Reading Assignments
Interactive Collision Detection, by P. M. Hubbard, Proc. of IEEE Symp on Research Frontiers in Virtual Reality, 1993.
Evaluation of Collision Detection Methods for Virtual Reality Fly-Throughs, by Held, Klosowski and Mitchell, Proc. of Canadian Conf. on Computational Geometry 1995.
Efficient collision detection using bounding volume hierarchies of k-dops, by J. Klosowski, M. Held, J. S. B. Mitchell, H. Sowizral, and K. Zikan, IEEE Trans. on Visualization and Computer Graphics, 4(1):2137, 1998.
Collision Detection between Geometric Models: A Survey, by M. Lin and S. Gottschalk, Proc. of IMA Conference on Mathematics of Surfaces 1998.
UNC Chapel Hill
M. C. Lin
Reading Assignments
OBB-Tree: A Hierarchical Structure for Rapid Interference Detection, by S. Gottschalk, M. Lin and D. Manocha, Proc. of ACM Siggraph, 1996.
Rapid and Accurate Contact Determination between Spline Models using ShellTrees, by S. Krishnan, M. Gopi, M. Lin, D. Manocha and A. Pattekar, Proc. of Eurographics 1998.
Fast Proximity Queries with Swept Sphere Volumes, by Eric Larsen, Stefan Gottschalk, Ming C. Lin, Dinesh Manocha, Technical report TR99-018, UNC-CH, CS Dept, 1999. (Part of the paper in Proc. of IEEE ICRA2000)
UNC Chapel Hill
M. C. Lin
Methods for General Models
Decompose into convex pieces, and take minimum over all pairs of pieces:
Optimal (minimal) model decomposition is NP-hard. Approximation algorithms exist for closed solids,
but what about a list of triangles?
Collection of triangles/polygons:
n*m pairs of triangles brute force expensive
Hierarchical representations used to accelerate minimum finding
UNC Chapel Hill
M. C. Lin
Hierarchical Representations
Two Common Types:
Bounding volume hierarchies trees of spheres, ellipses, cubes, axis-aligned bounding boxes (AABBs), oriented bounding boxes (OBBs), K-dop, SSV, etc.
Spatial decomposition BSP, K-d trees, octrees, MSP tree, R- trees, grids/cells, space-time bounds, etc.
Do very well in rejection tests, when objects are far apart
Performance may slow down, when the two objects are in close proximity and can have multiple contacts
UNC Chapel Hill
M. C. Lin
BVH vs. Spatial Partitioning
BVH:
Object centric
Spatial redundancy
SP:
Space centric
Object redundancy
UNC Chapel Hill
M. C. Lin
BVH vs. Spatial Partitioning
BVH:
Object centric
Spatial redundancy
SP:
Space centric
Object redundancy
UNC Chapel Hill
M. C. Lin
BVH vs. Spatial Partitioning
BVH:
Object centric
Spatial redundancy
SP:
Space centric
Object redundancy
UNC Chapel Hill
M. C. Lin
BVH vs. Spatial Partitioning
BVH:
Object centric
Spatial redundancy
SP:
Space centric
Object redundancy
UNC Chapel Hill
M. C. Lin
Spatial Data Structures & Subdivision
Uniform Spatial Sub
Quadtree/Octree
kd-tree BSP-tree
Many others
(see the lecture notes)
UNC Chapel Hill
M. C. Lin
Uniform Spatial Subdivision
Decompose the objects (the entire simulated environment) into identical cells arranged in a fixed, regular grids (equal size boxes or voxels)
To represent an object, only need to decide which cells are occupied. To perform collision detection, check if any cell is occupied by two object
Storage: to represent an object at resolution of n voxels per dimension requires upto n3 cells
Accuracy: solids can only be approximated
UNC Chapel Hill
M. C. Lin
Octrees
Quadtree is derived by subdividing a 2D- plane in both dimensions to form quadrants
Octrees are a 3D-extension of quadtree
Use divide-and-conquer
Reduce storage requirements (in comparison to grids/voxels)
UNC Chapel Hill
M. C. Lin
Bounding Volume Hierarchies
Model Hierarchy:
each node has a simple volume that bounds a
set of triangles
children contain volumes that each bound a
different portion of the parents triangles
The leaves of the hierarchy usually contain
individual triangles
A binary bounding volume hierarchy:
UNC Chapel Hill
M. C. Lin
Type of Bounding Volumes
Spheres
Ellipsoids
Axis-Aligned Bounding Boxes (AABB)
Oriented Bounding Boxes (OBBs)
Convex Hulls
k-Discrete Orientation Polytopes (k-dop)
Spherical Shells
Swept-Sphere Volumes (SSVs)
Point Swetp Spheres (PSS)
Line Swept Spheres (LSS)
Rectangle Swept Spheres (RSS) Triangle Swept Spheres (TSS)
UNC Chapel Hill
M. C. Lin
BVH-Based Collision Detection
UNC Chapel Hill
M. C. Lin
Collision Detection using BVH
1.
