Integrity determination and enhancement of multi-sensor navigation systems in maritime applications
- Integritätsbestimmung und -erhöhung von Multisensor-Navigationssystemen in maritimen Anwendungen
Liu, Shuchen; Abel, Dirk (Thesis advisor); Schön, Steffen (Thesis advisor)
Aachen : RWTH Aachen University (2023)
Dissertation / PhD Thesis
Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2023
For decades, academia and industry have pursued the vision of autonomous vessels in harbor areas for their potential efficiency and safety benefits. In safety-critical maritime applications, it is essential to estimate the vessel’s state with high accuracy and integrity. Conventionally, a navigation filter is applied to estimate the vessel’s position, velocity and attitude continuously and accurately, based on an inertial measurements unit (IMU) and a Global Satellite Navigation System (GNSS) receiver. The navigation filter can further integrate a Doppler velocity log (DVL) and a compass to improve its robustness against the system and environment disturbances. On this basis, an integrity enhancement and monitoring system is developed to improve the navigation security, including three modules to ensure the integrity of GNSS, IMU/DVL/compass, and navigation solution, respectively. The integrity of GNSS is monitored within an individual module that applies a hybrid strategy combining a statistic and a set-membership approach. IMU, DVL, and compass fault identification and suppression are achieved within a filter-bank structure. The two measures above significantly enhance the navigation system’s integrity and improve estimation accuracy. Finally, a zonotope-based protection level (PL) generation is developed to determine the integrity of the navigation solution, i.e., to bound the estimation error within the PL.The experimental validation is carried out with recorded data on the vessels MESSIN and DENEB. The designed integrity system is evaluated under ten simulated sensor faults and real-world GNSS interferences. The results show that 9 out of 10 sensor faults are identified within 2 s, and the impact of all sensor faults is suppressed. The root mean square of the horizontal positioning error does not exceed 0.4 m under any circumstances, while the error of a single extended Kalman filter (EKF) without integrity measures is up to 6.5 m. By applying three- and nine-dimensional zonotopes, the computation time is below 0.4 ms and 5 ms for 100 Hz navigation tasks, respectively. By evaluation with the Stanford-ESA Integrity Diagram, the zonotope-based PL generation achieves 100% nominal operation with the 25 m alert limit and is more robust against real-world disturbances than a statistic-based approach. As a result, the designed system satisfies the accuracy requirement for ocean, coastal, and port applications regarding the International Maritime Organization’s (IMO) standards, while a single EKF fails in the port application. Further, the designed integrity system satisfied the integrity requirement for the ocean and coastal applications.
- Chair and Institute of Automatic Control