Nonlinear Estimation for Gyroscope Calibration for the Inertial Pseudo Star

The gyroscope calibration problem is addressed for the Inertial Pseudo Star Reference Unit. A nonlinear scale factor error is described by two gyro error states (an amplitude and a decay constant) which enter the dynamics in a nonlinear fashion. A comparison is made between the extended Kalman filter (EKF) and the maximum likelihood system identification (MLSI) method for solving the resulting nonlinear estimation problem. The principal advantage of the MLSI over the EKF is improved convergence. By using the MLSI instead of the EKF, a filter designer developing a nonlinear estimation routine for gyroscope calibration would require less trial and error to achieve an acceptable design. The U-D factorization filter is combined with a separate bias estimation framework, and the results are extended to the nonlinear estimation algorithms under consideration. The separate bias U-D factorization filter requires fewer computations than the standard U-D factorization filter while retaining the advantageous numerical qualities of that approach. Thesis Supervisor: James H. Murphy Title: Charles Stark Draper Laboratory Thesis Advisor: Professor Wallace E. VanderVelde Title: Department of Aeronautics and Astronautics