A light non-monotonic knowledge-base for service robots

In this paper a Non-Monotonic Knowledge-Base (KB) for practical applications in service robots is presented. The KB is defined as a conceptual hierarchy with inheritance that supports the expression of defaults and exceptions. All classes and individuals, with their properties and relations, can be updated dynamically and the KB-System supports non-monotonic behavior. Non-monotonicity is handled on the basis of a specificity criteria, such that more specific properties and relations have precedence over more general ones. The system supports the expression of conceptual (or terminological) and factual (or assertional) knowledge, which are used in inference in a coherent and consistent way. The KB-System is embedded within the IOCA Architecture, where knowledge about how to communicate and interact with the world, and also knowledge of the particular interpretation situation are represented. The cognitive architecture is structured around a main communication cycle, and queries and conceptual inferences are performed on demand during the interaction of the robot with other agents or the world. The overall structure of the KB with its main interpreter and supporting utilities as well as the embedding of the KB-system in the robot’s architecture are also presented. The KB-System is illustrated with a case study in service robots scenarios, where a practical non-monotonic KB is required. Finally, the implementation of the KB-System in the robot Golem-III is described.