Improve collision mesh of axis 3

[VictorLamoine]

The collision mesh of the third axis really is too coarse; collisions are detected where the robot is clear from colliding.

This improves the mesh without adding a single vertex (I've done some simplification on other faces).

Should I modify the visual mesh as well?

[gavanderhoorn]

Yes, I didn't like them when I created them, and you're right, especially link_3 is way too 'blobby'.

This looks like a much better approximation, thanks.

re: visual mesh: yes, please do, they should essentially just be the same (so a copy would be fine).

could you rebase on indigo-devel also please?

[VictorLamoine]

Done.

I tried B-Spline fitting on a sampled point cloud (created from the visual model) but it's not good at all, so I can't think of an automatic solution for producing better meshes!

[gavanderhoorn]

Thanks.

Yeah, that doesn't look too good :).

[gavanderhoorn]

I've been eyeing V-HACD as a possible way to do this, but haven't had time to evaluate it properly.

[gavanderhoorn]

Played with it a little:

Not bad for an automated tool.

[VictorLamoine]

Too good! You could probably even reverse-engineer the robot with these aha!
It's integrated into MeshLab?

[gavanderhoorn]

Too good! You could probably even reverse-engineer the robot with these aha!

well, there are a lot of knobs to play with, resolution can probably be turned down even more.

But I'd say this is blobby enough. If you can reverse engineer a robot from this, you can probably make a lot of money.

It's integrated into MeshLab?

No, I just opened the .wrl in it.