Semantic mapping with a probabilistic description logic

Abstract

Semantic mapping employs explicit labels to deal with sensor data in robotic mapping processes. In this paper we present a method for boosting performance of spatial mapping, through the use of a probabilistic ontology, expressed with a probabilistic description logic. Reasoning with this ontology allows segmentation and tagging of sensor data acquired by a robot during navigation; hence a robot can construct metric maps topologically. We report experiments with a real robot to validate our approach, thus moving closer to the goal of integrating mapping and semantic labeling processes. © 2010 Springer-Verlag.