Interactive Continuous Collision Detection for Polygon Soups

27
Interactive Continuous Collision Detection for Polygon Soups Xin Huang 11/20/2007

description

Interactive Continuous Collision Detection for Polygon Soups. Xin Huang 11/20/2007. Introduction. Discrete CD Miss collision between sampled time instances Continuous Collision Detection (CCD). Introduction. Continuous Collision Detection (CCD) Consider continuous motion - PowerPoint PPT Presentation

Transcript of Interactive Continuous Collision Detection for Polygon Soups

Page 1: Interactive Continuous Collision Detection for Polygon Soups

Interactive Continuous Collision Detection for Polygon Soups

Xin Huang

11/20/2007

Page 2: Interactive Continuous Collision Detection for Polygon Soups

Introduction

• Discrete CD– Miss collision between sampled time

instances

• Continuous Collision Detection (CCD)

Page 3: Interactive Continuous Collision Detection for Polygon Soups

Introduction

• Continuous Collision Detection (CCD)– Consider continuous motion– Report first time of contact (TOC)

• Well applied in– Cloth simulation– Rigid body Dynamics– Local planning

Page 4: Interactive Continuous Collision Detection for Polygon Soups

Motivation

• CCD in Local Planning– CCD is used to check collision along the interpolating

trajectory between two free nearby samples – Perform the CCD query for samples near contact

space is vital in local planning

Page 5: Interactive Continuous Collision Detection for Polygon Soups

Related work (1)

• Schwarzer [2002]: Exact collision checkingUse adaptive bisection approachCan be used for polygon soupsCalculate un-directional motion bound Work not well when separation distance is very small

Page 6: Interactive Continuous Collision Detection for Polygon Soups

Related work (2)

• [Redon 2002]Use interval arithmetic to compute the motion boundContinuous OBB testWork not well when the motion has large rotationDoes not perform adaptive bisection, suffer from a large

number of bisection steps when objects are far apart

Page 7: Interactive Continuous Collision Detection for Polygon Soups

Related work (3)

• [Zhang 06]: Extend conservative advancementCompute continuous motion by linear interpolatingPerform hierarchy advancement based on convex hull

treeBenefit from motion coherence supported by swift++

As reported, perform more efficiently than [Redon 2002]

Page 8: Interactive Continuous Collision Detection for Polygon Soups

Challenge

• Exact CCD– Small separation distance

• [Redon 2002]– Large Rotation

• [Zhang 06]– Polygon soups– High number of convex pieces

Page 9: Interactive Continuous Collision Detection for Polygon Soups

Disassembly Plan (D-Plan)

• Efficient collision free path computation

• CAD/CAM models– non-manifold or polygon soups with no

connectivity or topology information– have tight spaces and multiple narrow

passages in configuration spaces

Page 10: Interactive Continuous Collision Detection for Polygon Soups

Some challenging benchmarks

200K+ polygons

Page 11: Interactive Continuous Collision Detection for Polygon Soups

12,236 vertices11,569 trianglesObstacle

Robot 16,781 vertices15,197 triangles

•Application in Assembly Maintainability: extract the part,

there are many non-manifold parts in the robot

Page 12: Interactive Continuous Collision Detection for Polygon Soups

Goal

• Perform fast continuous collision detection for polygon soups in local planning for part disassembly

Page 13: Interactive Continuous Collision Detection for Polygon Soups

Main Approaches

• Perform conservative advancement (CA) for polygon soups

• Directional motion bound computation

• Hierarchical CA for BVH

• Explore motion coherence to accelerate

Page 14: Interactive Continuous Collision Detection for Polygon Soups

Conservative Advancement

• CA for Convex Polytopes– computes an upper bound of TOC by

repeatedly advancing A by dt toward B while avoiding collision

– When close enough, if TOC<1, then report collision; else report collision free

Page 15: Interactive Continuous Collision Detection for Polygon Soups

Conservative Advancement

• CA for non-convex objects– Convex decomposition– Assume the closed mesh

• CA for polygon soups– Using SSV based on PQP– Compute the motion bound for SSV– Compute the motion bound for triangle– Can process polygon soups

