Fault Detection in AUV navigation: a computationally inexpensive approach

This paper proposes a computationally inexpensive approach for fault detection during AUV navigation. Thruster performances may degrade over time, or a complete failure may occur. Considering that currently most of AUV missions are surveys involving long linear transits, this paper focuses on analysis of line segment and detection of navigation errors with varying thruster performances degradation, using linear approximation and curvature analysis. Those mathematical approaches have been chosen due to their computational efficiency, an important aspect for low-cost vehicles and for a service watch module, which needs to run in the background without competing for resources with mission-critical modules. Extensive tests have been run with the Sparus AUV at the University of Girona facilities and at sea, confirming the validity of the proposed approach.