Nonlinear Filtering Technology And Its Application In The Deep Space Exploration Autonomous Navigation

The paper mainly studied the nonlinear filtering algorithm in the process of autonomous navigation of deep space exploration. In the process of deep space exploration, thedetector can not be tracked and controlled in real-time or in all-weather due to the influence of ground base, so that the necessity of autonomous navigating control on the detector comes out. However, the autonomous navigation of deep space exploration mainlyrelies on optical method. The precision of autonomous navigation of deep space exploration is influenced by hardware including star sensitive, optical navigation sensors on one hand, and by software,for example, navigating filtering algorithm, so that the improvement of precision of filtering algorithm plays a most important role. On the nonlinear problem of system, the paper mainly studied Unscented Kalman Filter (UKF), Particle Filter (PF)on the basis of traditional nonlinear filtering method---Extended Kalman Filtering (EKF), then analysed and optimally designed each filtering algorithm.For the sampling points of unscented Kalman Filter are so many that they affect thecomputing speed of system, hyper-Spheres Unscented Kalman Filter (SUKF) is chosen.The method which selects the limited quantities of points on the spherical surface of average point to approcah to the state of estimation, decreased the number of sampling points. the performance of which is better than unscented Kalman Filter (UKF) through analyzing.Then the improvement measures are promoted aiming at problems such as particledegradation, depletion and failure, and the particle filter sampling function isdesigned.Based on these above,EKF-PF, UKFPF and MCMC-PF are studied, andcompared with the conventional performance of particle filter, the system accuracy andspeed are analyzed. We conclude that the system accuracy and the performance ofreal-time can be improved effectively by improving the sampling function of theparticle filter and the resampling process.Finally, some other factors which effect the performance of UKF and PF such asthe number of particles, the system state covariance matrix, the measurement covariancematrix and the noise model on the precision of the filter are analyzed.