Key Technologies For FOG-SINS/GNSS Under GNSS-Denied Environment

With the proceeding of navigation under mulitiple satellite constellations and the reduced cost of receiver,GNSS has been widely applied in various fields of national production and living.However,when used in urban areas with high density population and battlefield that with complex electromagnetic environment,the reliability and continuity of GNSS can not be ensured.GNSS-denied environment is referred to as the failure of GNSS equipment when the satellite signal suffers from disturbances,shadowing and spoofing attack.Taking the land vehicle navigation in canyons for example,whose high precision orientation and location system would be challenged by the signal attenuation,dense multipath and short signal outage.The thesis investigates the key technologies of integration of GNSS and INS in land vehicle navigation,where the error processing of fiber optic gyroscope(FOG)and robust filtering method are the main focuses.The contents and contributions in this thesis are as follows:1)Aiming at the non-linearity and non-stationarity of the static drift of FOG,empirical mode decomposition(EMD)and its noise assisted analysis form are investigated to eliminate the random noise of FOG.To remove the supurious modes and residual error from noise assist,bounded iterated noise assist is utilized.Then the thresholding method of wavelet denoising is brought into the denosing of mode interval,where the continuity of the mode denoising is optimized according to the oscillation feature and extrema distribution of intrinsic mode functions.Finally,mode selection is realized by combining the sample entropy and Euclidean distance between two probability density function(PDF),which is expected to produce more accurate modes classification.Analysis based on test data indicates that,proposed method outperforms the method utilizing stationary wavelet transformation in reducing the bias instability and rate random walk of FOG random noise.2)In order to identify the features of FOG drift,the multi-scale prediction and modeling method based on the data-driven property of EMD are proposed.First,the characteristic of forward linear prediction(FLP)is analyzied,then by incroperating the accumulated generating operation from Grey theory,an algorithm that improves the accuracy of weights and without increasing the output delay of FLP is developed.Then the algorithms are used to denoise the modes from EMD,and finally the noise-free drift data is reconstructed by adding up all the denoised modes.Furthermore,noting that the temperature drift error is the coupling of multiple factors,a multi-scale drift modeling scheme is proposed,where the Shupe error resulting from the changing temperature gradient and bias error duing to the thermal stress are modeled and compensated separately by using extreme learning machine.Experiment results demonstrate that,in the application of temperature drift compensation of FOG in static environment,multi-scale modeling improves the prediction result by two orders compared with the random modeling that considers the drift as one scale.3)To solve the low efficiency problem of linear innovation update in nonlinear measurement update of integration navigation,the maximum a posterior(MAP)filtering framework is analyzed for tightly coupled GNSS/INS.Because the measurement noise has been incorporated into the state after the first measurement update,state augment is utilized in the later iterated update of MAP filter to alleviate the effect of state-dependent noise.Then a damping factor is used to accelerate the convergence and ensure a performance improvement of iterated filtering.Simulation and experiment results reveal that iterated update framework can improve the estimation accuracy of non-direct observable states,and when the posterior PDF is of unimodal Gaussian distribution the advantage is more significant.When a nonlinear measurement equation is involved,MAP filter produces smaller steady-state error compared with the non-iterated filter.4)In order to improve the robustness of cubature Kalman filter(CKF),a novel sigma update framework is developed based on model uncertainties analysis,and the updated cubature points are embedded into the 3 degree spherical-radial rule to form robust CKF(RCKF).To remove the dependence of cubature points on the dimension of state and the Gaussian hypothesis of state posterior PDF,the sigma-points error matrix from prediction stage is transformed into the posterior domain directly,which alleviates the sensitivity of CKF to observations missing and correlated noise.What is more,due to the improved prediction stage of filtering,the measurement model uncertainty has less effect on the estimation of attitude compared with that of position and velocity.Simulation results indicate that when the outage of GNSS reaches 2 minutes,the positioning error of extended Kalman filter(EKF)is more than 3 times of RCKF.Field test further demonstrates the effectiveness of RCKF,where the sensor outputs of land vehicle running in environment with frequent signal outages are used.Compared with EKF,RCKF improves the heading and position in east direction by 89.3%and 56.7%,respectively.