Diagnosis of Autosub 6000 using Automatically Generated Software Models

Abstract

Modern systems frequently consist of a complex mixture of hardware and software. Model-based diagnosis typically assumes that the effects of the software can be summarised by the commands sent to the hardware and thus the software can be left out of the model. In our effort to build a diagnosis system for an autonomous underwater vehicle (AUV) we have an example where this is not the case—commands sent to the hardware are not all available to the diagnosis system for a variety of reasons. In addition, the software controlling the AUV, the mission script is frequently completely changed from one mission to the next. Taking advantage of the fact that the mission script has a relatively simple structure that does not include loops we show that a diagnosis model of the mission script can be generated automatically that integrates with the model of the physical hardware. We show that this model allows us to diagnose faults that cannot be detected from the hardware model alone.