SandboxAQ and Acubed advance magnetic navigation 

As GNSS denial, jamming and spoofing threaten aviation safety, SandboxAQ and Acubed, the Silicon Valley innovation center for Airbus, have released real-world test results from a five-month, nationwide project designed to test the accuracy of AQNav.

AQNav is an artificial intelligence-driven magnetic navigation (MagNav) system. AQNav uses advanced quantum magnometers to read Earth’s crustal magnetic anomalies, like a geoohysical fingerprint, then employs large quantitative models (LQMs) to filter out electromagnetic interference and precisely determine an aircraft’s position without relying on satellite signals.

These new results come from a nationwide initiative with Acubed’s Flight Lab to test the navigational accuracy of AQNav. Meeting the aviation industry’s Required Navigation Performance (RNP) standards is necessary for deploying the system on military, commercial and civilian aircraft.

AQNav’s performance was tested under various opertional scenarios and demonstrated advanced precision, accoding to SandboxAQ. The goal was to determine whether magnetic anomaly-aided navigation could broadly meet navigation requirements for commercial aircraft. AQNav’s capabilities exceeded the accuracy required for en route travel between airports — even on the program’s longest flight.

Accuracy

To demonstrate how the real-time capable system would operate in real-world conditions, flight data was collected, reprocessed, and streamed in real time to produce statistical insights, offering representative capability data for joint team evaluation. 

Real-World Impact

SandboxAQ and Acubed focused on designing tests to mirror authentic, real-world aviation scenarios. For example: 

• Standard aircraft platform: AQNav was tested using publicly available magnetic maps aboard a standard Beechcraft Baron 58 – rather than a compensated geosurvey platform. This aircraft was modified only to accommodate the additional AQNav instrumentation – no extensive electromagnetic shielding or specialized noise isolation were used. All sensors were positioned inside the aircraft, powered by AQNav’s software to deliver a clean magnetic signal. 
• Use of a publicly available map. For all flights, AQNav researchers used the publicly available North American Magnetic Anomaly Map (NAMAM), which covers the U.S., Canada, parts of Mexico and surrounding oceanic regions. 
• Unfiltered flight paths: Flight operations spanned diverse, operationally relevant routes between 200 airports across the entire continental U.S. (Fig. 1), without filtering based on magnetic anomaly strength, magnetic map quality, or favorable geomagnetic gradients. More than 150 hours of flight data was collected.
• Diverse geophysical environments: Data was collected over a full range of conditions, from magnetically-rich mountains to sparsely featured plains, reflecting real-world geographies where aircraft might operate without GNSS. 
• True operational noise: Onboard, AQNav successfully filtered out the real-world interference generated by the aircraft, including electromagnetic, vibrational and other airframe-induced noise. 

Elijha Williams, AQNav’s technical engagement manager, said: “Our campaign was not about demonstrating proof of concept performance under ideal conditions, it was about proving AQNav’s viability under the noisy, messy, and unpredictable environments real pilots face every day.” 

During test flights exceeding two hours, AQNav outperformed the Inertial Navigation System (INS) without GNSS 100% of the time. During a one-hour flight over the challenging mountainous and forested terrain of California, AQNav achieved its best-observed accuracy of less than 74 meters, or roughly two-thirds the length of an American football field. 

Precision, Scale and Autonomy for the Future 

This campaign marks a significant step toward widespread adoption of AQNav in aviation. By consistently maintaining accuracy in an uncontrolled, national testbed, SandboxAQ demonstrated AQNav’s operational robustness under real-world conditions.

Andrew Sosa Sosanya, a quantum navigation machine learning engineer at SandboxAQ, highlighted the impact of the data collected: “Thanks to Acubed, the U.S. Air Force, and other partners, we’ve accumulated a highly relevant MagNav dataset. This creates a flywheel effect—the more data we gather, the faster we can improve model accuracy across diverse mission profiles.”

AQNav is also undergoing testing with Boeing, a U.S.-allied air force, and as part of NATO’s 2025 DIANA cohort.