[ODE] Re: trilist strategy

[Gary R. Van Sickle]

[snip]

> > S. Redon, A. Kheddar and S. Coquillart
> > "Fast Continuous Collision Detection between Rigid Bodies"
> > In proceedings of Eurographics. September 2002
> >
> > http://www-rocq.inria.fr/~redon/research.htm
>
> yeap, looks nice!
> I've just had a read about it (and sent a question (in french, it
> helps!) to the author).
> The problem is that it seems to be slower than discreet detections
> (hehe, we can't have both good detection and speed... too bad!).
> All samples also speak about only two bodies, and one stills, while the
> other moves.
>

In every collision detect scheme I've seen, the two-body problem is always
described in the "one moves, other stationary" manner.  This generalizes to the
"both are in motion" case by simply considering the motion of one body in the
also-in-motion frame of reference of the other.  E.g.:

V1 = <nonzero>;
V2 = <nonzero>;

Change frame of reference to V2 ==>

V1r = V1-V2;
V2r = V2-V2 = 0;

IIRC, this transformation applies to the technique presented in the paper.

> In a general rigid-bodies algo, I think it's a real problem, as often
>  there are many bodies...

His procedure is iterative, and handles multiple simultaneously-colliding bodies
in (IIRC) a proven-correct manner (given an arbitrary but finite number of
iterations).

--
Gary R. Van Sickle
Brewer.  Patriot.