PUBLIC INFORMATION OFFICE
JET PROPULSION LABORATORY
CALIFORNIA INSTITUTE OF TECHNOLOGY
NATIONAL AERONAUTICS AND SPACE ADMINISTRATION
PASADENA, CALIF. 91109. TELEPHONE (818) 354-5011
http://www.jpl.nasa.gov
Contact: Diane Ainsworth
FOR IMMEDIATE RELEASE
August 8, 1997
SOJOURNER'S 'SMARTS' REFLECT LATEST IN AUTOMATION
The Mars Pathfinder Sojourner, a lightweight machine on
wheels, is accomplishing a revolutionary feat on the surface of
Mars.
For the first time, a "thinking" robot, equipped with
sophisticated laser eyes and automated programming, is "thinking"
and reacting to unplanned events on the surface of another
planet.
"After a few days on the Martian surface, when we left the
rock named Yogi, we turned on Sojourner's hazard avoidance system
and asked it to start making some of its own decisions," said
Brian Wilcox, supervisor of the robotic vehicles group at JPL.
"This hazard avoidance system sets the rover apart from all other
machines that have explored space. Sojourner had to make that
trip to the next rock without the benefit of detailed information
to warn it of obstacles along the way."
Sojourner's intelligence is the product of many years of
research in the fields of telerobotics and automation. Beyond the
challenges of building a robot that can survive the frigid
climate of another planet, Sojourner is taking the robotic
exploration program one step further to a day when truly "smart"
machines will be able to explore hundreds of kilometers away from
a landing site all on their own.
"Because Sojourner is our first rover, we designed it to be
quite safe," Wilcox said. "She moves slowly (0.7 centimeter per
second or 0.3 inches per second) and stops a lot along the way to
sense the terrain and process information. Her IQ is probably not
as high as an insect's, but to put it in the proper perspective,
consider that even a house fly has more computing capacity than
the largest supercomputers today."
Sojourner's hazard avoidance system is comprised of five
laser stripes that project out to the ground. One laser is
located in the center of the rover body and points straight
ahead. Two of the laser beams are pointed outward at a modest
angle from the center beam, while two more project much farther
outward, like periphery vision, beyond the rover's body.
"These lasers cut across anything in the rover's path,"
Wilcox said. "By looking at the shape of the stripe with the
rovers cameras, we can detect rocks and build contour maps of the
terrain immediately in front of the rover. The rover then moves
three inches ahead and projects its laser beams on the ground
again. We take that data and continue to build our terrain maps
based on the shapes of the laser beams as they bend with the
shape of the ground."
To complement its laser eyes, the rover has three levels of
autonomy to use in choosing its path across the Martian sand. The
first level involves little risk: it instructs the rover simply
to avoid all but the flatest ground. The higher levels of risk
cause the rover to drive over bigger and bigger obstacles without
trying to avoid them, with the highest risk level allowing the
rover to climb over rocks almost as big as the wheels.
"If we were exploring a dry riverbed, which would be
relatively smooth and devoid of rocks, the first level of risk
would probably be all we needed to drive a rover," Wilcox said.
"But given the terrain we've landed in, which contains a huge
variety of rocks, we have to use the higher levels of risk.
Just like on the freeway, you would always drive around any small
object if you could. In more rugged terrain, we don't have that
luxury."
Another type of autonomy command tells the rover to find
rocks.
"We tested this mode when Sojourner made its way to the rock
named Souffle," Wilcox said. "We gave it only the x- and y-axis
coordinates, then told it to find a rock. It used its laser
hazard sensor to find, instead of avoid, the rock. If the
coordinates were about right, then it knew it must be at the
right rock. And even though the wheel slippage and gyro drift
would have caused the rover to be more than 30 centimeters (1
foot) away from Souffle, the rover instead parked itself within 4
centimeters (2 inches) of the spot we wanted near the rock."
Sojourner's intelligence is based on a mathematical model
that emulates animal behavior. The ability for animals to avoid
or flee danger, such as predators, is a low level reflex, Wilcox
explained, but it gets the job done.
"The rover can distinguish between obstacles that pose no
threat, such as a two- or three-inch-tall rock that can be driven
over, as opposed to a 10- or 12-inch-tall rock that might tip it
over," he said. "As soon as the rover recognizes the hazardous
rocks, it switches into the hazard avoidance behavior and begins
to turn until it no longer sees the obstacle. Then it moves
forward and returns to a course toward the current goal."
Sojourner's reasoning capabilities deviate from the strict
sequence of instructions used on prior space explorers. Previous
spacecraft have operated solely using sequences of instructions
created by human operators on the ground, along with
preprogrammed "safing" sequences which protect the spacecraft
from harm and re-establish contact with Earth if some failure
should occur. Sojourner has both of these capabilities, but also
has reflexes to avoid hazards and respond to unplanned events.
This is necessary because it would go much too slow if humans on
Earth had to make every decision when interacting with the
unknown environment.
"This is what allows Sojourner to take a stroll on its own,"
Wilcox said. "If it encounters new rocks that it had not
anticipated, the rover will use sensory information provided by
the lasers to circumvent the obstacle. If the wheels on one side
slip in the dirt, it will begin to turn until it is going back in
the right direction."
Sojourner's major limitations lie in its dependence on the
Pathfinder lander for communications. If it journeys much beyond
the lander's line of sight, it will become marooned. Already on
the drawing boards for future missions are rovers that will be
able to communicate with orbiters flying overhead a landing site.
Engineers hope to gain a much better understanding of Sojourner's
interactions in the Martian environment to make this concept a
reality.
"We are already learning a lot about the demands of driving
a rover," Wilcox said. "Right now we're developing a brand new
command sequence for the rover every day, but we're accomplishing
that with significantly fewer people than we did in the Voyager
days. In those days, it took as many as 700 people to fly the
spacecraft, whereas Sojourner only requires about seven people."
Powerful visualization tools not yet invented 20 years ago
have also advanced Sojourner's and future rovers' autonomous
capabilities. For instance, with a specially designed computer
program, stereo images taken by the lander or Sojourner are
integrated into a three-dimensional view of the rover's location.
Engineers can see the rover's tire tracks and know where it has
gone. Their view of the immediate surroundings also helps them
decide where Sojourner should go next.
"In 10 to 20 years, I'd like to see a rover that could go
anywhere on Mars," Wilcox said. "I'd like to launch a Lewis and
Clark expedition, where we would basically travel 1,000
kilometers cross-country with our rover."
Even in the shorter term, though, the strides made by
Sojourner's travels on Mars will be evident. The next rovers to
explore the red planet will have at least 1,000 times the
processing speed of Sojourner and will be able to travel much
farther much faster.
"Tomorrow's rover will be able to plan and think ahead,"
Wilcox said. "Whereas Sojourner can only plan 20 centimeters
ahead, our next rover will be able to plan 10 meters or more
ahead. And while Sojourner has no concept of what rocks look
like 20 meters away, her successor will be able to recognize
rocks that far away very easily."
#####
8/8/97 DEA
#9768