WAILEA, Hawaii — As both the number of satellites and the number of potential collisions grow, government and industry officials say they need to improve the ways satellite operators coordinate maneuvers.
The challenges of such coordination were on display earlier this month, when the European Space Agency announced it shifted the orbit of its Aeolus spacecraft to avoid a close approach to a SpaceX Starlink satellite. ESA made the decision after SpaceX didn’t respond to updated assessments that showed a higher probability of a collision, which the company blamed on a glitch in a paging system.
The event showed how much effort, and potential for miscues, goes into coordinating such maneuvers, something that experts say won’t scale as companies deploy megaconstellations of satellites in the coming years.
“It made us ask whether emails or late-night calls are the most efficient coordination mechanism, and would be advisable in a scenario with thousands more operational satellites,” said Francesca Letizia, an engineer in ESA’s Space Debris Office, during a Sept. 19 presentation at the Advanced Maui Optical and Space Surveillance Technologies, or AMOS, conference here.
She called for more data sharing about satellite positions among operators. Another step would be greater use of automation, including technologies like machine learning, to analyze orbital data and identify which potential close approaches warrant maneuvers. ESA plans to start a global competition on this issue in the near future, she said, giving researchers access to historical data from ESA missions on conjunctions for analysis.
During a later conference panel, SpaceX agreed that current approaches are not scalable. “We agree with ESA’s conclusions that it’s very difficult to be able to do this on an individual email basis,” said David Goldstein, director of special programs at SpaceX. “The personal relationships are super important, but we have to figure out some automated ways to do that.”
Goldstein said later that the lapse in communications was caused by a confluence of factors, not just a software bug. That included the timing of the event, over a three-day holiday weekend, as well as the fact that the Starlink satellite in question, in the process of being deliberately deorbited as a test, did not have activated an automated system intended to perform collision avoidance maneuvers. “We learned a lot from that,” he said.
The operational Starlink satellites do have the autonomous system in place. Goldstein said that, as of last week, those satellites had performed 21 collision avoidance maneuvers autonomously. All of them involved objects on track to make close approaches that were not maneuverable.
The Starlink satellites get conjunction data messages about potential close approaches uploaded to them. “We built the smarts into them to be able to make some good decisions” based on that data and the satellites’ own knowledge of their positions, he said, although there are people involved to doublecheck those analyses. “So far the satellites have made exquisite decisions with the information they have.”
Even with automation, he said there will still be some discussion when a close approach involves two operational, maneuverable satellites, specifically deciding which of the two satellites should maneuver. “Is it the satellite that got there first? Is it the satellite that has the most delta V available, or is it the satellite that can do the burn later?”