Navigation by Wreckage

NASA just flew one that keeps it.

FALCON — Fast Autonomous Lost-in-space Catalog-based Optical Navigation, an acronym doing visible violence to itself — ran aboard the Starling cubesat swarm, a NASA Ames demonstration mission launched in 2023, with flight software from EraDrive, a Stanford startup. Over three days it used its star tracker to spot satellites and orbital debris, matched them against an onboard catalogue of roughly twenty thousand objects, and worked out its own orbit from what it saw. No GPS. And then it did the part that should get your attention: for more than two hundred of those objects, it produced better orbital predictions than the ones the ground had uploaded.

It navigated by the garbage, and along the way it improved the garbage.

The stated motivation is a chokepoint. GPS is about thirty-one satellites, owned and operated by one country's military, and the signal that reaches you from twenty thousand kilometres up is faint enough to jam with equipment you can buy. It also thins fast above the constellation itself, which is a problem if you are going anywhere interesting. A spacecraft that can fix its own position optically is a spacecraft that does not have to ask permission, or hope.

But look at what it is asking instead. The reference frame FALCON used is the accumulated wreckage of everything we have ever put up there and not brought back. Roughly forty-six thousand tracked objects larger than ten centimetres. Over a million between one and ten. Something north of a hundred million fragments smaller than a centimetre, which nobody can see individually and which are modelled statistically the way you model weather. Most of it is not satellites. Most of it is pieces of satellites.

It is tempting to say that the property making that field usable as a coordinate system is exactly the property that makes it lethal. Density. It is not quite true, and the gap is worth holding open. FALCON needs objects that are large, bright, optically resolvable from a cubesat, and already catalogued — twenty thousand of them. Kessler syndrome runs on a different population. Donald Kessler's 1978 paper described the failure mode: enough objects in enough orbits and collisions start generating the fragments that cause the next collisions, until a shell of low Earth orbit becomes unusable for decades. That cascade is driven by the centimetre-scale shrapnel nobody tracks individually — the part of the field the navigator cannot see. You could remove most of the collision hazard tomorrow and cost FALCON almost nothing.

They are correlated, not identical; the same sixty years produced both. What is true is narrower and still strange enough: a field dense enough to steer by is a field dense enough to worry about, even though the things you steer by and the things that will kill you are mostly not the same things.

So this is not a story about trash becoming treasure. Nothing has been redeemed. Every object FALCON used as a navigational fix is still going at seven or eight kilometres per second, still capable of putting a hole through something, still going to be there long after everyone involved is dead. Making a hazard legible does not make it less of a hazard. It just means you now know where it is.

That is a real pattern and it is not confined to orbit. Lighthouses mark rocks. Every coastline chart in existence is a record of where ships were lost. The thing you had to avoid becomes the thing you navigate by, because avoiding it is what made you map it precisely. The difference — and it is the whole difference — is that nobody put the rocks there. This coastline is a shell around the entire planet, and we drew it by leaving things behind, deliberately, one launch at a time.

Which brings up the sovereignty question, and the answer is not the one the setup promises. The tempting version: GPS is designed infrastructure — placed with intent, funded, maintained, and therefore ownable, degradable, switchable — while the debris field is accidental infrastructure, unbuilt, unfunded, unjammable, with no single point of failure because there was never a single point of design.

Except FALCON is not navigating by debris. It is navigating by a catalogue of debris. Twenty thousand objects with known ephemerides, loaded aboard before the camera ever opened. You cannot match what you cannot identify — and the catalogue it matched against is the public one maintained by the U.S. Department of War. The same government whose ownership of GPS was the chokepoint four paragraphs ago. The dependency did not vanish. It moved: from a signal you can be denied to a list you can be denied, with the same hand near the switch.

Calibrate before you get carried away: this was three days, on cubesats, with two hundred objects out of twenty thousand catalogued out of a hundred million actual. It is a demonstration, not a deployed capability. It removes nothing from orbit.

Which is why the finding sitting quietly back in paragraph three is the one that matters. For more than two hundred objects, FALCON's own observations beat the predictions it had been handed. A spacecraft that corrects the catalogue it was given is a spacecraft that has begun producing the reference frame rather than only consuming it. That is a smaller claim than unownable infrastructure and a considerably better one, because it is a result rather than a mood.

Later this year, Starling's four spacecraft will point their cameras at the same crowded sky, pool what each of them sees, and work out their positions from one another. Not independence from the wreckage. Independence from the list.