Yet another physics model mystery

[ChinRey]

Just as I thought I understood how mesh physics models work, this came up. Can anybody explain why the uploader insists on turning this relatively simple model into seven hulls instead of just five?

And even more mysteriously, why does the number of hulls go up to eight if I remove the apparently superfluous edge across the face at the left front of the image?

[Aquila Kytori]

Hi

Ideally the physics model should be using either the *Not Overlapping box method which will be Analyzed in the uploader or the Triangle based (Planes) method which will not be Analyzed in the mesh uploader.

Your model looks a bit of a hybrid 

Note: When Analyzing the uploader can give unexpected results even for models using the Non overlapping box method.

         Generally for large meshes like parts of buildings the lowest Physics cost can be achieved using the Planes method.

        

*Just to add: Non Overlapping means leave a little space between each box.

[ChinRey]

---

Aquila Kytori wrote:

Hi

Ideally the physics model should be using either the *Not Overlapping box method which will be Analyzed in the uploader or the Triangle based (Planes) method which will not be Analyzed in the mesh uploader.

---

Thanks, Aquila, that did the trick. It also turned out I could simplify the physic a bit with no noticeable effect so by following your advice and simplifying it a bit I was able to get it down to three hulls. :matte-motes-big-grin:

But now I'm really confused. Separate boxes fow each hull was the first thing I tried. But my mother always told me never to leave holes in the physics model because Havok will make a havoc out of it so I placed them with no gaps between. Was my mother wrong??? :matte-motes-wink:

---

Aquila Kytori wrote:

Generally for large meshes like parts of buildings the lowest Physics cost can be achieved using the Planes method.

---

That's true. I had written that solution off in this case because I needed to define so many of the edges (there'll be doorways all around the house), not just the walking surface. But you got me to rethink the issue and I managed to find a way to do it with just 10 tris - physics weight 0.5.

30 different models tested to get the physics weight down from 1.8 to 0.5. Some people would say it isn't worth the effort but what do they know... :matte-motes-wink:

[Drongle McMahon]

The workings of the "Analyze" machine are rather obscure, and difficult to predict unless you provide it with a model that's already a set of convex hulls (and sometimes even tnen). I made as near a copy as I could of your mesh, with the same 27 vertex count, and it gave me the expected 5 hulls with "Analyze/Surface", which was the same as the best I could do for exact physics with separate boxes. There must be some difference between our geometry in the part hidden in your picture. It's also possible the viewer you are using mught use a different library for the "Analyze" decomposition. All 3rd party viewers had to use something else in the early days, although I think Firestorm has now licensed the same libraries as the LL viewer.

[Aquila Kytori]

---

ChinRey wrote:

---

But now I'm really confused. Separate boxes fow each hull was the first thing I tried. But my mother always told me never to leave holes in the physics model because Havok will make a havoc out of it so I placed them with no gaps between. Was my mother wrong???

---

Was your mother wrong when she told you "never to put anything smaller than your elbow in your ear " ?

No , your mother is never wrong so naturally anything she says will over-ride anything you are instructed to do by anonymous people on the Internet. 

---