New geometric algorithms and data structures for collision detection of dynamically deforming objects

摘要

Any virtual environment that supports interactions between virtual objects and/or a user and objects, needs a collision detection system to handle all interactions in a physically correct or plausible way. A collision detection system is needed to determine if objects are in contact or interpenetrates. These interpenetrations are resolved by a collision handling system. Because of the fact, that in nearly all simulations objects can interact with each other, collision detection is a fundamental technology, that is needed in all these simulations, like physically based simulation, robotic path and motion planning, virtual prototyping, and many more. Most virtual environments aim to represent the real-world as realistic as possible and therefore, virtual environments getting more and more complex. Furthermore, all models in a virtual environment should interact like real objects do, if forces are applied to the objects. Nearly all real-world objects will deform or break down in its individual parts if forces are acted upon the objects. Thus deformable objects are becoming more and more common in virtual environments, which want to be as realistic as possible and thus, will present new challenges to the collision detection system. The necessary collision detection computations can be very complex and this has the effect, that the collision detection process is the performance bottleneck in most simulations. Most rigid body collision detection approaches use a BVH as acceleration data structure. This technique is perfectly suitable if the object does not change its shape. For a soft body an update step is necessary to ensure that the underlying acceleration data structure is still valid after performing a simulation step. This update step can be very time consuming, is often hard to implement and in most cases will produce a degenerated BVH after some simulation steps, if the objects generally deform. Therefore, the here presented collision detection approach works entirely without an acceleration data structure and supports rigid and soft bodies. Furthermore, we can compute inter-object and intraobject collisions of rigid and deformable objects consisting of many tens of thousands of triangles in a few milliseconds. To realize this, a subdivision of the scene into parts using a fuzzy clustering approach is applied. Based on that all further steps for each cluster can be performed in parallel and if desired, distributed to different GPUs. Tests have been performed to judge the performance of our approach against other state-of-the-art collision detection algorithms. Additionally, we integrated our approach into Bullet, a commonly used physics engine, to evaluate our algorithm. In order to make a fair comparison of different rigid body collision detection algorithms, we propose a new collision detection Benchmarking Suite. Our Benchmarking Suite can evaluate both the performance as well as the quality of the collision response. Therefore, the Benchmarking Suite is subdivided into a Performance Benchmark and a Quality Benchmark. This approach needs to be extended to support soft body collision detection algorithms in the future.