Fault-diagnosis of subsea robots using neuro-symbolic hybrid systems

Abstract

We describe a diagnosis phase, part of a study currently performed on subsea robots health monitoring. The overall objective of this study is to allow a subsea robot with sub-system failure to keep executing its mission with a certain level of performance. The current development of the diagnosis phase is based on the use of neuro-symbolic hybrid systems. With this solution, the system is able to learn when it meets unforeseen failures.