In MeshcatVisualizer::load, for each link, the full link name (frame_name in the code) is assume to be of the form robot_name::link_name, which is true except for links in belonging to the world or default model instance, in which case their name is just link_name. (See GetScopedName in multibody_plant.cc.) This causes an assertion failure when _parse_name gets called on the frame_name.
@pangtao22 That code path is new from #12090 -- sorry I didn't catch it in review. I don't think this can happen unless geometry is added procedurally (via e.g. AddRigidBody), since I think parsing a URDF or SDF always adds a model instance. So this could be caught with one more visualizer test that adds procedural geometry.
I'm happy to take this over the next few days? I want to PR another small meshcat viz change anyway.
gizatt
commented
4 years ago
In
MeshcatVisualizer::load, for each link, the full link name (frame_namein the code) is assume to be of the formrobot_name::link_name, which is true except for links in belonging to the world or default model instance, in which case their name is justlink_name.(SeeGetScopedNameinmultibody_plant.cc.) This causes an assertion failure when_parse_namegets called on theframe_name.@pangtao22 That code path is new from #12090 -- sorry I didn't catch it in review. I don't think this can happen unless geometry is added procedurally (via e.g.
AddRigidBody), since I think parsing a URDF or SDF always adds a model instance. So this could be caught with one more visualizer test that adds procedural geometry.I'm happy to take this over the next few days? I want to PR another small meshcat viz change anyway.