#Visual and collision geometry
Every URDF link can have two kinds of geometry: <visual> (what you see)
and <collision> (what physics engines and planners use). They are often
different meshes, and mistakes in the collision layer are easy to miss
because most tools only show the visual one.
#Switch layers
Use the Visual menu in the top-left bar of the 3D view:
| Choice | Shows |
|---|---|
| Visual | Visual geometry only (default) |
| Collision | Collision geometry only |
| Both | Both, with collision shapes drawn see-through over the visual meshes |
Collision geometry is loaded the first time you ask for it, so the first switch takes a moment. Your pose and camera are kept.
Both is the quickest way to check that collision shapes cover the visual model — look for parts sticking out, offsets, or shapes at the origin that were meant to be moved.
To start every robot in collision view, change Default render mode in settings.
#Self-collision
Turn on Display → Self-collision to highlight links whose collision shapes overlap in the current pose. Overlapping links turn red, and a caption in the corner of the view names the colliding pairs.
Some contact is expected. Closed gripper fingers, for example, touch each other — robot SRDFs usually disable such pairs, and pairs disabled in the loaded SRDF are not reported.
The check updates as you move joints, so you can drag a slider and watch where the arm starts hitting itself. Links joined directly to each other are not reported: they usually touch by design.
To turn this into an SRDF disable_collisions list for MoveIt, see
Collision pairs.
#Wireframe
Display → Wire draws every mesh as a wireframe in any layer. It helps when a collision mesh is far heavier than it should be, or to see a part hidden inside another.