Revision history [back]

I've just opened a pull request against your repository, see aleksandaratanasov/urdf_collection#1.

A copy/paste of the PR description, so it doesn't get lost:

Basically, I haven't even tried to diagnose the original URDF, as there were too many issues with it, probably due to the conversion done by collada_to_urdf. I also suspect a discrepancy between the joint transforms in the original urdf and the meshes that were converted separately.

In the end, I:

• converted everything to a xacro macro (added top-level xacro as well)
• made naming consistent with ROS standards (base_link, etc)
• cleaned up all urdf joint transforms: based on data from a KR 3 CR datasheet
• removed collada->urdf conversion artefacts (unnecessary xml ns decls, etc)
• changed origins of all meshes to be relative to the joint zero, instead of to the global robot zero
• scaled meshes (ROS uses meters, meshes were in millimeters)
• fixed orientations of all meshes: X forward, Y to the left, Z up (ROS standard)
• created collision quality meshes (convex hulls)
• split out visual and collision quality meshes of KR 3 CR into separate directories

Note:

• origins of all intertial elements has been reset to zeros as I don't know what the real values should be
• masses, origins and values for inertia elements are incorrect
• effort limits of all joints have suspicious values (if unknown, just leave them at zero, unless you need to simulate with Gazebo obviously)
• dynamics element attributes seem to have standard values

I also added two convenience launch files to easily load and display the model in RViz (roslaunch urdf_collection test_kr3cr.launch).

Most of this is based on the Create a URDF for an Industrial Robot and Working with ROS-Industrial Robot Support Packages tutorials from ROS-Industrial.

I also created a test MoveIt configuration to verify the model and could successfully plan for it.