How does Gazebo internally handle multiple collision components for a link in the sdf?

asked 2020-11-26 06:39:43 -0600

rfn123 gravatar image

I have an sdf with a robot, and I want to generate collision meshes. I thought about using convex decomposition on the visual meshes to get a good approximation for the collision. My question is, would it be better regarding performance if I put each convex component in one seperate collision instance such that I have multiple collision components for a link? Or is it the same as specifying the combined convex component in a single mesh file?

Besically I want to know what's happening under the hood: Whether the collision detector checks if a collision component is convex, then does a fast collision check, and moves to the next collision component and checks again. Because if the collision detector _first_ combines multiple collision components and _then_ checks for collision, a convex decomposition would be useless since the combined components are not convex anymore.

Thanks in advance!

edit retag flag offensive close merge delete