Turns out, it can be done

A while ago I had what I thought was a fairly straightforward idea.

The Earth’s magnetic field is not perfectly smooth.

The planet is full of different rocks, structures, minerals and geological formations. Those should distort the magnetic field differently in different places.

So surely, I thought, the magnetic environment around a location should have something approaching a fingerprint.

Map the fingerprint.

Measure the field around you.

Compare the measurement with the map.

Work out where you are.

No GPS satellite required.

There are plenty of perfectly sensible reasons to be sceptical. The magnetic field changes. Sensors are noisy. Aircraft generate electromagnetic interference. A single magnetic reading isn’t guaranteed to be unique. Maps are imperfect and altitude complicates things.

But in September 2026 I came across a FlightGlobal report about a GPS-free test flight relying on the Earth’s magnetic field.

And not in the loose sense of “use a compass”.

The system does almost exactly what I had in mind: it measures tiny variations in the Earth’s magnetic environment and compares them with a known magnetic map to estimate location.

SandboxAQ describes its AQNav system in remarkably similar terms. High-sensitivity magnetometers measure distinctive magnetic patterns, while software separates useful geological signal from all the electromagnetic rubbish produced by the aircraft itself.

According to the company, AQNav has now been tested across more than 200 sorties and 500 flight hours. Earlier nationwide flight testing used the publicly available North American Magnetic Anomaly Map and demonstrated navigation across wildly different geological environments without GNSS.

Turns out, using the Earth’s magnetic fingerprint to work out where you are isn’t just possible. It’s already being flown.

But that isn’t actually the interesting part.

The interesting part is what happens when you take the same idea further.

Once you stop thinking of navigation as “receiving coordinates from GPS” and start thinking of it as identifying your position from naturally occurring fingerprints, the question gets much bigger.

Why stop at magnetism?

Why stop at Earth?

And, eventually:

Does the universe itself contain enough naturally occurring information to tell an observer exactly where — and when — they are?

That is a much more interesting rabbit hole.

GPS Has Quietly Changed How We Think About Position

GPS is so good that it has distorted our intuition about navigation.

Most of us now implicitly think position is something that comes from satellites.

It does not.

GPS is just one engineered way of answering a much older question.

A GPS receiver listens to multiple satellites whose positions and clocks are known. Each signal gives the receiver information about how far away that satellite was when the signal was transmitted.

With enough independent observations, the receiver solves for its own unknown position and clock error.

Conceptually:

known transmitters
+
known timing
+
signal travel time
=
position + time

It is brilliant engineering.

But there is nothing fundamental about satellites.

The deeper requirement is simply this:

the environment must contain observations that vary predictably with location.

If nature already provides those observations, the transmitters can disappear.

That changes the problem completely.

Earth’s Magnetic Fingerprint

A normal compass is not a positioning system.

It tells you the direction of the local magnetic field well enough to infer magnetic north.

That does not tell you whether you are in Lancashire, Alaska or the middle of the Atlantic.

The useful information comes from the fact that Earth’s magnetic field is not just a perfectly tidy dipole.

The crust contains different magnetic minerals in different concentrations and structures. Those geological differences produce local magnetic anomalies.

Instead of imagining a smooth field wrapping around the planet, imagine an invisible landscape containing bumps, troughs, gradients and irregularities.

That landscape can be mapped.

A sensitive magnetometer moving through it sees a changing signal.

The job then becomes a map-matching problem:

MEASURED MAGNETIC PATTERN
            |
            v
COMPARE AGAINST MAGNETIC MAP
            |
            v
FIND MOST LIKELY POSITION

This is magnetic anomaly navigation, usually shortened to MagNav.

And the reason it works is exactly the reason an engineer might initially worry that Earth’s magnetic field is “messy”.

The mess is the information.

If the field were perfectly uniform, it would be almost useless for localisation.

The imperfections provide the fingerprint.

SandboxAQ’s AQNav combines highly sensitive quantum magnetometers with algorithms designed to remove aircraft-generated noise and match the remaining crustal signal against magnetic maps.

That is a very different proposition from flying around with an expensive compass.

The system is asking something closer to:

Where on the known magnetic landscape would I have to be to observe this sequence of measurements?

That word sequence matters.

A single measurement does not necessarily uniquely identify one location.

