[Bf-committers] (no subject)

[Alexander Ewering]

Hi Kester! Glad to see you back :)

On Wed, 9 Feb 2005, Kester Maddock wrote:

> The collision sensors retriggering is because of a timing issue because the
> physics & logic run at different rates.

Well, if that were the case, it would need urgent fixing, too, as in this
state, the sensors are unpredictable. Many situations depend on clear positive
and negative pulses.

Even then, there are two totally distinct issues with sensors retriggering.

You're right that ONE of them is attributable to the different framerates.

Remember though that I rolled back the game engine in instinctive to the
*old* time system (without different frame rates), and there the problem
also occurs. It is indeed related to a sphere, no matter how close it is
to a surface, will still constantly bounce a tiny amount.

For me, as I need a stable physics system now for a commercial project,
I've rolled back the fixing to the original 2.25 code also. I've also
found an interesting way to improve [stacked] "Bounds" objects by
setting all their velocities to zero as soon as a certain impulse
threshold is not reached.

Another interesting way is to just suspend() dynamics then. But then
you need to find the appropriate place to resume() it, maybe when
collision state changes (from external influence).

> All in all, I think that ODE is the way to go.  An impulse based method, while
> simpler, has numerical precision & stability issues.
>
> I haven't had much time lately (I've been busy at work) as you might have
> guessed.

2.25 has had a quite stable, if simple, physics system.. I've managed to
rollback the system enough to be like 2.25 now... but for the official tree,
and for more complex physics, maybe another approach is necessary, yes.


| alexander ewering              instinctive mediaworks
| ae[@]instinctive[.]de   http://www[.]instinctive[.]de