A Robust Position Approach Based on Masked KalmanNet for GNSS/LEO/INS Integrated Navigation System

In the domains of the Internet of Things (IoT) and intelligent transportation, the importance of integrity monitoring has become increasingly prominent with the rapid development of autonomous driving. In complex environments, the redundancy of effective observations often fails to meet the stringent requirements of integrity monitoring. Moreover, due to the effects of multipath (MP) and Non-Line-of-Sight (NLOS), receiver noise frequently deviates from the Gaussian distribution, thereby significantly degrading the performance of traditional Global Navigation Satellite System (GNSS) / Inertial Navigation System (INS) integrated navigation algorithms. Capitalizing on the rapid deployment of Low Earth Orbit (LEO) satellite constellations, this paper proposes a novel integrity monitoring algorithm for GNSS/LEO/INS tightly coupled (TC) systems within the KalmanNet framework. First, LEO satellites provide enhanced observational redundancy, thereby improving the reliability of integrity monitoring. Second, the proposed algorithm relaxes the Gaussian noise assumption and instead utilizes a masked Convolutional Neural Network (CNN) - Long Short Term Memory (LSTM) - Attention (CLA) network to learn features from the observation data. Experimental results demonstrate that the proposed algorithm achieves root mean square error (RMSE) reductions of 55.53%, 51.74%, 36.42% and 29.50% compared to Standard Point Positioning (SPP), TC IGG3, RLAKF and BSRMCL algorithms, respectively.