Researchers at the Royal Belgian Observatory and the state's leading research institute of precision geodesics at the Chinese Academy of Sciences (CAS) have developed a new algorithm that can boost achievable accuracy from the existing Global Navigation Satellite System (GNSS).
“Our SVM [Support Vector Machine] The model represents a paradigm shift in ambiguity verification. “Co-author Jianghui Geng, who claims that no changes to satellite hardware or planetary receivers are required.” Using machine learning, we improved accuracy, but provided a scalable solution from the scientific monitoring of GNSS applications.
GNSS' new approach to “resolving ambiguity” could potentially provide more accurate positioning, researchers show. (📷: Guo et al)
The team's algorithm is designed for what is called “ambiguity resolution.” This is the process of resolving uncertainty in the carrier phase signal to improve the accuracy of GPS or other GNSS corrections. Compared to existing approaches, the team says that support vector machines improve both accuracy and reliability when providing accurate point positioning Ambiguity Resolution (PPP-AR) required for high-precision tasks such as autonomous vehicle navigation.
Trick: Integrate seven different diagnostic metrics into a single model. A machine learning model trained and further tested with real-world GPS data has increased the team to 92% success rate of vehicle navigation for 82% success rate (temporary loss or decomposition of GNSS signals) in kinematic scenarios.
The model of a team trained with real data takes into account seven different metrics. (📷: Guo et al)
“While machine learning-based models improve the success rate of ambiguity verification, we acknowledge that approximately 5% of misresolved ambiguity cannot be identified in the current model, particularly for solutions for convergence periods.
The team's work is published in the journal Satellite navigation Under open access conditions. The SVM model itself is available from the author of the paper upon request.
