Magnetic navigation flies on C-17 and airliner tests, but few results give error in meters

In March 2024 Daniel Patt of the Hudson Institute told Congress that the GPS-guided Excalibur artillery round in Ukraine “dropped from 70 percent effectiveness to 6 percent effectiveness over a matter of a few months”. The Government Accountability Office reported on October 7, 2026, that jamming in eastern Ukraine from 2023 to 2024 left one type of precision munition up to 80 percent less effective.

An uncrewed aircraft or boat has the same problem and no crew to notice it. Several alternatives to satellite navigation have now flown, so the question for a buyer has moved from whether they work to how well. Most announced results are stated as an improvement over an inertial system left to drift, and every method replaces dependence on a satellite signal with dependence on a reference map or a clear line of sight. Program offices should stop buying “GPS-denied navigation” as a feature and start specifying position error over time, under named conditions.

Inertial drift is the baseline every claim is measured against

An inertial navigation system measures acceleration and rotation and works out position from a known starting point. GAO notes that such sensors are self-contained, but their accuracy degrades the longer it has been since the last external fix; the gap between measured and actual position is called drift. Every other technique here corrects drift without GPS. A claim of “better than inertial” therefore means little unless the grade of the inertial unit, the time it ran unaided and an absolute error are reported with it.

The military M-code signal, which Breaking Defense describes as a jam- and spoof-resistant GPS signal in development since 1999, is a partial answer. It is still GPS, and the Army plans for its loss: GAO says the service’s satellite backup exists to supply receivers “if GPS, including M-code, is disrupted or denied.”

Every method has flown, but few tests report an error in meters

Visual navigation compares camera images with stored imagery of the ground. GAO reports it can be accurate up to 10 meters with the right reference data and is largely unusable over ocean, desert or dense forest. The Army’s JPADS guided cargo parachute shows the limits: the version that began fielding in 2026 has a daytime vision navigation capability, while night, weather and open water wait for an infrared version slated for 2028. A Ukrainian presidential adviser, Serhii Beskrestnov, said in remarks reported on October 11 that Russia has added image-based navigation to Geran-4 and Geran-5 attack drones.

Magnetic-anomaly navigation matches readings of the Earth’s crustal magnetic field against a map, which lets it work over water. The Air Force-MIT AI Accelerator flew it in real time on a C-17 during three sorties from Travis to Edwards Air Force Base in May 2023. In June 2026 a Defense Innovation Unit and Honeywell system flew four hours and 23 minutes over the Pacific on an Embraer 170 and, according to the unit, improved position accuracy by 89 percent over “traditional backup navigation methods.” SandboxAQ says its AQNav system “significantly outperformed” inertial-only navigation on Northrop Grumman’s Lumberjack drone this summer, where it was paired with visual navigation; the company declined to release performance data. None of these sponsor and vendor accounts gives an absolute error.

The exceptions come from developers’ own preprints. Researchers describing “proprietary quantum magnetometers” reported a best final accuracy of 22 meters in airborne trials. At sea, a Q-CTRL team reported gravity-map matching over an 83-kilometer sea route with error “at the nautical-mile level.” Neither is an independent test, and GAO’s summary is that magnetic and gravity navigation remain less accurate than GPS. Maps carry their own cost: Maj. Kyle McAlpin, the Air Force liaison to the MIT project, told Air & Space Forces Magazine that making them is “often time- and resource-intensive.”

Celestial navigation needs a view of the sky in place of a ground map. The Navy’s Automated Celestial Navigation System is in full production and, GAO notes, limited by clouds. On small airframes the precision is coarse: University of South Australia researchers flew a star-based system on a drone at night and located it to within about 4 kilometers.

Borrowed radio signals have produced the most precise airborne figure cited here. A team led by Zaher Kassas, flying with Air Force test pilots on a C-12 over Southern California at up to 23,000 feet, reached meter-level error using cellular towers and an altimeter. The authors did not study integrity or availability, and the method needs towers in range. The Army’s ALTNAV, which uses a commercial constellation, reached initial capability in late 2024, but GAO says full fielding depends on ground monitoring stations that several countries must first approve. The Space Force terminated its Resilient GPS small-satellite layer in January 2026.

Specify error over time, and name who supplies the map

GAO found in 2021 that difficulty setting clear performance requirements was holding back alternative navigation, and its 2026 report still describes the effort as fragmented across the services.

The requirement should state the largest position error the mission tolerates after a set number of minutes without GPS, over the terrain, weather and time of day the system will actually meet. Vendors should report absolute error against a truth source in place of a percentage gain. Because a daytime camera and a magnetometer fail under different conditions, a platform that must cross water at night needs at least two aiding sources whose blind spots do not overlap. The contract should also say who provides and updates the reference imagery or anomaly map for the theater, since a map-matching navigator without a current map is an inertial system with extra weight.