Abstract. We design compact and responsive kinetic data structures for detecting collisions between n convex fat objects in 3-dimensional space that can have arbitrary sizes. Our main results are: (i) If the objects are 3-dimensional balls that roll on a plane, then we can detect collisions with a KDS of size O(n log n) that can handle events in O(log n) time. This structure processes O(n 2)events in the worst case, assuming that the objects follow constant-degree algebraic trajectories. (ii) If the objects are convex fat 3-dimensional objects of constant complexity that are free-flying in R 3, then we can detect collisions with aKDSofO(n log 6 n) size that can handle events in O(log 6 n)time. This structure processes O(n 2) events in the worst case, assuming that the objects follow constant-degree algebraic trajectories. If the objects have similar sizes then the size of the KDS becomes O(n) and events can be handled in O(1) time. 1

If two closed polygonal objects with outfacing normals intersect each other there exist one or more lines that intersect these objects at at least two consecutive front or back facing object points. In this work we present a method to efficiently detect these lines using depth-peeling and simple fragment operations. Of all polygons only those having an intersection with any of these lines are potentially colliding. Polygons not intersected by the same line do not intersect each other. We describe how to find all potentially colliding polygons and the potentially colliding pairs using a mipmap hierarchy that represents line bundles at ever increasing width. To download only potentially colliding polygons to the CPU for polygon-polygon intersection testing, we have developed a general method to convert a sparse texture into a packed texture of reduced size. Our method exploits the intrinsic strength of GPUs to scan convert large sets of polygons and to shade billions of fragments at interactive rates. It neither requires a bounding volume hierarchy nor a pre-processing stage, so it can efficiently deal with very large and deforming polygonal models. The particular design makes the method suitable for applications where geometry is modified or even created on the GPU.