But suppose an aircraft measures:

A -> B -> C -> D -> E

while its inertial system simultaneously says it travelled roughly north-east, descended 400 feet and accelerated slightly.

Now the matching problem is much more constrained.

The system is not looking for one matching magnetic value.

It is looking for a matching path through the magnetic landscape.

That is much closer to geological SLAM than a conventional compass.

Difficult Is Not the Same as Impossible

This is the bit worth dwelling on because engineers — and apparently AI systems — are very good at accidentally killing good ideas by producing a sufficiently impressive list of caveats.

MagNav genuinely has difficult engineering problems.

An aircraft is an awful environment in which to measure faint magnetic variations.

There are electric motors, wiring, avionics, actuators, ferromagnetic structures, currents changing with load, vibration, heading changes and equipment being switched on and off.

The Earth’s external magnetic environment also varies.

Maps differ in quality.

Altitude affects the signal.

Some areas contain richer magnetic features than others.

None of those objections invalidate the physics.

They describe the engineering work required to extract the signal.

That distinction matters enormously.

A lot of innovation begins with someone saying:

In principle, the information should be there.

The next ten years are frequently spent figuring out how to hear it through the noise.

Then I Thought: Gravity Must Have a Fingerprint Too

Once the magnetic idea was established, gravity was the obvious next step.

Again, the useful concept is not simply measuring Earth’s average gravitational acceleration and announcing:

g = 9.81 m/s²
therefore I am somewhere on Earth

Outstanding work. Give the navigation team a biscuit.

Earth’s gravitational field also varies geographically because the mass beneath and around us is not uniformly distributed.

Mountains, rock density, sediment, crust thickness and geological structures alter the local gravitational field.

Those variations can be mapped as gravity anomalies.

A gravimeter can therefore, in principle, perform the same broad trick as a magnetic navigation system:

measure local gravity → compare with gravity map → constrain location.

This is already moving from theory into field demonstrations.

In August 2026 researchers reported a fully GNSS-independent marine navigation demonstration using a quantum gravimeter aboard a 29-metre vessel. The system used a gravity anomaly map to constrain an inertial navigation solution across an 83 km trajectory.

That is important because inertial navigation has a fundamental weakness.

An IMU is excellent at telling you how your motion is changing.

But tiny errors accumulate.

You gradually drift away from reality.

A natural map provides something inertial navigation desperately wants: an absolute external correction.

Magnetic and gravitational fingerprints can periodically tell the inertial system:

No. You think you’re there. The physical world says you’re here.

Gravity Is Beautifully Difficult to Lie About

Gravity has another appealing property.

It is exceptionally hard to jam.

GPS signals arrive at Earth extremely weakly. A transmitter can overpower or imitate them.

Magnetic sensors can be deliberately disturbed by nearby fields.

But spoofing the gravitational signature of several kilometres of geology is a somewhat more ambitious afternoon project.

You would need to rearrange meaningful amounts of mass.

At which point the person attempting the spoofing arguably deserves the position fix.

Gravity is also extraordinarily stable compared with many electromagnetic phenomena.

The mountain does not reboot.

The crust does not need batteries.

There is no subscription.

There is no antenna to destroy.

Nature continues operating the infrastructure with characteristically poor customer support but excellent uptime.

Full Gravity Gradients Make the Fingerprint Richer

There is an even richer version of gravity navigation.

Instead of measuring only the strength of gravity at one point, measure how gravity changes in different directions around that point.

That is gravity gradiometry.

A full gravity-gradient tensor contains multiple components describing the spatial curvature of the gravitational field.

You can think of it, loosely, as going from asking:

How steep is the hill here?

To asking:

What is the complete shape of the hill around me?

The latter contains much more identifying information.

Research into gravity-aided navigation has repeatedly found that richer gravity-gradient measurements can provide much stronger map-matching constraints than scalar gravity measurements alone.

So now we have two natural fields:

MAGNETIC TOPOGRAPHY
+
GRAVITATIONAL TOPOGRAPHY

They are produced by different physical properties of the planet.

That means their errors and ambiguities are not identical.

A place that happens to look like another place magnetically is less likely to also look identical gravitationally.

Fuse them with inertial motion and altitude and the fingerprint becomes significantly harder to confuse.

