[BREAKING][BUG FIX] Leverage diagonal mass of aligned free bodies.#2993
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 7fa3e466fe
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
duburcqa
force-pushed
the
align_free_body_diagonal_mass
branch
from
7fa3e46 to
7772265
Compare
duburcqa
changed the title
duburcqa
force-pushed
the
align_free_body_diagonal_mass
branch
from
7772265 to
a4c4923
Compare
duburcqa
changed the title
duburcqa
changed the title
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a4c4923134
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
duburcqa
force-pushed
the
align_free_body_diagonal_mass
branch
from
a4c4923 to
61ef114
Compare
|
|
duburcqa
force-pushed
the
align_free_body_diagonal_mass
branch
from
61ef114 to
b38dc3d
Compare
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b38dc3d9ca
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
duburcqa
force-pushed
the
align_free_body_diagonal_mass
branch
2 times, most recently
from
42559a7 to
324b7e5
Compare
|
|
duburcqa
force-pushed
the
align_free_body_diagonal_mass
branch
from
324b7e5 to
7f36c97
Compare
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 7f36c97d51
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
|
🔴 Benchmark Regression Detected ➡️ Report |
…rimitives to meshes and fixed children.
duburcqa
force-pushed
the
align_free_body_diagonal_mass
branch
from
e2fc843 to
72f637e
Compare
duburcqa
changed the title
1 participant
Description
Make the
alignmorph option reframe a free body's root link to its composite center of mass and principal axes for mesh and fixed-child bodies, not just primitives (and per heterogeneous variant). The anchor is computed at load from the parsed collision geometry / explicit inertials, so a kinematic entity and the rigid body it visualizes map the same qpos to the same world geometry. With the inertia exactly diagonal in that frame, the diagonal-mass optimization zeroes the free body's within-link off-diagonal mass coupling to tighten the sparse Hessian band.The composite anchor is density-independent only if a free body's links all specify a mass or all leave it estimated; a mixed subtree raises (for both rigid and kinematic entities), since otherwise the anchor — and thus the alignment — would depend on a material density a kinematic entity does not have.
How Has This Been Tested?
pytest tests/test_rigid_physics.pyon CPU and Apple Metal. Thetest_align_*family covers mesh/URDF alignment, the relative-getter frame stripping, heterogeneous per-variant inertia, rigid/kinematic vAABB parity, and the mixed-mass gate;test_convexifyandtest_heterogeneous_physics_parityexercise the anchored mesh free bodies.Checklist:
Submitting Code Changessection of CONTRIBUTING document.