[ODE] fxengine

[Kenneth Holmlund]

Anselm Hook wrote:

>>OpenSceneGraph seems to be the best one out there.  OpenSG is complicated.
>>None of the other Inventor clones seem very good.  Java3D requires Java.
>>Steve Bakers PLIB and SSG do not take enough advantage of C++.  For right
>>now I'm just using my own which although not doing culling is good enough
>>for me.  I use Lib3ds to load graphics and Cal3D loads puppets.
>>
Excellent! Robert Osfield of OSG arrives here tomorrow evening so this 
is certainly one of
the things we will discuss.

>>Well, I've married Lua to ODE so that I get a script based interface to
>>ODE.  That means that ordinary humans can fiddle with ODE environments
>>more easily.  But I'm still not happy with that.  What I think I really
>>need is a visual scripting tool that lets me drag and drop collision
>>hulls, dynamic bodies and graphic objects around and place them properly.
>>My schedule for that isn't determined yet.
>>
Havoks Reactor plug-in for 3DSMax is pretty cool and I hope the next 
version of Vertigo from CM-Labs
will do a similar job for Vortex. It would be very nice if ODE could 
become an alternative way of simulating
those ME files that Vertigo produces. Unfortunately noone has still 
managed to put together a platform
where the declarative models for the dynamics harmonizes with the 
scenegraph for the graphics.
There's actually an interesting proposal at: http://www.martinb.com/ (in 
particular
http://www.martinb.com/threed/animation/usescenegraph/index.htm).

>>Dynamically driving a biped is going to be nasty work - probably
>>impossible actually. 
>>
Russel should know all about this from his thesis I think :)

>>( Most people use a kinematics based model and
>>translate kinematic displacement into equivalent dynamic forces.)  I have
>>no aspiration to make a dynamics driven puppet - I'm just working on tools
>>for dynamics 'assisted' kinematics authoring - ultimately my code will
>>spit out a sequence of key-frames.
>>
>>-a
>>
I have to check out Lua. I don't know anything about it.

I should say that our ambitions for driving the character simulation 
with dynamics are
rather humble. We want to use procedural animation whenever it is 
possible since we
also want the system to scale up to crowds of people. Procedural 
animation is pretty ok in many
cases, but it becomes very messy when you realize that you're mainly working
on the procedures and not so much on animations...
Various types of simulation algorithms (forward, inverse, etc) can be 
very useful for
simulating specific movements, gestures and actions, but it is not 
realistic to think
of full real-time dynamics of a human... (not to mention the control 
system for it...).

/Kenneth
-- 
Kenneth Holmlund
Director VRlab
HPC2N, Umeå University, Sweden
holmlund at hpc2n.umu.se
T: +46-90-786 9655
C: +46-70-631 5520