LAUSR

This paper presents a dual-model solution for global positioning system (GPS)/inertial navigation system (INS) during GPS outages, which integrates with multiple-decrease factor cubature Kalman filter (MDF-CKF) and random forest (RF) that can be used for modeling and compensating the velocity and positioning errors. The prominent advantages of the proposed solution include: 1) filter divergence is restrained and robustness is improved with the proposed MDF-CKF method and 2) the error compensation accuracy of the RF-based dual model is higher than a normal artificial neural network-based single model. The process of the proposed solution contains: 1) the proposed MDF-CKF algorithm is employed for GPS/INS information fusion when GPS signal is valid; 2) in the meantime, the velocity, acceleration, and specific force data from IMU and INS are separately used to train the RF; and 3) when GPS outage occurs, position and velocity errors are predicted by the RF-based dual model. The experimental results show that: 1) the maximum improvement of the proposed MDF-CKF in position estimation accuracy against the traditional algorithm is 83.6%; 2) the RF-based dual model can effectively suppress the divergence than the radial basis function and INS-only mode; and 3) the dual model performs better than a single model for error modeling and compensation.