The design process for navigation Kalman filters: Striving for performance and quality

Abstract

A methodology for the design of navigation Kalman filters is discussed. The goal is to design a Kalman filter that can support sensor integration during its extensive life span with well-controlled performance. The idea is to relate the Kalman filter design with systematic performance evaluation. Using dual model and truth covariance analysis (TCA) approaches, a complete framework for modeling, evaluating and designing an arbitrary integration scheme based on inertial sensors is described. One important achievement is the separation of system-level decisions such as sensor selection or measurement policy from Kalman filter design. The second achievement is the effective procedure for Kalman filter design based on two, almost completely automated steps: state selection and reduced-order Kalman filter tuning. The last, but by no means least important accomplishment is the presentation of an integrated framework, with appropriate parameterization and interfaces, to support all design phases and to allow reuse, with minimal modification, for different projects. The methodology is illustrated with a case-study analysis of a low-cost vehicle INS/GPS system.