Collision detection for deformable objects Soft-body dynamics

realistic interaction of simulated soft objects environment may important obtaining visually realistic results. cloth self-intersection important in applications acceptably realistic simulated garments. challenging achieve @ interactive frame rates, particularly in case of detecting , resolving self collisions , mutual collisions between 2 or more deformable objects.

collision detection may discrete/a posteriori (meaning objects advanced in time through pre-determined interval, , penetrations detected , resolved), or continuous/a priori (objects advanced until collision occurs, , collision handled before proceeding). former easier implement , faster, leads failure detect collisions (or detection of spurious collisions) if objects move fast enough. real-time systems have use discrete collision detection, other ad hoc ways avoid failing detect collisions.

detection of collisions between cloth , environmental objects defined inside straightforward since system can detect unambiguously whether cloth mesh vertices , faces intersecting body , resolve them accordingly. if defined inside not exist (e.g. in case of collision mesh not form closed boundary), inside may constructed via extrusion. mutual- or self-collisions of soft bodies defined tetrahedra straightforward, since reduces detection of collisions between solid tetrahedra.

however, detection of collisions between 2 polygonal cloths (or collision of cloth itself) via discrete collision detection more difficult, since there no unambiguous way locally detect after timestep whether cloth node has penetrated on wrong side or not. solutions involve either using history of cloth motion determine if intersection event has occurred, or doing global analysis of cloth state detect , resolve self-intersections. pixar has presented method uses global topological analysis of mesh intersections in configuration space detect , resolve self-interpenetration of cloth. currently, computationally expensive real-time cloth systems.

to collision detection efficiently, primitives not colliding must identified possible , discarded consideration avoid wasting time. this, form of spatial subdivision scheme essential, avoid brute force test of



o
[

n

2


]


{\displaystyle o[n^{2}]}

primitive collisions. approaches used include:

bounding volume hierarchies (aabb trees, obb trees, sphere trees)
grids, either uniform (using hashing memory efficiency) or hierarchical (e.g. octree, kd-tree)
coherence-exploiting schemes, such sweep , prune insertion sort, or tree-tree collisions front tracking.
hybrid methods involving combination of various of these schemes, e.g. coarse aabb tree plus sweep-and-prune coherence between colliding leaves.




^ baraff, witkin & kass (2003). untangling cloth (pdf). 
^ van den bergen (1998). efficient collision detection of complex deformable models using aabb trees (pdf). 
^ teschner, heidelberger, müller, pomeranets & gross (2003). optimized spatial hashing collision detection of deformable objects (pdf). cs1 maint: multiple names: authors list (link)