Algorithms for Charismatic Robots

Rogers 226
Mon, 04/04/2016 - 10:00am

Heather Knight
PhD candidate, Carnegie Mellon

Abstract:
Robots are entering human environments in increasing numbers. These robots
perform functions in shared spaces (e.g. autonomous driving, vacuuming),
coordinate with humans (e.g., collaborative manufacturing, object delivery)
and leverage sociability (e.g., assistive robots, entertainment). Charismatic
robots can add value to human-robot interfaces by abstracting human social
rules and interaction strategies into the perception and/or behavior systems
of robots. Such systems can be both functional and expressive, but should add
social value to people’s experience with these machines.

Unfortunately, there are few generalized models of artificial social
expression, especially for the most common robot form-factors. Such models do
exist in human acting training, and they qualitatively describe how to
construct artificial characters and expression. My research seeks to
operationalize models from dramaturgical disciplines (e.g., movement
training) that aid in the construction of robot nonverbal behaviors. The
goals are: 1) to make the model algorithmic, 2) apply the model to low-degree
of freedom robots, and 3) refine our parameterizations using the expertise of
actors and performers.

This talk outlines how we have operationalized the Laban Effort System, a
qualitative set of expressive motion descriptives from dance and theater
training, to a set of quantitative motion features that robots can use to
convey their internal states to people. I will share examples of these
feature implementations on three low-degree of freedom robots, including two
robot heads and a mobile robot. During naturalistic deployment using the
autonomous omni-directional CoBot robots, we find that robot motion
characteristics help bystanders interpret the robot’s current state (e.g.,
rushed), impacting their likelihood of interrupting a robot's task. Expanding
on this finding, we are currently investigating additional strategies to
flexibly train robot state communications with people in order to effectively
include robot expressions into everyday robot tasks.

Bio:


URL: http://eecs.oregonstate.edu/colloquium/algorithms-charismatic-robots

_______________________________________________
Colloquium mailing list
[email protected]
https://secure.engr.oregonstate.edu/mailman/listinfo/colloquium

Reply via email to