Enhancing the integration of the GPS/INS during GPS outage using LWT-IncRGRU

Abstract

A common technique for navigation and positioning applications is the Global Positioning System (GPS)/Inertial Navigation System (INS) integration, which combines the strengths of GPS and INS to offer accurate and reliable information. However, the performance of the GPS/INS integration deteriorates during a GPS outage, which happens when natural or artificial factors block the GPS signal. The novelty of this paper is improving GPS/INS integration performance during GPS outages using Incremental Regularized Gated Recurrent Unit (IncRGRU) learning and Lifting Wavelet Transform (LWT). Incremental learning is a learning paradigm that can update the model parameters online from streaming data without forgetting the previous ones. Moreover, regularization is a technique that improves the network's generalization and avoids overfitting by adding some constraints or penalties to the model. In this way, the GPS signal is modeled by IncRGRU learning, and the Kalman filter corrects the INS output. Furthermore, LWT removes the noise from the sensors' signals. This algorithm has lower complexity and can work in real-time compared to conventional wavelet transforms. The performance of GPS/INS integration during GPS outages and the accuracy and robustness of GPS/INS integration is significantly improved by using LWT-IncRGRU on real-world datasets. The positioning errors are reduced by an average of 76% during GPS outages, and an average of 69% improves the GPS/INS integration compared to existing methods.

Keywords

• GRU
• Incremental learning
• GPS/INS integration
• Multi-objective regression
• Lifting wavelet transform
• Allan variance