Tremulous Forum
General => Feedback => Topic started by: Redman on February 17, 2008, 12:05:27 am
-
Hello. In Tremulous players and buildables have box collision model (bbox). All "trem-modders" are making Tremulous more hard (tjw...) or more easy (TremX). But why? Isn't that better to make Tremulous more realistic? I think trem devs can make more detail collision models for players and buildables and which rotates with player. That will be more funny if dretch can go under tyrant, not tyrant stand on him.
-
This feels more 'feedback'ish to me, so I moved it.
As for your question, when's the last time you saw a soldier carrying a Lucifer Cannon, or a marauder slashing and flying through the air? Realism leaves a lot to be desired :>
-
Non-rectangular bounding boxes are much much much more expensive to compute collisions for. Someone more knowledgable of the engine will be able to say more, but I believe I've been told that it's beyond the Q3 engine as a whole.
-
I believe I've been told that it's beyond the Q3 engine as a whole.
That's what I've heard.
-
Don't pick a Quake engine if you want rigid body physics. Even Quake 4 had bounding boxes to conserver CPU power, although they are no longer cubes like Q1-3
-
Don't pick a Quake engine if you want rigid body physics. Even Quake 4 had bounding boxes to conserver CPU power, although they are no longer cubes like Q1-3
q1-q3 are not cubes, any axis aligned rectangular prism is a valid bounding box, and i think even halflife 2 uses AABB's for collision detection.
-
I meant cubes as a general term for rectangular prisms, sorry.
-
Don't pick a Quake engine if you want rigid body physics. Even Quake 4 had bounding boxes to conserver CPU power, although they are no longer cubes like Q1-3
If want to see what happens if you don't do bboxes, but some crazy per nano-pixel hit-detection, look at DOOM3.
-
I don't think about 99999999999-faces collision models (picture: http://img225.imageshack.us/img225/7412/57577549mg2.jpg). I just think about some more details, like that: http://img222.imageshack.us/img222/8930/24177984cw2.jpg
-
If you were to do a model-based collision detection, you would likely want a low detail model. By low detail, I mean monochrome and 100-400, maybe 1000 at the most, vertices.
-
What would monochrome have to do with it?
-
Actually a very low definition model that is only used for collision detection on the server side (also client side for prediction) would be enough. You aren't rendering it, just using it internally.
-
even with current amount of 1 hitbox (a VERY low definition model huh?) servers are pushed to the maximum (64 slots, 999 bp etc.) with anything consisting of more than 6 faces to check the CPU load will go up a lot.
On the other hand, a WolfET mod : jaymod, has introduced an oriented hitboxes, and also had about 10 of them per player, moreover, they have moved, following the player animation, such as walking. But even that did increase the load a lot.
-
there is a reason that the quake 3 is free