Event Location: NSH 1305
Bio: Seth Teller is a Professor in the Department of Electrical Engineering
and Computer Science, and a Principal Investigator in the Computer
Science and Artificial Intelligence Laboratory (CSAIL), at MIT. He
received his B.A. in Physics from Wesleyan, and his M.Sc. and Ph.D.
in Computer Science from U.C. Berkeley. Teller’s research interests
include machine perception, autonomous mobile manipulation, human-robot
interaction and assistive technology for people with disabilities.

Abstract: Please note: The speaker has requested that this talk not be webcast or captured, it will not be available on videotape.

We describe a sliding autonomy framework in which a nearby
or remote human supervisor intermittently assists a robotic
mobile manipulator. Through a non-contact interface, the
supervisor breaks high-level tasks into suitable sequential
chunks, and provides to the robot perceptual cues and other
task guidance enabling it to perform each chunk autonomously.
Task execution occurs only after the robot has reflected its
mental model and plan and secured operator confirmation.

Successful realization of this approach requires addressing
a variety of interesting questions of perception, planning,
control, and human-robot interaction, as well as considering
issues of cultural acceptance of robots by people.

In a variety of real task domains, this refactoring of the
traditional human-robot division of labor will yield a number
of benefits: lower cognitive burdens for equipment operators;
increased safety and efficiency; and reduced system weight
and complexity, including relaxed performance requirements
for communications links. We substantiate these claims
through demonstration of two real-world systems that perform
non-trivial mobile manipulation tasks under intermittent
human supervision.