Real-Time Physics Simulation Forum

Posted: **Sat Feb 11, 2012 3:27 pm**

Hi, i have made some test program that uses btKinematicCharacterControllers to simulate humanoids that can walk around but I have some problems that I don't know why the occur.

When I increase the size of the charactercontroller, or more preciesly the size of the capsuleshape that the ghost object contains and performance goes down rapidally

When the capsule is 0.75*0.75 then the stepphysics takes approx 0.55 ms

and dispatchAllCollisionPairs takes 0.05 ms
and updateActions takes 0.435 ms

but when I let the capsule size be 30*30 then the difference is huge.

then in 1 frame the stepsimulation takes 103 ms

and updateActions takes 97.097 ms
and dispatchAllCollisionPairs takes 5.915 ms

is this normal?, nothing else is changed and the charactercontrollers are not colliding with eachother

very glad for as many responses as possible (as long as they have suqqestions and information)

/Per

Posted: **Sun Feb 12, 2012 6:42 am**

any1 got an idea about why?, would appreciate every suggestion

Posted: **Thu Feb 07, 2013 4:31 pm**

I'm having the same problem, I've searched the forum for it but this one is the closest with the problem I'm having now.
Does anyone have any ideia???

Posted: **Fri Feb 08, 2013 2:43 pm**

I speculate much more manifolds are generated. Mind checking your manifold count?

Posted: **Fri Feb 08, 2013 9:02 pm**

MaxDZ8 wrote:I speculate much more manifolds are generated. Mind checking your manifold count?

Sorry I forgot to tell that english isn't my first language and also...I'm a newbie on bullet (started two weeks ago) so...how do I do that???

Posted: **Sat Feb 09, 2013 8:02 am**

Call btCollisionDispatcher::getNumManifolds().

Posted: **Sat Feb 09, 2013 5:07 pm**

It's 2...it's always 2...even in the parts where I got slowdowns...

Posted: **Wed Feb 13, 2013 2:37 pm**

What broadphase are you using? I might want to investigate that.

Posted: **Wed Feb 13, 2013 5:57 pm**

Code: Select all

broadphase             = new btAxisSweep3(VECTOR3D(-200,-200,-200), VECTOR3D(200,200,200), MAX_PROXIES);
collisionConfiguration = new btDefaultCollisionConfiguration();
dispatcher               = new btCollisionDispatcher(collisionConfiguration);
solver                     = new btSequentialImpulseConstraintSolver();
dynamicsWorld         = new btDiscreteDynamicsWorld(dispatcher, broadphase, solver, collisionConfiguration);

Posted: **Tue Feb 19, 2013 6:44 pm**

I'm sorry I've grossly failed to explain myself.
I would like to replicate your experiment so I can investigate.
Do you mind sharing the code?

In the meanwhile, also try btDvbtBroadphase.

Real-Time Physics Simulation Forum

btKinematicCharacterController wierd when size gets larger

btKinematicCharacterController wierd when size gets larger

Re: btKinematicCharacterController wierd when size gets larg

Re: btKinematicCharacterController wierd when size gets larg

Re: btKinematicCharacterController wierd when size gets larg

Re: btKinematicCharacterController wierd when size gets larg

Re: btKinematicCharacterController wierd when size gets larg

Re: btKinematicCharacterController wierd when size gets larg

Re: btKinematicCharacterController wierd when size gets larg

Re: btKinematicCharacterController wierd when size gets larg

Re: btKinematicCharacterController wierd when size gets larg