Is there any fast (fast enough for real-time games) way of doing continuous angular collision detection without time steps?
I'm not sure whether to use ellipsoids or capsules for my player's bounding volumes yet but I was thinking about angular collision detection.
Currently I only do sphere to triangle tests for each triangle of the bounding volumes for my static environment, which I don't actually know if it's such a good idea because it's a whole lot of code for each triangle test. Do you have any thoughts on that? I'm trying to make a RPG game in which you fight with bow and arrow or sword and shield and I would like to create some physics effects to be able to add some more fun with boxes flying around and stuff like that, so I guess I could sacrifice some performance for quality since it's not a large amount of physics calculations I guess.
Anyway, I was thinking about how to go about doing continuous angular collision detection because I want a very robust technique which wouldn't eat too much computation power. I haven't yet found any information if there's even a way to do real continuous angular collision detection ("real" as in finding out the time of impact for the rotational motions) but I have found some material on using time steps for the angular calculations.
If there is a way to do continuous angular collision detection for rotations around any axis, could you please refer any material on this or summarize the technique?
Thanks in advance!
continuous angular collision detection?
-
dblack
- Posts: 1
- Joined: Mon Jul 25, 2005 5:58 pm
- Location: Zurich Switzerland
Re: continuous angular collision detection?
Why do you need angular collision detection?
You dont have to use the same collision geometry for everything. For example I have a capsule which is used to control player movement(it is constrained to always be upright). Then I have collision geoms for player limbs etc(ie a ragdoll). Debris, sword impacts, bullet imapcts etc are set to collide against the ragdoll. But the actual movement(except for when the player dies or for attached objects like clothing) is all done with linear capsule sweeps. Things which hit the ragdoll never move fast enough/tunneling isnt important enough to justify CCD for them.
In any case angular sweep tests are quite possible, although the math is a bit hairy in some casses and they can be expensive. (I have only experimented a little, but I am considering this for things like rotating platforms etc which can push/crush the player. But I will probably never need this and be fine with linear sweeps).
If you want somewhere to start, I have known people who have had success with methods described by Stephane Redon. Just do a google search.
David
You dont have to use the same collision geometry for everything. For example I have a capsule which is used to control player movement(it is constrained to always be upright). Then I have collision geoms for player limbs etc(ie a ragdoll). Debris, sword impacts, bullet imapcts etc are set to collide against the ragdoll. But the actual movement(except for when the player dies or for attached objects like clothing) is all done with linear capsule sweeps. Things which hit the ragdoll never move fast enough/tunneling isnt important enough to justify CCD for them.
In any case angular sweep tests are quite possible, although the math is a bit hairy in some casses and they can be expensive. (I have only experimented a little, but I am considering this for things like rotating platforms etc which can push/crush the player. But I will probably never need this and be fine with linear sweeps).
If you want somewhere to start, I have known people who have had success with methods described by Stephane Redon. Just do a google search.
David