Adaptive levels of detail for interactive Collision Handling

Abstract

Collision Handling has long been a major bottleneck in physically based animation. As scene complexity increases the problem becomes critical enough to prohibit real-time performance in the animation system. Previous approaches have attempted to achieve interactive rates by culling large parts of the scene or making drastic simplifications to the scene. We present here an adaptive level of detail approach to the collision-handling problem for interactive applications, which guarantees real-time rates irrespective of the complexity of the scene provided that the underlying hardware is able to cope with the lowest level of detail. This is achieved by using interruptible methods for the three problems of Collision Detection, Contact modelling and Collision Response. A multi-resolution bounding volume hierarchy is used to provide discrete levels of resolution in the data structures used for collision detection. We present a perceptually based collision prioritisation approach which makes graceful degradation of collision detail possible across a complex virtual environment. We present strategies for facilitating trade-offs between accuracy and speed in Collision Handling in the attempt to enhance the immersive nature of the virtual environment and optimise the quality of the dynamic scene.