Area Damage for Rockets, Grenades etc.

MickeyKnox · Post by **MickeyKnox** » Sat Jun 20, 2009 2:26 pm

I would like an opponent to take damage when i fired a rocket on him, that did not actually
hit him, but hit the ground close to him.

Thus i would like to have a list of all ISceneNodes in some sphere. I would then calculate for
all these SceneNodes there distance to the impact and apply damage accordingly.

Now, how do i get such a list of SceneNodes in some sphere?
I think the buildin CollisionSystem does not work for me in this case.
Any ideas?

JuicyLobster · Post by **JuicyLobster** » Sat Jun 20, 2009 2:59 pm

well a "sphere" could be thought as a distance to target really.

if "this node" is "this far" from "impact position" then apply "this damage"

You could then depending on the distance between positions of the node and impact spot set a fall off damage from the distance value.

edit: so to get all nodes within a certain spherical distance your just looking for the radius length between nodes in relation to the impact point of the rocket.

MickeyKnox · Post by **MickeyKnox** » Sat Jun 20, 2009 3:07 pm

But i still need that initial list of SceneNodes in range. Otherwise i would have to check the
distance to all SceneNodes in the Scene, which does not scale good: with lots af SceneNodes
this would take too long.

Sylence · Post by **Sylence** » Sat Jun 20, 2009 3:18 pm

Comparing calculating the distance between two nodes and comparing this distance with another value doesn't take ages.

If this is really too slow you could use a bsp tree or something similar. So you only have to check nodes next to the position.

Kalango · Post by **Kalango** » Sat Jun 20, 2009 5:04 pm

Yeah if you make some sort of quick sort algorithm, you can quickly select the nodes that are affected, and them, manage the damage, etc...
OR, you can make each enemy objetc to check the colision at every loop....
Anywaysm, you can always make siple apps to make tests and then choose the best way....

FuzzYspo0N · Post by **FuzzYspo0N** » Sat Jun 20, 2009 6:08 pm

Normally, a rocket has an on collision event. The players/enemies will have a bounding box, and so can the rocket itself.

So all you need is rocket bounding box and player bounding box. On rocket collision : (inside gamerocket.cpp for example)

if there are 16 players (which is a LOT) its still not a lot of checks on a bounding box.

if rocket.boundingbox contains/intersects with (any of the player boxes), calculate the distance and apply force.

Code: Select all

in void onRocketCollision()
{
    for each player
       if rocket.boundingbox intersects/contains players[i].boundingbox
          damage = getDistanceAndDirectionFromPlayer (players[i]);
          applyForceToPlayer(damage);
}

getDistanceAndDirection is obvious too,
dir = (rocketPos - PlayerPos) is the "direction" of the force, and
dir.normalise will give you a value between 0 and 1 for the force "amount". You can then lerp the values between max rocket damage, and that percent of damage.

MickeyKnox · Post by **MickeyKnox** » Sat Jun 20, 2009 7:34 pm

Ok, i just check the distance for all SceneNodes in the Scene. I thought that would be
foolish and another way would be preferable.

Anyway, thanks for your help.

sudi · Post by **sudi** » Sat Jun 20, 2009 7:44 pm

why are u checking for all scenenodes? is really every scenenode going to be affectd by the rocket?

MickeyKnox · Post by **MickeyKnox** » Sat Jun 20, 2009 8:11 pm

Well, i will check for those, that might be affected.
My question was, if i can exclude SceneNodes from distance testing that are way out of
range somehow, and if irrlicht offers some facilities to do so efficiently.

Currently, there aren't so many SceneNodes in my game, so testing them all is feasible.
In a later state however, the number of SceneNodes may forbid such a trivial approach.
I'm planning to check out physic engines later, maybe i'll find a better solution then.

FreakNigh · Post by **FreakNigh** » Sat Jun 20, 2009 10:34 pm

well.. you could test their all their x, y, z distances to make sure they are under a certain distance, and if all of them are then do the sqrt(x*x + y*y + z*z) so that you quickly check the masses but slowly check the probables.

a compare function like if(x<50) is like 3 or 4 cpu cycles, but a x*x is like 40.

sudi · Post by **sudi** » Sun Jun 21, 2009 12:04 am

ok something really weird happend i just wrote a octTree class to do the quick bounding box lookup search.
I tried with 10000 scenenodes....took 0 ms to find the nodes in range.
then i did a simple list traversal and checked all scenenodes. this also took 0 ms...mesured with clock().

BlindSide · Post by **BlindSide** » Sun Jun 21, 2009 12:39 am

You are probably not performing any "changes" to anything outside the loop, therefore it would get optimized out.

E.g:

Code: Select all

for
{
    var2 = 5;
}

vs

Code: Select all

var = 0;
for
{
    var2 = 5;
    var += var2;
}

a compare function like if(x<50) is like 3 or 4 cpu cycles, but a x*x is like 40.

Sadly this is only true for very old processors. These days the cost of a branch is much higher.

Here is a great resource for reference: http://www.agner.org/optimize/optimizing_cpp.pdf