Multiple Model Kalman Filtering for MEMS-IMU/GPS Integrated Navigation

Abstract

The conventional Kalman filtering algorithm requires the definition of a dynamic and stochastic model, and errors of low cost MEMS-IMU are likely to vary temporally. So the conventional Kalman filter exists limitation in MEMS-IMU/GPS integrated navigation. This paper presented the use of multiple model adaptive estimation(MMAE) where multiple Kalman filters were run in parallel using different dynamic or stochastic models in MEMS-IMU/GPS integrated navigation. And the modified multiple model Kalman filter was used in order to solve the limitation of multiple model adaptive estimation(MMAE). Using static tests, the algorithm designed was validated. The test results show that the modified multiple model Kalman filter can improve performance of MEMS-IMU/GPS integrated navigation system, compared to the conventional Kalman filtering algorithm. And using the designed algorithm, the positioning accuracy is better than 5m and velocity accuracy is better than 0.1m/ s ² , and the attitude errors are less than 0.5 degrees on the static condition.