Our methodology aims at producing autonomous artificial animals that not only look like, but also move like and behave like their natural counterparts. An important question is: How closely should our models attempt to emulate real animals? Clearly, a certain level of modeling fidelity is required in order to generate convincing results and, generally, the more faithful the models, the more realistic the results. Most real animals of interest are extremely complex both in terms of body structure and in terms of behavioral repertoire. Models of animals can therefore easily become excessively complicated. However, it is desirable for an animation system to be reasonably efficient so that it will run quickly enough on current graphics computers to allow interactive modification by the animator.
For the purposes of animation, we must strike a good compromise between model fidelity and computational efficiency. Striking the proper balance is a critical design issue, since inappropriate model accuracy can be counterproductive to the purpose at hand. For example, if we wanted to build a model of a tiger to show the effect of gait on the maturation of the bone in its legs, it may be necessary to model the cellular structure of the bone. However, this is hardly necessary if we are only interested in animating tiger gaits. Therefore we should keep the model complexity as low as is necessary to achieve the intended purpose--in our case, realistic appearance, locomotion and behavior.
|Xiaoyuan Tu||January 1996|