[ODE] Performance comparisons between different physics engines..

[Konstantin Voloshin]

Hello

There was a message by Russ a long time ago concerning possible
improvement of ODE approach to resolving penetrations:
http://q12.org/pipermail/ode/2002-April/005135.html

Briefly, that was about correcting bodies position instead of velocity
in case of penetration. A few months ago Adam Moravanszky
from Novodex posted that he has a working version of that idea.

Nate W.> Novodex is not prone to explosions after deep penetrations
Nate W.> - penetrations are resolved gradually rather than explosively.

This reminded me conversation with Adam a lot. I think, it's very
likely that the Novodex behaviour Nate has mentioned might come
from that algorithmic feature.

All this made me wonder, if that positional-correction approach
would fix StepFast problem with building stable walls of enabled
cubic bodies. Has anybody tested Novodex on this case (Nate,
Adam) ?

Konstantin


----- Original Message ----- 
From: "Nate W" <coding at natew.com>
To: <ode at q12.org>
Sent: Saturday, January 10, 2004 11:05 PM
Subject: RE: [ODE] Performance comparisons between different physics engines..


On Sat, 10 Jan 2004, Jon Watte wrote:

> > differences, for example Novodex is not prone to explosions after deep
> > penetrations - penetrations are resolved gradually rather than
> 
> Isn't that exactly what ERP controls in ODE? Reduce ERP a smidgen, and
> penetrations will be resolved over more than one step.

Kinda, but all other joint errors will be resolved gradually as well, so
it will affect all aspects of the simulation, not just penetrations.
  
It's almost as if the Novodex code allows one to adjust the error
reduction for contact joints separately from the error reduction of other
types of joints.  It seems to be different than ERP though - penerations
are resolved with constant velocity.

-- 

Nate Waddoups
Redmond WA USA
http://www.natew.com/