Nonlinear Filtering And Its Application In Navigation System

With the rapid development of control technology and computer technology, nonlinear filter technology is widely used in the signal processing, automatic control, computer vision, wireless communication, aerospace and target tracking and recognition areas. Based on the framework of Bayesian theory, particle filter has become the focus of optimal estimation for nonlinear and non-Gaussian dynamic system. With the background of practical scientific research, the dissertation mainly investigates particle filter method and its application in initial alignment of INS and INS/GPS integrated navigation system.Firstly, It describes the two typical approximate nonlinear filtering methods, namely Gaussian filter and particle filter in a unified way from the recursive Bayesian estimation approach. The application premise and implementation process of various nonlinear filtering methods are analyzed. Among many approximation solutions, Gaussian filter is simple and easy to implement, but it must be confined to the condition that the nonlinear system model can be assumed as Gaussian distribution. Thus, Gaussian filter has no reliable convergence guarantee for nonlinear and non-Gaussian system. However, particle filter as alternative to the Gaussian filter can deal with probability distribution of arbitrary form and indicates overwhelming advantages for the parameter estimation and state filtering problem, so particle filter obtains great development.Secondly, it concentrates on the drawbacks of particle filter, such as the particle degeneracy, sample impoverishment, and computation burden. The traditional particle filter is improved to remedy these drawbacks through different strategies, and simulation experiment validates the rationality and exactness of the new algorithms. The research on the improved method is achieved as follows: 1. a new type of particle filter is proposed by integrating the traditional Gaussian nonlinear filter as EKF and UKF with particle filter, leading to EKPF and UPF. The latest measurement is included in the important probability density function, so this type of filter properly decreases the severe particle degeneracy phenomenon. 2. a new type of particle filter is proposed by introducing the crossover and mutation operation of genetic algorithm, simulated annealing algorithm, particle swarm optimization algorithm into the conventional particle filter, leading to GPF, SAPF, PSOPF. It is proved that sample impoverishment phenomenon is effectively abated. 3. a new particle filter is presented by marginalizing out part of state analytically based on Rao-Blackwellization technique, leading to MPF. Owing to less dimension of state, it is possible that marginalization method can not only substantially reduce the computation complexity of particle filter, but also increase the estimation accuracy for the given filter.Thirdly, in order to handle the large initial misalignment angles of inertial navigation system for ship, it presents the velocity plus attitude matching of at-sea alignment algorithm based on rotation vector, quatemion(additive or multiplicative). Then, UPF is applied to the initial alignment of ship INS to provide the new solvable scheme of fast alignment. As a result, coarse alignment and fine alignment can be unified and traditional alignment mode will be changed. Comparison for estimation performance under the condition of three different models and different nonlinear filtering methods through computer simulation provides the support for selecting reasonable filtering algorithm and alignment model. Simulation results show that the quaternion matching algorithm can quickly estimate the azimuth misalignment angle as the same as the horizontal misalignment angle, which is not achieved using the at-sea alignment technique before. In practice, rotating table experiment and sea experiment further validate the feasibility and rationality in initial alignment of INS.Finally, in order to cope with nonlinear model and non-Gaussian measurement of INS/GPS integrated navigation system, it presents the new ideal of RPF-MCMC algorithm. Numerical simulation is implemented by the examples such as Univariate Nonstationary Growth Model (UNGM), Ballistic Target Reentry (BTR) and Bearing Only Tracking (BOT). Simulation results demonstrate that RPF-MCMC algorithm can effectively solve the nonlinearity of system model and non-Gaussian distribution of noise. Then, two mathematical model of INS/GPS are established, namely NPLM model and NPNM model, respectively. As a result, RPF-MCMC algorithm is applied to above models of INS/GPS integrated navigation system. Different model based on position/velocity integration and pseudorange/pseudorange rate integration is compared to analyze the effect of proposed algorithm on the system performance. Simulation results prove that the proposed algorithm can effectively manage the non-Gaussian measurement noise, and estimation accuracy is improved excessively. Vehicle experiment fully confirms the validity of RPF-MCMC algorithm in INS/GPS integration navigation system.