Another FCL bug that calls EPA while the origin is outside the polytope

[hongkai-dai]

Caught in the wild

terminate called after throwing an instance of 'std::logic_error'
[load_silverware_manual_test]   what():  Error with configuration
[load_silverware_manual_test]   Original error message: external/fcl/include/fcl/narrowphase/detail/convexity_based_algorithm/gjk_libccd-inl.h:(1612): validateNearestFeatureOfPolytopeBeingEdge(): The origin is outside of the polytope. This should already have been identified as separating.
[load_silverware_manual_test]   Shape 1: Box 0.61399999999999999023                       3 0.036999999999999998168
[load_silverware_manual_test]   X_FS1
[load_silverware_manual_test]                       1                       0                       0 -0.67500000000000004441
[load_silverware_manual_test]                       0                       1                       0                       0
[load_silverware_manual_test]                       0                       0                       1  0.91149999999999997691
[load_silverware_manual_test]                       0                       0                       0                       1
[load_silverware_manual_test]   Shape 2: Box 0.49399999999999999467  0.55200000000000004619 0.010000000000000000208
[load_silverware_manual_test]   X_FS2
[load_silverware_manual_test]                       1                       0                       0 -0.69199999999999994849
[load_silverware_manual_test]                       0                       1                       0                       0
[load_silverware_manual_test]                       0                       0                       1  0.93500000000000005329
[load_silverware_manual_test]                       0                       0                       0                       1
[load_silverware_manual_test]   Solver: GjkSolver_libccd
[load_silverware_manual_test]     collision_tolerance:      2.0097183471152322134e-14
[load_silverware_manual_test]     max collision iterations: 500
[load_silverware_manual_test]     distance tolerance:       9.9999999999999995475e-07
[load_silverware_manual_test]     max distance iterations:  1000

[sherm1]

😢

[hongkai-dai]

I cannot reproduce this error.

These two boxes are axis-aligned. It is easy to see that they are in perfect touching contact, along z direction.

[hongkai-dai]

I can reproduce this error with 09f846c. And the commit 2112037 which returns vertex as the support of the box also gets rid of the error reported in this issue.

But this could be an error in our code, that perfect touching contact can call either GJK or EPA. EPA might think that the origin is not strictly inside the polytope (EPA handles penetration contact).