NEURONAV: AI-Enhanced Maritime Navigation System

　　NEURONAV is an advanced AI‑enhanced positioning system designed for maritime navigation, aimed at improving the positioning accuracy of Global Navigation Satellite Systems (GNSS) in complex environments.

　　The system was jointly developed by Romania’s InSpace Engineering (RISE) and the Maritime Hydrographic Directorate (MHD), with funding from the European Space Agency’s NAVISP Element 2 program, and is designed to address key challenges in maritime navigation, including multipath errors, signal interference, and limitations in ground‑based positioning accuracy.

　　The development of this system draws on the accumulated experience from multiple European Space Agency projects, including large-scale data‑collection campaigns conducted in the Black Sea, the Aegean Sea, and the Danube River Basin. These operations, carried out using onboard instruments and dedicated vessels, involve the collection and analysis of GNSS data to assess performance, detect interference, and characterize multipath effects.

　　At a recent event hosted by the European Space Agency, Sergiu-Stefan Mihai and Ileana Mihu of RISE presented the project’s final results. Mihai stated that NEURONAV is centered around a machine-learning model—specifically, a convolutional neural network (CNN)—designed to predict and correct positioning errors caused by multipath effects.

Unique data processing approach

　　The system first collects GNSS-derived parameters, including satellite azimuth, elevation angle, carrier-to-noise ratio (C/N₀), and pseudorange residuals. These inputs are transformed into a Cartesian error-space representation and fed as the input matrix to the CNN. The network processes this data through convolutional layers, max-pooling, and fully connected regression layers to predict position corrections for the X, Y, and Z coordinates.

　　The NEURONAV hardware features an integrated Septentrio mosaic‑X5 multi‑constellation, multi‑frequency GNSS receiver and is equipped with a Jetson Nano single‑board computer for AI processing. Data acquisition, processing, and storage are fully automated, enabling continuous model training and validation without human intervention.

　　During test campaigns conducted in the Black Sea using an MHD hydrographic survey vessel, the results of multiple trials demonstrated consistent performance. Even under conditions of intense RF interference—marked by escalating GNSS jamming and spoofing following the outbreak of the conflict in Ukraine—the tests still revealed measurable improvements in positioning accuracy. For instance, in one validation exercise, the mean position error decreased from 1.507 meters in the receiver’s raw output to 1.268 meters after NEURONAV correction, while the 95th‑percentile error improved by more than 25% compared with uncorrected GNSS data.

　　NEURONAV can enhance the performance of existing GNSS receivers and ensure safety and reliability through on-board data processing, offering potential benefits to both professional and recreational mariners. Mihai stated that future work will focus on improving the system’s performance under strong interference, expanding the dataset, and integrating additional navigation sensors to bolster robustness.