Re: [Algorithms] Bsp and collision detection

[Neil M. <bi...@me...>]

 Perhaps Pixar does use polygons and I'm mistaken but Tony DeRose spoke =
about not getting so attached to texture maps, and said they don't use =
them.  Snyders work in 1992 rendered photorealistic scenes with direct =
ray tracing - something only possible in a scene composed of all =
implicit surfaces.

Reality is infinitely scalable - all objects are made of other things =
until they are too small to be seen.  Our eye
point samples the world to a certain resolution and certain number of =
colors.  I'd like to see a similar rendering approach eventually.  I'm =
not saying make a game today based on it.

If you want to represent an entire city or jungle, you can in a small =
amount of memory with the right implicit surfaces.  Then, just tesselate =
what you need for the immediate area.  This has strong internet uses as =
well as supporting collisions and photorealism.

Of course, non-polygons is a few years away - I'm not unhappy with the =
GeForce2 or OpenGL, but it's worth thinking about -=20
  ----- Original Message -----=20
  From: Matthew Newport=20
  To: Neil McLaughlin ; Gdalgorithms-List=20
  Sent: Friday, November 30, 2001 11:07 AM
  Subject: RE: [Algorithms] Bsp and collision detection


  There seem to be two issues here that are getting slightly mixed up.
  There's one question of what is the best representational format for
  graphics and collision detection in a computer simulation and there's
  another, not necessarily related question of what representation is
  closer to reality. While curves may prove to be a better system for
  computer simulation in the long term (though they're not very friendly
  to current PC 3D hardware when dynamically tesselated) they're =
certainly
  not any better a representation of reality than polygons. Neither =
bears
  much resemblance to the real world - there are no smooth curves in =
real
  objects, it all comes down to atoms in the end!

  As for Pixar not using polys, are you sure about that? I thought they
  used Renderman which is poly based. There's a common misconception =
that
  Pixar do ray tracing but that's not true. I seem to remember reading
  about Monsters Inc. recently and reading that they used loads of polys
  for the big furry blue guy's hair but I could be wrong about that...

  Matt.

  -----Original Message-----
  From: Neil McLaughlin [mailto:bi...@me...]
  Sent: 30 November 2001 13:41
  To: Gdalgorithms-List
  Subject: Re: [Algorithms] Bsp and collision detection


  An Oak Tree is a good example of the sort of stuff I'd like to see
  interactive - I want to climb the tree and/or explore the inside of
  arteries - walk along the wall, and when I come upon the branching =
area,
  smoothly enter.   Not just boxy 45 or 90 angle tunnels.  Does anyone
  know of such examples available?

  An oak tree IS a curved surface - a higher order curved surface - like =
a
  branching artery (seamless articulation).  It's a series of Swept
  surfaces that are stitched together. =20

  In  Computers & Graphics 22,6  (Fellner/Havemann 2/1999) they examined
  Max and Maya scenes and determined that to achieve the several orders =
of
  magnitude increase required for true reality simulations we need a =
more
  "compact and abstract format". =20

  This means implicit surfaces - for example a sphere is just an =
equation
  - why tesselate and carry around all those polygons, only to have to
  sort through them to do collisions?

  We need all shapes to use the same type of equation - so we have
  predictablity between any two shapes.=20

   John Synder did some work on Generative Modeling, mentioned in the
  latest Watt book. He went on to be chairman MS graphics research lab.
  Pixar doesn't use any polygons or textures. =20

  Compact descripition really do exist.  I've developed such a compact
  description.

  Neil McLaughlin

  ----- Original Message -----=20
  From: Gribb, Gil=20
  To: Gdalgorithms-List=20
  Sent: Friday, November 30, 2001 8:23 AM
  Subject: RE: [Algorithms] Bsp and collision detection


  In this day and age, I would say bsp trees are a fairly lousy solution
  to
  the collision detection problem. Modern geometry does not bsp well.
  People
  use it because they don't know any other way to do it.
  -Gil


  > Hi everyone,
  > i just can stop finding everyone talking about how fast, easy=20
  > and pleasant
  > to write was collision detection using bsp trees. But=20
  > unfortunately, even
  > though google helps, i couldn't find nothing interesting enough to =
get
  > started with it.
  > Does anyone have references, links, papers, websites, that=20
  > would help a lot
  > ?
  >=20
  > I was also wondering whether those algorithms could be used=20
  > also in a hybrid
  > engine (octree with mini-bsp inside for instance), or does=20
  > general collision
  > detection algorithms for the octree and then for the bsp have to be
  > performed ? How does one do it when using multiple data=20
  > structures like this
  > ?
  >=20
  > Thanks in advance
  > R=E9my Rakic
  >=20
  >=20
  > _______________________________________________
  > GDAlgorithms-list mailing list
  > GDA...@li...
  > https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list
  >=20

  _______________________________________________
  GDAlgorithms-list mailing list
  GDA...@li...
  https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list