Research On The Application Of Federated Filtering Algorithm In INS/GPS/CNS Integrated Navigation System

INS/GPS/CNS integrated navigation system combining the advantages of the three kinds of navigation,is a kind of superior,stable and safe navigation system.The technology of information fusion makes it easy to effectively utilize measurement information of each sensor,and obtain better results comparing with the single-mode navigation measurements.Federated filter as a new information fusion technology avoid tedious computation of multi-dimensional matrix operations comparing with traditional centralized Kalman filter.At the same time,federated filter has become one of the main method of information fusion in navigation system with design flexibility and fault tolerance.In this paper,a research of federated filtering algorithm and fault detection method in INS/GPS/CNS integrated navigation system will be established by using inertial navigation system as the reference system.The main research contents of this paper are as follows:1.The mathematical model of SINS/GPS and SINS/CNS subsystems are respectively analyzed in order that we obtain the state equations,measurement equations,discrete models and Kalman filter algorithms of them.Finally,simulations of the subsystems show the results of filtering.2.Principles of the federated filtering algorithm is analyzed,including structure of the federated filter and the principle information sharing.Then the federal filtering algorithm is introduced by four processes,determining initial information,distributing information,updating information and fusing information.Next the federated filter is divided into four basic structures with different performance,according to the different modes of information fusion between master filter and local filters.Last five kinds of information fusion methods are introduced,which are based on the characteristics of each subsystem,so that the federal filter can achieve a higher precision.3.Methods of fault detection based on state chi-square test,residue chi-square test and BP neural network were analyzed by the test algorithm with merits and demerits.4.The federal filtering algorithm for INS/GPS/CNS integrated navigation system is studied in the aspects of the selection of structure,the selection of common state,the determination of initial information and the realization of the information fusion.Then put forward the improvement scheme of a new type of integrated chi-square test aiming at the problems that they still exist when finishing the analysis of the three basic fault detection methods of the existing improvements.Finally,the simulation results show that the improvement scheme is feasible.