Last but not least, we would like to explore an intuitive way of directing the autonomous artificial creatures to a high degree. In the current implementation, it is easiest to create animations in which we do not demand highly specific control over what the creatures do and when they do it, as long as they behave naturally. However, it is not unusual to want to create an animation which shows, say, two or more events happening at roughly the same time and place. For example, when making the animation ``The Undersea World of Jack Cousto'' [Tu, Grzeszczuk and Terzopoulos1995], we wanted to show, in a period of less than one minute of animation, two fishes displaying mating rituals while a large school of fish passes by in the background stalked by a predator shark. The most difficult part of achieving this sort of scenario is to roughly synchronize the mating, the schooling and the preying and to make schooling/preying happen in roughly the right place (the background).
To achieve these sorts of synchronizations, the animator would have to spend a fair amount of time performing multiple trials, tuning the relevant parameters until the desired action sequence is achieved. An obvious way of simplifying the problem is to suspend the forward simulation of the school of fish and the shark and run a trial simulation to determine the time at which the mating occurs. Once this is known, we can start simulating the school of fish and the shark roughly around the same time. Of course, we may still need to influence the path of the school which is done by scripting the path of the `lead' fish. Currently, these detailed manipulations are done by hand and hence may be cumbersome.
It may be possible, however, to build a higher level controller on top of the animation system we currently have in order to accomplish similar tasks in a much more convenient way. A promising approach for accomplishing this level of control would be employ some descriptive language for monitoring the animation process and exerting direct control over the various aspects of the simulation; e.g. the start time of some specific creatures and their paths, etc. Therefore, one of the future research directions could be to define and develop such a language and its interface to the current animation system. Blumberg and Galyean Blumberg95 recently proposed and implemented a system that addresses interactive ``directability'' of animated autonomous creatures. Their work marked a first step towards the aforementioned future research direction.
|Xiaoyuan Tu||January 1996|