Distance fields for rapid collision detection in physically based modeling

In this paper we address the problem of rapid distance computation between rigid objects and highly deformable objects, which is important in the context of physically based modeling of e.g hair or clothing. Our method is in particular useful when modeling deformable objects with particle systems - the most common approach to simulate such objects. We combine some already known techniques about distance fields into an algorithm for rapid collision detection. Only the rigid objects of an environment are represented by distance fields. In the context of proximity queries, which are essential for proper collision detection, this representation has two main advantages: First, any given boundary representation can be approximated quite easily, no high-degree polynomials or complicated approximation algorithms are needed. Second, the evaluation of distances and normals needed for collision response is extremely fast and independent of the complexity of the object. In the course of the paper we propose a simple, but fast algorithm for partial distance field computation. The sources are triangular meshes. Then, we present our approach for collision detection in detail. Examples from an interactive cloth animation system show the advantages of our approach in practice. We conclude that our method allows real-time animations of complex deformable objects in non-trivial environments on standard PC hardware.