Modularized Filtering for Navigation

Abstract

This thesis presents a modular navigation filtering framework specialized for use in a research and development environment. The developed framework emphasizes flexibility and modularity of filter components over computational performance in order to minimize the engineering work required to reconfigure the design for a new application. Modularization is accomplished by exploiting the natural mathematical interfaces of the Kalman Filter and related variants, while exercising the design discipline to truly maintain filter component separation. This approach results in a framework that retains compatibility with many Kalman Filter variants as well as different system models. The filter framework is qualitatively tested with simple system and sensor model implementations in two types of Monte Carlo simulation. The first set of tests utilizes trajectory data generated from a crude random walk model to check system stability and tuning. The second set of tests evaluates system performance using more realistic trajectory data generated with the X-Plane flight simulator. The test results demonstrate adequate performance of the simple models and overall viability of the design. Lastly, several improvements are proposed to increase the utility of the framework.