Research On Key Models In SINS Dynamic Alignment Aided By GPS

Strapdown inertial navigation system (SINS) is an important autonomousnavigation equipment, and is of wide implication in military and civilian fields. Initialalignment is a key technology of SINS navigation and GPS/INS integrated navigation,and in-motion alignment aided by global navigation satellite system (GNSS) hasdeveloped into one of the important research fields. This dissertation mainly focuseson the theories and algorithms of SINS in-motion alignment aided by GPS, whichincludes the principles of SINS alignment, the random errors analysis and modeling ofINS, the initial alignment error model of SINS in earth centered earth fixed (ECEF)frame, the robust nonlinear filter for SINS in-motion alignment aided by GPS, theobservation model and robust nonlinear filter algorithms for GPS/INS tightly coupledalignment. The main works and contributions are summarized as follows:(1) The Power Spectral Density (PSD) and Allan variance are used to analyse theraw fiber optic gyro (FOG) data. And the result of PSD and Allan variance are similarto the parameters of specifications. The Allan variance could be used for modeling theFOG random drifts. The FOG output sequence must be tested and treated beforemodeling. The empirical mode decomposition (EMD) method is introduced forremoving the trend term of nonstationary random sequence. An inversed order methodis used to check the effect of EMD method，and the skew value and kurtosis value arealso introduced for normality test. The experiment indicates that the EMD method cannot only identify the trend term of nonstationary random sequence, but also improvethe normal natures of the nonstationary random sequence.(2) The sensitivity of low-cost SINS is not enough to finish the coarsealignment in statics base before navigation. Aiming at this problem, a generic inertialnavigation system alignment error propagation model which does not rely on smallmisalignment angles assumption is presented. The model is presented in the ECEFframe approach and is very suitable to implement the in-motion alignment with GPSaided. We also present the simplified model for large heading misalignment and threesmall misalignment angles. A vehicle test indicates that low-cost SINS with GPSaided can conform the dynamic alignment under the large misalignment angleassumption and the proposed model is practicable. And when the misalignment anglesare small, the simplified model is of high precision and efficiency. (3) The SINS error models based on quaternion, Rodrigues parameters andmodified Rodrigues parameters are proposed. Similar to the error model as describedbefore, the models are described under ECEF frame and no small angel assumption ismade in the model development. The formulae show that the three attitude descriptionmethods can be proved to be a unified form by high-order Cayley transformationbased on SO(3) theory. The actual tests show that the computational complexity andaccuracy of different SINS error models are analyzed through some simulations.(4) Based on singular value decomposition algorithm, a new H cubaturekalman filter algorithm is proposed to deal with the bias of SINS in-motion alignmentaided by GPS. The new algorithm can shrink the filter divergence caused by imperfectdesign parameter. And on the wider conditions for design parameter, the new filter ismore robust. The influence of different ineration threshold and step size of the optimaldesign parameter are analysed.(5) A detail derivation of state and measurement equations in tightly coupledGPS/SINS alignment is showed. The accuracy of estimation and the convergent timeare improved by applying the double difference pseudorange rate. A multiple fadingH adaptive cubature kalman filter is put forward. The actual test indicates that theproposed algorithm not only can improve the estimation accuracy of position andattitude states, but also have better robustness for outliers.