Machine-Learning to model GNSS systems

A decade ago, EUROCONTROL developed 'EURONOTAM', an EGNOS performance prediction tool to alert users to potential EGNOS under-performance. It uses a simplified 'meta model' based on satellite and RIMS geometries, limiting its accuracy as it didn't consider many performance-influencing factors.

To address these limitations, a new EGNOS performance prediction prototype has been developed, funded by the European Space Agency under a NAVISP contract.
This prototype uses state-of-the-art machine-learning techniques to forecast EGNOS performance by incorporating a wide variety of parameters, including weather and space-weather, thus moving beyond the constraints of traditional models.

Trained on extensive real EGNOS data, the ML-based prototype emulates the outputs of the EGNOS Central Processing Facility (CPF). It analyzes observation data and navigation broadcast files from RIMS along with system and simulation configurations to generate UDRE and GIVE values in the EGNOS Navigational Overlay Frame (NOF) message, offering significantly more accurate predictions.

Additionally, the ML-based prototype provides enhanced data analysis capabilities to examine the sensitivity of NOF outputs to individual input parameters, such as RIMS positions. This aids in design decisions and improves system performance across various scenarios.

Please check the Agenda and register for the event!