Research on intelligent avatar in VRML worlds

We present an intelligent avatar in VRML worlds which can understand users' intentions more precisely. First, the method customizing avatars is proposed to provide users with the ability to create their static avatars quickly and simply. Second, we use a hierarchical action model to efficiently convert the top level task actions of an avatar into combinations of concrete physical actions. Last, multimodal based interaction allows users to control their avatars in a natural way. Research on intelligent avatar is of much help to research on intelligent interactions among avatars as well as between avatars and virtual scenes in VRML scenes.