You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I see the leaf node type is triangle for building the BVH model in the old FCL lib (Another one is point, and I've never used). The reason might be that you must take the leaf as a convex shape when you use the GJK algorithm, but if I use the convex decomposition algorithm like VHACD for a concave object, it will give me several convex objects. If I use convex as the leaf node for the BVH, I think the traversal process will be faster than the triangle case. Did this new lib support this case?
The text was updated successfully, but these errors were encountered:
That is a very good question. Currently, the leaves of BVH Model are triangle shapes and the bounding volumes can be of different types.
We plan, in the very near with @lmontaut, to support VHACD within HPP-FCL for easy integration, as suggested in #428.
To do so, we plan to add the notion of Compound Geometry, as a collection of several geometries, like the one obtained from the convex decomposition of a nonconvex geometry.
Hi @zfl202,
As explained by @jcarpent, this is a feature we really want to add to hpp-fcl. VHACD has even been rewritten to be faster and more robust, and is also a one-file header-only library now.
I intend to integrate that in hpp-fcl in the near future, if you want to help don't hesitate to let us know.
I see the leaf node type is triangle for building the BVH model in the old FCL lib (Another one is point, and I've never used). The reason might be that you must take the leaf as a convex shape when you use the GJK algorithm, but if I use the convex decomposition algorithm like VHACD for a concave object, it will give me several convex objects. If I use convex as the leaf node for the BVH, I think the traversal process will be faster than the triangle case. Did this new lib support this case?
The text was updated successfully, but these errors were encountered: