Study On Optimization Based GPS Aided In-motion Initial Alignment

In-motion initial alignment plays an important role in an inertial navigation system(INS).The INS performance is affected by the accuracy of the initial alignment.The in-motion initial alignment is a hot research area these days;this thesis focuses on improving the performance of optimization-based,global positioning system(GPS)aided in-motion initial alignment.The performance of alignment is influenced by the noise of an inertial measurement unit(IMU)and the accuracy of an alignment algorithm.In this thesis,firstly,a novel calibration method and wavelet filter are designed to minimize systematic errors and random noise respectively.Secondly,two novel initial alignment algorithms are proposed to improve the accuracy of alignment and to reduce the computation power and time required for the alignment.The work in this thesis includes the following aspects:Firstly,a field calibration method for an IMU without any external devices is developed successfully.In the proposed calibration method,generalized nonlinear least-square(GNLS)is used to estimate deterministic errors.The results of GNLS are compared with two other commonly used algorithms such as Levenberg-Marquardt(LM)and Gauss-Newton(GN).Simulation results show that the proposed GNLS-based calibration method is slightly more accurate than the LM and GN.The convergence rate of GNLS is also faster than the LM and GN.The validity of the proposed method is examined by applying it to two separate accelerometers.Secondly,a real-time multi-levels discrete wavelet decomposition filter is developed as an effective pre-filter for the measurements of an IMU sensor.By implementing the pre-filtering method successfully,the random noise of IMU measurements is minimized,and reliable IMU data is obtained for the OBA algorithm.Thirdly,a novel optimization-based GPS-aided in-motion initial alignment is developed using a fast optimal attitude matrix(FOAM)and wavelet filter.The FOAM is used to compute a constant attitude matrix between an initial body frame and an initial navigation frame.The proposed FOAM-based initial alignment algorithm is more robust and computationally efficient than the existing q-method based alignment methods.A simulation framework is designed and a field experiment is performed to examine the validity of the novel alignment algorithm.The simulation and experiment results prove the efficiency and accuracy of the proposed alignment algorithm.The results verify that the proposed alignment method fulfils the demands of subsequent navigational operation.Finally,another novel initial alignment method based on a fusion of an extended Kalman filter(EKF)and an optimization-based alignment algorithm is designed.In the proposed approach,an EKF is implemented to estimate the drifts of a gyroscope,biases of an accelerometer,and errors in velocity and position.These errors are eliminated from measurements,and the corrected measurements are used for the further processing of optimization-based alignment.A recursive quaternion estimator(REQUEST)is used to estimate a constant initial attitude between an initial fixed body frame and an initial fixed navigation frame.The results of the REQUEST in computing the attitude matrix from vector observations are better than the results of the commonly used q-method.The simulation test is performed to examine the performance of the proposed method.The simulation results prove that the proposed method gives good alignment results.In-motion initial alignment is an essential part of INS.The contribution of this thesis would provide methods for improving the performance of in-motion initial alignment and lay a foundation for further research.