ИСТИНА

The paper concerns one simple method of calibration of an assembled navigation grade inertial measurement unit (IMU) using a low grade single axis turntable. The method was presented at ICINS 2010, ICINS 2013 [1,2], and is now being used in industry for 3-4 years. It appears to work well in practice. One of special points of this method is the situation, when the IMU is displaced from the axis of rotation. In some cases this fact can be neglected, but in some cases not. In [2] it was shown that parameters of this displacement need not to be measured prior to the experiment, but can be estimated automatically similar to the rest of IMU parameters. So the device can be placed arbitrarily onto the stand, and no care of displacement should be taken while conducting the experiment. This work concentrates primarily on aspects of observability and estimation accuracy of IMU displacement (which influence the accuracy of the calibration itself) depending on other parameters of the experiment, and shows more variety of experimental results than in previous publications. The idea of calibration is as follows. Please, see Fig. 1 for the schematic diagram of the calibration experiment. It consists of three cycles of rotation round a nearly horizontal axis of a low grade turntable, for about 10-20 minutes each. Note that there are no special requirements to these rotations except for a non-zero angular rate. Saying the turntable can be of low grade we mean the following: - it has one single axis of rotation; - no any sensors measuring the rate of the turntable (not used if present); - no rate stabilization required; - no predefined precise angular positions or angle measuring available; - no programmable actuator control required (but more convenient in practice); - no accurate alignment possible, neither for IMU with respect to a stand, nor for the stand to the horizon; - the axis of rotation is firmly fixed and no vibrating within the desired accuracy of IMU. For inertial sensor errors we use linear model in small terms, assuming that some pre-calibration step was performed to provide errors being small enough that second order terms can be neglected. This pre-calibration step can be based on the same experimental data (no additional operations required in a test bench), and it may be performed in different ways, but this is not the subject of this work. Thus, the model includes accelerometer and gyro small constant biases, errors of scaling factors and small angles of sensitive axes misalignment. In addition, the model can incorporate dynamic drift (g-sensitivity) coefficients and other parameters of an experiment according to an IMU type. Using error model and kinematic equations, we then can form a linear state-space system which becomes fully observable in the above experiment. This allows us to get an optimal estimation of all desired parameters using, for example, Kalman filter or other conventional estimator. Geometric parameters of the displacement of the accelerometer proof masses with respect to the axis of rotation can be present amongst other parameters in the system. Until recent time there were no full comprehension, whether these parameters can be estimated accurately or not in different situations. This work summarizes the experience of processing a variety of real experimental data regarding to observability issues of that displacement. Results incorporate estimation based on different IMU types (MEMS, FOG, RLG) and different rotation profiles. References [1] N. Vavilova, A. Golovan, N. Parusnikov, I. Sazonov, Proc. ICINS 2010, pp. 71-72 [2] A. Kozlov, I. Sazonov, N. Vavilova, N. Parusnikov, Proc. ICINS 2013, pp. 126-129