Position Based Dynamics: Patent

Issues with the forum and all license/patent related discussion
Dirk Gregorius
Posts: 874
Joined: Sun Jul 03, 2005 4:06 pm
Location: Kirkland, WA

Post by Dirk Gregorius » Thu May 17, 2007 1:25 pm

If they would have came up with really something new where they have spent money and manpower in I still wouldn't like it, but that is how it works. But if somebody so extremely obviously claims sombodies else invention as his own (in this particular case the work of the whole ODE community and the invention of Jacobsen - they even state in the paper that it is the same) this is impudent. Sorry I can't find any excuses for this pathetic behaviour.

Antonio Martini
Posts: 135
Joined: Wed Jul 27, 2005 10:28 am
Location: SCEE London

Post by Antonio Martini » Thu May 17, 2007 1:40 pm

Dirk Gregorius wrote:If they would have came up with really something new where they have spent money and manpower in I still wouldn't like it, but that is how it works. But if somebody so extremely obviously claims sombodies else invention as his own (in this particular case the work of the whole ODE community and the invention of Jacobsen - they even state in the paper that it is the same) this is impudent. Sorry I can't find any excuses for this pathetic behaviour.
if we want to be accurate the first time i saw the position based approach was in the older Provot's work and not in the Jacobsen's.

http://www-graphics.stanford.edu/course ... dcloth.pdf

and i have seen similar methods used in other papers as well. Verlet integration isn't necessary at all.

cheers,
Antonio

User avatar
Erwin Coumans
Site Admin
Posts: 4141
Joined: Sun Jun 26, 2005 6:43 pm
Location: California, USA
Contact:

Innovation: Free and Unencumbered by Patents.

Post by Erwin Coumans » Thu May 17, 2007 1:57 pm

Innovation, and the ability for companies and the game/movie fx industry as a whole to move forward is central to this discussion and relevant to everyone involved.
Having worked for Havok, and currently at Sony/SCEA, briefly on the Ageia Playstation 3 port and Bullet physics library at the moment (open source and Cell SPU parallel optimized version) I want to make following statements:
  • - Open source, and open standards encourage innovation and collaboration between companies. Even Havok and Ageia etc. widely use public resources for their benefit/innovation. This includes algorithms like GJK, EPA, SAT, SAP, Relaxation/iterative solvers, BVH hierarchies/aabb trees. Also most benefit from public implementations like Qhull, ODE, OPCODE, Bullet, SOLID etc. in various ways ranging from taking ideas, derive work or benchmark/comparison.
    - Open file formats/standards promote innovation. As game/movie fx production becomes more expensive and complicated, it's valuable to quickly combine available software packages to support novel tool pipelines rather then creating an entire pipeline from scratch.
So I fight for free open resources. That is why I setup this forum, participate in COLLADA Physics standard and develop Bullet physics library under the ZLib license. Individuals and (even large) coorporations collaborate, contribute, and that really stimulates innovation. Hopefully this encourages others to publically share and collaborate on innovations:

Free and Unencumbered by Patents.
Erwin

raigan2
Posts: 211
Joined: Sat Aug 19, 2006 11:52 pm

Post by raigan2 » Thu May 17, 2007 3:05 pm

AntonioMartini wrote: if we want to be accurate the first time i saw the position based approach was in the older Provot's work and not in the Jacobsen's.

http://www-graphics.stanford.edu/course ... dcloth.pdf

and i have seen similar methods used in other papers as well. Verlet integration isn't necessary at all.

cheers,
Antonio
I think Dirk was referring not so much to the position-based aspect, but to the "project positions along the gradient of the constraint function" which was explicitly given by Jakobsen, but now is patented by Muller.

raigan

raigan2
Posts: 211
Joined: Sat Aug 19, 2006 11:52 pm

Post by raigan2 » Thu May 17, 2007 3:15 pm

Dirk Gregorius wrote:I just remember that AGEIA holds another patent which is a description of the ODE Quickstep solver.
Has this had any effect on projects using ODE, or ODE development? Is the patent the reason that there hasn't been much development of ODE recently?

