Adaptive neural network prediction feedback for MEMS-SINS during GPS outage

The overall performance of standalone MEMS-SINS is dramatically degraded during GPS signal outages due to the highly nonlinear drift of the inertial sensors' measurements. A method of RBF-ANN prediction feedback for MEMS-SINS during GPS outage is presented in this paper. The RBF-ANN module is then trained to predict the MEMS-SINS error during GPS availability and provide accurate navigation data of the moving platform during GPS outage. The car test results indicate that the proposed adaptive neural network prediction feedback can efficiently provide corrections to the standalone MEMS-SINS predicted navigation error. During the car experiment, a total of 4 outages were intentionally introduced with intervals of less than 50 seconds. The average position errors are 3.8 m, average velocity errors are 0.6 m/s and average attitude angle errors are 0.5° during GPS signal outages.