Communications of the ACM

which is why that agency has been refining the technology of robotic humanoid teleoperation by minimizing the effects of signal latency, says Krueger. In October 2018, for instance, ESA astronaut Alexander Gerst aboard the ISS remotely controlled a four-wheeled, twin-armed, humanoid robot named Rollin’ Justin situated on a mocked-up Martian surface in the labs of the robot’s maker, aerospace firm DLR in Oberpfaffenhofen, Germany. Using a point-and-click tablet-based teleoperation control app developed by ESA, Gerst successfully retrieved antenna parts and replaced a burned-out computer circuit using the robot, while orbiting the Earth at 28,000 kph.

While such tests have validated ESA’s
teleoperation technology, the question
remains of how an actual space-qual-
ified robot would perform on Mars.
Rollin’ Justin was built for testing on a
smooth lab floor on Earth, with handy
QR codes on the walls telling it where
it is. When redesigned for the harsh
space environment (with less-capable
radiation-hardened electronics in-
stalled, and using image recognition
for navigation), Krueger concedes its
performance would be “downgraded.”
This illustrates the profound issues
faced by space robot designers: they
must test their designs on Earth before
deployment, but they also must ensure
doing so in a non-representative envi-
ronment does not breed false confi-
dence. With zero gravity to contend
with, plus cosmic radiation impacts
flipping bits in memory chips and mi-
croprocessors, and the extreme heat
and cold in space, there are many dif-
ferences between Earth and space en-
vironments with which to cope.

“For instance, we work against gravity here on Earth, but in space that constant force vector is not there. A robot arm’s gearbox or drive system has to counteract gravity on Earth, but not in space,” says Krueger.

Zero-gravity aircraft trips consti-
tute one option for testing in three di-
mensions. Another is an ultra-smooth
granite table on which NASA tested the
Astrobees: they placed each robot on
an “air puck” which fires CO2 down-
ward to make the Astrobee float atop
the smooth granite. “It gives you a re-
ally, really thin cushion of gas to float
on. It’s amazing; it’s very close to being
frictionless when you push it around,”
says Fong.

What is the most debilitating computer engineering challenge for space robots? “One of the big challenges we have, and will have for a very long time, is that the gap between the capabilities of space and terrestrial computers continues to actually increase,” says Fong. We may see strong progress here on Earth in self-driving cars and drones, he says, but the high-performance VLSI technology behind that does not reach the spaceflight community because it is not available in radiation-hardened form to protect it from the rigors of space.

Krueger agrees, “Space-grade processors cannot be as densely integrated as those used on Earth, and so are way slower, and that places computational limits on the image processing and machine learning techniques we can use in space applications,” he says.

For instance, says Fong, the Rad- 750, a common radiation-hardened 32-bit single-board computer for space applications, is based on the 1997 IBM/ Motorola PowerPC chip, and its top clock speed of 200MHz runs two orders of magnitude slower than the technology in our smartphones. “That limits the algorithms we can use,” Fong says.

Yet there are signs of hope. NASA’s Mars 2020 Rover, Fong says, will utilize field programmable gate arrays (FPGAs) that allow new, smarter processing logic circuitry to be configured on the fly as new algorithms are developed ( something Microsoft already does on its Azure cloud, allowing new cloud search algorithms to be uploaded without having to replace thousands of datacenter processors). The 2020 Rover also will deploy its own flying robotic assistant, called the Mars Helicopter, to map terrain on brief reconnaissance missions. Thanks to newer methods of tolerating radiation rather than blocking it completely, the Rover will use a modern Qualcomm Snapdragon processor.

Not all robotic solutions are going
to be popular, however. Cosmic radia-
tion, solar flares, space junk, and the
risk of spacesuit depressurization are
some of the things that make EVAs
(extravehicular activity, or spacewalks)
hazardous. Krueger recently found
that the ESA’s proposed robotic an-
swer to these risks did not go down too
well. The agency’s forthcoming Space
Exoskeleton Controller (SPOC) will be
worn by an ISS crewmember inside the
station to control a humanoid robot
out in space, perhaps swapping out
ISS batteries or maintaining its solar
arrays. Russia’s space agency, Roscos-
mos, has similar plans.
However, Krueger says, “When I told
an astronaut that with SPOC he would
not need an EVA, he was not excited
about it at all. In fact, he asked that we
don’t take the EVAs away from them.”

Just like many people here on Earth who don’t want machines taking their jobs, even those with the Right Stuff don’t want to lose the most talismanic job in spaceflight to a robot.

Further Reading

Apollo 17 Preliminary Science Report, National Aeronautics and Space Administration, 1973 https://www.hq.nasa.gov/alsj/a17/as17psr.pdf Astrobee Research Publications, National Aeronautics and Space Administration Ames Research Center, 2019 https://www.nasa.gov/content/ research-publications-0

Bualat, M., Smith, T., Smith, E., Provencher, C., Fong. T., Smith, E.E., Wheeler, D. W., et al. Astrobee: A New Tool for ISS Operations, Proceedings of AIAA SpaceOps, Marseille, France. https://arc.aiaa.org/doi/pdf/10.2514/

6.2018-2517

Smith, T., Barlow, J., Bualat, M., Fong, T., Provencher, C., Sanchez, H., Smith, E., et al. Astrobee: A New Platform For Free-Flying Robotics on the International Space Station, Proceedings of 13th International Symposium on Artificial Intelligence, Robotics, and Automation in Space https://ntrs.nasa.gov/archive/nasa/casi.ntrs. nasa.gov/20160007769.pdf

Coltin, B., Fusco, J., Moratto, Z., Alexandrov, O., and Nakamura, R. Localization from Visual Landmarks on a Free-flying Robot, Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems https://ntrs.nasa.gov/search. jsp?R=20160012286

Hyatt, P., Kraus, D., Sherrod, V,.

Rupert, L., Day, N., and Killpack, M.D. Configuration Estimation for Accurate Position Control of Large-Scale Soft Robots, IEEE/ASME Transactions on Mechatronics. https://ieeexplore.ieee.org/ document/8510913