This would seem to be a big liability as it would preclude ODE from being used in commercial games.. but it's been used in several (S.T.A.L.K.E.R is the most recent I can recall).

This makes me think that using a PBD-based solver in commercial game code (or as part of an open-source engine) may not be as big a problem as I had originally thought.

Are BloodRayne2/STALKER/etc all being suicidally risky? Or have they secured a license from Ageia? I'm pretty sure they're using the quickstep solver ;)

raigan
Last edited by raigan2 on Thu May 17, 2007 11:15 pm, edited 2 times in total.

Antonio Martini
Posts: 135
Joined: Wed Jul 27, 2005 10:28 am
Location: SCEE London

Post by Antonio Martini » Thu May 17, 2007 4:20 pm

raigan2 wrote:
AntonioMartini wrote: if we want to be accurate the first time i saw the position based approach was in the older Provot's work and not in the Jacobsen's.

http://www-graphics.stanford.edu/course ... dcloth.pdf

and i have seen similar methods used in other papers as well. Verlet integration isn't necessary at all.

cheers,
Antonio
I think Dirk was referring not so much to the position-based aspect, but to the "project positions along the gradient of the constraint function" which was explicitly given by Jakobsen, but now is patented by Muller.

raigan
from:

http://www.teknikus.dk/tj/gdc2001.htm

"The linear equations or code to solve the constraints can be obtained by deriving the Jacobian of the constraint functions"

cheers,
Antonio

mewert
Posts: 57
Joined: Sat Oct 08, 2005 1:16 am
Location: Itinerant

Post by mewert » Fri May 18, 2007 6:46 pm

Are BloodRayne2/STALKER/etc all being suicidally risky? Or have they secured a license from Ageia? I'm pretty sure they're using the quickstep solver ;)
I'm not lawyer, but if I were to implement a PBD-cloth solution or a quick-step style solver, I would put in it my documentation that it is an implementation of Jackobsens or ODE's techniques, which pre-date the patents claims. I think you'd have a pretty rock solid case there. I'm sure the authors of the patent know this too and would be too emabarassed to actually try to enforce their farcical claims.

raigan2
Posts: 211
Joined: Sat Aug 19, 2006 11:52 pm

Post by raigan2 » Fri May 18, 2007 7:43 pm

mewert wrote: I'm not lawyer, but if I were to implement a PBD-cloth solution or a quick-step style solver, I would put in it my documentation that it is an implementation of Jackobsens or ODE's techniques, which pre-date the patents claims. I think you'd have a pretty rock solid case there. I'm sure the authors of the patent know this too and would be too emabarassed to actually try to enforce their farcical claims.
That's definitely a good idea; still, given what I've heard about the way patent suits work, knowing a patent is totally stupid hasn't stopped anyone from using them against less-deep-pocketed defendants ;)

_Does_ anyone know about how commercial games using ODE have dealt with this? If they've been left alone by Ageia that's good news, but I feel that it's just as likely that the legal teams involved have come to some sort of non-publicized agreement..

raigan

billyzelsnack
Posts: 6
Joined: Wed Aug 03, 2005 7:29 pm
Contact:

Post by billyzelsnack » Sat May 19, 2007 3:38 am

Dirk Gregorius wrote:If they would have came up with really something new where they have spent money and manpower in I still wouldn't like it, but that is how it works. But if somebody so extremely obviously claims sombodies else invention as his own (in this particular case the work of the whole ODE community and the invention of Jacobsen - they even state in the paper that it is the same) this is impudent. Sorry I can't find any excuses for this pathetic behaviour.
Maybe more cases like this..

Netflix Sued Over Fradulently Obtained Patents

http://yro.slashdot.org/article.pl?sid=07/05/18/1947245

and we'll start seeing less citations in papers that later patents are based upon. Kinda hard to prove that you didn't know about prior art when you have a citation for it in your paper!

