Building out the Network
American City Business Journals
(news from 41 Business publications around the country)
|
A
robot built by researchers at the University of Massachusetts at
Amherst, left, uses a modified Segway Human Transporter for locomotion.
The modified Segway, called a Robotic Mobility Platform (RMP), lets
U.S. designers create robots that can compete with leg-based robots
produced by Japanese companies.
The Segways gyroscope system, designed to balance its rider and avoid tipping, enhances its use as a robotic platform.
|
UMass Segways into robot mobility
|
|
04/09/2004 14:34 PM
|
By Arthur Bright
Segway’s wheels may soon give UMass robots a leg up on their two-legged Japanese counterparts.
A
team of researchers at the University of Massachusetts at Amherst is
building robots using a modified Segway Human Transporter as the means
of propulsion. The performance of the modified Segway, dubbed a Robotic
Mobility Platform (RMP), allows U.S. designers to create robots capable
of matching those of Asian companies.
The RMP is a suitable
platform for a robot for a number of reasons, said professor Rod Grupen
of the UMass computer science department. First, the Segway is designed
to dynamically balance its rider to avoid tipping, using a system of
gyroscopes in its base.
“Because it can lean and balance, it can
go up a slope that might be up to 40 degrees, where a standard mobile
robot for indoor use would roll up and fall on its side,” Grupen said.
Also, the Segway’s engine is more powerful than that of the typical robot.
“It
has two horsepower per wheel,” Grupen said, “where a world-class
athlete peaks at three quarters of a horsepower. It’s a very
high-performance platform and can go in excess of 12 miles per hour. It
can do things most robots can’t do.”
When equipped with a laptop
brain and digital camera eyes, the UMass RMP robot can maneuver through
crowded areas without hitting other objects. In tests, the robot has
proved capable of reaching its destination while avoiding both immobile
and moving objects, like humans who walk and stop suddenly.
The
optical sensors also allow the robot to track and follow moving
objects. Robots with the ability to follow an assigned leader could
someday be used as mules in environments too dangerous for humans to
travel. Now, its tasks are somewhat simpler.
“We have it chasing
cars in parking lots and chasing running people,” Grupen said. “One of
the more ambitious goals we have, every year the department has an
event that includes a half marathon (13 miles). This year we’d like to
have the Segway participate.”
The UMass researchers intend to
combine the RMP with a two-armed, two-handed torso they’ve been
designing separately to make a mobile robot capable of visually and
manually interacting with its environment. While the Segway platform
might not be as sexy as the walking humanoid robots that companies like
Honda and Toyota have been building, the Segway’s dynamic balance gives
it an advantage over the Japanese robot legs and lets U.S. robot
developers better compete with their Asian counterparts.
The
leg-based systems “are getting very good, but they’re not at the high
end of the performance range. The Segway gives us an opportunity to
jump right back up into the game again, with its high performance and
small footprint.”
That opportunity is just what Segway, in
partnership with the Defense Advanced Research Projects Agency (DARPA),
hoped to spark when they distributed RMPs to 17 research centers across
the country (including UMass), which are employing the RMPs in a
variety of ways. An MIT group is building an armature-equipped Segway
that can navigate human environments and manipulate doors. Other
researchers around the country are developing groups of RMP robots that
can travel together like a mule train or work cooperatively to move
large objects.
The UMass team hopes to next develop the
infrastructure for producing the two-arm-equipped RMPs efficiently.
They are also planning new means of programming the robots by teaching
them to learn dynamically rather than relying on human-created code.
“Beyond
the five-year mark, we hope to create models of intelligence that grow
and mature by interacting with the world, because we think it will be
the only way to program such devices,” Grupen said.
Arthur Bright is a graduate student in journalism at Boston University.
|
|
|