Good afternoon all!
I hope this finds you well
Isaac Sim Version
4.5.0
Operating System
Ubuntu 22.04
GPU Information
- Model: 4070
- Driver Version: Latest
Screenshots or Videos
frames_2025-03-23_14.32.48.pdf (18.7 KB)
Topic Description
Detailed Description
I have been building an autonomous rover simulation and recently saved a flattened version of the original file after reorganizing some prim meshes under new xforms. Although the xforms are placed correctly, the TF tree appears to reference the meshes within those xforms rather than their new positions. As a result, when I visualize the simulation in RViz2, all the meshes seem to be constrained to the center of the world. This misalignment causes the costmap to stop updating once the robot moves far enough away, as the sensor is effectively “fixed” at the world center.
Steps to Reproduce
- Flatten the original file and move several prim meshes under new xforms.
- Launch the simulation and visualize the TF tree in RViz2.
- Observe that the TF tree reports the meshes as if they are centered, despite the xforms being correctly placed.
- Notice that as the robot moves, the costmap stops updating when the sensor appears to be out of range, since it remains fixed near the world center.
Error Messages
There are no direct error messages; however, the TF tree and RViz2 visualization clearly indicate that the transforms are not being updated as expected. Additionally, the costmap stops updating when the sensor (which should move with the robot) appears to be fixed.
(Please see the attached screenshots for my current TF tree and a demonstration of the issue during simulation.)
Additional Information
What I’ve Tried
- Verifying that the xforms in the scene are correctly placed.
- Checking the TF tree in RViz2 to ensure that all frames are published.
- Adjusting costmap parameters to see if that would influence the behavior.
- Reviewing transform settings in the simulation and the ROS bridge.
Related Issues
(If you have any related issues or forum posts, please include links here.)
Additional Context
I suspect that the issue may be due to the TF tree referencing the original mesh coordinates instead of the updated xform positions, which results in the sensor frame (and therefore the costmap) remaining fixed at the center of the world. Any suggestions to ensure that the TF tree accurately reflects the updated geometry would be greatly appreciated.
