Abstract: | Most navigation techniques rely on a model or map of the world. The drive for GPS-alternative navigation technologies has created a need for self building world models. Having a high quality world-wide model or map of a useful navigation signal such as the Earth’s magnetic field is often not feasible or practical. Data may be of varying quality, or missing entirely. Consequently, navigation systems need to be able to operate with incomplete models, and furthermore, contribute new data to these models whenever possible. In this paper we demonstrate the usefulness of applying Gaussian Process Regression (GPR) techniques to create self building world models to improve navigation performance. We demonstrate the power of GPR methods in creating covariance maps which can be utilized by a navigation filter. We test the GPR method with magnetic anomaly data using real maps and show the contributions possible with sparse data collects. Finally, we conduct a full navigation simulation in which sparse flight lines over a magnetic anomaly map constrain a magnetic anomaly navigation system to around 100 meters RMS error despite flight line data spaced 100 kilometers apart. |
Published in: |
Proceedings of the 2017 International Technical Meeting of The Institute of Navigation January 30 - 2, 2017 Hyatt Regency Monterey Monterey, California |
Pages: | 1083 - 1110 |
Cite this article: | Canciani, Aaron, Raquet, John, "Self Building World Models for Navigation," Proceedings of the 2017 International Technical Meeting of The Institute of Navigation, Monterey, California, January 2017, pp. 1083-1110. https://doi.org/10.33012/2017.14966 |
Full Paper: |
ION Members/Non-Members: 1 Download Credit
Sign In |