Which naturally suggests a more general architecture.

Earth Fingerprint Navigation

Forget “MagNav” and “gravity navigation” for a moment.

Imagine an Earth navigation system that treats every natural measurable field as another dimension of one giant fingerprint.

At a location we might observe:

magnetic vector
magnetic gradient
gravity anomaly
gravity gradient
barometric pressure
terrain elevation
celestial orientation
inertial motion history

The navigation problem then becomes:

F(x, y, z) = observed environmental fingerprint

Find the coordinates whose predicted fingerprint best matches the measurements.

And while moving:

F1 -> F2 -> F3 -> F4 -> F5

The trajectory itself becomes part of the fingerprint.

This is already much harder to jam or spoof than relying on one external radio service.

But there is a fairly obvious limitation.

It still sticks us to Earth.

Take the spacecraft to Mars and your lovingly constructed Lancashire gravity map becomes somewhat less useful.

So I asked the next question.

What Is the Version That Works Anywhere?

If Earth has naturally occurring fingerprints, the universe should too.

We are surrounded by astronomical objects emitting radiation.

Stars have identifiable spectra and positions.

Quasars are visible across enormous cosmological distances.

The cosmic microwave background provides a natural large-scale reference frame.

Transient astronomical events occur throughout the universe.

And then there are pulsars.

Pulsars are where this stops sounding like pub physics and becomes an actual navigation technology.

Pulsars: The Universe Accidentally Built Lighthouses

A pulsar is a rotating neutron star.

Some neutron stars rotate extraordinarily quickly and emit beams of electromagnetic radiation from regions around their magnetic poles.

If the geometry lines up with us, that beam sweeps across our detector once per rotation.

We observe pulses.

Some millisecond pulsars are extraordinarily stable rotators.

They effectively behave like natural clocks distributed across the galaxy.

This immediately gives them two characteristics navigation engineers adore:

  1. they come from known directions;
  2. their pulse timing can be modelled extremely precisely.

That makes them resemble natural GPS beacons.

Except nobody launched them.

Nobody owns them.

Nobody can conveniently turn them off.

And they are rather more difficult to shoot down.

ESA has described remote millisecond pulsars as providing the best known natural reference frame for autonomous deep-space navigation.

NASA went considerably further and actually tested the idea.

NASA Has Already Demonstrated Pulsar Navigation in Space

NASA’s NICER instrument on the International Space Station included the SEXTANT experiment — the Station Explorer for X-ray Timing and Navigation Technology.

SEXTANT used X-ray observations of millisecond pulsars to determine the station’s orbit autonomously.

In the 2017 demonstration, the software used observations from four pulsars and recovered the ISS trajectory to better than roughly 10 km in its worst direction using only pulsar measurements.

Ten kilometres sounds terrible if you are trying to find Tesco.

It is rather more impressive when your “satellites” are neutron stars potentially thousands of light-years away and your prototype navigation receiver is attached to the International Space Station.

The importance of the demonstration was not the final accuracy.

It was that the geometry and timing information genuinely contained enough information to determine position autonomously.

NASA now explicitly describes pulsar-based positioning, navigation and timekeeping as a route toward lunar, Mars and deep-space autonomous operations.

The basic principle is remarkably familiar.

GPS satellite clocks
       becomes
pulsar rotation clocks

The spacecraft measures differences between expected and observed pulse arrival times.

Those timing residuals depend on its position.

Observe enough pulsars distributed around the sky and you can solve for that position.

But Pulsars Give Us Something Even More Interesting: Time

Here is where the idea becomes bigger than navigation.

With GPS we normally solve for four unknowns:

x
y
z
receiver clock offset

That fourth value matters because distance is inferred from time.

A tiny timing error becomes a large ranging error at the speed of light.

Pulsars also provide timing information.

Suppose we have highly accurate models describing the expected pulse phase of several pulsars relative to a recognised reference time and reference frame.

Our observation then depends on both where we are and when we are.

Conceptually each pulsar gives us a constraint involving:

x, y, z, t

With enough independent observations, our unknown is not merely position.

It is our spacetime state.

Now we have reached the version of the original magnetic-fingerprint idea that I find genuinely fascinating.

A Spacetime Fingerprint

Imagine that instead of carrying a database containing Earth’s magnetic map, a spacecraft carries a catalogue describing the observable universe.

