Kinematic loop in anchored_world.world fails with dantzig LCP solver in DART

Issue #2608 new
Steven Peters
created an issue

The kinematic loop in test/worlds/anchored_loop.world fails to solve correctly with the dantzig LCP solver in DART. This was noted in the description of pull request #2756 when the test world was added, and was also noticed during the review of pull request #3086:

There is a similar issue reported upstream at dartsim but for a different model that has contact between links instead of a closed kinematic loop:

cc @Jeongseok Lee

Comments (5)

  1. Steven Peters reporter

    @Jeongseok Lee should we report this failure upstream as well? I think it's more of an LCP problem than an issue related to contact. I guess the comment about switching to pgs if dantzig doesn't get a good solution is relevant.

  2. Jeongseok Lee

    If you refer to ODE then I'm inclined not to report to the upstream because falling back to PGS from Dantzig is out of scope of Dantzig solver. It's rather how we use Dantzig and PGS solver I think. For example, we could use penetration based method as fallback instead.

  3. Log in to comment