DYNAMIC CALIBRATION OF ASSEMBLED STRAPDOWN INERTIAL NAVIGATION SYSTEMS ON A RATE TABLE

Abstract

The paper aims to present a review of inertial navigation system (INS) calibration technique developed for INS of different types and accuracy grades. The technique was developed at Moscow State University, in the laboratory of navigation and control. Several Russian INS manufacturers have been implemented this technique in their production process. The conventional parameters of inertial sensor’s output measurement model are estimated in a simple calibration experiment consisting of consequent rotations of the system for up to half an hour around horizontal axis. INS instrumental axes are consecutively aligned with the rotation axis. All conventional systematic INS error components produce specific attitude and acceleration errors when the latters are computed using inertial sensor’s measurements. As a result, these components can be separated from each other and estimated quantitatively. The principal advantages of the proposed approach are as follows: calibration can be conducted on a single-axis low-grade turntable (neither providing high accuracy of angular motion nor having angular or rate sensors); no predefined plan of operations in the calibration experiment is required, as well as no its specific properties are to be sustained; a unified estimation algorithm for the entire calibration dataset with no specific computations for any specific stage of the experiment; wide range of opportunities to modify models used, e.g. including additional parameters inherent to some types of INS and turntables. In our study, we formulate the estimation problem for INS calibration, and classify possible modifications of sensor’s error models and error equations that we have tested so far in real INS calibrations. Error models are represented as a linear dynamical system with measurements. Its state vector contains: attitude errors, a set of conventional INS error parameters and some additional instrumental errors. After that, the system state vector is estimated using an optimal algorithm (conventional Kalman filter). The observability of the problem considered depends on the rotation pattern of the inertial unit, and the experiment described above ensures that the system becomes observable.