Observability of integrated navigation system states under varying dynamic conditions and aiding techniques

The performance of integrated navigation systems not only depends on the quality of the used inertial measurement unit (IMU) and aiding sensor information, but also on the correct observation of the system's state vector. As a classical example, an integration filter shows a good performance if it manages to estimate the sensor errors properly. As the observability varies with the current system states as well as the quality of the aiding information, a meaningful characterization of the system's performance is difficult to obtain. The aim of this paper is to analyze the impact of the influences named above on the observability of the system model that is part of the navigation filter. For linear and linearized systems, e.g. Kalman Filter and Extended Kalman Filter, different measures of observability can be derived from control theory. This paper will show the necessary basic algorithms and methods to evaluate a system's observability. Based on these insights an evaluation of a standard filter model of an integrated navigation system is performed. Therefore, different dynamic conditions as well as aiding information are taken into account. The main focus of this work lies on the examination of real flight data and correlation between system states and their observability. Based on these investigations this paper presents a detailed view on the assessment and first results towards a better characterization of IMU performance.