Accelerating Line of Sight Computation Using Graphics Processing Units

摘要

We present a method to accelerate line-of-sight computation for computer generated forces (CGF) using graphics processing units (GPUs). GPUs have become commodity processors and they are part of every game console or PC system. Moreover, their performance has been increasing at a rate faster than CPUs and the trend is expected to continue in the foreseeable future. We present a hybrid algorithm that exploits the computational power of GPUs to perform visibility culling and combine it with exact visibility computations on the CPU. Our approach is directly applicable to dynamic terrains. It has been applied to complex terrain environments and our hybrid algorithm is able to perform line of sight computations in a few microseconds on a commodity PC.