Fully automated satellite-assembly lines? Not quite yet

While robots began assisting and replacing assembly line workers in automobile and airplane factories years ago, humans still reign supreme in satellite manufacturing. But that’s slowly starting to change.

“For routine assembly process, we didn’t see the return on that given the quantity of operations we were performing.…It wasn’t high enough to really drive the need for too many robotics” — Brian Holz, OneWeb Satellites chief executive

In contrast to the millions of cars and thousands of airplanes produced annually, satellites — and geostationary telecommunications satellites in particular —  are produced in much lower numbers. In a good year, the world’s satellite manufacturers might book a combined commercial 25 orders. That low volume limits the efficiency gained from industrial robots, at least on the ground.

“In terms of today, the uses for us are somewhat minimal,” said Tom Wilson, vice president of Orbital ATK’s Space Systems Group.

Orbital ATK mostly uses robotics to fit electronic boards with computer chips, Wilson said, which isn’t new for space or most other industries.

Like Orbital ATK, Space Systems Loral sees limited application for robots, despite averaging more telecom satellites per year.

“In our relatively low-volume, high-mix environment, the standard industrial robot doesn’t do you much good,” echoed Paul Estey, SSL’s chief operating officer. “You don’t have very many applications of it.”

Even OneWeb’s mega-constellation, whose first satellites are just now being built by the OneWeb-Airbus joint venture OneWeb Satellites in Toulouse, France, doesn’t provide the scale needed to justify the upfront expense of automating assembly.

“For routine assembly process, we didn’t see the return on that given the quantity of operations we were performing,” said Brian Holz, OneWeb Satellites chief executive. “It wasn’t high enough to really drive the need for too many robotics.”

To build OneWeb’s order of 900 identical 150-kilogram satellites, Holz said the company is focusing more on the automation of repeatable processes, like placing cells onto solar arrays. Automating some manual procedures reduces the amount of touch labor involved and consequently, the risk of human error, but doesn’t mean a full-blown takeover by robotic builders.

“You’d get into more robotics maybe if you had another order of magnitude in terms of quantity of objects to build. We are not building one, but we’re not building a million either,” he said.

Making the case for robotics 

These barriers to mainstream robotics don’t mean the industry isn’t investing in the technology, however. The European Space Agency has expressed great hope in bringing robotic capabilities to the clean rooms of satellite manufacturers.

“Within Europe, we see that the use of robots and automation is skyrocketing, especially in aeronautics,” said Paul  Robert Nugteren, ESA’s technology and strategy coordinator for the agency’s

Directorate of Telecommunications & Integrated Applications. “Robot suppliers specifically cater for the specific needs of these industries. The step to spacecraft manufacturing is happening now. Collaborative robots are good candidates to make their first appearance in spacecraft manufacturing. European industries are developing their capabilities in this area.”

Thales Alenia Space used a robot called Saphir to automate the installation of inserts for the spacecraft-bus panels of Bangladesh’s Bangabandhu telecom satellite. The French and Italian satellite manufacturer anticipates Saphir will slash the time for bonding an estimated 3,500 inserts per panel from three weeks to one week, and the number of employees needed for the task from two to one.

Lockheed Martin uses robotic arms to take some of the tedium out of solar array assembly. A $350 million satellite factory the company expects to open by 2020 near Denver could potentially employ robotic helpers that deliver tools to the humans building the satellites. “We’re also developing how larger robotic arms can 3-D print structures with both additive and subtractive capabilities at the same station,” Lockheed Martin spokesman Mark Lewis said. “[O]ne arm deposits material while the other can shave off and smooth surfaces so it all can be done in one station, verses two different processes currently.”

SSL, Estey said, has developed “an industrial robot-like device” to assist with building satellite subsystems.

“We are beginning to use that now,” he said. “If that works out as we hope, we are going to expand the use of that type of a robot elsewhere in the factory.”

Frédéric Teston, head of ESA’s Systems Department Directorate of Technology, Engineering and Quality, said the agency expects a lot of automation from the aviation and automotive industries will “spin-in” over time to the satellite industry. Despite the challenges, constellations of tens, hundreds, or thousands of satellites are viewed by the agency as “the first candidates for extensive automation and use of robotics.”

“Experience build-up in constellations will then make its way to lower-volume production cases,” he said.

Downstream currents

Suppliers of satellite parts and subsystems are also guided by volume in determining whether or not robotics have any meaningful application.

Nuvotronics, a Durham, North Carolina-headquartered provider of amplifiers, space-qualified phased-array antennas, filters, diplexers, and other components, has been rapidly introducing robotics to create more reliable products, circumventing human error, said President David Sherrer.

“Over the last two years, we have moved most of the key steps in our manufacturing to use robotic handlers,” he said. “Human operators do not need to touch the work in process except to move batches from operation to operation.”

Sherrer said robotic handlers operate without human intervention for several of Nuvotronic’s machining processes, such as metal filling, removing disposable design molds from parts, and assembly of smaller parts into hardware.

