AI-Powered GNSS Receivers for High Fidelity Measurements
Last Updated: 06/06/2023 13:32 Created at: 06/06/2023 13:28
Final Presentation of NAVISP Project EL1 034 now available:
On Tuesday, May 30th, 2023, GMV NSL Ltd and Thales Alenia Space presented the results of the NAVISP EL1 034 project "AI-Enabled Baseband Algorithms For High-Fidelity Measurements". Over 30 people from industry and research institutes followed the interesting presentation and the subsequent interactive Q&A session.
In the world of autonomous vehicles and machine control, precision is everything. But what happens when the environment throws a curveball? Despite advancements in Global Navigation Satellite Systems (GNSS), performance in tricky scenarios like urban canyons and low C/N0 conditions still leaves room for improvement. The challenge lies in obtaining clean measurements by managing transfer functions and minimizing environmental issues, which are vital for advanced Positioning, Navigation, and Timing (PNT) engines such as Kalman and particle filters, DPE, and PPP/RTK. Conventional approaches struggle in harsh environments, but leveraging GNSS-correlated 3D maps offers a promising solution. However, this introduces complexities in maintaining knowledge, adapting to dynamic environments, and managing computational resources. Artificial intelligence (AI) provides an effective solution by offering adaptive configurations, dynamic parameter adjustments, and real-time decision-making capabilities to bridge these gaps.
The initial phase of the activity involved conducting a comprehensive review of state-of-the-art techniques, evaluating algorithms at various stages of the receiver chain based on factors such as complexity, labelling techniques, integration capabilities, and alignment with real data. Three algorithms were selected in the preliminary design phase to enable accurate carrier phase measurements and mitigate the impact of multipath interference in developing a reliable tracking loop. The first algorithm used a supervised Multipath (MP) regression technique with a Convolutional Neural Network (CNN) to estimate pseudorange and mitigate multipath interference after correlation. The second algorithm employed post-correlation processing with supervised regression to generate an ideal Auto-Correlation Function (ACF) output, enhancing its understanding of signal features and characteristics through CNN analysis. The third algorithm, utilizing Gradient Boosted Trees, estimated pseudorange error relative to the reference Rx 1 by utilizing a RINEX format and applying supervised pseudorange regression on four receivers. Both variations of Algorithm 3 (with and without AI) surpassed the U-Blox solution by several meters in each ENU component at a 95% confidence level.
In conclusion, the AInGNSS project exceeded the state-of-the-art by applying various AI algorithms for GNSS improvement, collecting extensive data in challenging environments, considering multiple algorithms and signals, utilizing a multi-antenna system, while its unique focus on IQ data further distinguished it from the state-of-the-art. Although occasional significant improvements in PVT performance were achieved, particularly in controlled environments, the overall impact of AI on the algorithms' performance was not substantial compared to classical GNSS PVT methods. Nonetheless, the project's outcomes provide valuable knowledge for successful algorithm development and the establishment of a multi-purpose, multi-device testbed for data acquisition and processing.
The project was carried out in the scope of NAVISP Element 1, which is dedicated to technology innovation of the European industry in the wide PNT sector.
More detailed information can be found in the slides of the Final Presentation.