On Fri, 10 Dec 1999 22:41:11   Jeff Pierce wrote:
>As Andreas pointed out, implementing physics
>isn't necessarily straightforward.  As an extreme
>simplification, you can think about two sides to
>the spectrum:
>
>Fast,                                     Slow,
>Simple,  <----------------------------->  Complex,
>Wrong                                    "Correct"
>
>To use collision detection as an example, on the
>left side you have simple bounding box collision
>detection.  It's fast, it's easy, but unless
>everything in the scene is a cube it's a gross
>approximation. On the right side you have
>collision detection that uses the actual mesh.
>And then in between you have things like
>OctTrees.  The more computational and
>intellectual effort you're willing to spend, the
>more realistic your collision detection will be.

Doesn't this depend on the number of alternative
representations for the "world" one is willing to
support?  And upon the formulation of physics one
wants?

For instance, a basic result from polyhedral world
modeling is that there is a transformation one
can apply to a representation of a world with but
one mover such that the moving object can be
reduced to a point and all fixed, polyhedral objects
are systematically expanded so if the point is ever
contained by them, one can infer that the polyhedral
mover in the original world would collide with them.
The matter gets more complicated once rotations of
the objects are considered -- requiring one to
go into higher dimensions to model it -- but it
remains doable.

The challenge in a game world is updating the two
representations at once.

This idea is used in some robot planning work.

Similarly, although conceptually our canonical
formulation of physics considers frictionless
motion to be the basis for motion modeling,
and friction effects are applied on top of that,
one can formulate a friction-implicit version of
ground motion.

Over the years, there's been work done in this
area under the banner of "qualitative physics"
and that's a good keyword pair to search
literature on.

At a much simpler level, one thing which is
essentially in any model of a world is
representing rotational states of objects.
I could, for instance, work towards a
comprehensive representation of polyhedrals
with positional and rotational states,
possibly velocities and angular velocities.
This would serve as a concrete substrate to
test out my ideas on literate programming.

Or is that too basic and handled well by
what's out there already?

[snip]


---
___________________________________________
demiourgos at smalltalk.org        squeak.org/
home.stny.rr.com/algebraist/ 
www.smalltalk.org/ jtgalkowski at alum.mit.edu




Get your FREE, private e-mail account at http://www.salonmember.com.