Relax the tolerance in IpoptSolver. #22426
Conversation
|
@drake-jenkins-bot mac-arm-sonoma-unprovisioned-clang-wheel-experimental-release please |
Previously we set the tolerance to 1.05E-10, this tolerance is too tight and causing numeric problems. We switch back to IPOPT default tolerance.
hongkai-dai
force-pushed
the
relax_ipopt_tolerance
branch
from
638c8a9 to
20cb7cc
Compare
There was a problem hiding this comment.
+@RussTedrake for feature review please, thanks!
Reviewable status: LGTM missing from assignee RussTedrake(platform), needs at least two assigned reviewers, missing label for release notes (waiting on @hongkai-dai)
|
|
There was a problem hiding this comment.
Reviewed 4 of 4 files at r1, all commit messages.
Reviewable status: 1 unresolved discussion, needs at least two assigned reviewers, missing label for release notes (waiting on @hongkai-dai)
solvers/test/ipopt_solver_test.cc line 209 at r1 (raw file):
IpoptSolver solver; ConfigureIpopt(&solver); TestQPDualSolution1(solver, {} /* solver_options */, /*tol=*/1e-4);
btw -- i see. that is pretty loose. but i still think this is a good chance given all of the failures we've seen.
jwnimmer-tri
added
the
release notes: fix
There was a problem hiding this comment.
Reviewed 4 of 4 files at r1, all commit messages.
Reviewable status: 1 unresolved discussion
There was a problem hiding this comment.
Reviewable status:
complete! all discussions resolved, LGTM from assignees jwnimmer-tri(platform),RussTedrake(platform)
solvers/test/ipopt_solver_test.cc line 209 at r1 (raw file):
Previously, RussTedrake (Russ Tedrake) wrote…
btw -- i see. that is pretty loose. but i still think this is a good chance given all of the failures we've seen.
Agreed. I think it is better to have a solver that solves problems with inaccurate solution, rather than a solver that gives more accurate solutions sometimes but fails elsewhere.
Previously we set the tolerance to 1.05E-10, this tolerance is too tight and causing numeric problems. We switch back to IPOPT default tolerance.
This is motivated by #22361 (review)
This change is