With over 6 billion GNSS receivers in operation worldwide, ionospheric fluctuations can degrade GPS accuracy by 2–10 meters under moderate conditions and over 20 meters during severe space weather, costing industries up to an estimated $1–2 billion daily. Current solutions are limited due to their need for sophisticated ionospheric models. However, effectively modeling the ionosphere remains significantly complex and resource intensive due to its nonlinear and rapidly shifting nature, along with the high costs and technical difficulties of current multi-frequency and ground-based reference systems. By integrating advanced physics-informed ML on low-power embedded hardware and SDR systems, we aim for a 30–50% improvement in ionospheric corrections, reducing annual GNSS-related costs while enabling real-time sub-meter positioning.
Overcoming Challenges and Research Milestones
1
Software-Defined Radio (SDR) for Flexible GNSS Architectures
Develop SDR-based GNSS platforms with reconfigurable firmware, allowing seamless updates to correction algorithms and rapid integration of new features for enhanced security and accuracy.
2
Modeling of Ionospheric Delay and/or Scintillation Effects with Lagrangian and Hamiltonian Machine Learning Models
Leverage Lagrangian- and Hamiltonian-based ML models that intrinsically encode physical conservation laws, enabling real-time tracking of plasma flows, density gradients, and wave interactions. By embedding these physics-informed constraints directly into our ML frameworks, we achieve more robust and precise short-term corrections—particularly on low-power embedded hardware—ultimately improving accuracy in real-world GPS applications.
3
Scalability and System Integration
Design a modular system architecture that integrates diverse data sources and algorithms, enabling broad scalability—from small UAVs to large-scale aerospace fleets—and ensuring consistent, high-precision navigation across varied platforms.