[ODE] bbtree parent boxes constraints based on ode joints

[Piotr Obrzut]

Hi,

W Twoim liœcie datowanym 24 listopada 2004 (12:15:52) mo¿na przeczytaæ:

PO> Hi,

PO>    i'm making mine own col. det. system which is variation on BBtree
PO>    algorhitm system for mine hierarchical model which consist of some
PO>    parent-children related boxes (this model is based in fact on cal3d
PO>    model, this boxed model is build in such a way which allow me to know after
PO>    collision which mesch, and which bone was affected by this
PO>    collision).
   
PO>    It is based on simple idea: when it check collision between
PO>    2 object (this object may have some hierarchy) when the collision occure
PO>    it will check collision between children of collided objects
PO>    (of course when this object have some descendants). Collision
PO>    checking should continue recursive while both objects will have
PO>    some children. Therefore every parent box should "contain" all
PO>    descendants boxes in every situation - position of box depends indirectly
PO>    on rigidbody position (it change cal3d bone position and this
PO>    position defines box position), so every parent box should be so
PO>    big that it will contain all child boxes in every posible relative
PO>    pos to the parent.

PO>    The question is: How can i make parent box as small as possible
PO>    when it should contain child box in all its posbile positions?
PO>    (positions of both boxes depends on 2 rigid bodies pos. constrainted
PO>    by known joint)



More precisily mine problem is situated in joints errors,
so the regenerate question is:

It is possible to get from joint the max. relative position of bodies
attached to this joint when I have to consider every possible
relative position of mentioned bodies (including those position which
come off joints inaccuracy)?
If yes then how can I estimate the max error situation?

btw. i have some problem with mine model, when 2 bodies are conected with ball
joint (its anchor pos is set to 0) they shouldn't change their relative translation length, but
mine does so - i set the global ERP to 0.8. I know that many people
was writing about joints problem but i haven't notice any suitable.
I could send a picture of that if it isn't trivial problem.


-- 
greetings,
 Piotr Obrzut