Final Presentation NEURONAV: AI – Augmented Positioning System for Maritime Navigation

Last Updated: 07/08/2025 11:57     Created at: 05/08/2025 14:53

On 5th August 2025, Romanian InSpace Engineering (RISE) delivered the final presentation of the NAVISP EL2-154 “NEURONAV” project, an AI-augmented positioning system for maritime navigation. The objective was to enhance GNSS accuracy in multipath-prone environments by using machine learning to correct receiver output in real time.

The system leverages vessel geometry to predict and mitigate multipath errors, using a convolutional neural network (CNN) trained on satellite signal features such as azimuth, elevation, C/N₀, and pseudorange residuals. The AI model operates onboard, automatically preparing datasets, training, and validating without human supervision.

Testing was conducted on Romanian hydrographic vessels, with data collected across multiple campaigns. Despite challenges such as GNSS interference and RTK signal loss—largely attributed to regional conflicts—the model demonstrated consistent performance improvements. Validation showed a 25.78% reduction in mean positioning error compared to the receiver alone.

The project was supported by NAVISP Element 2 The slides of the final presentation are available here.