Risk and reliability modelling for multi-vehicle marine domains

It is well-known that autonomous underwater vehicle (AUV) missions are a challenging, high-risk robotics application. With many parallels to Mars rovers, AUV missions involve operating a vehicle in an inherently uncertain environment of which our prior knowledge is often sparse or low-resolution. The lack of an accurate prior, coupled with poor situational awareness and potentially significant sensor noise, presents substantial engineering challenges in navigation, localisation, state estimation and control. When constructing missions and operating AUVs, it is important to consider the risks involved. Stakeholders need to be reassured that risks of vehicle loss or damage have been minimised where possible, and scientists need to be confident that the mission is likely to produce sufficient high-quality data to meet the aims of the deployment. In this paper, we consider the challenges associated with risk analysis methods and representations for multi-vehicle missions, reviewing the relevant literature and proposing a methodology.