Robotics Reference
In-Depth Information
pleases the robot, the pleasure variable is increased by one and whenever
something happens that the robot does not like, the pleasure variable is
reduced by one. Things that please the robot could be playing a game
with its owner, or singing a song. Things that displease it could include
sensing that the room temperature is below its level of comfort (so the
robot “feels” cold), or losing a game to the user.
Clearly, the more pleasant things that happen to the robot the higher
will be the value of its pleasure variable, and the more unpleasant things
that happen to it the lower will be the value of the pleasure variable.
There could be a table of values that specify how joyous or distressed the
robot feels, depending on the value of the pleasure variable. For example,
Pleasure Variable Mood
+3 or greater
ecstatic
+2
very happy
+1
moderately happy
0
neither happy nor unhappy
-1
moderately unhappy
-2
very unhappy
-3
thoroughly miserable
The robot could be programmed to exhibit a certain range of responses,
depending on its mood as determined from the above table. For example,
it might laugh uncontrollably when ecstatic, it might let out a cheer when
very happy, ...and itmight burst into tears andwail loudlywhen it is
thoroughly miserable.
The Oz Model of Emotion
The Oz project was a computer system developed at Carnegie Mellon
University from the late 1980s up to 2002. Work on the project has
been and is being continued elsewhere, including Georgia Institute of
Technology and Zoesis Studios in Newton, MA.
4
Oz allowed authors
to create and present interactive dramas. The system design includes a
simulated world, several simulated characters, a theory of dramatic pre-
sentation and a simulated drama manager. Each character had to be able
to “...display competent action in the world, reactivity to a changing
4
The ideas behind the Oz model of emotion are a joint collection of many in the Oz group, but
the credit for most of the development work on the project, including designing and building the
system, belongs to Scott Reilly. It is the subject of his PhD thesis, for which another of the Oz team,
Joseph Bates, was advisor.
