Robin Hood's Algorithm for Time-Critical Level of Detail
Eduardo Hern · 2005
We describe a novel method for time-critical rendering of complex scenes populated by numerous distinct objects having an intricate geometry. During the preprocessing step the time consumed by logic and rendering processes is measured. For each frame we estimate the rendering time and distribute it between each potentially visible object which chooses its best representation within the time restrictions. Finally we take advantage of the unused time to improve the rendering process of the remaining objects. This method adjusts image quality adaptively and aims to maintain a constant bounded frame rate even for scenes which complexity may change drastically between frames. In contrast to previous solutions our algorithm can be used in combination with discrete and continuous LOD techniques and it does not impose limitations on the scene conditions. We demonstrate the benefits of our method in a virtual ecosystem composed by trees.