Collision shape is congruent with object, just avatar walks in the air


#1

The blender export of the fbx with collision shape works well now. I just recognized, that the avatar is not walking on the surface of the object. Is that something it will be fixed in an upcoming HF version, or maybe something I can fix on my side?

the test object: http://www.skimi.org/HF_objects/landscape7.fbx

http://www.skimi.org/HF_objects/vid/Skimi_physics_HF.mp4


#2

Right now you have to make sure the collision hull meshes are convex.

There are discussions of making the collision hulls better and doing concave hulls as well, hopefully someday. For now, we have to stick with convex only.

More info in this, tad bit dated video, which demonstrates a really quick hack of doing a group of convex only mesh, that doesnt require using tools such as V-HACD it start around 4 minutes in when @Alan_ discusses about how to do it quick:

Here is one for using V-HACD as a tool for creating collision hulls for non landscape objects

Also as a suggestion, instead of using the FBX world up and the sort, use obj export instead for physical hulls.


#3

I will say again that all this unnecessary complexity could be avoided by supporting static trimesh for physics shapes.


#4

Yep, which is why i said hopefully support them come soonish


#5

In this pinball machine, everything is static trimesh except for the dynamic pinball. The flippers and bumpers are moved kinematically. PhysX handles that sort of thing nicely. Id imagine bullet would too since the static meshes are basically respositioned when position shifted.

Another dynamic to trimesh collision example:

And yet another good example. Here the skis (and avatar capsule) are dynamic while everything else, the mesh slopes, even the terrain are trimesh physics. You can see a little bit of inaccuracy because the default quantization is 0.05m.

Finally, mesh objects rezzed, dynamic, so convex hulls, but everything else is physics trimesh, except for the air kraken and bullets which are dynamic convex hulls.


#6

Hi Menithal, thanks for your support. I’ve did it like the tutorial shows - and okay, it is better and I can work with that. But would be cool, if HF support a better solution soon :slight_smile: