12/13 Meeting Notes

[Ryan-Gibbons]

TA Meeting

• For matching meshes to physics without doing it manually, add colliders to each part?
• Most meshes are triangle meshes -> have whole mountain be a trinagular mesh with collisions for each triangle
• Split up road and mountain, might not want to have to check the collisions with all of the scene every frame
  - Avoid really big single models, go for smaller models
• Our physics engine might cull the collision checks to within a bounding box for us, might only have to do that manually if we're having performance issues
• LOD should have different vertex counts: have both models loaded into scene, and just switch out which we're rendering based on the distance. This way, the non-visible objects never enter the rendering pipeline
  EX:

      Scene
  /    |     \ 
     Group
       |
  Object/Model
       |        \
     gltf        gltf
   model1     model2
 
model1.visible = true, model2.visible =  false 

• FirstPersonControls is the move for camera -- if we're going to do it manually, make cursor lock in place while still taking user mouse input. Restrict view based on relation to car's axis rather than world view

• Make 3d text a child of the car so that it's fixed in front of it

• OR, have text be rendered somewhere else in the world with a second camera with an orthographic view, could count for a second view for our advanced features.

• Look to previous assignments for how the GLSL interfaces with js

• To start on fog, look toward using a fragment shader similar to the special material in the raytracer assignment

• Have particles and shade those (individual vertices), animate it using the fragment shader (apply noise)

• Look at THREE js particle systems