Techniques for Soft-Failure Detection in a Multisensor Integrated System

Abstract

A navigation sensor integration architecture is proposed to improve navigational reliability and accuracy on vessels having two inertial navigators plus other aiding sensors, such as GPS, Loran-C, Omega, and speed log. This hierarchic alliance of Kalman filters is a systematic structure of dynamically configured parallel filters, sharing measurements but not state information. Unlike a centralized, or even a federated architecture, this scheme can always provide an uncorrupted optimal solution, even after the occurrence of a subtle sensor or software fault that was not immediately detected. Thus it substantially simplifies reconfiguration and eliminates the need to “back out” of a failure. It also facilitates the detection and isolation of sensor and software faults by allowing an additional level of tests, which use sensitive intrafilter fault detection techniques with interfilter fault isolation. These techniques are investigated using high-fidelity simulations and sea trial data, with encouraging results.