“What used to take 10 minutes of human touch-time to fixture a substrate to grow metal into a layer is now done hands-off, including the pre-cleaning and post wash and dry steps,” he said. “The ability for these machines to run 24/7 without making errors translates into reduced cost and higher yield.”

That process used to take 10 minutes; it now takes about 30 seconds, he said.

In contrast, Melbourne, Florida-based Harris Corp.’s Space and Intelligence Systems division, a provider of large, unfurlable satellite antennas, builds only a couple of such antennas every year. Over the past 50 years, the company has produced around 80 antenna reflectors, of which around 50 are unfurlable.

Tom Campbell, Harris Corp.’s director of program management for antennas and structures, said a typical antenna project takes 14 to 30 months to complete. Half of that time is design, he said, while the other half is the actual build process.

Campbell said automation has led to some decrease in the number of humans working on a project, but added that Harris has “been able to grow the business to take up the need for additional people.”

“It’s important to have a critical mass so that when customers call on us, we are ready to go,” he said.

Harris is using a lot of automation on the design side by adding antenna modeling code to computer-aided-design tools in order to better define final products, Campbell said. The company is also “very bullish” on 3-D printing to create structures “that are maybe an order of magnitude more optimized for the purpose at hand,” he said.

Robotics have limited use so far, but have found some applications  in areas such as carbon-tube wrapping, he said, and in test equipment. Scale remains the biggest challenge.

“The instances of true high-volume manufacturing are still few and far between,” he said. “I think the industrial engineers are still favoring lean processes in most cases as we still tend to build most satellites one at a time.”

Robotics in space

Both Orbital ATK and SSL are developing robotic servicers for in-orbit repair and life extension. Toward the end of 2018, Wilson said, Orbital ATK subsidiary Space Logistics plans to launch its first mission-extension vehicle, or MEV-1, which will serve Intelsat for an initial five years. SSL parent company MDA Corp. is creating a company called Space Infrastructure Services, or SIS, that expects to launch a robotic servicer in 2021 based on a project with the U.S. Defense Advanced Research Projects Agency.

The two companies have different visions of how to approach in-orbit servicing as a business, and have quarreled in legal disputes over the technology in recent years, but both envision their products as kicking off what would rightly be called a paradigm shift if fully realized.

“Things like deep space gateways or large telecommunications satellites that you couldn’t fit in a launch vehicle fairing because of their size and the amount of power, [that’s what] we are talking about,” said Orbital ATK’s Wilson. “The idea is you can have a very-high-power system that can last for multiple decades that can be assembled on orbit with some of the technology we are developing. It’s more of a future look.”

Some of that critical technology is being produced through a NASA Tipping Point contract for the Commercial Infrastructure for Robotic Assembly and Services program, which Orbital ATK has been working on since September 2016.

Joe Anderson, leader of business development for Orbital ATK’s Advanced Programs Group, said NASA extended the project through to a ground demonstration in fall 2018, which will practice, among other things, how to add, remove and relocate structures such as solar panels.

“We are hopeful that next year NASA would come out with their phase two, which would be [ a Request for Proposal] for an in-orbit demo,” he said. “We are looking forward to that. NASA seems to be very interested.”

In February, DARPA picked SSL to build the platform for the Robotic Servicing of Geosynchronous Satellites, which SSL’s  Space Infrastructure Services subsidiary will market commercially after completing demonstrations for the agency. SSL also has a NASA Tipping Point contract for a program called Dragonfly meant to advance technology for the in-space assembly and repair of satellite antennas.

“Certainly in 10 years we are going to be using robotic assembly in orbit,” SSL’s Estey said. “The persistent platform is a good example where you’ve got modules you launch separately but are put together on orbit. That definitely in our opinion is going to happen.”

The persistent platform is a concept where a satellite bus stays in orbit indefinitely, and is updated using new payloads and technologies launched as parts. Robotic servicers would then plug in those payloads and remove broken or obsolete components to “refresh” the platform as needed.

“The application of robotics is much more applicable in the on-orbit assembly world and in [satellite] servicing … once we get to being able to be very capable of that, we see a complete change in the satellite architecture,” said Estey.

Persistent platforms could grow to be substantially larger than what can fit in a rocket’s payload fairing, enabling gigantic satellites or other spacecraft to be constructed on orbit. SSL, 3-D printing startup Made In Space, and Orbital ATK are all working on robotic in-space manufacturing systems that could enable such structures.

Anderson said Orbital ATK is evaluating a structure it calls the “GEO tower,” which would be a “multiple-decade-type application that would provide all of the backbone for communications payloads… just like a cell tower today.”

Multiple telecom operators could install payloads on the tower, and swap those payloads based on changes in technology and demand. GEO towers could also be used for Earth observation, assembling in orbit an aperture large enough to provide the resolution needed from the far away vantage point of GEO.

Wilson said Orbital ATK is still developing the technology and possible business plans for GEO towers, which could then be maintained using robotic servicing spacecraft. SN