作者: Joel Auslander , Alex Fukunaga , Hadi Partovi , Jon Christensen , Lloyd Hsu
关键词: Control theory 、 Computer vision 、 Evolutionary computation 、 Motion (physics) 、 Computer animation 、 Animation 、 Computer science 、 Artificial intelligence 、 2D to 3D conversion
摘要: We extend an earlier automatic motion-synthesis algorithm for physically realistic articulated figures in several ways. First, we summarize incremental improvements to the original that improve its efficiency significantly and provide user with some ability influence what motions are generated. These techniques can be used by animator achieve a desired movement style, or they guarantee variety synthesized over runs of algorithm. Second, report on new mechanisms support concatenation existing, automatically generated motion controllers produce complex, composite movement. Finally, describe initial work generalizing from 2D 3D figures. Taken together, these results illustrate promise challenges afforded controller-based approach synthesis computer animation.