Том 8, № 4 (2017)

Errors of a strapdown inertial navigation system (INS) based on navigation grade fiber-optic gyros (FOG) are studied, with the INS measurement unit performing modulation rotation about the axis orthogonal to the vehicle deck. The main focus is made on the problem of reducing the time of gyro in-run drifts calibration, using the GNSS data when the INS is turned on. In addition to velocity and position measurements taken by the GNSS, phase measurements from antennas spaced on a certain baseline are included.

A Novel Adaptive Unscented Kalman Filter (NAUKF) has been developed and applied to fuse the outputs of strap-down IMU, the measurements of GPS satellites (pseudo-range and Doppler), strap-down magnetometer and a barometric altimeter, using tight coupling architecture. The proposed filter NAUKF considers the residual unmodeled noises of process and measurement as non-zero mean Gaussian white noises, estimates and compensates for the mean (bias) and covariance of the noise online, based on the principles of adaptive filtering, even if they are time-varying quantities. Employing the adaptive filtering principle into UKF, the nonlinearity of system can be restrained and the NAUKF is obtained. The noise statistic estimators designed in the proposed algorithm are built on the basis of forgetting factors which are usually calculated empirically. In this research a new calculation concept based on “Genetic Algorithm” as an optimization tool is utilized to determine the optimal values of forgetting factors. The performance and convergence of noise statistic estimators used in NAUKF are checked by Monte-Carlo simulation. The utilization of NAUKF for Tightly-Coupled INS/GPS Integration (TCI) has shown a superiority against the UKF for TCI especially during the GPS outages. The comparison is done experimentally by means of real flight trip of anUAV.

It has been demonstrated that variation in the flat shape of a MEMS gyro inertial body under inertial loads results in gap variation in capacitance transducer of sense oscillations. An algorithm for improving the design parameters of moving electrodes is proposed to keep the conversion factor of the gyro insensitive to translational vibrations directed normally to the disc plane of the inertial body, taking into account the flat shape loss.

Comparative analysis of quality has been performed for the models of total electron content which could be applied in single-frequency navigation receivers of GPS/GLONASS. The analysis is based on the evaluation of total electron content modeling errors versus its experimentally measured values represented by global ionosphere maps (GIM). The GEMTEC model jointly developed by Irkutsk State University (IGU) and Irkutsk Branch of Moscow State Technical University of Civil Aviation (MGTU GA) has proven to show the least averaged errors.

A PZT type telescope for observations of gravity gradient and lunar rotation was developed, and a Bread Board Model (BBM) for ground experiments was completed. Some developments were made for the BBM such as a tripod with attitude control system, a stable mercury pool and a method for collecting the effects of vibrations. Laboratory experiments and field observations were performed from August to September of 2014, in order to check the entire system of the telescope and the software, and the results were compared to the centroid experiments which pursue the best accuracy of determination of the center of star images with a simple optical system. It was also investigated how the vibrations of mercury surface affect the centroid position on Charge Coupled Device (CCD). The results of the experiments showed that the effects of vibrations are almost common to stars in the same view, and they can be corrected by removing mean variation of the stars; and that the vibration of mercury surface can cause errors in centroid as large as 0.2 arcsec; and that there is a strong correlation between the Standard Deviation (SD) of variation of the centroid position and signal to noise ratio (SNR) of star images. It is likely that the accuracy of one (1) milli arcsecond is possible if SNR is high enough and the effects of vibrations are corrected.