KenB
Posts: 53
Joined: Sun Dec 03, 2006 12:40 am

We should tell them what we think...

Post by KenB » Tue Jun 12, 2007 5:50 pm

I think this is a good place to express to Ageia and other companies in this area what we think about patents on algorithms, in particular algorithms that, with small variations, have been known for decades.

I have a hard time taking them seriously.

Also, the scientific publishing/patenting combo really stinks. I think it would be a good policy if scientific journals and conferences simply refused to accept papers on algorithms, that also are associated with software/algorithm patents. But I suppose it boils down to the usual CG bragging rights, rather than science.

Dirk Gregorius
Posts: 874
Joined: Sun Jul 03, 2005 4:06 pm
Location: Kirkland, WA

Post by Dirk Gregorius » Tue Jun 12, 2007 6:19 pm

Actually it seems that there was a mistake in the paper. So I wonder how this effects the patent. Not that they patented some trivial, they patented it also wrong. This says a lot about the quality of Nx...

http://www.matthiasmueller.info/

raigan2
Posts: 211
Joined: Sat Aug 19, 2006 11:52 pm

Post by raigan2 » Tue Jun 12, 2007 6:29 pm

Hi,
Actually for me both the original and revised weighting method doesn't work as well as my home-made one -- both produce oscillations/over-corrections when the error is high and there are large mass ratios.

To update, we've been communicating with Muller and some other Ageia people to see about this. I actually had emailed Muller about the weighting terms several months ago, and since I failed to use an alias on forums, I've attracted lawyer attention. I presented (I think) a convincing argument for the invalidity of the patent, I'm hoping that we can come to some sort of agreement..

The main thing is that they're attempting to patent a technique which is a _generalization_ of an existing technology, which seems sort of backwards -- if a specific instance of the technique is known, the more general method should also be considered know too.. since their position-based-dynamics contains Jakobsen's method as a subset, Jakobsen would appear to be covered by their patent, which is obviously wrong.

Anyway, thanks everyone for the feedback and help, I'll let you know how things develop. Many indie games use patented tech (any real-time fluid sim that uses Stam's "Stable Fluids" for instance) however most patent-holders seem to be reasonable about this -- it's not like we're announcing 3rd party AAA middleware based on this stuff, we just want to make a simple 2D platformer!

raigan

Oscar Civit Flores
Posts: 25
Joined: Fri Jan 20, 2006 2:24 pm
Location: Barcelona

Post by Oscar Civit Flores » Thu Jun 14, 2007 10:51 pm

Hi all,

This patent thing is looks like the arms race during the cold war to me... they've all gone MAD.

http://en.wikipedia.org/wiki/Nuclear_arms_race

That said, I don't think Müller or any other author of the paper is to blame, most of them have solid non-patented (AFAIK) research careers. Patenting looks like a "bussiness" decision and I doubt they had any chance to say "no". So maybe we can just stop buying PhysiX cards if they troll you :twisted:

Boycott is not patented, so count on my support if required.

Oscar

Evan Christensen
Posts: 3
Joined: Sat Jul 23, 2005 12:58 am
Location: Bellevue, Wa

Post by Evan Christensen » Fri Jun 15, 2007 12:25 am

Sadly a boycott seems like it will probably backfire. Ageia's inability to sell their physX cards makes it all the more likely that they will need to rely on groundless patent suits for income.

raigan2
Posts: 211
Joined: Sat Aug 19, 2006 11:52 pm

Post by raigan2 » Fri Jun 15, 2007 12:26 am

Hi,
We've been speaking with Muller and other people at Ageia, they're quite sympathetic. So, hopefully this won't be a problem, I'll post once the whole thing is resolved.

The main thing is that, since this paper is a generalization of Jakobsen's method, it actually covers his method.. which doesn't make _any_ sense. But that's what happens once you get into software/algorithm patents I suppose.

raigan

Post Reply