A Novel System-Level Calibration Method For Gimballed Platform Imu Using Optimal Estimation

An accurate calibration of inertial measurement unit errors is increasingly important as the inertial navigation system requirements become more stringent. developing calibration methods that use as less as possible of imu signals has also become a challenge. in the present work, first a mathematical model of a 6-dof gimballed imu in space-stabilized mode is presented. it is considered as held stationary in the test location incorporating 15 different error sources, including accelerometers bias, scale factor error, gyros drift, initial alignment error, and imu case installation error. using kinematic relations between imu platform, imu body, and imu platform centered inertial reference frame, six differential equations of the only system-level imu velocity and gimbal angle are derived. then the extracted model is validated for error-free case using sim-mechanics matlab simulink tools to evaluate the introduced mathematical model. simulation results for 24 hours point out the correctness of the developed model in error-free case. the imu error analysis methodology incorporates 21-state extended kalman filter and uses only system-level imu velocity and gimbal angle measurements taken during one and a half hour with 9 platform attitudes test to estimate imu error sources. without the need to install imu at rotating table, different platform attitudes are achieved using consequent rotations of gimbals. imu error sources estimation is accomplished off-line. this paper describes the design and test results of a new gimballed imu calibration method without using a rotating table, and error model development methodology formulated to support the design and test of ekf algorithm and two optimal smoothers: forward-backward andrts. results obtained from ekf implementation indicate that the technique is comprehensive and accurate, and requires less specialized test equipments. also, results show that constant states are not smooth-able. additional measurements are still helpful for refining an estimate of a constant state.