An unscented Kalman filter for in-motion alignment of low-cost IMUs

This paper describes the alignment of low-cost inertial measurement units (IMUs) using an unscented Kalman filter (UKF), which allows large initial attitude error uncertainties. The state vector includes position, velocity, attitude, and sensor biases and scale factors. Position information from the differential global positioning system (DGPS) solutions is used as measurements. Test results with a micro-electrical-mechanical-systems (MEMS) IMU showed that the alignment converged within 50 s with RMS values of 0.093/spl deg/, 0.094/spl deg/ and 0.388/spl deg/ for roll, pitch and heading, respectively. The UKF works well even in cases of large initial attitude errors (about 30/spl deg/) not only for heading but also for roll and pitch. Therefore, the UKF is a unified approach to handle large and small attitude errors of an inertial navigation system (INS) seamlessly.