How do I use Bullet Raycasting for character vision? I want the character to be able see so when the character see enemies, the character should run/fire/etc...
I thought about the following method:
m_dynamicsWorld->rayTest();
But I am not sure how it should work for character vision.
8 replies to this topic
Sponsor:
#2 Crossbones+ - Reputation: 3548
Posted 12 November 2012 - 12:09 AM
No, it should work, I guess, but you'll probably need more than one ray to sample the character's field of vision. You might also want to take into account peripheral vision (at the edges, humans react better to visual stimuli, but discern less). So for instance, you could try and cast a ray from the character's eye to a bunch of enemies that are potentially in sight (perhaps throw a few rays for each enemy, to estimate how much of the enemy is actually visible), multiply this score by the enemy's rough position with respect to the character's field of vision (this can be done approximately with a dot product of the ray with the character's eye vector, i.e 1 if the enemy is directly in front of the character, falling smoothly to 0 if he is out of his field of vision), and trigger a response when the result is above a certain detection threshold.
Edited by Bacterius, 12 November 2012 - 12:10 AM.
#4 Members - Reputation: 353
Posted 12 November 2012 - 08:07 AM
You can use this to cast your ray
If you want a field of view, you can just cast a bunch of rays in a cone or something around the direction of your camera.
Note you should probably cast the user pointer to whatever you are storing with it before returning.
void* CastRay(btVector3& to, btVector3& from, float distance)
{
btCollisionWorld::ClosestRayResultCallback raycallback(from, to * distance);
m_dynamicsWorld->rayTest(from, to, raycallback);
if (raycallback.hasHit())
{
return m_collisionObject->getUserPointer();
}
return nullptr;
}
If you want a field of view, you can just cast a bunch of rays in a cone or something around the direction of your camera.
Note you should probably cast the user pointer to whatever you are storing with it before returning.
Edited by rocklobster, 12 November 2012 - 08:11 AM.
#8 Members - Reputation: 226
Posted 22 November 2012 - 07:27 AM
You could also have something like a btGhostObject , with a cone collision shape which represents the characters vision, you'd then be able to detect 'collisions' created during the simulation step . (depends on wether you're justing using raycasting or are using a bit more of the bullet library)
#9 Members - Reputation: 358
Posted 22 November 2012 - 10:07 AM
@xexuxjy: That's the EXACT idea that I have been trying to implement in the past few days!
The only problem that I'm having with it is that If I have a cone mesh, how do I load it into bullet physics? The cone mesh data will be stored into LPD3DXMESH.
The only problem that I'm having with it is that If I have a cone mesh, how do I load it into bullet physics? The cone mesh data will be stored into LPD3DXMESH.