For many pulsars it knows:

  • sky position;
  • pulse profile;
  • rotation frequency;
  • spin-down behaviour;
  • binary motion where relevant;
  • timing model;
  • distance estimate.

It also carries a star and quasar catalogue.

Potentially models of additional large-scale observables.

The spacecraft wakes up with no GPS.

No Deep Space Network.

No radio beacon.

No friendly planet telling it where it is.

Its sensors look outward.

They identify astronomical sources.

They measure directions.

They measure pulse phases.

They measure Doppler shifts.

They measure relative timing.

And then they ask:

At what location, velocity, orientation and time would an observer see this exact sky?

That is the universal version of magnetic map matching.

The map is simply much larger.

The Sky Is Not the Same Everywhere

This point is easy to miss because humans barely move on astronomical scales.

We tend to think of the night sky as a fixed wallpaper.

It is not.

Move far enough and geometry changes.

Nearby stars exhibit parallax.

Apparent directions change.

Signal travel times from pulsars change.

Doppler shifts depend on relative velocity.

Gravitational potential changes clock rates and signal propagation.

Different observers occupy different positions on the incoming wavefronts of distant sources.

So an astronomical observation is not merely describing the source.

It also contains information about the observer.

That is the key inversion.

Astronomy normally asks:

What can this light tell us about the thing that emitted it?

Navigation asks:

What can this light tell us about the person receiving it?

Exactly the same photons.

Different question.

Use More Than Four Pulsars

The most obvious implementation would mimic GPS and choose a small number of excellent pulsars.

That is probably sensible for early systems.

But once sensing and computation become sufficiently capable, there is no fundamental reason to stop at four.

Suppose a future navigation system can observe dozens of pulsars simultaneously.

Each one contributes another constraint.

Then add stars.

Then quasars.

Then other useful astronomical references.

The problem becomes massively overdetermined.

Instead of:

4 observations
4 unknowns

we might eventually have:

hundreds of observations
+
a comparatively small state vector

The system can then ask which possible state best explains all of those observations simultaneously.

That allows bad measurements to be rejected, uncertainty to be estimated and ambiguous solutions to be eliminated.

This is how I would want to build it.

Not celestial GPS.

Celestial sensor fusion.

Quasars Give Us an Almost Fixed Background

Quasars are extraordinarily distant active galactic nuclei.

Because they are so far away, their apparent angular motion across the sky is tiny.

Radio astronomy already uses distant extragalactic sources to define the International Celestial Reference Frame.

That means quasars provide something useful for our hypothetical system: orientation.

A spacecraft capable of identifying a set of quasars can determine how it is rotated relative to a very stable celestial reference frame.

So pulsars help with timing and position.

Quasars help with orientation.

Stars provide additional geometry.

Different phenomena solve different parts of the state-estimation problem.

Which is exactly what we want.

What About the Cosmic Microwave Background?

The cosmic microwave background is radiation left over from the early universe.

It is not perfectly uniform.

Among other things, it contains a strong dipole caused largely by our motion relative to the CMB rest frame.

Measure that pattern accurately enough and it gives information about your velocity relative to a cosmological reference frame.

This does not magically give you a unique address in the universe.

But it gives another independent constraint.

And this is the recurring theme:

we don’t necessarily need one magical field.

We can combine several independent natural observations.

Gravitational Waves Could Eventually Join In

Gravitational-wave events provide another potentially useful signal.

When black holes or neutron stars merge, disturbances propagate through spacetime.

If the same event is observed at separated locations, differences in arrival time contain geometric information.

For a single spacecraft this is not nearly as convenient as persistent pulsars. Events are transient, detection is difficult and the hardware requirements are currently enormous.

But future systems could use sufficiently well-characterised events as occasional additional constraints or calibration points.

Again, the important idea is not that every sensor must be used.

It is that the universe gives us multiple independent information channels.

So What Are We Actually Solving For?

A sophisticated system would estimate considerably more than latitude and longitude.

Its state might include:

X = [
    position_x,
    position_y,
    position_z,
    velocity_x,
    velocity_y,
    velocity_z,
    orientation,
    clock_offset,
    clock_drift
]

The navigation engine predicts what an observer in state X should see.

It compares that prediction with what the sensors actually see.

