[ODE] Re: ODE

Russ Smith russ at q12.org
Mon Jan 7 23:45:01 2002


Tomas Edsö <tomas.edso@cpen.com> wrote:
> 
> Hi again.
> 
> Got the new version of ODE and it is working nicely! Have been
> planning to do some contributions of my own and I would like some
> advice.
> 
> I have noticed that boxes that get stacked on top of each other start
> to "boil" or shake in an stange non-physical way. If this is caused
> by the box-box collision handling, then this may disappear in the
> next version, What do you think? Is it worth looking into or will the
> job be done soon anyway?

this problem is caused by dCollideBB() not fully handling the case
where two boxes are aligned more face-face than they are edge-egde
or edge-face. only one contact point is generated when up to three
may be needed. this is definitely worth fixing, and i have no
immediate plans to do it. help would be welcome.

> If so I would instead like to work on the implementation of
> cloth-simulation.

sure - but bare in mind that cloth simulation is not so easy to
fit within ODE's API (it that's what you had in mind). it might be
useful to create a separate API and simulation engine for cloth, then
have a method to integrate this with ODE rigid body simulations
(special constraints perhaps?).

> I was thinking to try to use the method described by Baraff/Witkin in
> their paper "Large Steps in Cloth Simulation". Is this algorithm one
> that you would recommend or should I perhaps try another one?

i have not looked at this field recently, though i'm aware there are
a lot of different techniques.

> Also, I have noticed that there is no spring-and-damper joint. Will
> this be implemented in later versions?

i have no plans. you're welcome to have a go.

> Best regards
> 
> /Tomas Edsö
> Lund, Sweden

russ.

--
Russell Smith
http://www.q12.org