Design and Performance Enhancement of a Point Mass Filter for a Data-based Referenced Navigation System

Abstract

An integrated Strapdown Inertial Navigation System (SDINS) with Data-Based Referenced Navigation (DBRN) system can provide an accurate autonomous, jamming proof, and covert navigation capability in Global Navigation Satellite System (GNSS) degraded and denied environments, because it is independent of any GNSS or radio navigation. The previous research has shown that a terrain DBRN system based on Extended Kalman Filter (EKF) has some limitation when operated in a relatively flat or repetitive area owing to the problems inherent in EKF. To solve these problems, an experimental study for a DBRN system adopting a Point Mass Filter (PMF) has been reported, in which the filter is applied in the terrain DBRN and bathymetric DBRN systems for an air vehicle and an underwater autonomous underwater vehicle navigation, respectively. The main goal of this dissertation is to improve the performance of conventional PMFs in order to apply for the multi-geospatial DBRN system. For the purpose, firstly, the divergence problem, which is sometimes occurred in grid adaptation method of Bergmans PMF, should be reduced. Also, the PMF should be improved with revising the grid design scheme of the conventional PMF that is developed by Šimandl et al. in order to make the filter suitable for the DBRN systems. Secondly, a better navigation solution than a single geospatial DBRN system should be obtained by developing a PMF scheme that is adequate for fusing multi-geospatial data. This dissertation proposed new algorithms to solve the above issues of the conventional PMF, when applied to DBRN systems. These algorithms are developed as follows. Firstly, the new PMF scheme reduces weakness of the conventional Bergmans PMF by applying the grid redefinition algorithm that feeds the position output of PMF back to SDINS. In addition, re-initialization method reduces the calculation cost required for computing the a priori probability density function (pdf) by revising prediction step of conventional PMFs. This method is proper to the linear system model. Furthermore, the proposed PMF scheme gives a specific criterion to set the number of grid points and the grid distance for a linear system model such as a model of SDINS. It is verified that the proposed PMF algorithm is suitable for a DBRN system having linear system models and highly nonlinear measurement models from the Monte Carlo simulation. Therefore, the proposed algorithm could reduce the divergence problem when a vehicle has large initial position error, and flies over flat or repetitive terrain for a long time. Secondly, this dissertation proposes a scheme of multi-geospatial DBRN system by using a new peakedness measure and the proposed PMF. From the Monte Carlo simulations, it is verified that more robust and stable navigation solution is obtained than single geospatial DBRN systems by using a switching method which selects a likelihood function that has maximum value of peakedness measure in measurement update stage. In conclusion, while the proposed PMF takes more time than EKF, the results of this study indicate that a navigation solution of DBRN system using the proposed PMF is superior to that of EKF in an area having relatively unchanged or repetitive geospatial data. If we use the proposed switching method, the multi-geospatial DBRN system gives more robust and stable solution than a single geospatial DBRN system.