A cold, dry run: Robot tests well before expedition to Antarctica
Sunday, October 11, 1998
By Byron Spice, Science Editor, Post-Gazette
http://www.post-gazette.com/healthscience/19981012nomad2.asp
It's almost 10 degrees below zero as the wind blasts across the skin of
Nomad, a four-wheeled robotic vehicle a bit shorter and a bit higher than a
Volkswagen Beetle. In this deep freeze, ordinary wiring cracks, motors seize
and electronics fail.
But the Carnegie Mellon University robot, newly outfitted for Antarctic
meteorite hunting, withstands the subzero onslaught.
Then again, Nomad is no closer to Antarctica's Patriot Hills during this
trial than Cranberry.
Plans call for packing up Nomad this week for shipment to Antarctica, where
a five-member team from Carnegie Mellon's Robotics Institute will spend six
weeks testing its ability to find meteorites. But Dimi Apostolopoulos,
manager of the Nomad program, first wanted to check whether his team's
efforts to harden the robot against the cold had worked.
"He wanted Antarctica, so I brought Antarctica to him," said John Kauer,
manager of New Federal Cold Storage, a division of Robert Wholey Co. Last
week, Nomad spent 24 hours in Blast Cell No. 1 of New Federal's
110,000-square-foot, refrigerated warehouse in Cranberry.
Normally, the three-story blast cell would be stacked with pallets of meat
or other foods being flash frozen. But for Nomad, Kauer's crew laid down a
sheet of ice over the floor and cleared out the scaffolding so that three
large fans in the ceiling could direct their blast on the robot.
So 1,200 pounds of frigid robot joined the 10 million pounds of frozen
turkey, 10 million pounds of ham and 4 million pounds of ice cream already
packed into New Federal's 3.8 million cubic feet of cold storage. The
ammonia-charged refrigeration system, driven by 1,200-horsepower condensers,
kept Nomad's temperature at minus 8 degrees Fahrenheit, roughly the
temperature it will encounter in Antarctica in the coming season.
"This year is critical for us," Apostolopoulos said. Early this year,
researchers went to Antarctica to test technology that might be used aboard
a "meteorobot," but this will be the first time that they will use a robot
to look for meteorites. The outcome will determine what direction they will
take next year, the last year of what was expected to be a three-year
program sponsored by the National Aeronautics and Space Administration.
Initial plans called for sending a small, solar-powered robot especially
designed for meteorite hunting to Antarctica by the program's third year.
That may not prove possible, however. More likely, Apostolopoulos said,
Nomad will make another trip to Antarctica next year. What remains to be
determined is whether a fourth trip can be justified, or whether to redirect
the program to other uses besides meteorite hunting.
"If the Carnegie Mellon group finds a meteorite this year, that would be an
achievement," said Ralph Harvey, a geologist at Case Western Reserve
University in Cleveland. He heads the human team sponsored by NASA that has
looked for meteorites in Antarctica for more than 20 years. "The robot is
still having a hard time figuring which way is up."
Harvey noted that he can teach a person how to find a meteorite within a few
hours. "There's just no hope that any kind of robotic beast is going to be
able to match that," he added.
Apostolopoulos doesn't argue the point. Using computer vision, light
spectroscopy and magnetic sensors, a machine might be able to discern rocks
that are probable meteorites, or it might be able to say whether a given
rock field is likely to have a significant number of meteorites. Equipped
with ground-penetrating radar, a robot might sweep through a snow-covered
field, finding meteorites that would be invisible to humans.
But Nomad or any robot is unlikely to approach the 95 percent accuracy rate
of human spotters, he admitted.
Antarctica has proven to be a treasure trove of meteorites. These bits of
planetary matter erode rapidly in most climates, but in cold, dry Antarctica
they can survive indefinitely. Collection began in earnest 22 years ago,
when University of Pittsburgh geologist William Cassidy began directing
annual searches sponsored by NASA.
