added a unit test comparing, in one randomly sampled configuration for one robot description, against the current Jacobian calculation. If it is correct, future implementations shouldn't calculate a different matrix. And if they do, we get a free chance at checking that Jacobian again :wink