Clock synchronization is vital for navigation systems that require a stable, available, and resilient common time reference to ensure the positioning and timing performance for all users. The composite clock approach, adopted by systems such as GPS (Global Navigation Satellite System) and EGNOS (European Geostationary Navigation Overlay Service), addresses these needs by ensembling multiple physical clocks through Kalman filters. Such algorithm can efficiently detect anomalies in individual clocks and can then adapt each clock’s contribution to the overall system time. Yet, while clock behaviour can be seen as non-linear, non-Gaussian and non-stationary, conventional filtering methods implemented today often assume the opposite. There are also significant challenges related to the initialization of such filters and handling the clock ensemble reconfiguration.
To address these challenges, this study proposes to introduce the concept of deep learning assisted GNSS composite clock, leveraging supervised learning methods to dynamically estimate filter parameters in real-time and non-linear environment, also considering partial information. Based on recent advances (based on Revach et al., KalmanNet: Neural Network Aided Kalman Filtering for Partially Known Dynamics, 2022), the proposed approach embeds a dedicated recurrent neural network within existing composite clock filters enhancing their ability to capture complex clock behaviours and ensemble reconfigurations.
By comparing the performance of traditional filtering methods with this novel deep learning integrated system, it will be demonstrated how this hybrid approach can yield more reliable, precise, and robust timescale. The resulting composite clock solution not only mitigates the limitations of the classical Composite Clock algorithms, but also provides a powerful tool for dealing with uncertainties in clock dynamics. Ultimately, this work highlights the potential of machine learning to enhance clock ensembling strategies, opening the usage of artificial intelligence in next-generation navigation systems.
