Application Of Robust Square Root UKF In SINS/GPS

Thanks to the rapid development of computer technology and the progress of modern control theory,the navigation system has started to change from a single navigation system to a combined navigation system,and the integrated navigation system has become an important development direction of the current navigation technology.SINS and GPS are two common single navigation systems.SINS has the advantages of high navigation precision,strong concealment and complete navigation information in a short time,but navigation accuracy decreases with time.The GPS to maintain a certain degree of accuracy for a long time,but need to receive signals from the outside world,vulnerable to interference.SINS/GPS integrated navigation system can achieve the performance of the two sub-navigation system complement each other to focus on the advantages of two subsystems,is a common integrated navigation system.How to integrate the data of two sub-navigation systems to obtain a more accurate and better-performing navigation system is the key to the integrated navigation technology and is also the focus of current research on integrated navigation technologies.Kalman Filter(KF)is a kind of minimum variance estimation algorithm,which is estimated by iterative recursion and is very suitable for computer implementation.And its simple structure,the optimal performance,easy to grasp,so once put forward in the integrated navigation system has been applied.KF is a linear filtering method,but most of the nonlinear system,in fact,has proposed the nonlinear form of KF,EKF and UKF,UKF without solving the Jacobian matrix,for the strong nonlinear and non-differentiable system has better The effect of more than EKF advantage.This paper mainly improves the existing problems of UKF,making it better applied to SINS/GPS integrated navigation system.The main contents of this dissertation include:First introduced several commonly used inertial navigation system coordinate system and these types of coordinate system conversion between the methods.The ellipsoid model of the earth is established,and some key model parameters of the ellipsoid model are introduced.The basic principles of the inertial navigation system are introduced.The mechanical layout equations of the inertial navigation system are deduced,including the basic equations of inertial navigation,the velocity update equation,the position update equation and the attitude update equation.The components of GPS and its positioning principle are briefly described,and the main errors in GPS positioning are analyzed.Briefly introduces the existing SINS and GPS combination mode,and gives the structure diagram of each combination mode.Secondly,in the UKF filtering process,the accuracy of the filter is reduced when the measured data is out of range.The anti-difference algorithm is incorporated into the UKF.This paper introduces several algorithms commonly used in anti-error estimation,and introduces the M estimation emphatically,and presents some commonly used methods to solve the M equivalence.Aiming at the anti-error problem under the condition of inaccurate system noise,an adaptive factor is constructed to expand one-step prediction covariance to restrain the influence of inaccurate system noise on filtering.In the filtering process,the equivalent value is adjusted according to the residual error Solving order and solving method of weight matrix and adaptive factor.When the transportation environment is plains,rivers and oceans,the operating environment is taken as a large priori information to change the solution of the equal weight of the component of height direction.The truncation error in the computer implementation of UKF will make the error of the filtering error and the one-step prediction covariance lose the positive definiteness,which will reduce the filtering accuracy of the UKF and even cause the filtering divergence.The paper integrates the square root filtering algorithm into the UKF to solve the problem of computing divergence.Finally,the recursive process of the square root of robustness of variance root mean-square with inaccurate system noise is given.The direct method model and the indirect method model with large heading angle error angle are established respectively.MATLAB simulation is carried out using simulated data and measured data respectively Experimental,filtering algorithm choose traditional UKF,anti-poor UKF,improved anti-square root of UKF.The experimental results show that both the improved square root of variance and the experimental data have better robust performance in both the simulated data and the measured data,which can well solve the robust problem and the divergence problem in the case of inaccurate system noise.