When I model my vehicle there's three things I have to do:
In the same way, I also keep a mel/Python script for aligning the rotation and scale world pivots. The physical shapes can (currently) only be one of:
- Make sure that the pivots are in the places where the joint anchors lye.
- Keep all shape data (can't kill history).
- Hang physical shapes onto mesh shapes (swapped by my importer).
In the same way, I also keep a mel/Python script for aligning the rotation and scale world pivots. The physical shapes can (currently) only be one of:
- box
- sphere
- capsule
The pink, transparent stuff you see in the models are the physical shapes. They're of course not rendered during normal execution, just used for collision detection and to some degree for moment of inertia.
Here is a typical scene graph when modelling. All nodes named i_xxx are ignored. My importer flips the physical and mesh nodes, yielding a pure hell of linear algebra. Also, I have made some simplifications for the modeller [...me?] that in no way eliviated the importer burden. (Since the linear algebra involved is far from "placed here, rotated so", I had a hard time getting this far, and I still have a few boogs.)