Everything that does not work
It is worth listing what is unavailable, because the absence is the whole problem.
| Method | Works on Earth | On the Moon | Why |
|---|---|---|---|
| Satellite navigationGPS, Galileo | Yes | Not usefully | No lunar constellation yet. Earth's global navigation satellite system (GNSS) signals reach the Moon, and the LuGRE receiver computed a fix from GPS and Galileo on the surface on 3 March 2025, but the geometry is poor: in lunar orbit its position accuracy was about 1.5 km. |
| Magnetic compassmagnetometer | Yes | No | No global dipole. Crustal anomalies are local and weak, tens to a few hundred nanotesla at the surface, which is worse than nothing. |
| Visual landmarksterrain | Yes | Poorly | Grey regolith under a hard sun with no atmosphere. Scale and distance judgement fail badly. |
| Dead reckoningwheel odometry | Yes | Partly | Wheel slip. Apollo's rover navigation assumed 1.85 percent and closed its traverses within 100 m, but a climb up North Ray Ridge on Apollo 16 slipped enough to put the readout 250 m off. |
| Star trackingcelestial | At night | Yes | The sky is always black. This one actually gets better. |
What the base uses instead
Surface beacons
Four surveyed radio beacons at the corners of the working area, ranging to anything inside it by time of flight on the same LTE network that carries data (LTE is the 4G cellular standard; Nokia flew one to the surface on IM 2 in 2025). This is the primary source inside the mast's 5.4 km horizon. The ranging accuracy is a design allocation of a few metres, not a measured figure: no lunar cellular network has published ranging performance yet.
The beacons are surveyed once, at installation, by a long baseline observation against the star tracker and a direct to Earth ranging session. After that they are the fixed reference everything else is measured from.
| Beacon | Position | Function | Power |
|---|---|---|---|
| BCN 1habitat mast | 0, 0 | Primary, colocated with comms | From the bus |
| BCN 2pad north | 44, minus 48 | Landing approach reference | Solar and battery |
| BCN 3plant west | minus 58, 38 | Excavation face coverage | Solar and battery |
| BCN 4array ridge | minus 46, minus 30 | Fourth corner, geometry | From the field |
Terrain relative navigation
Outside the beacon network a rover matches what its camera sees against the Lunar Reconnaissance Orbiter (LRO) terrain model, a method called terrain relative navigation (TRN). It is the technique Mars 2020 used to land Perseverance within about five metres of the spot its lander vision system chose, against a 40 m requirement, and the one Blue Ghost used on 2 March 2025 to pick a hazard free spot and, in Firefly's words, land directly over its target in Mare Crisium. It works well on the Moon for the same reason it works badly for a human: craters all look alike to us but their size distribution is highly distinctive to a matcher.
LRO's narrow angle camera (NAC) resolves 0.5 m per pixel from its 50 km orbit, and the Lunar Orbiter Laser Altimeter (LOLA) gives elevation. The twenty metre fix and the thirty seconds to converge below are design allocations for a rover camera at ground level, not flight results: the flight heritage is from descending landers looking down.
Error budget
Why ten metres is the requirement
Accuracy requirements should come from an operation, not from a wish. Ten metres is set by three of them.
A lander has to touch down on a 22 metre apron, so terminal guidance needs better than ten metres to have any margin. A rover returning to the airlock in a dust storm of its own making needs to find a two metre hatch. And a crew member walking home on a low battery needs to be pointed at the base and not past it, because on a body with a 2.4 km horizon, walking past your habitat is fatal.
A crew member on extravehicular activity (EVA) has a finite consumable budget. The walk back rule is that no one goes further out than they can walk home on the reserve, and that calculation is only as good as the position estimate it starts from. Navigation on this base is a life support system with a different name.
Failure modes
| Failure | Effect | Fallback | Response |
|---|---|---|---|
| Beacon offlineone of four | Geometry degrades | Three beacon fix | Accuracy falls to about 6 m. Acceptable, logged, repaired next EVA. |
| Inertial measurement unit (IMU) driftgyro ageing | Between fix error grows | Beacon and TRN | Calibrated against beacons on every return to base. |
| Camera obscureddust on the lens | No TRN fix | Inertial and odometry | Cleaning is part of the post EVA routine. |
| Total network lossmast down | No beacons at all | TRN and star tracker | Rovers hold about 20 m unaided, which is enough to come home. |
| Lost crew memberthe case that matters | Safety critical | Suit beacon and buddy | Suit transmits position continuously on a separate channel. Two crew minimum, always. |
What is not yet known
How much a wheel slips on polar regolith is not known. Apollo's rover analyses assumed 1.85 percent on the level and the navigation system closed its traverses within 100 m, but that was on mare and highland soil at low latitude with a crew aboard, and one ridge climb on Apollo 16 put the readout 250 m off. Odometry is never trusted alone for that reason. The beacon ranging accuracy is a design allocation: Nokia's LTE network operated on the surface on IM 2 in 2025, but no ranging performance has been published. Terrain relative navigation from a rover camera at ground level has no flight record; the heritage is from landers looking down during descent, where Mars 2020 landed within about five metres of its chosen point and Blue Ghost on its target. LuGRE's accuracy on the surface has not been published; in lunar orbit it was about 1.5 km. When the relay constellations will broadcast a navigation signal is not settled: the first of Intuitive Machines' relay satellites under NASA's Lunar Communications Relay and Navigation Systems contract is planned to fly with the IM 3 mission in late 2026, ESA's Lunar Pathfinder is planned for launch no earlier than November 2026, and Moonlight's initial service is planned for the end of 2028 with full service by 2030. The dependency on the local survey stays until then.
Increment plan
Increment 1, inertial and star tracker
Enough to land and to survey the site. Position is established by direct to Earth ranging during the first days, which is slow but absolute.
Increment 2, beacons surveyed and live
Four beacons installed and surveyed. This is the step that makes routine EVA and rover operations safe.
Increment 3, terrain relative navigation
Terrain matching against the LRO model, which extends useful navigation beyond the beacon network and out to the traverse limit.
Increment 4, LunaNet AFS
The LunaNet Interoperability Specification defines an Augmented Forward Signal (AFS), a GNSS style broadcast in S band from the relay satellites that a surface receiver can use for position, velocity and time. When the constellation broadcasts it, ranging against it gives an independent absolute fix and removes the last dependency on the local survey.
Sources and further reading
Every figure in this document traces to one of these. Agency documents and peer reviewed work first, reporting only where it is the primary record.