> Correct me if I'm wrong, but don't game engines "usually" use an invisible 
> skeleton rather than the actual visible polygon mesh for collision 
> detection purposes? (Precisely to simplify the problem and make it 
> tractable in realtime.)

No, you're completely correct, in fact I'm already doing that in my game. 
The problem I have now is that because my physics loop runs at 100 Hz and 
some objects are moving quite quickly (like a metre per physics step) it's 
hard to determine exactly where the collision took place, and hence 
calculate the correct impulse to stop them intersecting at the next step.  I 
don't have enough CPU cycles spare to roll back the simulation with a finer 
time step, so I'm just using some hackery to estimate it for now.  It works 
ok most of the time, and things bounce and spin off nicely in 3D, but 
occasionally one object visibly goes inside another for a bit.  I'm sure 
with a bit more hackery I can make it more robust without having to do more 
physics steps.  Then I need to optimise the collision *detection*, as it 
gets a bit slow when 10 or so objects are colliding simultaneously.

Post a reply to this message