Research On Nonlinear Filtering Algorithm For Autonomous Navigation

Deep space exploration is a subject of important scientific and economic significance as well as more challenges.Since it is difficult to obtain high-precision navigation information for deep space exploration missions using traditional non-autonomous navigation methods,it is necessary to develop autonomous navigation technology for deep space exploration.The main factors affecting the accuracy of autonomous navigation are the autonomous navigation methods and the navigation filtering algorithms.Since the navigation systems are the nonlinear systems,the study of how to use the nonlinear filtering algorithms to achieve high-precision state estimation has become a research hotspot in this field.The design of autonomous navigation technology and advanced nonlinear filtering estimation theory and methods has great research significance for ensuring the reliability and stability of the autonomous navigation system.Therefore,this dissertation mainly studies the design of autonomous navigation to construct the autonomous navigation platform,and conducts in-depth research on the nonlinear filtering algorithms,and introduces improved nonlinear filtering algorithms to achieve the optimal state estimation of the navigation system information,finally designs a series of performance evaluation indicators to judge the state estimation performance of different filtering algorithms and the navigation performance of different navigation methods.The main research works of this dissertation are as follows:1.In view of the high requirements of autonomous navigation for accuracy and real-time performance,the working principles and methods of inertial navigation,deep space probe transfer orbit autonomous celestial angle measurement navigation and celestial velocity measurement navigation are mainly investigated in this dissertation.And,the design methods of inertial/celestial integrated autonomous navigation and celestial angle/velocity measurement integrated autonomous navigation are also introduced.The autonomous navigation module can be constructed by establishing the corresponding system model and measurement model.In addition,the orbit simulation module is built and the selection criteria for navigation celestial bodies are analyzed.2.Aiming at the problem of poor adaptability of a single model and unknown system parameters,the theory of multi-model adaptive estimation is studied,and then the unscented Kalman filter and the unscented particle filter based on multi-model adaptive estimation are proposed to improve the adaptive ability of the system and the system state estimation accuracy.The proposed algorithms are compared with the standard Kalman filter and particle filter algorithm in two nonlinear systems.The simulation results show that the filtering effect of multi-model adaptive estimation is obviously better than that of non-adaptive estimation,and the proposed filtering algorithms have higher navigation accuracy,which can satisfy the requirements of integrated navigation.3.In view of the particle degradation problem and the computational complexity of filtering algorithm,two nonlinear transformation methods are studied,and two improved nonlinear filtering estimation algorithms are proposed.The first is the spherical simplex unscented particle filter,which mainly uses the spherical simplex unscented transformation strategy to deal with the nonlinear transfer of mean and covariance,reducing the amount of calculation.Then,simulations for deep space probe celestial angle measurement navigation are carried out to verify the availability of this proposed algorithm;the second is the adaptive minimum skew simplex unscented particle filter,which adjusts the error covariance matrix in the filtering process by using an adaptive method based on innovation sequence correction to reduce the estimation error,thereby enhancing the filtering accuracy,and uses the minimum skew simplex sampling strategy to improve the calculation efficiency.Simultaneously,the proposed filtering algorithm is applied to the typical system and celestial integrated navigation system for simulations.The simulation results show that the proposed algorithm significantly reduces the amount of calculation and improves the navigation accuracy.4.Aiming at the problem of particle depletion caused by resampling,first of all,based on the two ideas of particle selection and weight optimization,the traditional random resampling method is improved to increase the diversity of particles,and an improved unscented particle filter is proposed to achieve nonlinear state estimation and improve the estimation accuracy.Simultaneously,the effectiveness of the proposed algorithm is verified through the implementation of celestial angle measurement navigation simulation scheme.Secondly,for the problem of uncertain or unknown system parameters,combined with the improved resampling technology,a new two-step adaptive particle filter method is presented.The proposed algorithm and traditional particle filters are applied to the typical system and integrated autonomous navigation system respectively,to verify the availability of the proposed filtering algorithm.5.In order to verify the effectiveness of the nonlinear filtering algorithms and the rationality of the autonomous navigation system scheme design,this dissertation studies how to evaluate the performance of the navigation algorithms for automonous navigation systems,and constructs an evaluation system including four evaluation indexes,such as accuracy,real-time performance,availability and continuity.A series of specific evaluation methods are designed to evaluate and verify the navigation performance of nonlinear filtering algorithms and autonomous navigation methods.Finally,different evaluation graphical interfaces are presented.