gdalgorithms-list

Re: [Algorithms] Broad-phase collision detection for dynamic objects

[Jon W. <jw...@gm...>]

Stuart Golodetz wrote:
>
> before.) Basically I started out thinking about a hierarchical grid-type 
> scheme (the sort you'd find in e.g. Christer Ericson's Real-Time 
> Collision Detection book) but my understanding was that as written it 
> works by looking only at the objects' positions at the point of 
> collision detection, and doesn't take account of their movements over 
>   

You could put bounding volumes of the moving objects into the grid (or 
loose octree, or sweep-and-prune). That would identify possible 
overlapping pairs. It still would be sufficiently tight, unless you have 
multiple objects moving diagonally really fast.
Also, I believe the Bullet library uses continuous (swept) collision 
detection, so you can find a bunch of code there to read up on.

Sincerely,

jw


-- 

  Revenge is the most pointless and damaging of human desires.

Re: [Algorithms] Broad-phase collision detection for dynamic objects

[<chr...@pl...>]

Jon Watte wrote:
> You could put bounding volumes of the moving objects into the grid (or
> loose octree, or sweep-and-prune). That would identify possible
> overlapping pairs. It still would be sufficiently tight, unless you have
> multiple objects moving diagonally really fast.

Jon is right (again). For the broad-phase, the simplest
solution is to work with volumes that bound the objects
under motion, and insert these volumes into your
spatial partitioning (hgrid, etc).

This breaks down for very fast moving objects (e.g.
projectiles) which you may want to handle another way,
but only if they actually show up as a bottleneck
(which is unlikely).


Christer Ericson, Director of Tools and Technology
Sony Computer Entertainment, Santa Monica