Automatic fault detection and execution monitoring for AUV missions

Advanced AUV's, particularly those capable of long duration missions, need increasing amounts of autonomy in order to carry out sophisticated missions without requiring constant support from a ship. One important aspect of this autonomy is fault detection and execution monitoring. In this paper we describe the application of the Livingstone 2 diagnosis system on Autosub 6000 and examine in particular the issue of ensuring the mission is being executed correctly. Most AUV operations require a fast turnaround between retrieving the vehicle and redeploying it, typically constrained only by the time required to charge batteries. In many circumstances missions are planned very quickly so an automatic way to monitor mission execution is required. We describe an approach to automatically generate diagnosis models that correspond to the mission instructions which can then be combined with pre-built models of the physical components of the AUV to improve fault detection and monitor execution. We show that by incorporating this program model we are able to diagnose faults that were previously undiagnosable, and demonstrate that an actual fault that occurred on Autosub 6000 can successfully be detected.