Page 183 - Designing Sociable Robots

breazeal-79017 book March 18, 2002 14:11

164 Chapter 10

witness and interpret each action. It is also important that each action last for a sufficiently
long time span for the observer to read it. Given these two guidelines, Kismet expresses only
one emotion at a time, and each expression has a minimum persistence of several seconds
before it decays. The time of intense expression can be extended if the corresponding
“emotion” continues to be highly active.
The transitions between expressive behaviors should be smooth. The build-up and decay
of expressive behavior can occur at different rates, but it should not be discontinuous like
throwing a switch. Animators interpolate between target frames for this purpose, while
controlling the morphing rate from the initial posture to the final posture. The physics of
Kismet’s motors does the smoothing for us to some extent, but the velocities and acceler-
ations between postures are important. An aroused robot will exhibit quick movements of
larger amplitude. A subdued robot will move more sluggishly. The accelerations and decel-
erations into these target postures must also be considered. Robots are often controlled for
speed and accuracy—to achieve the fastest response time possible with minimal overshoot.
Biological systems don’t move like this. For this reason, Kismet’s target postures as well
as the velocities and accelerations that achieve them are carefully considered.
Animators take a lot of care in drawing the audience’s attention to the part of the scene
where an important action is about to take place. By doing so, the audience’s attention
is directed to the right place at the right time so that they do not miss out on important
information. To enhance the readability and understandability of Kismet’s behavior, its
direction of gaze and facial expression serve this purpose. People naturally tend look at
what Kismet is looking at. They observe the expression on its face to see how the robot is
affectivelyassessingthestimulus.Thishelpsthemtopredicttherobot’sbehavior.Iftherobot
looks at a stimulus with an interested expression, the observer predicts that the robot will
continue to engage the stimulus. Alternatively, if the robot has a frightened expression, the
observer is not surprised to witness a fleeing response soon afterwards. Kismet’s expression
and gaze precede the behavioral response to make it understandable and predictable to the
human who interacts with it.
Expression is not just conveyed through face, but through the entire body. In general,
Kismet’s expressive shifts in posture may modify the motor commands of more task-
based motor skills (such as orienting toward a particular object). Consequently, the issue of
expressive blending with neck and eye motors arises. To accomplish successful blending,
the affective state determines the default posture of the robot, and the task-based motor
commands are treated as offsets from this posture. To add more complexity, the robot’s
level of arousal sets the velocities and accelerations of the task-based movements. This
causes the robot to move sluggishly when arousal is low, and to move in a darting manner
when in a high arousal state.