Seamless Interfacing: Situation Awareness through Action Recognition and Spatio-Temporal Reasoning

Abstract

Intuitive means of human-machine interaction are needed in order to facilitate seamless human-robot cooperation. Knowledge about human posture, whereabouts, and performed actions allows interpretation of the situation. Thus, expectations towards system behavior can be inferred. This work demonstrates a system in an industrial setting that combines all this information in order to achieve situation awareness. The continuous human action recognition is based on hierarchical Hidden Markov Models. For identifying and predicting human location, an approach based on potential functions is presented. The recognition results and spatial information are used in combination with a Description Logics-based reasoning system for modeling semantic interrelations, dependencies, and situations.