Similarity recognition for OWL-based action recipes in RoboEarth

Abstract

Sharing and integrating heterogeneous knowledge from different robots requires finding an agreement between the underlying abstract, hardware independent ontologies, saved in the RoboEarth cloud framework in the World Wide Web. A variety of methods from the literature may be used for this task, by means of similarity computation of two ontologies. While most of them come from other fields such as sequence alignment methods from the field of bioinformatics, they can be extended for ontology alignment purposes. While such methods basically perform a pair wise comparison of ontology entities, another approach called the OWL-Lite Alignment method uses a variety of different ontology alignment methods to integrate many ontology comparison techniques in one common framework. Both the sequence alignment based and OWL-Lite based solutions are presented and their suitability to conduct a similarity check for robot task descriptions upon an upload into the RoboEarth platform discussed.