|
|
Robotics Institute
Seminar, January 23
Time
and Place | Seminar Abstract | Speaker
Biography | Speaker Appointments
Activity Modeling and Coordination in Complex Domains: Robot Teams, Humanoids, and People
Maja Mataric
|
Time and Place |
Mauldin Auditorium (NSH 1305)
Refreshments 2:15 pm
Talk 2:30 pm
This talk will summarize
our activities in developing models and
algorithms
for activity modeling and coordination of complex behavior
in
individuals (specifically humanoid robots) and groups (specifically
teams
of robots and/or assemblies of people).
The
first part of the talk will focus on explicit/intentional and
implicit/emergent
coordination methods. We will describe a
framework
for
explicit multi-robot coordination, cast in the formalism of
distributed
task allocation, and used to provide a taxonomy of
multi-robot
coordination methods, existing algorithms (including our
work
on auction-based task allocation), and outstanding problems. We
will
then address implicit coordination and describe a framework for
minimalist
distributed algorithms (relative to internal state and
communication
requirements) and an associated set of methods for
automatically
deriving minimalist controllers (with experimental
examples
from the multi-robot construction domain).
The
second part of the talk will focus on activity modeling and
control
in humans and humanoids. We will describe a method for
automatically
extracting structure from activity data using
multidimensional
scaling (a spatio-temporal adaptation of Isomap) and
describe
how structured activity (individual or group) can be
automatically
derived/learned. We will show how to derive efficient
movement
controllers for complex humanoid agents and robots and how to
enable
humanoid skill learning by imitation.
Activity recognition,
classification,
and imitation will be brought together in a unified
generative
model, and demonstrated on humanoids as well as mobile
robots
engaged in human-robot interaction.
As
with any proper robotics talk, a plethora of videos will be shown.
Research
details are found at http://robotics.usc.edu/interaction/
Speaker
personal page is found at http://robotics.usc.edu/~maja
Speaker
bio is found at http://robotics.usc.edu/~maja/bio.html
|
Speaker Biography |
Maja Mataric'
is an associate professor in the Computer Science
Department and Neuroscience Program
at the University of Southern
Embedded Systems (cres.usc.edu), and
co-director of the USC Robotics
Research Lab (robotics.usc.edu). She
received her PhD in Computer
Science and Artificial Intelligence
from MIT in 1994, her MS in
Computer Science from MIT in 1990,
and her BS in Computer Science from
the
Award, the MIT TR100 Innovation
Award, the IEEE Robotics and
Automation Society Early Career
Award, the
Junior Research Award, and is
featured in the documentary movie "Me &
Isaac Newton." She is an associate editor of three major
journals and
has published over 30 journal
articles, 17 book chapters, 4 edited
volumes, 87 conference papers, and 22
workshop papers, and has two
books in the works with MIT Press.
Her research is aimed at endowing
robots with the ability to help
people and involves systems ranging
from individual assistans
(for convalescence, training, education,
companionship, etc.) to cooperative
robot teams (for habitat
monitoring, emergency response,
etc.) and problems of intelligent
control and learning in complex,
high dimensional/high degree of
freedom systems that integrate
perception, representation, and
interaction with people. Research details are found
at http://robotics.usc.edu/interaction/
|
Speaker Appointments |
For appointments, please contact Kevin R Dixon
The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.