Knowledge Representation for Anchoring Symbolic Concepts to Perceptual Data

Perceptual anchoring is the process of creating and maintaining a connection between the sensor data corresponding to a physical object and its symbolic description. It is a subset of the general symbol grounding problem and has been investigated over the past years. In this chapter we present a method for grounding sensor data of physical objects to the corresponding semantic descriptions, in the context of cognitive robots. Specifically we investigate the challenge of establishing the connection between percepts and concepts referring to objects, their relations and properties. We examine how knowledge representation can be used together with an anchoring framework, so as to complement the meaning of percepts and support better linguistic interaction with the use of concepts. This implies that robots need to represent both their perceptual and semantic knowledge, often expressed in different abstraction levels while originating from different modalities. We focus on the integration of anchoring with a large scale knowledge base system and with perceptual routines. This integration is applied in a number of studies, which span from high-level to commonsense reasoning. A specially interesting application of anchoring is in the context of smart home and we consider it as an example of the applicability of the anchoring framework.