Page 16: Interactive Continuous Collision Detection for Polygon Soups

Directional motion bound computation

• Calculate linear motion interpolation– : U = V + W*R

• Project the motion along the direction of the closest distance d

• Compute the directional motion for BV and Triangle in Leaf BV

d

U

Page 17: Interactive Continuous Collision Detection for Polygon Soups

Hierarchical CA for BVH

• Bound volume traversal trees– Given two BVHs (HA, HB), starting from the root

nodes, recursively perform pairwise TOC computation– If TOC(na, nb) < TOCcurrent, the recursive traversal

continues; otherwise it stops

Page 18: Interactive Continuous Collision Detection for Polygon Soups

Explore motion coherence

• Temporal coherence in contact space sampling and constraint motion

• Motion coherence during each step of CA

• Compute the initial TOC according to the closest features stored in last iteration

Page 19: Interactive Continuous Collision Detection for Polygon Soups

Explore motion coherence• Further benefit motion coherence by exploring

local tree containing the closest features• The initial small TOC will help culling many BV

pair Tests and Triangle pair test

Page 20: Interactive Continuous Collision Detection for Polygon Soups

Algorithm

1. Build the Bound Volume traversal tree for object A and B2. Compute the initial TOC using local tree motion coherence3. Traverse the BVH tree to compute TOC4. If the current node in traversal tree is leaf node, compute

the distance and directional motion bound between the two triangles to calculate TOC

5. If the current node is not leaf node, go to step 3 to traverse its child node if the TOC of the two BV is less than current minimal TOC

6 .Advance the object A by TOC until TOC > 1 or distance between the two objects is less than threshold

7. Return TOC

Page 21: Interactive Continuous Collision Detection for Polygon Soups

Demo

Page 22: Interactive Continuous Collision Detection for Polygon Soups

Experiments analysis

• The alpha model (1K triangles)

Page 23: Interactive Continuous Collision Detection for Polygon Soups

Experiments analysis

• Test motion coherence

No motion coherence Enable motion coherence

Page 24: Interactive Continuous Collision Detection for Polygon Soups

Conclusion

• Implement the CCD for polygon soups using conservative advancementPerform the hierarchy conservative advancement for

polygon soups based on PQPCalculate the directional motion bound for SSV using

linear interpolatingExplore motion coherence

Page 25: Interactive Continuous Collision Detection for Polygon Soups

Future work

• Compute the motion bound for constrained motion

• Perform continuous collision detection for sliding motion

Page 26: Interactive Continuous Collision Detection for Polygon Soups

References

• Xinyu Zhang, Minkyoung Lee, Young J. Kim: Interactive continuous collision detection for non-convex polyhedra. The Visual Computer 22(9-11): 749-760 (2006)

• Gottschalk, S., Lin, M., Manocha, D.: OBB-Tree: A hierarchical structure for rapid interference detection. In: H. Rushmeier (ed.) SIGGRAPH 96 Conference Proceedings, Annual Conference Series, pp. 171–180. ACM SIGGRAPH, Addison Wesley (1996).

• Kim, B., Rossignac, J.: Collision prediction for polyhedra under screw motions. In: ACM Conference on Solid Modeling and Applications (2003)

• Larsen, E., Gottschalk, S., Lin, M., Manocha, D.: Fast proximity queries with swept sphere volumes. Tech. Rep. TR99-018, Department of Computer Science, University of North Carolina (1999)

• Schwarzer, F., Saha, M., Latombe, J.C.: Exact collision checking of robot paths. In: Workshop on Algorithmic Foundations of Robotics (WAFR) (2002)

Page 27: Interactive Continuous Collision Detection for Polygon Soups

References

• Lin, M., Manocha, D.: Collision and proximity queries. In: Handbook of Discrete and Computational Geometry (2003)

• Mirtich, B.: Timewarp rigid body simulation. SIGGRAPH 00 Conference Proceedings pp. 193–200 (2000)

• Mirtich, B.V.: Impulse-based dynamic simulation of rigid body systems. Ph.D. thesis, University of California, Berkeley (1996)

• Redon, S., Kheddar, A., Coquillart, S.: Fast continuous collision detection between rigid bodies. Proc. of Eurographics (Computer Graphics Forum) (2002)

• Xinyu Zhang, Stephane Redon, Minkyoung Lee, Young J. Kim: Continuous collision detection for articulated models using Taylor models and temporal culling. ACM Trans. Graph. 26(3): 15 (2007)