Hybrid Collision Detection for Interactive Large-Scale Environments

Real-time collision detection is a challenging problem in large-scale virtual environments. We present a hybrid bounding volume hierarchy (HBVH) based framework aiming to reduce numerous unnecessary overlap tests in collision detection. Temporal coherence is also exploited to accelerate algorithm. Furthermore, we optimize our algorithm through SIMD (single instruction, multiple data) instruction-level parallelism. Our experiment shows that the algorithm works well for the purpose of interactive simulation of massive dynamic scenes, not only for rigid bodies, but also deformable objects.