Research And Simulation Of Nonlinear Filtering Algorithms For Autonomous Navigation Of Spacecraft

Autonomous navigation of spacecraft has become a key technology in deep space exploring fields. Especially, the design and application of navigation algorithms are of great importance in the autonomous navigation system. With the supports of Tenth Five-Year 863 Program-Autonomous Navigation Technology of Deep Spacecraft (2004AA735080-5 ) and The Test of Deep Spacecraft Autonomous Navigation Simulation System (2005AA735080-2) , this dissertation makes a research on two representative methods—stirling's interpolation filter and particle filter(PF). This paper applies the second-order nonlinear interpolation filtering algorithm and improved particle filtering algorithm in cruise and encounter mission phases of small body exploration respectively, and the feasibility of these algorithms is validated by simulation. The main contents of this paper are as follows.Firstly, the basic problem involved in the spacecraft autonomous navigation is studied and the measurement types are analyzed. The orbit dynamic models during the cruise and encounter phases for small celestial body exploration missions are established. Proper navigation schemes are given for different navigation mission phases.Secondly, a spacecraft autonomous navigation model is established aiming at flying characters of the spacecraft in cruise mission phase. And this paper implements a new method—second-order nonlinear interpolation filtering algorithm to state estimation of the navigation system in cruise phase. This algorithm takes second-order of multinomial as function approximation based on interpolation formula. It adopts recursion structure of the traditional EKF(Extended Kalman Filter). Calculation is greatly simplified due to avoiding computing Jacobian matrix and the new method reduces the model error caused by Taylor linearization in traditional EKF. According to state and observation functions of navigation system in the cruise phase, EKF, Stirling's interpolation filter and second-order nonlinear interpolation filtering algorithm are adopted to determine the autonomous orbit of spacecraft. The simulation result indicates that the second-order nonlinear interpolation filtering algorithm achieves higher numerical stability and accuracy.Thirdly, this paper establishes an autonomous optical navigation scheme based on the information of the limb feature points. To ensure the accuracy and robustness of the system, the improved particle filtering algorithm is put forward to implement in the encounter mission phase to deal with the nonlinear and non-Gauss characters. The improved particle filtering algorithm put forward in this paper adopts UD-EKF to update particles. In comparison to PF algorithm, the new algorithm improves accuracy in certain degree by considering the newest observing information; In comparison to EPF algorithm, this algorithm has advantages of small data storage and high calculation speed because of Upper-triangular and Diagonal decomposition, also the influence of calculation rounding error is greatly reduced, and the nonnegative definiteness and symmetry of the covariance matrix is guaranteed, so the calculation of the new algorithm is reduced while ensuring it's accuracy, and this algorithm is appropriate to autonomous navigation mission in the encounter phase. Simulation results demonstrate the validity and efficiency of this method, in comparison to PF and EPF algorithms.Further more, this paper studies the impact on the accuracy of optical autonomous navigation system caused by the distance between the spacecraft and the celestial body with simulation.