From free-flying droids to humanoids, from crawlers to inflatable torsos, space robots of myriad types are now being considered for missions in low Earth orbit, on interplanetary spacecraft, and on other worlds.
It might sound like a prop list from a Star Wars movie, but space agencies and their contractors are developing a panoply of robotic assistants with a serious aim in mind: to boost the productivity and safety of astronauts.
The idea behind robot assistants is multifaceted: one aim is to offload time-consuming repetitive tasks like space station cleaning and inventory making from crew members to free-flying or humanoid robots. Ground robots controlled from, say, spacecraft orbiting the Moon or Mars could construct human habitats ahead of a landing, or perform reconnaissance ahead of human exploration missions.
In addition, the dangers of space junk, as well as the risks of cosmic radiation exposure and depressurization during spacewalks, could be quelled if a humanoid robot does the work, controlled by an exoskeleton-wearing astronaut more safely ensconced within a spacecraft.
At the heart of this drive for robotic assistance is the fact that robots need few of the quickly depleted resources astronauts burn up so very readily—principally oxygen, water, and power. The hope is that by taking on the drudge work of space, robots should be able to save spacefarers time, giving them the chance to focus on the parts of missions requiring human intelligence.
What kind of missions are being envisioned for robots? Terry Fong, director of the Intelligent Robotics Group at the NASA Ames Research Center near San Francisco, sees a signature example in a startling discovery made on the lunar surface in December 1972. "On Apollo 17's second extravehicular activity (EVA), astronaut Jack Schmitt found orange volcanic glass at the Shorty crater. It was certainly one of, if not the most interesting, discoveries made by Apollo 17," says Fong.
"But when they found it, they were three-quarters of the way through their excursion, and they had limited time because of their oxygen and power levels, and they had to get back to the lunar module," he says. As a result, they had to rush their investigation.
However, if the Apollo 17 crew had been able to deploy some robotic reconnaissance to scout the area ahead of time, says Fong, "They could have changed the timing, they could have changed the route, they could have gone there first and spent more time there," he says.
This should inform similar future missions, he says. "This would allow the humans to be a lot more efficient when they do go and do their work."
NASA is not alone here. At the European Space Research and Technology Center (ESTEC), the research center for the European Space Agency (ESA) in The Netherlands, engineers are planning to develop ground robot assistants that could be teleoperated by astronauts orbiting a moon or planet or based in a space station like the ISS, or NASA's planned Lunar Orbital Platform-Gateway. Such robots could perform the kind of reconnaissance Fong suggests, but equally, says Thomas Krueger, a robotics systems and software engineer in ESTEC's Human-Robot Interaction Lab, they could be used to build and maintain infrastructure for a planetary habitat.
For the moment, however, such planetary surface robotic assistants are research projects, and the drive for greater robotically fueled efficiency in crewed spaceflight has begun at a much more modest level. It kicked off in May 2006 with the introduction to the International Space Station of three free-flying robot assistants labeled Synchronized Position Hold, Engage, Reorient, Experimental Satellites (SPHERES).
The SPHERES are actually 18-sided polyhedrons just 21-cm across and with a mass of 4kg (8.8 lbs.). Functionally, at least, they resemble Luke Sky-walker's light sabre training droid in the first Star Wars movie. Developed by the Space Systems Laboratory of the Massachusetts Institute of Technology (MIT), they use tiny puffs of CO2 gas to dart around the ISS. The SPHERES have ultrasound sensors that receive inaudible chirps from ultrasound beacons placed around any module they work in, allowing the robots to compute their position in three-dimensional space.
In addition to carrying gadgets such as smartphones and cameras for astronaut experiments, the SPHERES are also a cheap satellite research platform allowing swarm flying and rendezvous and docking experiments to be performed inside the ISS. Crucially, the robots have provided the testbed for a more-powerful successor: another free-flying astronaut assistant robot called the Astrobee, developed by Fong's team at NASA Ames.
Having arrived on the ISS in April, the three Astrobees (dubbed Honey Bee, Bumble Bee, and Queen Bee) are cubes with 30cm-long sides. These droids' propulsion in zero gravity comes from electric fans, while their navigation is based on image recognition AI. "The Astrobees have louvred vanes that direct the air from the fans and so, in some senses, are more like drones," says Fong.