2. 3. 4.
5. 6. 7. 8. 9.
Check for collision between two parent nodes (starting from the roots of two given trees)
If there is no interference between two parents, Then stop and report no collision
Else All children of one parent node are checked
against all children of the other node If there is a collision between the children
Then If at leave nodes
Then report collision
Else go to Step 4
Else stop and report no collision
UNC Chapel Hill
M. C. Lin
Evaluating Bounding Volume Hierarchies
Cost Function:
F = Nu x Cu + Nbv x Cbv + Np x Cp
F: total cost function for interference detection
Nu: no. of bounding volumes updated
Cu: cost of updating a bounding volume,
Nbv: no. of bounding volume pair overlap tests
Cbv: cost of overlap test between 2 bounding volumes Np: no. of primitive pairs tested for interference
Cp: cost of testing 2 primitives for interference
UNC Chapel Hill
M. C. Lin
Designing Bounding Volume Hierarchies
The choice governed by these constraints:
It should fit the original model as tightly as possible (to lower Nbv and Np)
Testing two such volumes for overlap should be as fast as possible (to lower Cbv)
It should require the BV updates as infrequently as possible (to lower Nu)
UNC Chapel Hill
M. C. Lin
Observations
Simple primitives (spheres, AABBs, etc.) do very well with respect to the second constraint. But they cannot fit some long skinny primitives tightly.
More complex primitives (minimal ellipsoids, OBBs, etc.) provide tight fits, but checking for overlap between them is relatively expensive.
Cost of BV updates needs to be considered. UNC Chapel Hill
M. C. Lin
Trade-off in Choosing BVs
Sphere AABB OBB 6-dop increasing complexity & tightness of fit
decreasing cost of (overlap tests + BV update)
Convex Hull
UNC Chapel Hill
M. C. Lin
Building Hierarchies
Choices of Bounding Volumes cost function & constraints
Top-Down vs. Bottum-up speed vs. fitting
Depth vs. breadth branching factors
Splitting factors where & how
UNC Chapel Hill
M. C. Lin
Sphere-Trees
A sphere-tree is a hierarchy of sets of spheres, used to approximate an object
Advantages:
Simplicity in checking overlaps between two
bounding spheres
Invariant to rotations and can apply the same transformation to the centers, if objects are rigid
Shortcomings:
Not always the best approximation (esp bad for
long, skinny objects)
Lack of good methods on building sphere-trees
UNC Chapel Hill
M. C. Lin
Methods for Building Sphere-Trees
Tile the triangles and build the tree bottom-up
Covering each vertex with a sphere and group them together
Start with an octree and tweak
Compute the medial axis and use it as a
skeleton for multi-res sphere-covering
Others
UNC Chapel Hill
M. C. Lin
k-DOPs
k-dop: k-discrete orientation polytope a convex polytope whose facets are determined by half- spaces whose outward normals come from a small fixed set of k orientations
For example:
In 2D, an 8-dop is determined by the orientation at +/-
{45,90,135,180} degrees
In 3D, an AABB is a 6-dop with orientation vectors determined by the +/-coordinate axes.
