Three rules for building a megaconstellation
This article originally appeared in the June 10, 2019 issue of SpaceNews magazine.
Six years ago, Planet began launching smallsats for a constellation that today numbers 140, making it the largest commercial satellite system in the world.
By the end of this year, Planet may no longer hold that title. OneWeb and SpaceX, which began deploying broadband satellite constellations in low Earth orbit this year, both stand to surpass Planet’s total by the end of 2019.
OneWeb, which in February launched the first six of at least 648 planned satellites, aims to have 150 satellites in orbit by year’s end by launching 35 satellites at a time when its launch campaign resumes in late summer or early fall.
SpaceX, which launched its first 60 Starlink satellites May 23 and has two to six more launches on tap this year, expects to soon be launching 1,000 or more satellites annually as it seeks to deploy nearly 12,000 broadband satellites.
“In a year and a half, maybe two years, if things go well, SpaceX will probably have more satellites in orbit than all other satellites combined — a majority of the satellites in orbit will be SpaceX,” SpaceX founder Elon Musk said last month.
Planet’s constellation, comprised mostly of camera-equipped cubesats, differs from SpaceX’s and OneWeb’s fledgling constellations, which will use bigger satellites for broadband connectivity.
But OneWeb, Plane and SpaceX claim similar priorities when it comes to building satellites in far greater numbers than anyone has done before.
RULE NO. 1: BUILD NEW TECH FAST
Planet’s motivation for rapid satellite manufacturing stems from its origin as a startup beholden to the whims of launch providers who set launch dates based on bigger, primary customers and not secondary cubesat payloads, according to Chester Gillmore, Planet’s vice president of spacecraft development and manufacturing.
“We knew we had to very, very quickly change and make things adaptable,” he said during a May 28 panel discussion on agile production and launch hosted by The Aerospace Corporation in Merritt Island, Florida, not far from where OneWeb is building its satellites. “We had to adapt because we had these set launch dates that we didn’t control.”
In September, Planet opened a manufacturing facility in San Francisco that can produce up to 40 satellites a week. Gillmore said Planet doesn’t face the same launch pressures it did early on, but learned to value having the ability to quickly build satellites regardless.
“We’ve overbuilt capacity and our capabilities to such a great extent not because we are going to build thousands of one thing, but because we want to be able to build at the rate of thousands to build many, many smaller batches of 20 to 25,” he said. “Every time we go into a production cycle, we are actually building new builds.”
OneWeb and SpaceX, in contrast, aim to build thousands of satellites — up to 2,000 for OneWeb and up to 12,000 for SpaceX.
David Goldman, SpaceX’s director of satellite policy, said the company is taking lessons learned building rockets and applying those to its Starlink constellation.
“Rapid iteration is the DNA of the company… a lot of that has to do with [the fact] that we design and manufacture most of the components ourselves,” he said. It’s very integrated. If the technicians who are putting things together find something that can be improved, they can go straight to the engineer who designed it and that can be fixed on the fly. That’s how you can iterate as quickly as we can do today.”
OneWeb is relying on a large base of suppliers for its constellation, including Ruag, Teledyne and Sodern, but is also integrally involved in the manufacturing of its spacecraft through OneWeb Satellites, a 50-50 joint venture it formed with Airbus Defence and Space.
Chris Winslett, OneWeb Satellites program director, said the joint venture’s new $85 million factory Florida’s Exploration Park is set up to build two satellites a day when it starts full-scale operations this year.
RULE NO. 2: AUTOMATE SELECTIVELY
Building satellites by the dozens, hundreds, or thousands still requires lots of human labor, but manufacturers say they have found places where automation makes sense.
SpaceX is uploading tracking data for satellites and space debris from the U.S. Air Force’s Combined Space Operations Center and other sources, Goldman said, so Starlink satellites can autonomously fly around hazards — not unlike the self-driving electric cars Musk is building at Tesla.
Winslett said OneWeb plans automated processes for flying its satellites, since scaling up the number of people who fly a few geostationary satellites to manage a constellation of hundreds would require “a small village.”
Winslett said OneWeb views automation as most useful in test campaigns and for spacecraft flight operations. OneWeb wants automation to reduce human-induced errors, he said, but not to replace human workers entirely.
“We consciously did not design our factory such that it was just a bunch of robots putting satellites together,” he said. “We did want the human person there to make the judgments.”
Winslett said an advantage of building a large constellation is that the spacecraft, as they enter operation, provide enough collective telemetry to do “big data analytics” to glean lessons that can then be incorporated into production.
“You are actually developing a statistically meaningful number very quickly,” he said. “Analyzing the data, looking for trends, looking for areas where we can make adjustments, either eliminating tests or growing tests — those are areas where automization is more prevalent for us.”
Gillmore said Planet ultimately strives for “as little physical automation as we can,” and to introduce any automation “as late as possible.”
Relying on automation too early risks cementing a flawed procedure into production that could be very difficult to undo, he said.
RULE NO. 3: LEAVE ROOM FOR FAILURE
Megaconstellation builders don’t want mistakes with their satellites, but said they have learned not to expect perfection, either.
Goldman said SpaceX sought with its first big Starlink launch to establish that failures with something so new shouldn’t come as a surprise.
“We are not going to have another launch until we watch these for a little while to see what works and what doesn’t, and then we can build that into the next models to make sure that they are better,” he said. “I think the best thing you can do is just be honest with the public and tell everybody what you are up to, and don’t say you are going to have 100 percent reliability if you can’t accomplish that.”
“You have to be willing to set a culture where you are willing to fail early and learn from it,” Winslett added, highlighting SpaceX as an example. “We have to lay those expectations out there and let people understand that it’s OK to fail. It’s OK to have an issue or a mistake, learn from it, correct it and move on.”
Planet’s philosophy is that failures should happen “as quickly as possible to find out what works and what doesn’t work,” Gillmore said. He draws a line, though, between planning for failure and pursuing it.
“Failure is never the goal,” he said. “When you spend a lot of money designing something, launch it into space, and it doesn’t work, it’s very bad. When we say be somewhat tolerant to failure, the idea is not that we are expecting it, it’s more about taking a different approach to mission assurance than was previously done.”
That approach includes focusing hard on getting critical systems like cameras and antennas right, Gillmore said.
“We don’t have to fully ground qualify every single subsystem,” he said. “We can ground test the things that are required and safety critical.”
Gillmore said that for every 20 spacecraft Planet launches, it budgets for two to be technology demonstrators. That way, the company can evaluate new gear in space rather than “overtest” each spacecraft on the ground.
Gillmore said he is looking forward to seeing launch costs continue to decline so Planet can prototype more technology for future generations of satellites.