[ODE] Re: Trimesh-Trimesh updates ?

Marco Grubert mgrubert at conitec.net
Mon Jan 24 13:29:36 MST 2005


> I am considering to build a pure tri-mesh based
> dynamic environment -- for example, bowling pins
> bumping into each other, etc.  Is the current ODE stable
> enough to handle the trimesh-trimesh collisions?

Short answer: no.
One of the shortcomings is that the algorithm expects penetrations to be 
resolved within 1 timestep. If A penetrates B by 0.5 units and gets pushed 
outward only by 0.2 units then it will still be 0.3 units deep inside B. The 
algorithm then compares total movement to depth, notices that A moved 0.2 
units, yet is indicated as 0.3 units inside B: 0.2 < 0.3 thus this contact 
is discarded as an anomaly with the result that the objects will sink into 
each other.
The trimesh-trimesh collider has a number of heuristics which are good in 
dealing with impacts but break down for resting contacts. In my tests 
(box-stack made out of trimeshes) objects explode, get stuck or sink into 
each other once they come to a rest.

If you don't need a general purpose solution and know details about your 
world before-hand (masses, dimensions of geoms, maximum velocity) you might 
be able to tweak the collider to deal with those.

- Marco Grubert 



More information about the ODE mailing list