Hide/Show Apps

Development and comparison of the extended kalman filter and unscented kalman filter for both tightly coupled and loosely coupled INS/GNSS integration by using MEMS IMU

Güreş, Onur
Navigation technologies evolved with improvements for the last two decades. Navigation devices are easy to buy and satellite navigation is available all around the World. The key part for designing a navigation system is the algorithms used inside of it. This thesis focuses on the comparison between the proven reference unit and author designed navigation algorithms by using same IMU and GNSS inside of the reference unit with collected field data. In this thesis, both Extended and Unscented Kalman Filter estimation techniques were developed by applying tightly and loosely coupling methods in MATLAB. As the reference unit NovAtel SPAN IGM-A1 was selected. This unit is a self-proven unit which also has one of the market’s best integration algorithm SPAN. The unit includes ADIS-16488 MEMS IMU and a NovAtel OEM6 GNSS Receiver with single antenna. The selected standalone units for developed algorithms’ confirmation were also ADIS-16488 as MEMS IMU and NovAtel OEM6 GNSS receiver. Hence, a comparison between the reference unit and the author designed filters was made. Car navigation trial results were compared for different estimation and integration methods. Results showed that tightly coupled EKF was the best among developed algorithms. The results also showed that tightly coupling integration outperformed both UKF and EKF loosely coupled INS/GNSS integration. The unstability and effects of scale factors on UKF were also inspected.