Calibration and Compensation of Scale Factor Non-linearity and Non-Orthogonality Errors for Dynamically Tuned Gyroscope (DTG)

Abstract

Inertial navigation system (INS) is utilized in several applications such as missile guidance, space navigation, and marine navigation. An efficient calibration method for improving the inertial navigation system accuracy is presented. As the vital error sources in the inertial navigation system are associated with the deterministic errors of the inertial measurement unit (IMU), the proposed technique precisely determines the calibration parameters to reduce these errors, especially the gyro’s scale factor, and non-orthogonality error. In recently proposed calibration methods, the scale factor is determined by the output/input relationship linear fitting. Although the determined scale factor meets the requirement of various navigation systems to some extent it doesn’t fit the high accurate ones, such as guided missiles and marines. That’s because the gyro damping effect is changed with different input angular rates which causes the gyro scale factor varying. The presented calibration method tackles this phenomenon by assigning different weights to each input rate through a weighted linear regression fit. Moreover, the gyro nonorthogonal error which comes from the imperfection gyro mounting is merely determined by the lateral coupling signal. But the fact is that the lateral coupling signal is not induced from the gyro non-orthogonal error only but also comprises the gyro signal which is directly proportional to the centripetal acceleration caused by the applied angular rates. Even though this signal is tiny but it will be accumulated for the long-time navigation system and degrades its accuracy. The presented calibration method utilizes a lateral accelerometer to realize that signal and tear out it to accurately obtain the nonorthogonal error. Finally, a laboratory test for the proposed method was carried out to ensure its effectiveness. Where the actual applied rates are determined twice, once with the built error model by the presented calibration method and the other by the traditional one.