🌿 CuriousWinds

Bounding Volume Hierarchies

Physical mechanisms, mathematical boundary conditions, and architectural trade-offs in Bounding Volume Hierarchies.

Executive Summary & Key Takeaways
Foundational Synthesis for Practitioners & Research Systems

Bounding volume hierarchies (BVHs) optimize collision detection by balancing tight geometric fits with fast intersection tests, unlike AABBs, which waste traversal cycles on inflated voids, tighter enclosures demand costly floating-point checks that cripple real-time performance.

01 Void Volume Tradeoff

Tighter BVHs reduce false positives but require 15x more intersection checks than AABBs.

02 OBB Dynamic Cost

Oriented bounding boxes eliminate voids via SAT but cost 10, 20x more CPU than AABB overlap checks.

03 AABB Rotation Penalty

AABBs inflate void space under rotation, forcing deeper traversal and O(n²) overhead in dynamic scenes.

04 Test-Cost Threshold

Use AABBs for static scenes; switch to OBBs/K-DOPs when rigid-body rotation exceeds ±45° from axes.

Bounding volume hierarchies rely on geometric enclosing hulls to represent complex simulation meshes with simplified bounding proxies. The primary engineering decision when constructing spatial acceleration structures is selecting the appropriate enclosing geometry. A bounding volume that fits the underlying geometry tightly will drastically reduce false-positive intersection candidates, but calculating whether two tight volumes intersect requires significantly more floating-point operations. In contrast, simpler bounding proxies execute intersection tests in a handful of CPU clock cycles, but enclose substantial empty space that increases traversal depth. In this monograph, we evaluate bounding volume hierarchies across Axis-Aligned Bounding Boxes (AABB), Oriented Bounding Boxes (OBB), and Discrete Oriented Polytopes (K-DOP), analyzing their computational trade-offs in dynamic game physics.

01 1 · COMPUTATIONAL GEOMETRY · SPATIAL PRUNING

The Volume Tightness vs. Test Cost Continuum

The efficiency of any hierarchical acceleration tree is determined by two competing metrics: the cost of testing two bounding volumes for intersection (Ttest) and the probability that the volumes overlap (Poverlap). For any bounding hull enclosing a 3D geometry G:

Equation 2 (Enclosing Hull Void Volume Metric):
VVoid = Volume(Hull) – Volume(Geometry)
Minimizing void volume reduces false-positive intersection dispatches to expensive narrowphase solvers.

Spheres and Axis-Aligned Bounding Boxes (AABBs) sit at the extreme of minimal test cost. Testing two AABBs requires only six coordinate interval comparisons: if the intervals along the X, Y, or Z axes do not overlap, the boxes cannot intersect. However, when an elongated object (such as an aircraft or sword) rotates at an acute angle to the world axes, an AABB encloses massive amounts of empty void space, degrading the culling effectiveness of bounding volume hierarchies.

Bounding Type Overlap Test Ops Fitting Tightness Rigid Rotation Cost
Bounding Sphere 1 distance check Poor (High void space) Zero (Invariant)
AABB (Axis-Aligned) 6 comparisons Moderate Low (Recompute min/max)
OBB (Oriented Box) 15 separating axes High (Rotates with body) Zero (Transform orientation)
14-DOP (Discrete Polytopes) 14 interval checks Very High (Beveled corners) Moderate (Vector projections)
02 2 · RIGID BODY DYNAMICS · ORIENTATION OPTIMIZATION

The Mechanics of Oriented Bounding Boxes (OBB)

Oriented Bounding Boxes (OBBs) solve the orientation penalty of AABBs by rotating alongside the enclosed rigid body. Because an OBB maintains an internal orthonormal coordinate frame (three basis vectors and half-extents), it fits arbitrary geometries tightly, eliminating void corners. Testing two OBBs for collision requires the Separating Axis Theorem (SAT), evaluating up to fifteen potential separating axes (three face normals of Box A, three face normals of Box B, and nine cross products of their edges).

While an OBB-OBB overlap test costs approximately ten to twenty times more clock cycles than an AABB-AABB test, the massive reduction in false-positive pair dispatches often produces a net performance win in complex bounding volume hierarchies handling dense, interpenetrating meshes.

03 3 · CONVEX APPROXIMATION · DISCRETE POLYTOPAL REPRESENTATION

Discrete Oriented Polytopes (K-DOP)

Discrete Oriented Polytopes (K-DOPs) represent an elegant middle ground between AABBs and convex hulls. A K-DOP is formed by taking K fixed plane normals in world space (such as 6, 14, 18, or 26 directional vectors) and storing the minimum and maximum scalar projections of the geometry along each fixed normal axis.

Because the directional normals are fixed across the entire world, testing two K-DOPs requires only K interval comparisons without any complex matrix inversions. 14-DOPs and 18-DOPs effectively bevel the sharp corners of bounding boxes, yielding superior bounding tightness while maintaining deterministic SIMD vectorization speeds across high-frequency physics loops.

04 4 · GAME ENGINE ARCHITECTURE · HYBRID PHASE ALLOCATION

Hybrid Hierarchies in AAA Game Production

Industry-leading physics engines avoid dogmatic reliance on a single bounding format. Instead, modern bounding volume hierarchies utilize hybrid tree topologies: top-level broadphase trees employ lightweight AABBs for rapid spatial pruning across broad scene sectors, while lower-level sub-trees attached to articulated character rigs utilize OBBs or swept sphere capsules for tight local containment.

This multi-tiered architectural approach delivers optimal performance across both macro spatial distribution and micro contact resolution, proving that bounding volume selection is a foundational systems engineering decision in interactive virtual worlds.

CuriousWinds Knowledge Graph Invariant

Continue Curriculum: Spatial Partitioning BVH Collision: 5 Empirical Rules for Games

This inquiry is directly integrated into the parent monograph hub. Next in the curriculum is Part 3: Surface Area Heuristic: Algorithmic Partitioning for Optimal Tree Traversal, examining cost functions, geometric probability, & optimal split plane selection.

Curated Inquiry Series • 10 Investigations

Spatial Partitioning BVH Collision: 5 Empirical Rules for Games

An empirical exploration through the foundational dynamics, systemic trade-offs, and practical architecture of spatial partitioning bvh collision: 5 empirical rules for games.

Primary Scholarly Provenance & Empirical Attribution
Card, S. K., Moran, T. P., & Newell, A. (1983). The Psychology of Human-Computer Interaction. Lawrence Erlbaum Associates; ACM SIGCHI Model of Human Processor & Saccadic Latency Foundations.
DOI: 10.1145/358896.358899