Integrating 3D models with photography
Interested in integrating your 3D work with the real world? This might help
# 1 01-03-2004 , 06:08 PM
dragonfx's Avatar
Registered User
Join Date: Jul 2003
Posts: 1,257

Is maya non-deterministic?

I mean, i have made a simple setup: 2 active hardbody spheres with some initial velocity, affected by a gravity field colliding with a pasive hardbody plane.
When i hit play none of the two spheres will have exactly the same trajectory...(at the begining yes but let it run 600 frames and it starts to differ)
I havent noticed that before and Im no adept of dynamics but... is that a normal behavior?

(at the beginning i tought it was that i turned the second sphere into a soft but repeating the setup to before that point the trajectorys remained non-deterministic...)


Last edited by dragonfx; 01-03-2004 at 06:18 PM.
# 2 01-03-2004 , 06:56 PM
james_s's Avatar
Subscriber
Join Date: Jan 2004
Location: Connecticut
Posts: 65
I've had the same question -- just hadn't asked it yet.

I would have thought the dynamics would be the same every time, else how could you rely on your simulation construction?

However, I've seen the same thing -- the simulation behaves slightly differently each time it is run.

Why?

:ermm:

# 3 02-03-2004 , 03:14 AM
Pony's Avatar
Subscriber
Join Date: Mar 2003
Location: with PonysGirl
Posts: 2,573
you can rely on your simulation other wise you would see weird behavior right off. I think the slight variation after 600 frames has to do with it not using super high precision, that inserts enough randomness after a long while for the chaos effect to take place. They dynamics would be much slower if it was made to be exact even after 600 frames. After all even after 600 frames, a shear will still have the same velocity after falling X number of feet with in 3 or 4 decimal places, as it did at frame 10.

# 4 02-03-2004 , 03:15 AM
mtmckinley's Avatar
The Maya Mountain
Join Date: Aug 2002
Location: Seattle, WA
Posts: 8,245
you can always bake the simulation and then it will play the same each time.

# 5 02-03-2004 , 09:18 AM
dragonfx's Avatar
Registered User
Join Date: Jul 2003
Posts: 1,257
well pony, decimal after decimal sometimes the balls collide wich each other... and sometimes they dont, and sometimes they fall form the plane and lately they dont!... at the beginning they satisfactorily go bumping and rolling in the same way, but at the ending the variations are big, otherwise i wouldnt have noticed em...

weird...


Last edited by dragonfx; 02-03-2004 at 09:25 AM.
# 6 02-03-2004 , 10:07 AM
rich's Avatar
Subscriber
Join Date: Oct 2003
Location: Paris, France
Posts: 418
I must admit I dont know alot about Mayas implementation of dynamics but if it is non-deterministic, it would have to have been developed this way because a computer program is inherently deterministic.

This means that Alias would have included some 'randomizer' to simulate chaotic behviour. It follows that, knowing Alias, they would allow you to toggle this 'randomizer' on and off.


That'll do donkey... that'll do...
# 7 17-03-2004 , 05:54 AM
Registered User
Join Date: Jan 2003
Location: saudi arabia(Indian)
Posts: 4
is it falling on a horizontal velocity or is it falling under gravity

# 8 17-03-2004 , 05:57 AM
Registered User
Join Date: Jan 2003
Location: saudi arabia(Indian)
Posts: 4
dragonfx!
every particle of the universe attaract every other pariticle f the universe with a force directly propotional to product of masess and inversely propotional to square of distance between them!

# 9 17-03-2004 , 07:29 AM
mascal's Avatar
Registered User
Join Date: Dec 2003
Posts: 16
Dynamics calculations are computed in a very complex way.
Many tasks start when dynamics is needed and it's impossible to know which one will terminate first.
So the results will never be outputted in the same order when you recalculate the animation.
Furthermore many calculations are interdependent and the random completion will actually put a small amount of randomness in the final result.
This is a small bit of noise, but it will be added for every frame, so the longer the animation and the more different could be the animation if recalculated.
To this you have to add the fact that every object in Maya is made of triangles, even nurbs, so a cg sphere will behave different from a real sphere and so on...

Caching has been created also for this problem (not only for speed reason). So if you should create a cache for an animation with dynamics and keep it if you are satisfied or delete and recreate it if you don't like what happened.

Bye!!

# 10 17-03-2004 , 01:45 PM
james_s's Avatar
Subscriber
Join Date: Jan 2004
Location: Connecticut
Posts: 65
I completed a "rocket" simulation that I found in my "Maya Complete" book. I spent quite a bit of time fiddling around with things like the position of the impulse to get the rocket to do what I wanted.

I also had an Aim Constraint on the ground-based camera so I could watch the rocket from launch to crash-landing.

After the rocket engine "burned out", it would fall back to the ground and bounce around a bit.

I wanted to be able to position the camera so that at the end, when the rocket stopped rolling around, it stopped basically in front of the camera. However, I was never able to get the rocket to behave the same way twice in a row, so it was impossible for me to ever get that camera position exactly right.

Since I didn't have any turbulant fields or other intentional randomness I don't know what caused the problem.

Is this the kind of randomness everyone is experiencing?

# 11 17-03-2004 , 02:25 PM
mtmckinley's Avatar
The Maya Mountain
Join Date: Aug 2002
Location: Seattle, WA
Posts: 8,245
You can bake dynamics to get the same result over and over.

# 12 17-03-2004 , 02:57 PM
james_s's Avatar
Subscriber
Join Date: Jan 2004
Location: Connecticut
Posts: 65
The problem is more fundamental. If I can't get the simulation to duplicate behavior each time it is run, then things like baking or building a particle cache are kind of "run it and hope it comes out like I want" events -- which is exatly what I've seen.

I was under the impression that baking a simulation is done to save processor time or allow the simulation elements to be adjusted with traditional keyframe animation tools. Similarly, I understood that the purposes behind creating a particle disk cache are to reduce processing time, allow accurate scrubbing in the timeline, and for rendering.

The results from the dynamics system should be exactly the same every single time it is started with identical elements and parameters. Any random elements not purposely introduced by the user should be considered errors and be eliminated.

# 13 18-03-2004 , 08:40 PM
dirack's Avatar
Registered User
Join Date: Oct 2003
Location: Cage #42 of the Asylum
Posts: 396

you can always bake the simulation and then it will play the same each time.

You can bake dynamics to get the same result over and over.

I can guess what you're going to say next :p

Oh, and I was just thinking of something... over on IRC some time ago, kbrown I believe, said it was possible with some mel record<something>input, and create a motion path with using a joystick in real time. Isn't it possible to create a mel for recording a one time input like that, only now the input is the used dynamics instead of the joystick? user added image user added image

Posting Rules Forum Rules
You may not post new threads | You may not post replies | You may not post attachments | You may not edit your posts | BB code is On | Smilies are On | [IMG] code is On | HTML code is Off

Similar Threads