How to support user-defined robots and environments?

[marip8]

I'm interested in testing this repository on my own robots and motion planning environments and have a few questions about how to get started (apologies if these are documented somewhere in this repo and I missed them):

1. How to import a user-defined robot?

Given a spherized URDF of an arbitrary robot, how do I generate the corresponding FK solver required by VAMP? The paper mentions a tracing compiler, but I didn't see it in this repo. Is that tool included? If not, is there a roadmap for making it available (open-source or otherwise)?

2. How to "spherize"?

The documentation says:

To generate code for a new robot, we require a spherical decomposition of the robot's collision geometry. See the *_spherized.urdf files for examples of these spherical decompositions for each robot. We also use a secondary hierarchical decomposition: see the high-level single sphere URDF used for each robot in the *_spherized_1.urdf files. These decompositions are conservative estimations of the robot's mesh geometry.

What method do you suggest using to spherize arbitrary robot and environment representations (specifically ones currently defined by URDFs with mesh files)? Is this something you do manually or is there a common tool you use to do this?

[zkingston]

For 1., we're in the process of cleaning up the compiler code that generates the FK/CC solver for VAMP and will bundle it in a separate manuscript, for publication and eventual release. If there is a specific robot you'd like to have support for, email @wbthomason and I and we might be able to help out.

For 2., we used an adaptation of https://github.com/mlund/spheretree (also described in this publication [1]) to spherize the robot geometry.

[1] https://dl.acm.org/doi/pdf/10.1145/966131.966132

[marip8]

Thanks for the quick reply. I don't have a specific robot in mind right now; I was mostly thinking about how I could use this repo for projects in the future. I'll keep an eye out for the release of the compiler. Looking forward to testing this new planner; the results look really promising