Improved Gaussian process regression-based method to bridge GPS outages in INS/GPS integrated navigation systems

•In order to improve the performance of the INS/GNSS integrated navigation system during GNSS outages, the strong GPR predictor is designed to provide pseudo-GNSS signals.•Experiments show that the strong GPR predictor outperforms the weak GPR (that is the traditional GPR) predictor in terms of estimation accuracy and stability. During GNSS outages, the performance of the INS/GNSS integrated navigation system is significantly reduced. To provide the pseudo-GNSS signals for the integrated navigation system, an improved Gaussian process regression (GPR)-based method is investigated. The relationship between the outputs of the inertial measurement unit (IMU) and the GNSS position information is modeled by the improved GPR. Rather than the traditional GPR algorithm, the proposed GPR algorithm establishes a strong predictor, which improves estimation accuracy. Field test data was collected on a land vehicle navigation system to evaluate the proposed method. The comparison results indicate that the proposed method can effectively improve the performance of the INS/GNSS integrated navigation system during GNSS outages.