Skip to content

Latest commit

 

History

History
89 lines (78 loc) · 3.75 KB

lec6.md

File metadata and controls

89 lines (78 loc) · 3.75 KB
Raw

Graphics and Computation COMP30019 Lecture 6

Physics

  • They don't exist in an virtual environment by default
  • We have to implement these ourselves
  • Newton's Laws of Motion

Collision Detection

  • As the name implies, collision detection is all about the detection of an intersection between different objects in a virtual environment

Why is this important?

  • Prevent objects from clipping through each other
  • Produce realistic environments by having objects react accordingly
  • Bad collision results in things like this (warning, loud!)
  • Two different approaches to collision detection

Continuous (a priori)

  • We predict if a collision is going to occur based on the trajectories of the objects
  • This works pretty well but is also really expensive since you need to be checking every vertex for a collision
    • If you have thousands of collisions you're gonna have a bad time
    • If you have 1 object with 100 triangles, and another with 200, then there are 20000 different potential intersections to calculate
  • Very computationally complex

Discrete (a posteriori)

  • Allows objects to actually penetrate each other
  • The physics engine pushes it back to where the collision happened initially
  • You do not see the penetration (i.e. middle of the diagram)
  • This works okay but there's a chance that it'll suffer from tunnelling
    • If your object is travelling really fast then it might go through objects

Bounding Volumes

  • Used to simplify collision detection
  • Volumes enclose the whole object, and should be as small as possible
  • Simpler calculations
  • But you may end up with that grenade issue as shown earlier

Types of Bounding Boxes

Axis Aligned Bounding Box (AABB)

  • Bounding box remains oriented with respect to the main axes
  • Pros:
    • Translation invariant
    • Does not change if you move every point of the object by the same amount in a given direction
    • Simple to compute
  • Cons:
    • Needs to be recomputed every frame
    • Changing volume, so in some cases it might hit depending on how the tank is oriented

Oriented Bounding Box (OBB)

  • Box is oriented with respect to the tank
  • Pros:
    • Translation invariant
    • Volume is more compact than AABB, so less empty space
  • Cons:
    • Computation of intersections is complex as box orientation/rotation is dynamic

Bounding Spheres

  • Sphere around the object
  • Pros:
    • Transformation invariant
    • Simple collision detection
  • Cons:
    • More empty space in the bounding volume

Hierarchical Bounding Volumes

  • We can split a model into smaller bounding volumes to improve efficiency
  • Broad phase (simple, low cost)
    • Used as a first pass to identify potential collisions
  • Narrow phase (complex, high cost)
    • Once we find a potential collision, we do more complex collision tests
  • Once you have a collision you can do a lot of things after that collision with more complex series

Hitboxes

  • Collision detection but in one way
  • Don't need to figure out if a person hits a bullet - only the other way around
  • Here hierarchical bounding volumes are used to calculate damage based on where the character is hit
  • Larger hitboxes results in stuff like Hanzo in Overwatch
    • Apparently Gorge is not a Hanzo main
    • "This character is hard to play"

"Ethical developers - make sure you kill your chicken properly."