Behavioral models are used in games and computer graphics for
realistic simulation of massive crowds. In this paper, we present a
GPU based implementation of Reynolds [1987] algorithm for simulating
flocks of birds and propose an extension to consider environment
self occlusion. We performed several experiments and
the results showed that the proposed approach runs up to three
times faster than the original algorithm when simulating high density
crowds, without compromising significantly the original crowd
behavior.