UNC Chapel Hill
M. C. Lin
Choices of k-dops in 3D
6-dop: defined by coordinate axes
14-dop: defined by the vectors (1,0,0), (0,1,0),
(0,0,1), (1,1,1), (1,-1,1), (1,1,-1) and (1,-1,-1)
18-dop: defined by the vectors (1,0,0), (0,1,0), (0,0,1), (1,1,0), (1,0,1), (0,1,1), (1,-1,0), (1,0,-1) and (0,1,-1)
26-dop: defined by the vectors (1,0,0), (0,1,0), (0,0,1), (1,1,1), (1,-1,1), (1,1,-1), (1,-1,-1), (1,1,0), (1,0,1), (0,1,1), (1,-1,0), (1,0,-1) and (0,1,-1)
UNC Chapel Hill
M. C. Lin
Building Trees of k-dops
The major issue is updating the k-dops:
Use Hill Climbing (as proposed in I-Collide) to update the min/max along each k/2 directions by comparing with the neighboring vertices
But, the object may not be convex Use the approximation (convex hull vs. another k-dop)
UNC Chapel Hill
M. C. Lin
Building an OBBTree
Recursive top-down construction: partition and refit
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
Given some polygons, consider their vertices
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
and an arbitrary line
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
Project onto the line
Consider variance of distribution on the line
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
Different line, different variance
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
Maximum Variance
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
Minimal Variance
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
Given by eigenvectors of covariance matrix of coordinates
of original points
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
Choose bounding box oriented this way
UNC Chapel Hill
M. C. Lin
Building an OBB Tree: Fitting
Covariance matrix of
point coordinates describes statistical spread of cloud.
OBB is aligned with directions of greatest and least spread
(which are guaranteed to be orthogonal).
UNC Chapel Hill
M. C. Lin
Fitting OBBs
Let the vertices of the ith triangle be the points ai, bi, and ci, then the mean and covariance matrix C can be expressed in vector notation as:
where n is the number of triangles, and
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
Good Box
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
Add points: worse Box
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
More points: terrible box
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
Compute with extremal points only
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
Even distribution: good box
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
Uneven distribution: bad box
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
Fix: Compute facets of convex hull
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
Better: Integrate over facets
UNC Chapel Hill
M. C. Lin
Building an OBB Tree
and sample them uniformly
UNC Chapel Hill
M. C. Lin
Building an OBB Tree: Summary
OBB Fitting algorithm:
covariance-based
use of convex hull
not foiled by extreme distributions
O(n log n) fitting time for single BV O(n log2 n) fitting time for entire tree
UNC Chapel Hill
M. C. Lin
Tree Traversal
Disjoint bounding volumes: No possible collision
UNC Chapel Hill
M. C. Lin
Tree Traversal
Overlapping bounding volumes:
split one box into children
test children against other box
UNC Chapel Hill
M. C. Lin
Tree Traversal
UNC Chapel Hill
M. C. Lin
Tree Traversal
Hierarchy of tests
UNC Chapel Hill
M. C. Lin
Separating Axis Theorem
L is a separating axis for OBBs A & B, since A & B become disjoint intervals under projection onto L
UNC Chapel Hill
M. C. Lin
Separating Axis Theorem
Two polytopes A and B are disjoint iff there exists a separating axis which is:
perpendicular to a face from either or
perpedicular to an edge from each
UNC Chapel Hill
M. C. Lin
Implications of Theorem
Given two generic polytopes, each with E edges and F faces, number of candidate axes to test is:
2F + E2
OBBs have only E = 3 distinct edge directions, and only F = 3 distinct face normals. OBBs need at most 15 axis tests.
Because edge directions and normals each form orthogonal frames, the axis tests are rather simple.
UNC Chapel Hill
M. C. Lin
OBB Overlap Test: An Axis Test
L
s ha
hb
L is a separating axis iff:
s >h+h ab
UNC Chapel Hill
M. C. Lin
OBB Overlap Test: Axis Test Details
Box centers project to interval midpoints, so midpoint separation is length of vector Ts image.
