To build the best bots, NASA happily looks to others here on Earth

NEW ORLEANS—“People who’ve met me keep asking ‘Hey, why is NASA here? You’re not a startup, not an investor,’” Terry Fong recalls. The lead for NASA’s Intelligent Robotics Group took the stage at the recent 2018 Collision Conference in between people preaching their coffee business models and others promoting everything from cloud services to Vespas. Fong’s organization may obviously be different, but he absolutely had his recruitment pitch as ready as the next attendee. Industries everywhere—NASA very much included—want to better leverage autonomous and intelligent systems to automate tasks and make new initiatives possible. So this senior scientist for autonomous systems found himself on the showroom floor in search of potential collaborators, just like everyone else.

“Tech development doesn’t exist in a bubble, and NASA doesn’t do everything end to end,” Fong tells Ars. “We exist in an ecosystem. There are things we want to pull in, whether from a startup or a large corporation, and there are things we’re trying to push out to industry. For me, it’s important to understand what NASA can reuse and not make ourselves, or what we can work with people to adapt in ways that are useful for our missions.”

Intelligent robotics may feel new to most of us, but NASA has naturally been experimenting within this space for decades. For example, its most famous “recent” effort—a humanoid housed on the International Space Station called Robonaut—is part of a longterm R&D project that has been going on for roughly 18 years, according to Fong. Robonaut's original purpose was to alleviate a lot of the repetitive, manual tasks astronauts must complete so they are freed up for other initiatives, but it ran into some public issues within the last few years. Offline since 2014, NASA finally called it in for repairs this spring. And to help the bot move on to its 3.0 phase, NASA has been opening things up to external ideas.

“The current plan is to bring the unit on ISS back down to Earth, fix a few things, probably make some upgrades as well, and then some time in the next 1-2 years to send it back up to ISS,” Fong says. Old hardware and software needs occasional refreshing, of course. So over the past few years, NASA has been running the Space Robotics Challenge (similar to the DARPA one, but for space!) to encourage competition in developing new algorithms and software that can help upgrade the agency’s current robots. “There’s a lot more development, testing, and demonstrations that can be done on ISS,” Fong says of Robonaut. “So one target of [the Robotics Challenge] has, of course, been Robonaut—can you develop software that can be upgraded onto a system like Robonaut?”

New bots, same teamwork

NASA’s push for robotic collaborators extends to its newer intelligent systems, too. Fong’s big project of late has been something called Astrobee, a free-flying trio of autonomous robots that will travel to the ISS in November. Astrobee may not look like the pop culture epitome of a robot in the same way as Robonaut, but it will potentially better serve a similar purpose. Simple tasks that used to occupy astronaut time—monitoring air quality, light levels, and sound, or scanning RFID-tagged inventory of drawers and hatchways—will suddenly become automated.

“Currently, astronauts have to check and scan barcodes or read off numbers; it’s a very manual process. Now we’re starting RFID tagging things going up, so Astrobee can fly around and do inventory instead,” Fong says. “It sounds like a tedious thing, but knowing where things are is critically important. Astronauts have really packed days. When they start an activity, we say, ‘For this, you need this tool, which is in this drawer on this module.’ If they go to the drawer and it’s not there, that messes up the rest of the day. So having the ability to verify where things are and check ahead of time—that’s a great use of a robot.”

Astrobee did not materialize out of thin air, and it did not materialize solely out of NASA developments. Fong points to an MIT creation called SPHERES—free-floating brightly colored bots from about 12 years back—as an origin point. Roughly the size of a volleyball, each of these bots initially relied on an end-of-life TI-DSP. “You could only find spare parts on eBay, basically,” says Fong.

So a few years ago, NASA decided to try upgrading them. “We modified an off-the-shelf Android phone and set that up as a brain and sensor upgrade—cell phones have good processors, high-bandwidth communications, touch screens, cameras, IMUs—it was a big brain upgrade,” Fong says. These tweaks created what NASA soon called Smart SPHERES, and those devices led to a number of experiments and R&D on ISS. Essentially, that laid the ground work for the design and development of Astrobee. As Fong puts it, “It’s been this long path from SPHERES to Smart SPHERES to Astrobee.”

Even once Astrobee materialized as an initiative, NASA’s robotics group did not silo itself off. The operating system Astrobee runs, ROS (robot operating system), comes from the Open Source Robotic Foundation, for instance.

“[ROS] is designed so people can develop and make use of that. So Astrobee on ISS isn’t just a NASA thing; it’s a community resource—a research platform,” Fong says. “If you’re familiar with ROS, you can write software for Astrobee. If you want to do an experiment on ISS, you can run an experiment on Astrobee. If you’re doing a robotics contest—like Zero Robotics, in the same tradition as Botball or Vex—people develop sims and get to run their software on the ISS. The main way we do that now is through ROS, and we didn’t develop it, but now we help develop it.”

Bots and humans in future space harmony

This communal development ethos seems to mirror Fong’s and the Intelligent Robotics Group’s overall attitude toward increasingly autonomous bots themselves: work done together is simply better than work done alone.

“Our robots today, try as they might, can’t do everything,” Fong told the crowd during his Collision presentation. “So how can we merge humans with autonomous robots and free humans up to do more or work better together?”

One of the tasks Fong envisions this bot-astronaut partnership really thriving within is planetary surface missions. An astronaut would be able to stay in orbit, dispatch an autonomous bot to a planet, and then communicate with the device remotely as information and situation dictates. Fong emphasizes this isn’t simple joysticking—“It’s the way you and I would work together; I’m not joysticking; we’re peers or partners”—which should lead to higher-level, more abstract work capabilities.

“It’s exactly the thing we saw in Avatar the movie—we’re not trying to immerse a person inside, but we want humans in a spacecraft operating an interface to interact and operate a robot on a planetary surface,” Fong noted during his presentation. Such a setup has already been tested multiple times on the ISS, in fact.

NASA’s robotic initiatives will only increase from here, both on Earth (like its notable autonomous vehicle partnership with Nissan) and off it. And like everyone else, that means NASA now has some philosophical robotic questions in the back of its mind alongside all the technical ones. As autonomous technology becomes more advanced and pervasive, how should it be used by an agency at the very forefront of technology and exploration?

Fong doesn’t necessarily leave conferences like Collision with answers in the same way he does business cards. But just as the agency constantly develops and seeks to improve its technical capabilities, NASA seems to be already thinking through the philosophical challenges, too.

“To the extent that bots become more autonomous (or bigger systems become autonomous), how do we trust them to do what we want them to do? Are they going to work within the bounds we created? What if those bounds are fuzzy?” Fong says. “A lot of what NASA does in space involves going to places that are unknown, uncertain, undiscovered. By definition, we don’t know what to fully expect. So it’s OK not to just color inside the lines; in fact, the lines don’t exist. That means if we’re looking at how to treat a fully autonomous robot—especially if it works with humans, with astronauts—there are questions we’ll have to ask if we allow that system to be more independent, self-reliant, and able to make its own decisions.”

Listing image by NASA