Hi, thanks for the great project.
I noticed that during the URDF-to-MJCF conversion process, the collision data is removed and only the visual data is kept:
# get rid of collision data, assuming the visual data is much better resolution.
# not sure if this is the best move...
xml_data = mrc.remove_tag(xml_data, "collision")
I am a bit confused about this design choice. In many URDF models, the collision geometry and visual geometry are intentionally different. For example, the visual mesh may be high-resolution and used only for rendering, while the collision geometry may be simplified, convex, or manually designed for stable and efficient physical simulation.
Could you explain the reason for removing the collision data in this conversion pipeline?
Hi, thanks for the great project.
I noticed that during the URDF-to-MJCF conversion process, the collision data is removed and only the visual data is kept:
I am a bit confused about this design choice. In many URDF models, the collision geometry and visual geometry are intentionally different. For example, the visual mesh may be high-resolution and used only for rendering, while the collision geometry may be simplified, convex, or manually designed for stable and efficient physical simulation.
Could you explain the reason for removing the collision data in this conversion pipeline?