Elon Musk's Starlink in firing line over near misses

Night lights: 60 of the Starlink Internet communication satellites seen in the evening sky. Photo / Getty Images

Travelling at nearly 17,000mph, 340 miles above the Earth, a small satellite from Elon Musk's Starlink shot past another spacecraft at a distance of 160ft.

The width of a rugby pitch might seem like reasonable room for manoeuvre on Earth. But in orbit, where objects are normally separated by hundreds or thousands of miles, it was much too close for comfort

The near-miss on March 30 has led to renewed scrutiny of an avoidance system built by Musk's broadband satellite company. In 2019, Musk announced Starlink's constellation, which is set to include 12,000 satellites, would "use their thrusters to manoeuvre automatically around anything that [the US military] is tracking".

But the near-miss late last month has infuriated executives at rival companies as an example of SpaceX, which owns Starlink, putting the safety of competing projects in jeopardy. "It was very close. An unacceptable distance to OneWeb," says Chris McLaughlin, head of regulation at OneWeb, which is partly owned by the UK taxpayer.

McLaughlin says OneWeb's engineers received a red alert from the US Space Force after the company's latest launch of 36 satellites. As these ascended into orbit, they passed through a mesh of Starlink satellites orbiting at 340 miles high. "Our engineers tried to contact SpaceX via email. They didn't get a response at first. When they did, we were told not to worry. Our engineers thought they had to be kidding."

Eventually, in spite of Starlink's in-built automated avoidance system, OneWeb agreed to move its satellite slightly to avoid a crash. McLaughlin adds: "Our engineers said we were not prepared to accept for them to go fully automated. We had no idea how it worked"

It is not its first near miss. In 2019, SpaceX once again shut off its automatic avoidance system after it came close to a $560m (£406m) European Space Agency climate observation satellite, which fired its thrusters to avoid a crash. After a delayed initial response to the collision warning, Starlink put it down to a "bug" in its alert system for its satellite engineers.

Space collisions are extremely rare but not impossible. In 2009, an Iridium satellite collided with a derelict Russian probe at a speed of 26,000mph.

'Nobody knows if the AI will work'

Rivals say they have no idea what Starlink's AI will do if allowed to automatically try to avoid a collision. Starlink's satellites operate using Linux powered processors combined with GPS data and tracking data from the US military that monitors thousands of space objects of 10cm (4in) or larger. But there is little public information about how its algorithms work.

Yet just days before the March 30 incident, Starlink signed a memorandum with Nasa promising a "deeper level of co-ordination, co-operation and data sharing" between launches and Starlink's constellation. In the agreement, Nasa says that it will not attempt to move in the event of a near-miss event with Starlink under the promise Starlink will automatically get out of the way "to ensure the parties do not inadvertently manoeuvre into one another".

"The truth is nobody knows anything about this autonomous collision avoidance mechanism," says John Janka, head of regulatory affairs at Viasat. He adds it would be useless if a satellite failed for another reason. "If you are doing 70mph on a motorway and the brakes fail, what good is your system?"

Prof Marek Ziebart, of University College London, says the danger of a collision could go far beyond just the cost of replacing the satellites. He says: "If you have two objects about the size of a car converging in space, when they collide they create an uncontrollable shower of debris in orbit, like a shower of grenades. Removing that detritus is an impossible task."

'People really start talking to each other'

Ziebart says it should be possible to solve some of these issues. Modern air traffic control makes plane collisions almost unheard of, but this requires deals between companies and even international treaties between various space agencies. McLaughlin, of OneWeb, says SpaceX has since offered to open up its algorithm so the UK-backed company will have a better idea of how it works. Viasat argues SpaceX should be forced to keep to tighter orbits in space. Starlink should also reduce the failure rate of its satellites, which he says is as high as 2pc.

But disclosing its algorithm is not likely to be quite so simple if it were to a Chinese state-backed low-earth orbit constellation. SpaceX did not respond to requests for comment.

The company is currently applying for a further 2,700 extra satellites with the US Federal Communications Commission, amid vocal protest from OneWeb, Viasat and Amazon, with a decision expected to be reached in the coming days.

"This is a point for people to really start talking to each other," Ziebart says.

With tens of thousands more satellites whizzing around the Earth, the increasing space jam makes a potential crash ever more likely.

- Telegraph Media Group