A model for semantic localization

Abstract

We propose a model for Semantic Localization, i.e. establishing positional relations on meaningful objects, to enable the principled integration of heterogeneous localization clues -- such as those derived from ubiquitous sensors in the Internet of Things. Our approach is two-pronged: we consider relation-structured Phenomenal Maps alongside spatially-organized Physical Maps. Phenomenal Maps may be used to answer semantic queries about the relative position of objects without necessarily resorting to physical coordinates. Physical Maps are not restricted to purely Euclidian spaces, to the contrary we identify useful applications for topological, and metrical maps among others. We give the framework for a structured mechanism through which localization information in all these representations may be reconciled.