Joint preview's cyan bounding box and rotation angle doesn't match `Body 1`

[d-kleine]

Description

When creating a revolute joint using rigid body selection mode, the cyan preview (bounding box) does not match the size and shape Body 1's actual bounding box. Instead, it appears generic and oversized.

Reproduction steps

1. Create a new file
2. Add ground plane via Create -> Physics -> Ground Plane
3. From Isaac Sim Assets, add so101_new_calib at the origin in the viewport
4. Enable Rigid Body Selection Mode
5. Create a revolute joint between two connected rigid bodies (e.g., base as Body 0, and shoulder as Body 1)
6. Select the newly created RevoluteJoint in the Stage view

Observed Behavior:

• The cyan preview (bounding box) does not match Body 1 (shoulder)
• The preview appears generic and oversized

Expected Behavior:

• The preview should match the size and shape of Body 1 (shoulder)

The same incorrect preview behavior also occurs when Rigid Body Selection Mode is disabled and All Model Kinds is used.

---

Comparison Case (for reference only):

• Remove the joint
• Disable Rigid Body Selection Mode
• Enable All Prim Types
• Create a revolute joint between the corresponding meshes (e.g., mesh of base and mesh of shoulder)

In this case, the cyan preview correctly matches the size and shape of Body 1 (shoulder).
(Note: Creating joints between meshes is not the correct workflow; this is only used here to demonstrate how the preview is expected to look.)

Isaac Sim version

5.1.0

Operating System (OS)

Windows 11

GPU Name

RTX 3080 Ti

GPU Driver and CUDA versions

GameReady Driver 591.86, CUDA 13.2.1

Logs

No response

Additional information

This issue has been observed with revolute joints, but it may also affect other joint types.

[PeterL-NV]

Hi, Thanks for the reporting. This is a known issue from OpenUSD BBoxCache. This is a cosmetic/UI-only issue — it does not affect simulation behavior or joint physics.

[d-kleine]

This is a cosmetic/UI-only issue — it does not affect simulation behavior or joint physics.

That's right, the physics are correct. But it’s not only the size and shape of the cyan bounding box that don’t match Body 1, but also the displayed rotation behavior in the cyan preview for revolute joints.

In my case, I just implemented a robot with multiple revolute joints, and this incorrect preview caused significant confusion.

For example, using the so101_new_calib asset and the existing joints in the joints folder:

• The first joint, shoulder_pan, rotates around the z-axis with limits of approximately -110° (lower: -109.99987) to +110° (upper: 109.99987) by default.
• The cyan rotation preview appears to visualize the external angle range.
• In reality, the joint behaves according to the internal angle, so the actual motion is effectively the opposite of what the preview suggests.

This discrepancy is also visible in the Physics Inspector, where the shoulder_pan joint behaves contrary to the cyan preview angle around the z-axis.

So, in short, while the simulation behaves correctly, the rotation preview for revolute joints is misleading because it visualizes the joint rotation in the opposite direction of the actual motion, making it appear inverted relative to the real behavior.

[PeterL-NV]

I would recommend to check with Isaac Sim 6.0 EA to see if you still see the same.

[d-kleine]

Just tested in v6.0.0-dev2, same behavior

[PeterL-NV]

Thanks for the update. As this needs fixes from OpenUSD BBoxCache, it's out of our control. I am closing the issue. Feel free to file new issue if you need other help.

[d-kleine]

Thanks for the clarification, I appreciate it. 👍🏻 I have to admit it’s a bit frustrating, since having correct rotation behavior along with an accurate bounding box preview would make creating joints much easier and more reliable.

Would you happen to have a link to the OpenUSD BBoxCache issue you mentioned? I’d like to take a look at it as well to better understand the current limitations.

[PeterL-NV]

Would you happen to have a link to the OpenUSD BBoxCache issue you mentioned?

Unfortunately I only have an internal ticket on this. It's closed too as we don't track OpenUSD issues there.

[d-kleine]

I see – thanks!