AI-enhanced magnetic navigation for military planes gets capabilities trialStory
June 09, 2023
A team of researchers at the Air Force Research Labs (AFRL) and the Massachusetts Institute of Technology (MIT) reported a successful demonstration of the capabilities of an aircraft outfitted with artificial intelligence (AI)-enhanced magnetic navigation, known as MagNav.
U.S. Air Force materials describe MagNav as a system that enables aircraft to determine their position using Earth’s geomagnetic field as a reference. The system, installed in an AgilePod pod-based modular sensor package, uses a magnetometer to read Earth’s local magnetic field and determine the aircraft’s location. The system is regarded as a potential replacement for satellite-based GPS, which although widely used, is a feared single point of failure for those military personnel depending on it for navigation.
Earth’s geomagnetic field is an invisible field generated by the planet’s core and crust. The field is well-known and mappable by detecting local variations in direction and intensity, often caused by things like iron deposits and even human-made objects. Unlike satellites, the geomagnetic field can’t be shot down and is ever-present. The Air Force gave early, non-AI-assisted MagNav a test in 2017, when an aircraft equipped with a magnetometer continuously determined its location within 13 meters over an hour-long flight.
The Air Force/MIT team conducted the demo – during the May 2023 Exercise Golden Phoenix out of Travis Air Force Base (central California) – using a C-17A Globemaster III aircraft leveraging AI and machine learning (ML) capabilities from the Air Force-MIT AI Accelerator (AIA). The AIA is a joint project between universities, small businesses, traditional defense, and nontraditional commercial ventures working with the U.S. Air Force and Space Force that seeks to develop and ensure the ethical use of AI technologies for the U.S. and society as a whole. (Figure 1.)
[Figure 1 ǀ A formation of 21 aircraft taxi on the flightline as part of Exercise Golden Phoenix at Travis Air Force Base, California, which hosted the AI-assisted MagNav demo. U.S. Air Force photo by Senior Airman Alexander Merchak.]
The joint Air Force and MIT announcement states that during the trial, the MagNav team powered up the AIA’s calibration and positioning neural network, which was trained during flight in just a matter of minutes on a regular laptop computer. The AIA’s neural network architecture removed magnetic noise generated by the aircraft to derive position by comparison to a known magnetic map.
The Air Force/MIT team reported mixed but encouraging results: The system tested on the C-17 was accurate to around one kilometer (0.62 miles), a notable downgrade from the more accurate GPS, which can navigate a craft to within a few feet of its destination. The MagNav team pointed out a positive side in its report, though, noting that for a majority of missions, navigation to within a single kilometer is sufficient before traditional tools like visual navigation or landmark recognition kick in and are available to guide the aircraft to its final destination.
Perhaps the most promising advantage of magnetic navigation is the fact that it is virtually impossible to jam or spoof, especially from any significant distance. “Every pilot fears single points of failure,” explained Maj. Kyle McAlpin, AIA MagNav liaison, in the Air Force’s MagNav test announcement. “Our strategy documents lament the DoD’s over-reliance on GPS, a single point of failure in our ability to navigate precisely.” McAlpin pointed out that every flight demands what he called “unassailable positioning and navigation.
“We can achieve that by augmenting GPS with alternatives like celestial navigation, signals of opportunity, visual navigation, and magnetic navigation,” he asserted, saying that the MagNav demonstration was an important step toward making that a reality.
Also, unlike GPS, the MagNav system relies solely on the equipment carried on the aircraft and not on external equipment like satellites or communication equipment, both of which can be targeted or jammed.
Additional benefits of a future, more mature MagNav system would include the fact that the earth’s crustal magnetic field remains fairly stable over time, plus they can also be found in the world’s oceans, which cover two-thirds of the earth’s surface; this feature could make MagNav even more beneficial to the military, which operates at sea at least as often as over land.
The gains seen in the AI-MagNav trial, according to the Air Force announcement, are potentially the first critical step toward implementing the system across such military platforms as aircraft, submarines, uncrewed aerial systems (UASs), and hypersonic glide vehicles.
Personnel from the Air Force Research Laboratory Sensors Directorate, Air Force Institute of Technology Autonomy and Navigation Center, MIT, and MIT Lincoln Laboratory supported the AI-outfitted demonstration.