[ODE] does ODE need convex polyhedra collision?

[Sergio Valverde]

>> I think (only guessing now - I haven't even started implementinganything)

>> the final contact(s) can be found from the final simplex pair in GJK
>>case. As

While it is true that the solution must be found from the simplex pair
(or closests features) returned by the algorithm, this is not a simple task!
Would be great if some people on the list could point us to relevant
references/source code providing _detailed_ explanations about this.

In general, it is interesting to investigate these closest features
algorithms but ALL these algorithms in their original form are "not"
suitable to compute the full list of penetrating contacts. 

I wonder if the Gino's book contains some information on the subject.

Sergi Valverde

-----Original Message-----
From: nearaz [mailto:nearaz at interamotion.com] 
Sent: lunes, 13 de octubre de 2003 20:36
To: ode at q12.org
Subject: RE: [ODE] does ODE need convex polyhedra collision?


> In general, every contact consists of SEVERAL contact points. 
> The algorithms you commented they only return ONE contact point, 
> which is not enough.

I think (only guessing now - I haven't even started implementing anything) 
the final contact(s) can be found from the final simplex pair in GJK case.
As 
I understood, GJK in it's "plain form" doesn't even return closest points, 
only separation distance.

> Do you have more information on how to compute the entire set
> Of contact points from the information returned by these algorithms
> (v-clip, solid, gjk) ? 

None yet :(

BTW, Solid seems to use enhanced GJK (in one way). I've also stumbled upon
"A 
new algorithm for computing minimum distance" by Sundaraj, d'Aulignac and 
Mazer - it seems to be a blend of Lin-Canny and GJK.

Anyway, I think I'll just try to implement "something" and see if any
results 
are possible. I have to implement something for my university course! :)


Aras Pranckevicius aka NeARAZ
http://www.gim.ktu.lt/nesnausk/nearaz/


_______________________________________________
ODE mailing list
ODE at q12.org
http://q12.org/mailman/listinfo/ode