Roland Hübscher
Interactive visual computer animation is becoming an important tool for science education in grade school.
Unfortunately, students and teachers cannot easily create their own animations, because programming these
systems tends to be too hard for non-professional programmers. I present an approach that simplifies the
description of complex interactions of objects by describing interactions with declarative, temporal
constraints. A system that describes animation in terms of the actions of the objects and
the interactions between the objects is being built on top of a grid-based, graphical programming
environment.
The need for languages that allow grade-school students to create
interesting, non-trivial animations is therefore of great importance. The
domain of visual animation suggests to use visual programming languages,
which would allow the end user to directly manipulate the animation instead
of some text which is separate from the animation. Having a textual
description of a visual animation has the problem of mapping the visual
and textual concepts to each other.
Rules describe what actions an object can execute. As long as it is clear what the actions of the objects are
and which action is chosen in what
situation, no problem arises. But this is not always so simple. An
interaction involves several objects, and it is not always clear which of
the objects acted on which other objects. For instance, when on the pool
table the (white) cue-ball collides with the resting eight-ball, the
cue-ball stops and the eight-ball accelerates as Figure 2
shows. Did the cue-ball accelerate the eight-ball or did the eight-ball
stop the cue-ball, or did both happen -- and do we really want to make
this decision anyway? A physicist would probably prefer the physically correct description of the collision
process. The chances are, that he or she needs to figure out exactly how the balls react to each other which
may require additional parameters about mass, impetus, spheres and so on. On the other hand, the user
might only be interested in a description of the interaction of the two balls since, in an animation, all the
balls have to do is behave correctly when bumping into each other.
FIGURE 2: The white cue-ball collides with the black eight-ball.
FIGURE 3: The temporal constraint describes the collision of the
cue- and the eight-ball.
A temporal constraint describes an interaction by constraining each
state in a sequence of states. Constraining a state means that the
constraint, a condition on the state, allows only certain states to be at
this place in a sequence. There is no need to figure out the actions and
conditions that make the interaction emerge from the actions of the
individual objects. As the name interaction suggests, the
interaction can be viewed as an action between the objects. Of
course, the user is allowed to use just actions to describe an interaction,
since this can sometimes be easier than describing a complex interaction
resulting from simple actions. A good example for the latter
case is the Braitenberg vehicles (Braitenberg 1984).
A temporal constraint (see Figure 3) describes a
sequence of states (for instance, the ones in Figure 2) and
looks similar to a rule. There are a few important differences, though.
Whereas a rule manipulates the picture of the animation, a
constraint chooses one of possibly many different picture sequences
generated by the rules. Furthermore, a temporal constraint restricts an
arbitrary long sequence of states, although they are rarely more than three
states long, whereas a rule refers only to two states. An additional
advantage is the declarative of constraints which allows the user to ignore
implementation details.
The following analogy may help to make the differences of between
temporal constraints and rewrite rules clearer. The possible states the
human body can be in and how the body can go from one state into another
depends on the laws of physics and the body's structure. This is
captured by the rules that define the space of states and state
transitions. If the person, that is, the owner of the body decides to use
the body, for instance to reach for the coffee cup, then all the actions of
many body parts must be coordinated. This is exactly what a temporal
constraint does. By constraining the possible action sequences, only
action sequences are allowed that result in the hand grasping the cup.
An implementation of a prototype is being implemented on top of Agentsheets
(Repenning 1993), a grid-based graphical programming environment
suitable for the construction of visual programming languages. Using
Agentsheets has the advantage of reaching a quite large community of users
simplifying the user testing which has not been done yet.
Simplifying the creation of visual animation is of great importance
since it increases their value to science education. Not only can the
students create their own animation of the real world, they can interact
with it, trying out different assumptions and ideas.
The frequent discussions with Clayton Lewis, Alex Repenning and the
members of the NAP group have been very helpful.
Abstract
Keywords:
visual animation, science education, visual programming, rewrite rules, temporal constraints
Introduction
Visual animation can be used to display dynamic processes and are an important tool in grade school
science education. Animation is even more useful if it can be programmed by the end users, that is, by the
students and their
teachers. Unfortunately, the complexity of
programming animations can keep the end users from creating their own
animation and thus, the interactional value of the tool is diminished.
VISUAL PROGRAMMING LANGUAGES
I will restrict myself to rule-based visual languages since they have
proven to be useful expressing a wide variety of animations
(Bell 1992, Furnas 1990). A system based on rewrite rules consists of three
components, namely the picture of the animation as seen by the viewer
(the state of the computation), a collection of rewrite rules
(the program), and a rule interpreter. A rewrite rule consists of a
before- and an after-picture. The rule interpreter tests whether
there is a rule whose before-picture describes a part of the animation
picture. If the test is successful, the described part of the animation is
replaced by the after-picture as is demonstrated in
Figure 1.
ACTION AND INTERACTION
FUTURE WORK AND CONCLUSIONS
Acknowledgments
