Updating autonomous underwater vehicle risk based on the effectiveness of failure prevention and correction

Autonomous underwater vehicles (AUVs) have proven to be feasible platforms for marine observations. Risk and reliability studies on the performance of these vehicles by different groups show a significant difference in reliability, with the observation that the outcomes depend on whether the vehicles are operated by developers or non-developers. We show that this difference in reliability is due to the failure prevention and correction procedures - risk mitigation - put in place by developers. However, no formalisation has been developed for updating the risk profile based on the expected effectiveness of the failure prevention and correction process. In this paper we present a generic Bayesian approach for updating the risk profile, based on the probability of failure prevention and correction and the number of subsequent deployments on which the failure does not occur. The approach, which applies whether the risk profile is captured in a parametric or nonparametric survival model, is applied to a real case study of the ISE Explorer AUV