|Abstract:||Safety-of-life applications often rely on the PNT solutions provided by GNSS receivers, as well as the informed uncertainty of the estimated states. According to the 2008 US Federal Radionavigation Plan, integrity can be defined as the "measure of the trust that can be placed in the correctness of the information supplied by a navigation system". Therefore, measurements of integrity show, in general lines, the similarity between the provided uncertainties and the actual error of the solution. In this context, this paper approaches the often over-optimistic GNSS error model employing machine-learning algorithms in order to improve integrity measurements for the end-user by maximizing the information output of the filter. The possibility of achieving high-integrity solutions in a low-cost mass-market platform creates an opportunity for a large realm of applications, such as beyond visual line of sight UAV operation, vehicles in precision agriculture, marine applications, and others. This paper makes use of Allystar’s HD8040D dual-band GNSS receiver RTK solution, even though the proposed methods can be applied for any processing strategy such as single-point positioning, differential GNSS, or precise point positioning. As results, even though the EKF-derived integrity measurements outperformed the machine learning strategy in terms of nominal operation, the information derived from the EKF was lower than the other methods. This indicates that information from the measurements is lost at the filter level when attempting to increase the nominal integrity behavior. The Neural Network (NN) model was able to increase the information metric by almost 20-fold, as well as reducing the unavailable points when compared with the other machine learning models. For applications that rely on information from the filter, the Decision Tree algorithm presents a middle ground between the EKF and NN solutions. All machine learning models considered are suitable for real-time, and count with hardware acceleration in some embedded platforms.|
Proceedings of the 35th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2022)
September 19 - 23, 2022
Hyatt Regency Denver
|Pages:||2694 - 2700|
|Cite this article:||
Mendonca, Marco, Jokinen, Altti, Yang, Ryan, Hau, Gary, Tseng, Yi-Fen, "Improving Integrity and Information Output on a Low-Cost GNSS Platform Using Machine-Learning Algorithms," Proceedings of the 35th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2022), Denver, Colorado, September 2022, pp. 2694-2700.
ION Members/Non-Members: 1 Download Credit