Final Presentation available for NAVISP EL1-104 bis “Q-PAD: AI for Anomaly Detection in Multi-Sensor PNT”

On 20th October 2026, TeleOrbit GmbH and RobNav GmbH presented the results of NAVISP EL1-104 bis. The activity investigated how artificial intelligence could detect anomalies in quantum magnetometers and help determine how much their measurements should be trusted within multi-sensor positioning, navigation and timing solutions.

The consortium developed the Q-PAD (Quantum-assisted Positioning and Anomaly Detection) end-to-end simulation environment, covering UAV, aircraft and low-Earth-orbit scenarios. It combined quantum- and classical-sensor models with machine-learning anomaly detection and an adaptive navigation engine. The trained models produced a continuous anomaly score, allowing the sensor-fusion algorithm to reduce its reliance on magnetometer observations when their estimated quality deteriorated. 

The approach was assessed using simulated datasets and a hardware-in-the-loop testbed incorporating a nitrogen-vacancy magnetometer and a three-dimensional Helmholtz coil. The results showed that anomaly-aware sensor fusion improved navigation performance during magnetic disturbances. However, models trained using synthetic data alone did not transfer successfully to the available physical instrument, highlighting the need for real measurements, improved simulation fidelity and more mature sensor hardware. 

The activity generated a reusable simulation platform, trained anomaly-detection models, an adaptive sensor-fusion module and a hardware-in-the-loop testbed. Recommended improvements include greater sensor sensitivity, reduced temperature dependence and space-certified supporting electronics. 

The activity was fully funded by ESA NAVISP Element 1, whose objective is to innovate PNT technologies, products and services.