When one attempts to animate advanced animals, such as humans, it becomes necessary to incorporate more abstract action selection mechanisms, such as mechanisms based on reasoning. This approach involves using AI techniques and is termed task-level motion planning (a good reference book on this subject is the book by Magnenat-Thalmann and Thalmann Nadia90); The most representative animation work along this line is that by Badler and his group who animated a human figure 'Jack' [Badler, Phillips and Webber1993]. The planner takes as input Jack's initial state and a 3D representation of his world and generates as output a series of actions necessary for accomplishing an assigned task, such as ``go and get some ice cream''.
While planning ability certainly is one of the most important characteristics in human behavior (and hence is important to model for animations of humans), it is not known as a common feature in most animals lower on the evolutionary ladder than primates. Rather, animal behavior is believed to rest on the more primitive and more fundamental faculty of reactive or adaptive behavior [Tinbergen1951, Manning1979, McFarland1993a]. Adaptive behavior enables animals to be autonomous and to survive in uncertain and dynamic environments. Our approach to behavioral animation reflects the adaptiveness of animal behavior. (Note that we are not referring to adaptiveness in the sense of learning nor evolutionary adaptation, but rather, to the ability to select appropriate behaviors according to the perceived situation.) In our approach, we gain high level control through the construction of a model of adaptive behavior where the actions of an artificial animal result from its active interaction with the world as guided by its perception.
|Xiaoyuan Tu||January 1996|