Week 5 - collisions

Question 1

What makes implicit surfaces useful for collision detection?

Answer 1

It is quick to tell whether a point is on the surface or, or inside, the surface

Question 2

How does backtracking help if we are too late detecting when a collision has occurred and a point is inside the surface?

Answer 2

Backtracking, we compute the intersection point by going back in the direction that we entered the surface, and then finding the exact position of the surface

Very useful for ray object intersection

Question 3

How does quick dirty fix (projecting) help if we are too late detecting when a collision has occurred and a point is inside the surface?

Answer 3

This is a very inaccurate hack where we project back to the object’s closest point.

The detected collision position is inaccurate but much simpler to compute

Question 4

What is the complexity of collision detection for n dynamic objects?

Answer 4

O(n^2)

Question 5

What methods exist to make collision detection between n dynamic objects faster?

Answer 5

Bounding boxes (simplifying collision calculations?)

Recursive / hierarchical tests also.

Only spend time on parts that are close to eachother. (Don’t find collision between two parts of the scene that are very far away)

Question 6

When does the bounding box algorithm not perform well?

Answer 6

Predicting the intersection of a collision

Question 7

What are the types of bounding volume (box)?

Answer 7

AABB (Axis aligned bounding boxes)

Sphere
- Placing a sphere around the volume at different resolutions, if the sphere’s don’t intersect neither do the objects. Spherical collision test is much simpler. Much faster to find intersection between bounding spheres rather than the detailed objects

OBB (Oriented bounded boxes)
K-DOP (K discrete polytopes)