12 / 2023-03-20 15:48:59

GA-LSSVR Assisted GNSS/INS Positioning In Occlusion Region

integrated navigation; least squares support vector regression; genetic algorithm; IGGIII

The accuracy of navigation and positioning in urban buildings is seriously affected by satellite multipath and non line-of-sight signal (NLOS) errors. Satellite signals are easily lost under high kinematic motion states, and the position and attitude errors reckoned by pure inertial sensor can rapidly diverge over time. This paper proposes a navigation scheme for occluded areas based on least square support vector regression (LSSVR), and genetic algorithm (GA) is selected to seek the global optimal solution of regularization parameters and bandwidth of kernel function. The proposed scheme uses the position increment output from the GNSS receiver and INS to train the model parameters during normal observation. When the signal loses lock, the integrated filtering estimation of the GNSS receiver pseudo position and pseudo velocity predicted by the prediction model and the position and velocity reckoned by the INS is adopted, and a IGGIII robust factor is introduced to suppress the impact of the coarse difference in the predicted position of the model on the adopted navigation resolution. The results indicate that the algorithm can accurately predict the position increment of the GNSS receiver, and adjust the covariance of the outliers in the predicted values to achieve effective suppression, thereby maintaining a continuous, stable, and high-precision navigation and positioning of the GNSS/INS integrated navigation system in areas where satellite signals such as urban buildings are blocked.