The Astrobees' image-based navigation, which uses cameras to create a dense three-dimensional feature map of space station module interiors, frees the ISS crew from having to set up indoor GPS beacons, as they did with SPHERES. "This is a critical feature as it allows Astrobees to be very flexible. It makes the whole world of the ISS its navigation system," says Fong.
What can Astrobees do? They each sport three payload bays on which useful tools can be plugged in, such as a robot arm, a novel sensing system, or a QR/barcode reader for inventory taking.
Fong says NASA has "explicitly gone out of their way" to allow third-party developers to invent for Astrobee, as the data, mechanical, and electrical interfaces are openly published and the software is open source and based on the Robot Operating System (ROS). "It's an extensible open platform," he says.
Designers test their space robots on Earth before deployment, but must be sure that doing so in a non-representative environment does not breed false confidence.
While free flyers have their place on many tasks, however, in a space station replete with humans, some tasks will need human-shaped dexterous robots to work alongside people. However, this is a work in progress, and no humanoid robot is currently in regular operation on a spacecraft.
Way ahead in the research stakes, however, is the NASA Robonaut program that began in 1996 at the NASA Johnson Space Center in Houston, TX. NASA Johnson engineers introduced Robonaut 1 to the world in 2002: a humanoid torso with dexterous, gripping hands and soft, compliant arms that cannot hurt humans, along with a raft of sensors. That initial testbed did not fly to the ISS, but in 2011, Robonaut 2 was flown to the ISS aboard Space Shuttle Discovery, and was used in multiple dexterous robot tests there.
NASA says Robonaut 2 was "good at cleaning the many handrails inside the station," allowing astronauts to focus on key science and repair work. The droid, controlled by an astronaut in a VR helmet, also proved dexterous enough to become proficient at flipping switches and pushing buttons.
Some of those tests were conducted by Fong and his colleagues at NASA Ames. "Robonaut 2 was on the space station for several years and we learned quite a bit about it. And then it had some issues with its power system and had to be brought back down to Earth," he says.
Says ESA's Krueger, "NASA had the Robonaut on board the station and it's a nice device but it is not really a system you can tell: 'hey, go to that module and give me that box'. We are not there yet. But it might be possible that we can control that robot better from the ground via remote operation."
More recently, the NASA Johnson Space Center has developed a newer humanoid space robot called R5, perhaps better known as Valkyrie, which builds on the experience with Robonaut 2. With new electronics, actuators, and sensors, it is now being further engineered for "break-throughs in humanoid control, motion planning, and perception" at the University of Edinburgh in the U.K., according to that institution's School of Informatics, which is working on the 125-kg, 1.8-meter-tall robot in a joint research program with nearby Heriot-Watt University.
Still another innovation in space-based humanoid assistants, and one receiving advanced concept seed funding from NASA, is the inflatable upper-torso robot being developed by Marc Killpack, a professor of mechanical engineering, and his colleagues at Brigham Young University. This robot addresses two of spaceflight's enduring problems: payload launch mass and storage space.
A robot whose torso and limbs can be inflated to a useful stiffness would have far less mass than a similarly sized robot with metallic arms and chest. When not in use, the droid can deflate its torso and limbs to take up less storage space.
Inflatable robots are not as crazy as they might sound; Fong's team previously worked with R&D contractor Otherlab on such a concept. "I'm actually quite excited about them," says Fong. "They have a much higher strength-to-weight ratio than traditionally designed rigid assemblies. That's because of the power of using fluid pressure. And the fabric-based manipulators, some of them can be much more intrinsically safe in terms of impact because they are very low inertia."
ESA's main focus is on controlling planetary surface robots from orbit, 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-qualified 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 installed, 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 environment does not breed false confidence. With zero gravity to contend with, plus cosmic radiation impacts flipping bits in memory chips and microprocessors, and the extreme heat and cold in space, there are many differences between Earth and space environments 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 constitute one option for testing in three dimensions. Another is an ultra-smooth granite table on which NASA tested the Astrobees: they placed each robot on an "air puck" which fires CO2 downward to make the Astrobee float atop the smooth granite. "It gives you a really, 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 radiation, 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 answer 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, Roscosmos, 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.
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
©2019 ACM 0001-0782/19/12
Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and full citation on the first page. Copyright for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers, or to redistribute to lists, requires prior specific permission and/or fee. Request permission to publish from [email protected] or fax (212) 869-0481.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2019 ACM, Inc.
No entries found