Hochgenaue und robuste GNSS-gestützte Zustandsschätzung für die autonome Schifffahrt

  • Highly accurate and robust GNSS-aided state estimation for autonomous shipping

Gehrt, Jan-Jöran; Abel, Dirk (Thesis advisor); Jeinsch, Torsten (Thesis advisor)

Aachen : RWTH Aachen Univsersity (2021, 2022)
Dissertation / PhD Thesis

Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2021


The subject of the present work is the research and development of a navigation system for highly accurate and robust estimation of the navigation state, consisting of position, velocity and orientation. For this purpose, a Kalman filter-based sensor fusion of inertial sensors with observations from global navigation satellite systems (GNSS) is developed. A nonlinear tightly coupled state space is designed, which contains error states of the applied sensor system in addition to the navigation states. Common Kalman filters for nonlinear systems, extended Kalman filter (EKF) and unscented Kalman filter (UKF) are implemented and experimentally compared. It is found that in maritime applications, there is no advantage to either architecture for regular maneuvering operations or for large initialization errors. With differential GNSS corrections, average position accuracies of less than 30 cm with a standard deviation of less than 10 cm are achieved. The yaw angle estimate achieves an average accuracy of less than 1 ◦ with a standard deviation slightly above 1 ◦. To further increase position accuracy, differential methods are integrated into the filter to implement tightly coupled real-time kinematics (RTK). For this purpose, GNSS carrier phases are processed in the filter, which, however, are perturbed by an unknown bias to be estimated, the so-called ambiguity. To increase the quality of the ambiguity estimation, a novel sea level integration is presented to support the GNSS carrier phase ambiguity estimation. It is shown that in a scenario where the navigation filter without sea level support cannot achieve a valid RTK fix, by sea level support, the RTK fix is achieved after short 5 s. With the developed tightly coupled RTK, the navigation filter achieves average position accuracies of less than 3 cm with a standard deviation of 2 cm. Robustness to local GNSS perturbations is significantly increased by integrating measurement quantities from an onboard velocity sensor into the filter. For this purpose, both loose coupling and tight coupling of the sensor are investigated. It is shown that, thanks to integration of the velocity sensor, a sufficiently accurate navigation state can be estimated even in the presence of Global Navigation Satellite System (GNSS) failures of more than 8 min. The robustness of the differential approaches is significantly increased by developing a novel estimator that allows RTK to estimate highly accurate positions even in the absence of differential data. It is shown that the ambiguities can be reliably estimated even after 5 min without differential data. In the future, the filter will be extended by integrity verification methods and used for guidance, navigation and control of ships in autonomous operation.