Filter-Bank Approach within Tightly-Coupled Navigation System for Integrity Enhancement in Maritime Applications

Abstract

Navigation systems, estimating the vehicle states with high accuracy, robustness and integrity, are crucial for guidance system in autonomous applications. This publication presents a filter-bank approach designed for a tightly-coupled navigation system, aiming at improving the integrity of the navigation system and the accuracy of the estimated navigation solution under sensor failures. In particular, this navigation system is based on an inertial measurement unit (IMU) and is aided by pseudorange and deltarange-observation of satellite systems, and 3D velocity measurements from a Doppler velocity log (DVL). The designed filter-bank approach proposes a new structure, which handles the challenge of different update rates and measurement delays of the involved sensors. Under IMU failures, the estimated states from the filter bank are reversed to estimate the vehicle acceleration, which improves the IMU bias and state estimation performance at next epochs. The designed approach is validated in a real-time test campaign with an unmanned surface vehicle (USV) in the harbor of Rostock, Germany. During the real-time test, several forms of simulative faults are added to the DVL measurements to evaluate the designed approach. Meanwhile, all necessary sensor data is recorded for further evaluation in post-processing environment. The experimental results show that the designed approach is capable of identifying the simulated sensor failures. Compared with conventional tightly-coupled extended Kalman filter (EKF), the designed approach improves the average horizontal positioning accuracy and its standard deviation under simulative DVL failures by 167% and 771%, respectively.