Page 440 - Sensing, Intelligence, Motion : How Robots and Humans Move in an Unstructured World

EXAMPLES 415

human. Notice the difference between the positions of the master arm and the
slave arm: The operator clearly intended the arm to go much lower than it did. 7
The value of this experiment is more than a joke exercise. Imagine a robot
assistant working next to an astronaut, controlled via teleoperation by a remote
operator. If the operator is on the ground, a signal transmission delay due to large
distance will introduce a bias into the operator’s perception of the scene compared
to the actual scene at the moment. Imagine that the operator makes a mistake,
guiding the robot directly onto the astronaut. The robot behavior demonstrated
in the Figure 8.11 experiment would save the situation.
Finally, the pictures shown in Figure 8.12 relate to the autonomous robot
operation in the setting 3 above. A good way to demonstrate robot interaction
with a nearby moving human is dance. If we can make the robot behave the
way two human partners expect each other to move in a dance, we can count
on the robot’s adequate behavior in a human–robot crew. During the dance
the partners stay close to each other. They continuously react to each other’s
movement. One partner does not have to look intently at the other partner at all
times; he or she is confident that their partner will make sure that no collision
takes place. For each of them it is not enough to know that his/her head or hand
is safe: There is an expectation that one’s whole body is safe from unpredictable
collisions. Hence the interaction involves their whole bodies. A demonstration
of this kind of interaction is a demonstration of a highly coupled robot–human
team operation.
With these ideas in mind, we have carried out a special project between the
University of Wisconsin (UW) Robotics Laboratory, on the one hand, and the
UW Dance Department, on the other hand. Professor Tibor Zana from the UW
Dance Department, who is also Artistic Director of the Wisconsin Dance Ensem-
ble, choreographed the dance. The video frames shown in Figure 8.12 are from
the resulting videos. Again, still pictures are not a good medium for showing
motion: A color video looks much more interesting than these black-and-white
still pictures.
The robot motion planning shown in these pictures was fully autonomous.
The robot was not programmed for any specific paths. (Tests with predefined
paths, which the robot would modify on the fly when reacting to the ballerina’s
movement, have also been tried.) The robot was only programmed to stay out of
the ballerina’s way and to move toward her when losing the proximity contact
with her. In other words, the actual motion was in response to the ballerina
movement. In a typical pair dance (e.g., waltz, tango, foxtrot, swing), one partner
is the leader and the other partner is the follower. In our robot–ballerina dance
the ballerina was the leader. This is admittedly not a typical dance convention
today, but aren’t robots the sign of the future!
The robot behavior in these experiments looks convincing and somehow
“alive.” We humans are not used to seeing machines behave like humans or

7 For those romantically inclined after reading Isaac Asimov’s robotic laws, the same would happen
if the obstacle was not a human but a chair.