Then it changes X until prediction and observation agree as closely as possible.

Conceptually:

GUESS WHERE/WHEN I AM
        |
        v
PREDICT WHAT THE UNIVERSE SHOULD LOOK LIKE
        |
        v
COMPARE WITH WHAT I ACTUALLY SEE
        |
        v
UPDATE THE GUESS
        |
        +--------------------+

That is essentially state estimation on a ridiculous scale.

Relativity Is Not an Optional Extra

Once distances become astronomical and clocks become sufficiently precise, Newtonian approximations stop being enough.

Signal arrival times depend on observer motion.

Gravity affects clocks.

Light propagation is affected by gravitational fields.

And there is no single universal clock ticking identically everywhere.

So “where and when am I?” requires a defined coordinate system and reference time scale.

That does not kill the idea.

GPS already requires relativistic corrections to work accurately.

A universal navigation system simply has to take the same problem much further.

The answer it produces is not some mystical absolute location outside physics.

It is position and time within a defined physical reference frame.

That is enough.

Could It Work With No Starting Position?

This is the version I find most interesting.

Most navigation systems work best when they already have a rough idea where they are.

But imagine switching this one on with no useful prior location.

It sees a sky.

First it identifies distinctive sources and determines orientation.

Then pulsar phases and source geometry generate candidate positions and times.

Doppler observations constrain velocity.

Additional observations eliminate incompatible hypotheses.

The solution space gets smaller and smaller until one region of spacetime fits the observed fingerprint overwhelmingly better than the alternatives.

Conceptually:

I DO NOT KNOW WHERE I AM
          |
          v
IDENTIFY THE SKY
          |
          v
GENERATE POSSIBLE STATES
          |
          v
COMPARE TIMING + GEOMETRY + MOTION
          |
          v
ELIMINATE IMPOSSIBLE STATES
          |
          v
CONVERGE

Whether that can genuinely be made to work from anywhere in the universe is a much harder claim.

Source visibility changes.

Our catalogue has a finite epoch.

Light takes time to travel.

Sources evolve.

Beyond sufficiently large distances the sky itself can differ radically from the one our catalogue describes.

Cosmological horizons exist.

So I would not claim we have discovered a magical universal coordinate reader.

But within large useful regions of space, the principle is compelling.

The Catalogue Becomes the Infrastructure

There is still infrastructure in this system.

It just changes form.

GPS requires satellites that are launched, maintained, synchronised and continuously operated.

Natural navigation does not require us to operate the beacons.

But it does require us to understand them.

The critical infrastructure becomes a catalogue:

source identity
position
motion
pulse model
spectral signature
distance
timing behaviour
uncertainty
reference epoch

Future spacecraft could receive updated catalogues while communication is available and then navigate independently when it is not.

Eventually they could improve parts of those models from their own observations.

At that point the navigation system starts to look less like a GPS receiver and more like an autonomous observatory.

The Universe Is Dynamic — Which Might Actually Help

At first, source evolution sounds like another problem.

Pulsars spin down.

Stars move.

Binary systems orbit.

Transient events appear and disappear.

But change contains time information.

A sufficiently rich observation may not merely be spatially distinctive.

It may be temporally distinctive too.

The catalogue becomes something closer to:

F(x, y, z, t)

The expected fingerprint depends on both position and time.

Which is exactly what we want if the goal is to determine a spacetime coordinate rather than merely a position.

Modelling that accurately is extremely difficult.

But the information is there.

There Is No Universal “Now”

There is an important qualification here.

Relativity does not give us one absolute universal time shared by every observer.

Different observers can disagree about simultaneity. Clock rates depend on velocity and gravitational potential.

So the system cannot discover some objectively correct cosmic wall clock.

What it can do is determine its spacetime coordinates relative to an agreed reference frame and coordinate time.

For navigation, that is what matters.

What Would the Hardware Look Like?

A future system might combine several sensor classes:

X-ray detector / telescope
        -> millisecond pulsar timing

optical telescope
        -> stars + quasars + parallax

high-stability local clock
        -> short-term timing reference

IMU
        -> high-rate relative motion

radio astronomy sensors
        -> additional pulsars / quasars

possibly CMB sensing
        -> large-scale velocity reference

The software is probably at least as important as the hardware.