Interest jumped two years ago when NASA scientists announced that one of the
meteorites collected by Cassidy's crew was a Martian meteorite that included
what looked to be signs of primitive life. Many scientists have raised
doubts about the claim, but it boosted interest in Antarctic meteorite
searches and NASA subsequently approved the meteorobot program.
"Lots of people were jumping on the bandwagon," Harvey recalled. "I started
out exceptionally skeptical of the whole idea" of using robots. But, as
Carnegie Mellon and NASA readily admit, the plan had always been to use the
project to test technology that might be used to explore other planets.
Advancing the science of meteorology has always been a secondary concern.
Robotics experts expected the task to be difficult and they were not
disappointed, Apostolopoulos said. And though Nomad will never be the equal
of human meteorite hunters, it marks a step toward robots with
science-driven autonomy.
"For years and years, we were building robots that just drove around," he
explained. Robots such as Sojourner, which landed on Mars last year, have
been dependent on humans to control them. Nomad, however, will have enough
sensors and on-board smarts to be able to detect areas of potential interest
by itself, even if it would be dependent on humans to study the fine details
of any rock or interesting feature it might find.
On a distant planet, as in Antarctica, constant human supervision of a robot
explorer would be difficult or impossible, he noted, so robots will have to
exercise autonomy if they are ever to explore large expanses.
The goals for the coming experimental season in Antarctica will be somewhat
less ambitious. Nomad will attempt to use on-board analysis to determine
which rocks are likely to be meteorites, based on their size, shape, color
and spectral features.
Unfortunately, as the Carnegie Mellon team learned earlier this year, the
Patriot Hills area where they will camp is bereft of meteorites. The area,
Apostolopoulos suggested, might be just warm enough to cause the meteorites
to melt into the ice.
But it is the only area available to them this year through the Chilean
Antarctic Institute and Chilean Air Force, which are providing
transportation and support services. So the group will take meteorites with
them to "salt" the rocky areas that Nomad will explore.
Meanwhile, Pitt's Cassidy, who now works with the Nomad group, will be
surveying other areas where a robot might hunt for meteorites next year.
The other goal is to test the robot's ability to navigate. The conditions,
Apostolopoulos noted, are similar in many ways to the polar areas of the
moon, where William "Red" Whittaker, director of Carnegie Mellon's Field
Robotics Center, has proposed sending a robot to look for lunar ice.
Nomad, which traversed 125 miles of Chile's Atacama Desert in June and July
of 1997, uses a stereo pair of video cameras and a laser rangefinder to
steer itself. That may not work in Antarctica, where the white expanses give
the stereo cameras little nearby to focus on and where laser light might
scatter off the ice.
One alternative, Apostolopoulos noted, might be to use the robot's
panospheric camera, a video camera that takes in a 360-degree view of the
horizon. By focusing on distant landmarks or the sun, Nomad might be able to
use this camera to navigate.
Nomad also will test "blind navigation" in which it will monitor angle of
its body and the speed of each wheel to detect whether it is venturing into
unsafe areas.
Though Nomad was able to negotiate the Chilean desert, it took a year of
rebuilding to prepare it for Antarctica. Cameras and other sensors have been
swathed in insulation and each has its own heater. The main electronics box
in the robot body has been insulated and has two heaters installed.
Electrical cables and connectors have been replaced with those designed for
subzero temperatures. Lubricants have been replaced with those rated for
extreme temperatures. Metal tires have been replaced with rubber.
During last week's test, Nomad was left in the blast cell for 25 hours and
then its gas-powered generator was cold started. Heaters were then turned on
and, within an hour, the computers were warm enough to operate.
"Overall, it was a very successful day," Apostolopoulos said. One electrical
connection failed, leaving the scientific camera unusable, but the robot
otherwise performed flawlessly and the crew that will accompany Nomad to
Antarctica got some experience working with their testbed in frigid
conditions.
"They got their first chill of the year," he added.