[Algorithms] Re: SSV and other bounding volumes

[Brian M. <bma...@ra...>]

    I know that there has been a lot of discussion about the merits of OBB's
and whether or not the code for them works but has anyone tried out a wider
range or bounding volumes or combinations?
    For my target platform memory is a scarce resource so OBB's are not so
appealing. I know there are a lot of ways to optimize memory usage for OBB's
but the paper tests I've run suggest that it's still going to need too much
memory.
    Again due to memory issues against polygon performance a lot of geometry
is going to be created on the fly from base representations so I'm trying to
come up with a solution that allows the collision structure to be built on
demand as well. I've been thinking of storing the decisions required to
build a good tree rather than the tree itself - the theory being that
deciding splitting points etc is more time consuming than just classifying
triangles.
    There are 2 papers that seem to suggest approaches:

    'Fast Proximity Queries with Swept Sphere Volumes'. and
    'Efficient Collision Detection of Complex Deformable Models using AABB
Trees'

   Anyone any thoughts? The traditional academic approach has both more
memory available and less triangles to test... so I'd welcome any insights
so hopefully I won't have to implement a whole load of options and time them

-Brian.