It would need source identification, precise timing, relativistic propagation models, probabilistic state estimation, catalogue management, outlier rejection and continuous uncertainty tracking.

This is not a better compass.

It is effectively an observatory whose main subject is itself.

Accuracy Depends on Where You Are and What You Can See

There is no honest single accuracy figure for this concept.

Near Earth, GPS remains vastly more convenient and accurate for ordinary navigation.

Current pulsar navigation is nowhere near phone-GPS accuracy. X-ray detectors are not phone components. Observation periods can be significant. Useful pulsars are not distributed perfectly across the sky.

But the comparison changes completely in deep space.

GPS disappears.

Earth-based radio navigation depends on ground infrastructure and increasing communication delay.

A spacecraft that can independently establish its own state to kilometre-level accuracy — and eventually perhaps substantially better — becomes extremely valuable.

The useful question is not:

Is this better than the GPS in my phone?

It is:

How accurately can a spacecraft navigate when nobody else is available to tell it where it is?

The Same Idea Works at Different Scales

What I like about this is that the same pattern keeps appearing.

On Earth:

magnetic anomalies
+
gravity anomalies
+
terrain
+
IMU
=
localisation

In the Solar System:

pulsars
+
stars
+
planetary observations
+
IMU
=
autonomous deep-space navigation

At larger scales:

pulsars
+
quasars
+
stellar geometry
+
cosmological references
+
relativistic modelling
=
spacetime localisation

The sensors change.

The underlying idea does not.

Observe a sufficiently distinctive natural fingerprint and solve backwards for the state of the observer.

This Is Really an Information Problem

Strip away the individual bits of physics and this becomes an information problem.

If two possible locations produce exactly the same observations, those observations cannot distinguish them.

Add another independent measurement and perhaps they can.

Then another.

Then observe how all of those measurements change as you move.

Eventually the combined fingerprint can become sufficiently distinctive for useful localisation.

That is why sensor fusion is the important part.

Magnetism alone has ambiguities.

Gravity alone has ambiguities.

Pulsar timing has limitations.

Star geometry has limitations.

The useful system is the intersection of all of them.

Each observation removes possibilities left open by another.

The More Interesting Question

The magnetic-navigation flight is interesting on its own.

GPS denial is a real problem and passive geophysical navigation gives aircraft another way to constrain position without depending entirely on satellite signals.

But it raises a much bigger question.

We normally think navigation infrastructure is something civilisation has to build.

Beacons.

Lighthouses.

Radio towers.

Satellites.

But perhaps the physical universe already contains most of the raw information necessary for navigation.

We have simply spent most of our history without sensors capable of reading enough of it.

A magnetic anomaly map turns geology into navigation infrastructure.

A gravity map turns mass distribution into navigation infrastructure.

A pulsar catalogue turns neutron stars into navigation infrastructure.

A quasar catalogue turns distant galaxies into an orientation system.

None of those things exists for our benefit.

We are just learning how to use the information they happen to provide.

From Magnetic Fingerprints to a Universal Address

The thought that started this was simple:

The Earth’s magnetic field varies from place to place. Surely that variation can act like a fingerprint.

Turns out it can.

Then comes the next question:

If magnetism can help identify where we are, what other natural fields can do the same thing?

Gravity can.

Then:

Can we combine several natural fingerprints?

Yes — and doing so should make localisation considerably more robust.

And finally:

What is the version that does not depend on Earth at all?

That takes us to pulsars, quasars, stellar geometry and potentially other astronomical observables.

Not one magical invisible field.

Something better:

a fingerprint assembled from the universe itself.

The ambitious version is a machine that can wake up somewhere in deep space, look around, listen, measure, and eventually answer four questions without asking Earth:

Where am I?
Which way am I facing?
How am I moving?
What time is it in my reference frame?

We are nowhere near putting that complete system in a pocket.

But individual pieces already exist.

Magnetic fingerprint navigation is flying.

Quantum gravity navigation is being demonstrated.

Pulsar navigation has worked aboard the ISS.

Quasars already define our celestial reference frame.

So this is not an argument that universal navigation has been solved.

It is an observation that the ingredients are beginning to line up around a surprisingly simple idea:

location may not need to be broadcast to us.

It may already be written into what the universe looks like from where we are.