Skip to content

Fix issue where when a geom penetrates a height field and sinks, it stays sunk and there is no force exerted to bring it out #831

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
surajvh-ga wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Fix issue where when a geom penetrates a height field and sinks, it stays sunk and there is no force exerted to bring it out #831

surajvh-ga wants to merge 2 commits into from

Conversation

@surajvh-ga
Copy link
Contributor

@surajvh-ga surajvh-ga commented Nov 3, 2025

  1. Mujoco has code to handle the flipping of the normal direction in the above specified case: link
  2. MJX Jax backend also has this implemented: link

Here is the test.xml used to reproduce this issue.
You can run this with mjwarp-viewer test.xml

Before:

before.mp4

After:

after.mp4

@surajvh-ga
Fix the issue of geoms sinking into the height field. The root cause …
d4550f9 
…is that there is not ejecting force being exerted when the geom has penetrated the height field. In the mujoco and mjx jax implementation, the normal direction is field to generate this force
@surajvh-ga
Copy link
Contributor Author

surajvh-ga commented Nov 3, 2025

This probably fixes the sinking issue reported in #736

@surajvh-ga
Copy link
Contributor Author

surajvh-ga commented Nov 3, 2025

@thowell would appreciate you checking it out

@thowell thowell requested review from kbayes and thowell November 3, 2025 16:46
@thowell
Copy link
Collaborator

thowell commented Nov 4, 2025 •
edited
Loading

@surajvh-ga thanks for submitting a pr!

i think we also need to consider additional cases, for example where collisions occurs on the opposite side of the height field. something like this scene:

<mujoco>
  <option gravity="0 0 1"/>
  <asset>
    <hfield name="hfield" nrow="16" ncol="16" size="1 1 1 0.1"/>
  </asset>
  <worldbody>
    <light type="directional" pos="0 0 1"/>
    <geom type="hfield" hfield="hfield"/>
    <body>
      <geom type="box" size=".1 .2 .3" rgba="1 0 0 1"/>
      <freejoint/>  
    </body>
  </worldbody>
  <keyframe>
    <key qpos="0 0 -1 1 0 0 0"/>
  </keyframe>
</mujoco>

initialized at keyframe 0.

@thowell thowell requested a review from btaba November 5, 2025 09:27
@kbayes
Copy link
Member

kbayes commented Nov 5, 2025

The mjc_fixNormal function doesn't have to do with height fields. That's related to fixing issues with libccd and is obsolete. I'm not really a fan of doing hacks like this, even if it improves the behavior a little. The way we currently handle height fields is flawed and the contacts we produce for deep penetrations are nonsense. I'm actively looking into how to fix this at a lower level.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@thowell thowell Awaiting requested review from thowell

@kbayes kbayes Awaiting requested review from kbayes

@btaba btaba Awaiting requested review from btaba

At least 1 approving review is required to merge this pull request.

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@surajvh-ga @thowell @kbayes