Probabilistic approaches in ontologies: Joining semantics and uncertainty for AUV persistent autonomy

This work starts from the need of incorporating uncertainty into ontological representations, given that its presence is an inevitable fact for any real field experience. Applications in the underwater domain are very relevant, especially in the oil&gas field. A vehicle able to maintain such a level of cognition, incorporating uncertainty and semantics, is more reliable, can address different situations and safely operate on underwater infrastructures. The paper presents ways to incorporate a Bayesian framework into traditional owl-based ontologies, and analyses its implications for the overall system. A novel solution is proposed, in order to dynamically update the belief on the world model, providing at the same time means for the other modules - especially the planner - to access discretised values. The full architecture and proposed system have been integrated in Nessie VII AUV and results from post-processing of mission data are presented.