1.Possibilities of AI Algorithm Execution in GNSS.pptx

Possibilities of AI Algorithm Execution in GNSS

1. Possibilities of AI Algorithm Execution in GNSS

2. Contents 1. Possibilities of AI Algorithm Execution in GNSS 2. Abstract 3. Introduction 4. Possible Approaches Based on Intelligence in GNSS 5. Stage I: At RF Front End 6. Stage II: At Pre-correlation Level 7. Stage III: At Post-correlation Level 8. Stage IV: At Navigation Level 9. Threats and Risk-factors Using AI Algorithms in GNSS 10. Conclusion 11. Acknowledgements 12. References

3. Possibilities of AI Algorithm Execution in GNSS Writer 1.Darshna Jagiwala , Woman Scientist, DST and 2.Shweta N. Shah, Assistant Professor, SVNIT, India. Event: URSI - RCRS 2022, IIT (Indore), India, 1 - 4 December, 2022.

4. Abstract This study explores the feasibility of using AI algorithms in GNSS. Two primary AI methods, Machine Learning (ML) and Deep Learning (DL), are discussed, with SVM and CNN highlighted as key algorithms for enhancing GNSS position accuracy. The research examines the integration of AI at various stages of GNSS receivers and discusses the associated risks and threats.

5. Introduction GNSS System Technologies GNSS, including GPS, GLONASS, GALILEO, and BeiDou , provide global real-time services using Doppler effects and tracking loops. Advanced technologies like DGPS, AGPS, RTK, and e-Dif enable centimeter-level positioning. Benefits of AI in GNSS AI, particularly ML and DL, offers solutions to enhance GNSS position accuracy. Smart models mitigate range errors caused by atmospheric conditions and other factors. GNSS System Technologies GNSS, including GPS, GLONASS, GALILEO, and BeiDou , provide global real-time services using Doppler effects and tracking loops. Advanced technologies like DGPS, AGPS, RTK, and e-Dif enable centimeter-level positioning.

6. Possible Approaches Based on Intelligence in GNSS Introducing AI into GNSS can significantly improve its performance. Four stages of GNSS receiver integration are discussed: RF Front End, Pre-correlation, Post-correlation, and Navigation levels.

7. Stage I: At RF Front End DL-based Antenna Beam forming Advanced beam forming techniques use NNs to detect signal features even at low SNRs, providing an alternative to classical methods. Neural processors can suppress interference and enhance signal quality.

8. Stage II: At Pre-correlation Level GNSS Software Receivers and AI Algorithms AI algorithms process Intermediate Frequency (IF) signals in GNSS software receivers. Various studies demonstrate improvements in signal classification and multipath detection. Literature Review at Pre-correlation Stage Studies address issues such as indoor NLOS detection and multipath interference using NN, SVM, and clustering algorithms. Improved classification accuracy and robustness have been reported.

9. Stage III: At Post-correlation Level Data Formats and AI Algorithms RINEX, NMEA data, and raw correlation outputs are used as inputs for AI algorithms. CNNs and SVMs are used for multipath detection and ionosphere modeling, enhancing GNSS accuracy. Literature Survey at Post-correlation Stage Research highlights include CNN-based multipath detection and LSTM-NN for ionosphere prediction. AI algorithms show significant improvements over traditional methods.

10. Stage IV: At Navigation Level RAIM and INS Approaches AI enhances GPS/INS integration, improving performance during GNSS signal outages. Various AI models, including RBFNN and CNN, are used for applications such as hurricane tracking and water detection. Literature Review at Navigation Level Studies review applications like hurricane tracking using CNN models and improving INS accuracy during GNSS signal outages. AI provides robust solutions in these scenarios.

11. Threats and Risk-factors Using AI Algorithms in GNSS Geographical Diversity AI models trained on regional data may not perform well globally due to variations in atmospheric conditions. This is a major issue when merging AI with GNSS. Data Security and Storage AI's data-driven nature raises concerns over data security and storage. Massive amounts of data require robust infrastructure and secure storage solutions. Infrastructure High-end processors and computing speeds are needed for AI-based solutions. Adequate planning and investment in infrastructure are essential. Choice of AI Module Inputs/Outputs and Transparency The quality of AI outputs depends on input data quality. Standardized data collection is crucial. Additionally, AI systems must maintain transparency and accountability, often referred to as the 'black box problem'.

12. Conclusion AI algorithms are increasingly used across various sectors, including space exploration, agriculture, and healthcare. This paper reviewed AI applications in GNSS at different receiver stages, highlighting the benefits and challenges. SVM and CNN are frequently used to enhance GNSS accuracy. However, implementing AI in GNSS poses challenges such as data quality, transparency, and high computational requirements.

13. References 1. Zidan , J. et al. (2021). GNSS Vulnerabilities and Existing Solutions. IEEE Access, 9, 153960-153976. 2. Roddy , D. (2006). Satellite Communications. 4th ed. McGraw-Hill. 3. Siemuri , A. et al. (2021). Machine Learning Utilization in GNSS. 2021 International Conference on Localization and GNSS. 4. Jagiwala , D.D. et al. (2021). NavIC Performance Observation on Low Latitude Region. 2021 2nd International Conference on Range Technology. 5. Ramezanpour , P. et al. (2020). Deep-learning-based beamforming . IET Signal Processing, 14(7), 467-473.