Study On The Initial Alignment And Fault Repair Of GNSS/INS Integrated Navigation System

The Global Navigation Satellite System(GNSS)can provide users with high-precision and low-frequency positioning and navigation services in an unshaded environment,however,in complex environment,GNSS signals are easily blocked or interfered.Inertial Navigation System(INS)can independently provide high frequency and continuous position,speed and attitude information after initialization.By the reasonable integration of the two system,GNSS/INS integrated system can provide users with continuous and reliable navigation solution in the shelter or semi-shelter environment.This dissertation mainly focus on GNSS/INS integrated navigation system initial alignment methods and fault detection and repair techniques,so the main researches include high precision SINS rapid self-alignment,large misalignment angle fault treatment,magnetometer aided MEMS IMU attitude fusion adaptive robust filter,neural network aided GNSS/INS system fault recognition and repair,inertial aided BDS three-frequency cycle slip detection and repair.The work and contributions can be summarized as follows:(1)The accuracy and speed of initial alignment will directly affect the performance of Inertial Navigation System.Aiming at the problems of low accuracy,slow convergence speed and poor observability of the static base,the model of static base alignment is establishedm,and the observability of the static base alignment is analyzed with PWCS method,then the optimal multi-position alignment scheme by using the transposition mechanism is put forward to improve the observability.The results show that the proposed optimal three-position alignment scheme can make all state variables observable,improve the estimation accuracy of state variables,and effectively shorten the alignment time.(2)The large yaw misalignment angle greatly increases the nonlinear degree of the SINS error model,which leads to the decrease of the estimation accuracy of the traditional EKF filter or even divergence.According to large misalignment angle problem,an arbitrary misalignment angle error model of SINS is derived,then the simplex sampling sigma points strategy is introduced to reduce the complexity of the UT transformation.At the same time,in order to guarantee the accuracy,the scaled minimal skew is put forward to dynamically adjust the distance between sigma points and their center,so high order error can be avoided.Finally,the proposed SSUKF algorithm is used to process large misalignment angle nonlinear equations,and the results show that the SSUKF algorithm has the approximate estimation accuracy asthe SUKF algorithm,but the computational complexity is obviously reduced,which is beneficial to reduce the computational burden.(3)The low-cost MEMS IMU attitude initialization needs the assistance of other sensors,such as magnetometer,but the external magnetic interference will lead to the measurement failure of magnetometer.Aiming at this condition,a simplified six-parameter calibration model is proposed,then the traditional LM algorithm is improved and iteration strategy is optimized,and the field quick magnetometer calibration algorithm is established under the body frame.On this basis,a magnetometer/IMU adaptive robust fusion attitude determination model is proposed based on the robust estimation theory of correlation observation.The results show that this proposed model can effectively weaken the attitude fusion anomaly caused by magnetometer failure,and the robust attitude fusion estimation of low-cost IMU under short-term vibration is realized.(4)In complex urban environment,GNSS/INS integrated navigation system is easily affected not only by observation gross error but also dynamic model failure.In view of the disadvantages that traditional fault detection method cannot identify the two kinds of fault,the overall fault detection method based on mahalanobis distance is established,and the optimal RBF neural network training strategy is put forward to assist integrated navigation fault identification,then a dual-adjustment robust factor and an adaptive forgetting factor are established according to the different source of failure,so the two types of fault in the integrated system can be effectively identified,separated and decreased.The results show that the optimal RBF neural network can achieve a recognition success rate of 92% for small dense GNSS observation gross errors.In addition,when the GNSS signal is completely unlocked,the optimal RBF neural network can also predict the navigation solution according to the output of INS,and provide continued high-precision position information in a short time.(5)In severe multipath environment,the severe pseudorange multipath effect will lead to the cycle slips or satellite outage,and the labile residual of pseudorange multipath between epochs will also pollute the estimated value of cycle slips,and then leading to the failure of the traditional three-frequency pseudrange-phase combination cycle slip detection model.In order to realize the accurate detection and repair of cycle slip in a severe multipath environment,the inertia adied BDS three-frequency combination method is proposed,then the INS aided cycle-slip detection monitoring value is established and the effect of INS positioning error on cycle-slip capacity is analyzed.The results show that the inertia adied cycle slip detection method can effectively improve the success rate and repair rate for BDS triple-frequency cycle-slip detection in severe environment with multipath effects,and the detection and repair of dense small cycle jump are not affected by the residual of multipath effect.