New approaches to culling and LOD methods for scenes with multiple virtual actors

Despite the ever-increasing power of graphics workstations, rendering and animating virtual actors remains a very expensive task. Current scenegraphs are clearly oriented toward the management of static scenarios instead of toward scenes with multiple dynamic elements guided by complex behaviors, as in the case of virtual actors, human or not. Applications that deal with the management of multiple actors require the development of specific methods that reduce not only the number of polygons sent to graphics hardware but also the calculations involved in the management of multiple reference systems and the behavior of actors. In this paper we propose a culling method based on the use of different types of bounding spheres, which minimizes the number of calculations related to the behavior of actors and allows better use of the capacities of present-day graphics hardware. With the same aim in mind, we also propose a method for managing levels of detail which acts not only on a geometric level but also on skeletal and behavioral levels. These ideas can be implemented on top of traditional scenegraphs, resulting in significant improvements in computational costs. This improvement is analyzed, as well as the results obtained when managing scenes with thousands of virtual actors.

Graphical AbstractExample of a simulation scene of 6000 virtual actors moving in real time after applying the new culling and LOD techniques.Figure optionsDownload high-quality image (56 K)Download as PowerPoint slideResearch Highlights
► New method for managing level of details at skeletal and behavioral levels.
► Culling method based on bounding spheres that minimizes the number of calculations related to the behavior of actor.
► Culling method that allows a better use of the capacities of graphics hardware.
► Improvement of computational cost managing scenes with thousands of virtual actors.