Initial Alignment Of Strapdown Inertial Navigation System Based On Anti-disturbance Filtering

At present,inertial navigation system(INS)plays an increasingly important role in military and civilian applications.INS has the advantages of being free from external disturbance,unaffected by terrain and bad weather,good concealment,high data update rate,good short-term precision,good stability,and the ability to provide the position,speed and attitude information of the carrier at the same time.Initial alignment is the key technology of strapdown INS(SINS),which directly affects the accuracy of the navigation system.In this thesis,the working principle,error characteristics and disturbance characteristics of the SINS are analyzed.The error model of the SINS under static base is established.The stability and robustness of the traditional Kalman filter(KF)will be degraded for SINS with model uncertainty,unknown statistical characteristics of noise and multiple disturbances.Therefore,it is necessary to design a class of initial alignment methods based on anti-disturbance filtering to improve navigation accuracy.The main work and innovations of this thesis are as followsFor an INS,it is difficult to accurately measure the model error.In most existing initial alignment methods,the model error is supposed to be a Gaussian white noise.In this thesis,a predictive iterative Kalman filter is proposed.The prediction filter is used to estimate the model error caused by the inertial sensors,and the iterative Kalman filtering technique is used to prevent the one-step prediction error from being too large,resulting in unstable results.The predictive iterative Kalman filter has better stability and higher alignment accuracy than the traditional KF by predicting the model error and adjusting the weight of the model error.Aiming at the problem that the inertial alignment method based on KF is difficult to estimate the noise and external disturbance and the anti-disturbance ability is weak,a smooth variable structure filter(SVSF)with forgetting factor is addressed.Adding a forgetting factor to the filter limits the memory length of the SVSF,making full use of the current measurement data.This can increase the weight of the current data in the state estimation,and then can avoid filter divergence The SVSF with forgetting factor has higher precision and strong anti-disturbance ability than the traditional KF when the inertial navigation system encounters unknown disturbanceThe smooth variable structure filter is essentially a suboptimal filter.When the system noise is a Gaussian white noise and the system is healthy,the accuracy may be worse than the optimal KF.Aiming at the problem that the noise statistical characteristics in the INS are unknown,a combined Kalman filter and smoothing variable structure filter is proposed.Optimal estimation is performed by defining a fully smooth bounded layer.When the smooth boundary layer is smaller than the constant boundary layer,the traditional KF is used to increase the accuracy of the initial alignment.When the smooth boundary layer is larger than the constant boundary layer,the SVSF is adopted to increase the robustness of the initial alignment.In this way,the robustness of the smooth variable structure filtering algorithm and the accuracy of the Kalman filtering algorithm are combined.The anti-disturbance filtering algorithm of this thesis is verified by simulation and experiment results.It can be concluded that the proposed methods can improve the accuracy and stability of the initial alignment with anti-disturbance ability and robust performance,which provides the basis for the design of high precision INS.