In 1989, when Lynne E.
Parker left ORNL's Center for Engineering Science Advanced Research
(CESAR) to pursue a Ph.D. degree in computer science at the Artificial
Intelligence (AI) Laboratory of the Massachusetts Institute of Technology
(MIT), she planned to focus on a new fieldcooperative robotics.
"CESAR had only
one robot but a second one was under construction," she says, noting
that she had been offered a job at CESAR after taking a University of
Tennessee course in AI from Chuck Weisbin, then CESAR director. "Because
I was thinking of returning to CESAR, I was interested in finding ways
to get two robots to work together. At that time, people thought that
getting one robot to perform a task was all you could hope for. But
it became clear that the advantage of two or more robots working together
on a complex task is redundancy and fault toleranceone can take
over for another if it fails."
The subject of
her Ph.D. thesis was heterogeneous multirobot cooperation. Based on
her MIT research and her work after she returned in 1994 to ORNL's Computer
Science and Mathematics Division (where CESAR resides), she wrote the
computer program ALLIANCE, which enables several robots to jointly perform
a task. The achievement has made Parker a pioneer in the infant field
of cooperative robotics.
Because of her
groundbreaking research, this East Tennessee native who graduated from
Powell High School in Knox County and earned a B.S. degree in computer
science from Tennessee Technological University, received a prestigious
award on April 12, 2000, at the White House. Parker was one of five
scientists from DOE's national laboratories and among 60 university
and government researchers to be honored with a Presidential Early Career
Award for Scientists and Engineers. Parker, who received her award from
Neal Lane, the President's science adviser, was recognized "as a shining
example to future generations of researchersthe best of the group
of scientists and engineers who will be responsible for America's 21st
century greatness." On the same day she also received a DOE Office of
Science Early Career Scientist Award from Secretary of Energy Bill Richardson
in a ceremony at Washington's Forrestal Building.
April 12, 2000, Lynne Parker received a Presidential Early Career
Award for Scientists and Engineers and a DOE Office of Science
Early Career Scientist Award. The second award was presented to
her by Secretary of Energy Bill Richardson.
robots today are stand-alone industrial robots that carry out single
tasks such as cutting, bending, or welding metal for automobiles. But
future intelligent machines in factories and other environments are
likely to include cooperative robots.
robotics," Parker says, "more than one robot performs a task that cannot
be done by one robot alone." Through use of on-board software written
in C and C++ and attached sensors and effectors, Parker has "trained"
small, mobile robots to work together to manipulate objects. For example,
Ada, Alexandra, Edith, and Grace, CESAR's four "Nomad technology" robots
named after female pioneers in computer science, recently passed a baton
over a series of barriers to reach a goal. The four robots, she says,
can get the job done faster than one robot can.
"We have demonstrated
that robots can move in formation," Parker says. This capability could
be useful for mowing a sports field, sweeping a gym floor, or scraping
an ice rink.
program coordinates the movement of robots so they don't interfere with
each other," Parker says. "It also allows them to cooperate through
communication. They share information about their intentions so the
other robots can adapt as they work to achieve a common goal."
Parker prepares to place a baton in the gripper of one robot,
which will pass it across the wooden barrier to another robot.
These are two of ORNL's four "Nomad technology" robots, which
are named Ada, Alexandra, Edith, and Grace after female pioneers
in computer science. The two smaller robots are named after Roman
emperors Hadrian and Augustus in honor of CESARORNL's Center
for Engineering Science Advanced Research.
Parker cites several
advantages for cooperative robotics. A team of robots can accomplish
more complex tasks than a single robot can working alone. The team is
more reliable and robust; if one robot fails, the other robots can take
over and continue the mission. By working in parallel, the team can
complete the task much faster than one robot. Because the individual
robots will have a simpler design, a team of robots may cost less to
construct and maintain than one robot built to carry out a complex task.
Teams of robots could
be used to perform complex tasks in areas too dangerous or otherwise
undesirable for humans. Such tasks could include cleaning up hazardous
waste sites, exploring planets, mining in unpopulated areas, participating
in search and rescue missions, and decommissioning nuclear power plants,
as well as taking part in such activities as automated manufacturing,
industrial maintenance, and surveillance for threats such as biological
and chemical warfare weapons.
For Caterpillar Inc.,
Parker and her colleagues are performing computer simulations of a cooperative
robotics system in which unmanned, automated bulldozers would remove
coal from surface mines in remote areas. "Our goal," Parker says, "is
to enable a human in one vehicle to direct the actions of the unmanned,
robotic vehicles that remove and collect the coal."
For a Defense
Applied Research and Development Agency (DARPA) project involving ORNL
and Science Applications International Corporation (SAIC) in Littleton,
Colorado, Parker is developing software that will coordinate the movements
of DARPA's sensor-equipped, tactical mobile robots, which were developed
to do surveillance in urban areas. "Military officials want to be able
to determine safely whether terrorists are staying in a suspect building,"
Parker says. "They envision man-portable robots called packbots that
toss over a fence several smaller robots, or 'throwbots' that resemble
bowling balls with spikes. Such throwbots might crawl through sewage
pipes, surreptitiously enter a building, and capture and relay images
of people and weapons in the building."
One of Parker's
biggest challenges is to enable robots to learn from each other. In
one multirobot learning project, the Nomad technology robots were assigned
this goal: Using all your capabilities, figure out how to move as a
team to keep as many moving targets under view as possible. The robots
have vision, compass, infrared, and two-dimensional (2D) laser sensors,
as well as odometric, tactile, and sonar sensors. They have an indoor
laser-based 2D global positioning system that allows them to locate
themselves and each other precisely in the room space. Thanks to a radio
ethernet system, they can communicate with each other and with a host
"So far, we've
found that the learned approach is better than random movement of the
robots, but it's not as effective as my hand-generated computer solution,"
Parker says. "When I give the robots my force vector field model, which
specifies that robots are attracted to targets and repelled by each
other, the robots quickly spread out and get close to the targets as
a team. This is the fastest solution."
Later this year,
Parker will write programs for four smaller, even more mobile, robots,
which are named after Roman emperors, in honor of CESAR. She will try
to "teach" these robots how to learn by giving them "hints" and "global
positive reinforcement" as feedback.
When asked why
she chose computer science as a field of study, Parker answered, "I
always liked math and science. My father, who is a civil engineer, suggested
that I might like computer science because it combines both science
and math. Computer science is a relatively new field that presents new
challenges. I like to work on and solve new problems, and the field
suits my independent nature."
is to make robots more autonomous and independent by giving them the
ability to learn. She knows that she and they have a long way to go,
but she says, "That's what makes my job fun."
Center for Engineering Science Advanced Research