Krishna R. Pattipati, Life Fellow, IEEE

Large low-Earth-orbit (LEO) constellations make satellite navigation much harder to jam and much faster to recover after an outage, but they do not by themselves deliver “1 cm” or “unjammable” navigation; for an unmanned aircraft, resilience comes from combining independent navigation sources, a good inertial measurement unit, the right route and in-flight replanning.

The Question

Satellite navigation signals reach the ground weaker than background noise, so a cheap jammer can deny them over kilometers. Constellations of thousands of LEO satellites promise much stronger signals, and their promoters speak of “1 cm” and “unjammable” navigation. Two companion papers ask what such constellations can actually deliver, and what an unmanned aircraft needs to keep its navigation within a corridor when signals may be jammed.

• Part I asks what accuracy, recovery time and jamming resistance a large LEO constellation can provide, and under what assumptions.
• Part II asks which combination of navigation sources, inertial measurement unit, route and in-flight decisions keeps an aircraft on course, nominally and after losing any one source, and at what cost.

Part I: What LEO Constellations Buy

LEO buys signal strength and speed of recovery, not better geometry per satellite. Its satellites are 10 to 40 times closer than GPS satellites, so their signals are hundreds of times stronger, and they cross the sky in minutes rather than hours.

Starlink, the largest constellation today (about 11,000 satellites), carries no navigation signal. Researchers position the receivers with it to within several meters, limited by unpublished orbits and unsynchronized satellite clocks. Part I shows what a constellation of that size could deliver if it broadcasts proper navigation signals.

Part II: What Keeps an Unmanned Aircraft on Course

Resilience is a property of hardware, estimation, route and in-flight decisions together; getting any one of them wrong undoes the others. The studies use a simulated 150 × 100 km theater with randomly placed jammers and a requirement to stay within a 50 m corridor.

1. Measure inertial drift in flight, not in the lab. Datasheet values underpredicted in-flight drift by a factor of three to five. A route that looks safe on paper can fail for that reason alone. The drift model should be re-estimated continuously, over a window matched to how fast conditions change (about 15 minutes in our tests).
2. Judge routes by their longest outage, not total exposure. Several short gaps are harmless; one long gap is fatal, because inertial error grows faster than linearly with time.
3. Plan on the threat’s statistics, not on sampled scenarios. Routes optimized for a few sampled jammer layouts threaded gaps that existed only in the samples and did worse on new fields.
4. Sense and replan in flight. Detecting jammers and replanning every 5 km raised the probability of meeting the requirement from 0.70 to about 0.90. A simple re-planner with an exact estimate of where undetected jammers may be matched to the performance of elaborate look-ahead searches at one hundredth of their computation.
5. Survive the loss of any one source. Among the designs evaluated, there are only three independent sources (a LEO satellite layer, the ground-based eLoran system and a thermal camera), together with a navigation-grade inertial measurement unit and a route chosen for the degraded case, met this test. Two sources were not enough: when one is lost, the other is itself unavailable in about one mission in ten.
6. The route matters as much as the hardware. The same equipment failed 21 of 200 test missions after losing satellite navigation on the direct route, and only one on a route 22 km longer that avoided a 49 km stretch of featureless terrain.

Cost: navigation that meets the nominal requirement was cheap, about $22 million a year for 1,000 aircraft using GPS, eLoran, a thermal camera and a navigation-grade inertial unit. Adding centimeter capability with a 1,000-satellite LEO layer raised this to about $370 million a year, unless the constellation’s cost is shared with other users.

What It Means

• For constellation programs: the value of a large LEO layer is jamming margin and fast recovery. Centimeter accuracy depends on precise orbit and clock products, signals stable over a satellite pass, and two frequencies; these deserve as much investment as satellite count.
• For evaluating claims: ask whether “1 cm” means a predicted or an achieved error, float or fixed, static or flying, and with which orbit and clock products. Ask what antenna “unjammable” assumes.
• For aircraft designers: plan with drift measured in flight, choose routes for the degraded case, carry three independent navigation sources if losing anyone must be survivable, and replan in flight with a sound estimate of the threat.
• For operators of a navigation system: sources that share components or depend on one operator can fail together. India’s NavIC regional system lost standalone positioning in 2026 after repeated clock and launch failures left it with three active satellites, one fewer than positioning needs; a replacement satellite launched in October 2026 is meant to restore that minimum.

What the Results Do and Do Not Show

All results come from simulation under stated models; they are not demonstrated service levels. The constellations are idealized, the jammer fields, terrain and costs are synthetic or illustrative, and the failure measure is the aircraft’s own navigation uncertainty, not an integrity guarantee. Within those limits, the findings were checked against held-out data and alternative model choices, and the code that generates every table is available.

The two papers are working drafts: K. R. Pattipati, “What Low-Earth-Orbit Constellations Buy for Jam-Resilient Positioning” (Part I) and “Resilient Navigation Architectures and Proactive Routing for Unmanned Aircraft Under Uncertain Interference” (Part II).

You can read Part I’s full essay here.

You can read Part II’s full essay here.