Re: [Algorithms] GJK

[Will P. <wi...@cs...>]

> > I think all descriptions I have read of it have
> > been just hacky ways to determine how to convex-ly combine (by finding the
> > best lamda values to use) the points in the previous simplex and the new
> > support point to find the minimal convex simplex that contains that new
> > support point.  Does that seem correct?
> 
> I think so - as far as I can tell, and including the word "hacky" :)

I think it's just hard to understand because the recursive solutions
provided have been simplified already from cramer's rule.  I'm not seeing
why the new "closest point in simplex to origin" is perpendicular to the
affine combination of the points in simplex, so I've stopped at that point
to figure it out.  Once I understand that, I think the rest of the math
should fall out nicely.

At any rate, it's hacky because gino van den bergen (note my ignorance on
how to properly capitalize his name) saves the relevant dot products for
points that are in the simplex.  I wonder if this level of optimization is
necessary given the cost of everything else that is typically contained in
a rigid body physics simulator, but I guess I'll see. :)

Will

----
Will Portnoy
http://www.cs.washington.edu/homes/will