B A TB
T
s
s = (TA TB )n
TA
n
UNC Chapel Hill
M. C. Lin
OBB Overlap Test: Axis Test Details
Half-length of interval is sum of box axis images. B
rB
n
UNC Chapel Hill
M. C. Lin
r =bRBn+bRBn+bRBn B11 22 33
OBB Overlap Test
Typical axis test for 3-space.
s=fabs(T2*R11 T1*R21); ha=a1*Rf21 + a2*Rf11; hb=b0*Rf02 + b2*Rf00;
if (s > (ha + hb)) return 0;
Up to 15 tests required. UNC Chapel Hill
M. C. Lin
OBB Overlap Test
Strengths of this overlap test:
89 to 252 arithmetic operations per box overlap
test
Simple guard against arithmetic error
No special cases for parallel/coincident faces, edges, or vertices
No special cases for degenerate boxes
No conditioning problems
Good candidate for micro-coding
UNC Chapel Hill
M. C. Lin
OBB Overlap Tests: Comparison
Test Method
Speed(us)
Separating Axis
GJK LP
6.26
66.30 217.00
Benchmarks performed on SGI Max Impact, 250 MHz MIPS R4400 CPU, MIPS R4000 FPU
UNC Chapel Hill
M. C. Lin
Parallel Close Proximity
1 1
Two models are in parallel close proximity when every point on each model is a given fixed distance () from the other model.
Q: How does the number of BV tests increase
as the gap size decreases?
UNC Chapel Hill
M. C. Lin
Parallel Close Proximity: Convergence
1
UNC Chapel Hill
M. C. Lin
Parallel Close Proximity: Convergence
1 /2
1 /4
UNC Chapel Hill
M. C. Lin
Parallel Close Proximity: Convergence
1
UNC Chapel Hill
M. C. Lin
Parallel Close Proximity: Convergence
1 1/
/2 4
UNC Chapel Hill
M. C. Lin
Parallel Close Proximity: Convergence
1
UNC Chapel Hill
M. C. Lin
Parallel Close Proximity: Convergence
1 /4
1 /
16
UNC Chapel Hill
M. C. Lin
Parallel Close Proximity: Convergence
1 /4
1 1/
/4 4
UNC Chapel Hill
M. C. Lin
Performance: Overlap Tests
k
O(n)
OBBs
2k
O(n2)
UNC Chapel Hill
M. C. Lin
Spheres & AABBs
Parallel Close Proximity: Experiment
3 2
106 5 3
2
105 5 3
2
104 6
43 2
103 5 3
2
102 6
43 2
Log-log plot
101
10-4 2 3 456710-3 2 3 456710-2 2 3 456710-1 2 3 4567100 2 3 4567101
Gap Size ()
OBBs asymptotically outperform AABBs and spheres
UNC Chapel Hill
M. C. Lin
Number of BV tests
Example: AABBs vs. OBBs
UNC Chapel Hill
M. C. Lin
Approximation of a Torus
Implementation: RAPID
Available at: http://www.cs.unc.edu/ ~geom/OBB
Part of V-COLLIDE: http://www.cs.unc.edu/ ~geom/V_COLLIDE
Thousands of users have ftped the code Used for virtual prototyping, dynamic
simulation, robotics & computer animation
UNC Chapel Hill
M. C. Lin
Hybrid Hierarchy of Swept Sphere Volumes
UNC Chapel Hill
M. C. Lin
PSS LSS RSS
[LGLM99]
Swept Sphere Volumes (S-topes)
UNC Chapel Hill
M. C. Lin
PSS LSS RSS
SSV Fitting
Use OBBs code based upon Principle Component Analysis
For PSS, use the largest dimension as the radius
For LSS, use the two largest dimensions as the length and radius
For RSS, use all three dimensions UNC Chapel Hill
M. C. Lin
Overlap Test
One routine that can perform overlap tests between all possible combination of CORE primitives of SSV(s).
The routine is a specialized test based on Voronoi regions and OBB overlap test.
It is faster than GJK.
UNC Chapel Hill
M. C. Lin
Hybrid BVHs Based on SSVs
Use a simpler BV when it prunes search equally well benefit from lower cost of BV overlap tests
Overlap test (based on Lin-Canny & OBB overlap test) between all pairs of BVs in a BV family is unified
Complications
deciding which BV to use either dynamically or
statically
UNC Chapel Hill
M. C. Lin
PQP: Implementation
Library written in C++
Good for any proximity query
5-20x speed-up in distance computation over prior methods
Available at http://www.cs.unc.edu/ ~geom/SSV/
UNC Chapel Hill
M. C. Lin
Reviews
There are no reviews yet.