Can Starlink Signals Help With Navigation Beyond GPS?

GPS and other global navigation satellite systems (GNSS) underpin maps, transport and precise timing. Their signals can be disrupted or spoofed, so researchers are investigating complementary sources of positioning, navigation and timing (PNT). Low-Earth-orbit (LEO) communications satellites are one possibility: their wideband signals can support precise measurements of how long a signal takes to arrive. But that potential alone does not make a reliable navigation system. A receiver also needs to know where a satellite was when it transmitted and how its clock relates to a reference time.

This study asks whether Starlink signals could support pseudorange-based PNT when paired with orbit and clock information supplied through a coordinated service. A pseudorange is an estimate of signal travel time expressed as a distance. The researchers describe their approach as a mock “fused” system, in which a communications constellation takes on a secondary navigation role. The experiment tested elements of that idea; it did not demonstrate an operational Starlink navigation service.

What the field test did

The researchers set up a control station at a known location with a large steerable dish and GPS-based timing. It recorded live Starlink Ku-band signals, estimated frame-clock corrections and sent information to a separate receiver at an unknown location. That user receiver measured signal arrival times and used the corrections in its position and clock estimates.

For the experiment, the researchers used publicly accessible Starlink ephemerides—the data describing satellite orbits—rather than meter-accurate orbit models based on onboard GNSS observations. The control station estimated frame-clock corrections from its live measurements. The paper discusses how ephemeris errors affected modeled signal travel time, and how the short distance between the two ground stations helped limit some differential errors.

Signal assistance was another important part of the setup. The control station supplied decoded payload data that the user receiver could use to build signal replicas and improve detection and timing measurements. This was experimental aiding, not simply a broadcast of orbit and clock corrections already available to ordinary users. The paper leaves other signal-aiding approaches for future work.

The user equipment was a prototype PNT-only receiver, not a standard consumer device. It used a feedhorn antenna and a 25 MHz receiver bandwidth, while the reference station used a large dish and much wider signal capture. The team recorded signals over 13 minutes in January 2025, tracking six satellites. Because the equipment could observe only one satellite at a time, measurements were collected sequentially as satellites passed, rather than as a simultaneous snapshot. An oven-controlled crystal oscillator helped maintain receiver timing through the long, hardware-induced gaps between observations.

The researchers compared three methods for measuring signal arrival time. Their super-resolution method estimated arrival time beyond sample-level resolution. That describes the technique used in this test; it does not show that ordinary consumer receivers can use it or achieve the same performance. The measurements were combined in a batch estimator that modeled changes in receiver clock bias and drift.

Results—and what they do not establish

For the super-resolution method, the paper reports formal 95% bounds of 2.33 metres for the horizontal error ellipse’s semi-major axis, 3.37 metres vertically, and 11.19 nanoseconds in timing. The actual solution errors fell within those reported bounds. These are results and estimator-derived bounds for this particular experiment, not an independently validated guarantee that 95% of future fixes will fall within those values.

The findings show what a carefully aided setup achieved, not what unaided Starlink signals or a typical device can do. The test was brief, involved six satellites and used sequential observations. Its results depended on the control station’s measurements, decoded-data assistance and an oven-controlled clock that bridged gaps. The authors discuss possible future services in which a constellation operator distributes orbit and clock models, or a third party provides corrections through reference stations. Neither was demonstrated here as a working public service.

This campaign does not establish performance across different locations, devices or conditions, nor does it demonstrate navigation during jamming or spoofing. Its evidence is a field demonstration using live Starlink downlinks and a specialized experimental arrangement. The source is an arXiv preprint, and the supplied material does not establish its peer-review status. The results are best read as evidence that the approach merits further testing—not proof that Starlink currently provides GPS-like navigation or a ready-made backup to GNSS.