warp.bvh_query_capsule#

warp.bvh_query_capsule(
id: uint64,
start: vec3f,
dir: vec3f,
radius: float32,
root: int32,
) → _BvhQueryCapsule#

  • Kernel

Construct a conservative capsule sweep query against a BVH.

Iterates over every BVH item whose stored bounding box overlaps the swept capsule. Each node’s bounds are inflated by radius before the ray-slab test (an axis-aligned box inflation, not a true sphere cap), so the query never misses a primitive within radius of the segment but may return extra candidates near box corners.

To sweep a closed capsule from p0 to p1, pass dir = p1 - p0 (unnormalized) and max_dist = 1.0 in bvh_query_next(); contact at both endpoints is included. A zero-length segment (p0 == p1) is not supported — use bvh_query_sphere() instead. A negative radius is clamped to zero. Advance results with bvh_query_next().

Parameters:

  • id – The BVH identifier

  • start – The segment start point (p0), in BVH space

  • dir – The segment direction (p1 - p0), in BVH space

  • radius – The capsule radius; negative values are clamped to zero

  • root – The node to begin the query from, or -1 (default) for the BVH’s global root

Returns:

A warp.BvhQuery. It is opaque; pass it to bvh_query_next().

Example

@wp.kernel
def capsule_sweep(bvh_id: wp.uint64, p0: wp.vec3, p1: wp.vec3, radius: float,
                   count: wp.array[wp.int32]):
    query = wp.bvh_query_capsule(bvh_id, p0, p1 - p0, radius)
    item = int(0)
    while wp.bvh_query_next(query, item, 1.0):
        wp.atomic_add(count, 0, 1)

lowers = wp.array([[0.75, -1, -1]], dtype=wp.vec3)
uppers = wp.array([[2.0,   1,  1]], dtype=wp.vec3)
bvh = wp.Bvh(lowers=lowers, uppers=uppers)
count = wp.zeros(1, dtype=wp.int32)
wp.launch(capsule_sweep, dim=1,
          inputs=[bvh.id, wp.vec3(0.0, 0.0, 0.0), wp.vec3(0.5, 0.0, 0.0), 0.3, count])
print(count.numpy()[0])

1